diyAudio Forums Archive > Top > Other Stuff > Power Supply Design Power Supply Design for a solid state amplifier - Click HERE for Original Thread elaar Hi, I was hoping someone wouldn't mind looking at this design and commenting on it's suitability. I'm making a 140w/8ohm amplifier (Randy Sloanes favourite design in his book using L-Mosfets). I don't need to say much about it because i've written quite a bit on the design pic. This is the first time i have built a psu and so am a bit bewildered, I have made this design by basically copying from about 4 different designs I have studied. Things i was wondering include are the capacitors the best performance values, and what type of capacitor should each one be? Have I also missed out any components that could improve the filtering? Should the capacitors be bypassed, should the diodes have bypass caps and with a 500VA transformer is soft start recommended? Thank you to those that take the time to look at it and comment. Thanks, Andy pinkmouse The ground from the caps should go straight to the chassis ground, as should your speaker returns. Save your HQ ground for signal levels. elaar Ahh okay, thanks for the input I will change the design. I also meant to ask, I have quite alot of copper sheeting, and as the case isn't that large and the toroidal is large, I was firstly going to put the toroidal into an old pc psu case, and use the copper shielding to make a faraday cage for the amplifier pcbs. Should the copper cases then have a route to ground through a cermaic capacitor? (I have seen this done in some amplifiers). Cheers, Andy richie00boy In a slight contrast to Pinkmouse, I advise you use the centre of the caps as your star point, so each bridge has its own connection to the caps 0V, not shared like you have drawn it. Then a wire off this goes to chassis. All your speaker returns go to the star point not the chassis. The amp board will have a connection also to the star point which you will then take input screen/ground from, and any other 0V references on the board. You might wish to also have a 'dirty 0V' trace on the board which uses another wire back to the star point, for zobel and decoupling networks. You need another cap across the other switch. I would use 4.7nF Y-rated caps for this purpose as they fail safe. Your input and output sockets should be isolated from the case. I would not bypass very large caps with very small ones as this creates resonant circuits. A soft start is essential for 500VA, usually over 300VA you would use one. I would personally put the fuses in the amp rails not before the rectifiers. Thus each amp is fused safely then. If there is a problem with the bridge or caps the primary fuse will blow, if sized correctly and protected with a soft start. Up to you on the faraday cage, I'd be surprised if you needed it or to even put the thing in a PC PSU box, all hassle for nothing IMO. AndrewT Hi, just to offer yet another option on your PSU. Fit a soft start. F1 should be in the Live connection from the mains. Reduce the fuse rating to as low as will allow the amp to start up and run at full power. It may blow after a week or two due to fatigue from the continual reheating. If so go up one rating i.e from T3.1A to T4A. Add a second cap across the 2pole mains switch or fit real RC snubbers. 500VA will support about 300W maximum. Is your amp being designed to suit 4ohm speakers? Move F2 & F3 to after the smoothing caps. F5A in the supply rails will allow about 10Apk into your speakers. Are you designing for 4ohm speakers? F3.1A (about 50Vpk into 8r) will better suit 8ohm speakers. Single rectifiers perform just as well as dual rectifiers. Less heating with a dual but more voltage loss. 16.6mF smoothing caps in combination with 8ohm speakers (RC=133mS) give a low frequency roll off -3db at 1.2Hz. This should be about one octave lower than the input roll off frequency and about half an octave below the NFB roll off frequency. Leaving a useable LF -3db = 2.4Hz and -1db =12Hz. If you use 4ohm speakers all these frequencies should be doubled or you can double the smoothing caps to 30mF to maintain RC=120mS. Personally I would use 180mS for a bass amp or wideband amp. The PCB 100nF should be located at the highest current users on the amp PCB to bypass current glitches to ground along the VERY shortest routes. Consider adding an interference suppression cap from live to neutral, Xrated 47nF to 220nF and surge absorber (or parallel absorbers) to suit your maximum mains voltage. Consider adding interference suppression from live to ground and neutral to ground, Yrated 2n2F to 47nF. Keep the safety earth and make it PERMANENT. No option! Keep the disconnecting network from safety earth to Central Star Ground. Consider adding a ground lift switch in parallel to the other components. You will probably never need it. Connect your PSU common direct to Central Star Ground. Do not move the Central Star Ground onto the PSU common. It can be adjacent if it suits your PSU and amp locations (see later advice). Connect Amp PCB power ground to Central Star Ground. Connect all dirty returns on the PCB to the power ground. These are decoupling caps, Thiel network or Zobel if not located on the speaker terminals and bypass caps. Connect all clean returns on the amp PCB to the signal ground. Connect PCB signal ground to PCB power ground using 10r. Connect RCA input return to PCB signal ground. RCA return should be isolated from chassis. Connect signal ground to Central Star Ground. Connect speaker return direct to Central Star Ground. Connection advice:- Keep the PSU rectifier/s flow and return, transformer feed and centre tap, smoothing cap flow and return as compact as possible or alternatively keep the loop area in each circuit small by combining wires or twisting. Run a short thick wire from a PSU common to the Central Star Ground , alternatively connect a bolt through the the PSU commons and nut it down tight. On the OTHER SIDE of the nut attach all the high current returns and attach a further nut. Finally attach the signal returns to the bolt and nut tight. The bolt and intervening nuts act as VERY SHORT wires but keep the high currents of the charging side separate from the power returns and the next nut keeps the high currents of the power returns separate from the signal ground reference. Any other dirty returns (relays etc) can be conneted to the wire from CSG to the disconnecting network richie00boy Lots of good advice from Andrew, with my only disagreement being this: quote: Originally posted by AndrewT Single rectifiers perform just as well as dual rectifiers. Less heating with a dual but more voltage loss. Single rectifiers can perform well, but they can never match a dual setup. In a high power setup (anything over 300VA I would say) then I would always use dual rectifiers because of their advantages. The load up the transformer more efficiently (so it runs cooler/you can squeeze more power out of it) but most importantly they allow individual returns and a true star ground, which will result in a higher quality ground. elaar Hi, many thanks for all of the great replies, they're very helpful. I'll draw up a modified diagram showing in detail how i think all of the common and signal grounds should connect up from what you've said (I've had to read some of the replies multiple times and I think i understand it). The design in question uses the (now a bit tricky to get hold of ) 2SJ162/2SK1058 L Mosfets, this can do 140w(@8ohms) with 55v rails and 200w(@8ohms) with 68v rails witout any circuit alterations. I'm aiming at the 140w with 55rails, because: a) I'm using this as a learning exercise and have no need for a very powerful amplifier (I already have 2 huge power amps), infact I would have been happy with 60w amplifier, but... b) This is sloanes favourite amplifier and so in my eyes is a good one to do. c) I already have a 40v+40v 500VA toroidal. This means I should get a fraction under 60v (59.8v roughly) with no load, which is good because I have 8 good 10000uF capacitors already which are only rated at 63v. They won't have a particularly long life with that high a voltage but 100v ones would be too expensive for me currently. AndrewT: that is one of the reasons why I have used dual rectifiers, just to purposely drop the voltage that little bit more. Also to answer your question about whether it will be used for 4ohms, the answer is probably not. I rarely use 4 ohm speakers, and even in the off chance I do like i said the amplifier will never be used at high volumes so i think it would be okay. When it comes to soft starts, I was wondering what method people generally design? thyristor soft-start with an output muting relay, or paralleled power resistors and relays? I was thinking of incorporating a small 30-50VA transformer into the case to run auxillary circuits, maybe opamps, clip detection circuits etc.. What if I connect that transformer directly to the power switch and use power from that to activate the relay (maybe with the use of a 555) which will short paralleled resistors to the 500VA transformer? Hopefully you won't mind having a quick look at the grounding picture/modified circuit I will post a bit later to make sure I have understood all of your advice correctly. Many thanks, Andy PS sorry, the picture should not include F2+3, they are present on the amplifier pcbs, so that's that part sorted. richie00boy Soft start by relay and resistors, powered off aux supply should be fine. A simple RC charging network and transistor would do the job. jackinnj quote: Originally posted by richie00boy I would not bypass very large caps with very small ones as this creates resonant circuits. creates a circuit which will resonate if there is a sharp di/dt -- as in a digital circuit. analog is a bit different and there some things definitely to be avoided -- the bypass capacitors across the reservoir cap prevent radiated EMI (from the rectifiers) and inductively coupled EMI and RFI (from the line connection) from making its way onto the positive and negative rails. Shredly I would regulate the power supply; 140W at 40V should be well under 4A, even with 20W of overhead dissipation, and Pedja Rogic's regulated design here will actually give better performance than all those huge capacitors as well as giving you 5A to play with, which means your PSU would be capable of delivering 5A on each rail, for a total output of 400W to power your 140W RMS continuous output amplifier, which surely will peak at under 300W; thus giving you a safety factor of 33%. With a 500VA transformer, you'll have plenty of safety factor there as well, 25%, and you can probably save some money and time using a regulator instead of trying to use mondo caps to filter the ripple out; those big caps are expensive, and the LM338s are about $5 US each, compared to probably $15 or more for those great big 10,000uF caps. ST makes the 338 too, if you prefer European. ;) In addition, the LM338 regulator will probably do a better job than caps; the ripple rejection ratio is 75dB with a $2 tantalum 10uF cap bypassing the ADJ terminal to the local negative; if greater rejection is desired, you can go as high as 20uF, but National Semiconductor states that additional rejection is minimal above 20uF. You'll need to adjust the value of the 2.2K resistors from the ADJ terminals of the 338s to the local negative, to get the voltage you need. In case it's not clear, by "local negative," I mean ground on the positive half, and negative on the negative half. You should have a 3300uF or 4700uF cap to provide the peak voltage to the regulator, where you now have two 3300uF and a 10000uF. Here are the design considerations for those resistors: Assuming your toroid puts out +/- 40Vrms, and has two secondary windings, both of which are normal (output measured RMS, and providing two secondaries) for toroids, you should be getting 56.57VP (Vrms x 2^0.5), so 55VDC might be a bit high; you would have the dropout voltage of the regulators to deal with, and the drops across the regulator diodes. The dropout is 1.3V maximum, and the diodes will drop 0.4V to 0.7V depending on the type. You'll most definitely want to check to make sure that there is adequate voltage drop (1.3V absolute minimum, and many designers would recommend at least twice this, in case of temporary voltage drops from the mains) across the regulators; if there is not, they can drop out and the ripple rejection will be defeated. If you design with 2.6V, your maximum conservatively designed value will be (56.57 - ((2 x 0.7) + (2 x 1.3))) = 52.87VDC. I would recommend you shoot for 52V, which will slightly reduce your power output; you'll get an output of 132.4W RMS continuous. Your drop across the regulators is now (2.6 + .87) = 3.47V. Dropping this 3.47V, and sending the masiumu 5A, the regulators will dissipate 17.35W; there are 20W heatsinks readily (and cheaply!) available for TO-3 casings, and this would be reasonable design. The voltage output of the LM338 is 1.25V(1 + (R2/R1) + Iadj(R2). R1 is recommended to be 120 ohms; Iadj(max) (which is an error term) is 100uA, and Iadj(typ) is 45uA, but delta Iadj is stated as less than 5uA from 10mA to 5A output current, and from 3V to 35V dropped across the regulator; no minimum value is given, and since it is an error term we will set it to zero and 100uA to find our range. The output voltage is thus dependent upon R2. The ideal resistance of R2 for your 52V is therefore: (1.25V(1 + (R2/120)) + 0uA(R2) = 52V or R2 = 120((52V/1.25V) - 1) = 4872 ohms Refactoring 4.7k into the equation, we get 50.2V, a bit lower than your desired level; but if we go with 5.1k, the next higher 5% value, we get 54.375V, and we're into the dropout danger zone. At this level, 50.2V, you'll get 127W. The error term of 100uA x 120 ohms will give an additional 12mV, for a total of 50.212V; a miniscule increase. On the capacitors, one thing you should be very aware of is that overloaded capacitors can explode. I have seen a 4700uF aluminum electrolytic cap explode with sufficient force to drive pieces of the aluminum case an inch or more deep into acoustic ceiling tiles ten feet above the bench, and that's at 12V, not 55V. When we made our first linear supplies in school, we were required to use a containment around the bench, and wear a face shield and welder's apron, before applying power to them for the first time. To this day I wear at least eye protection and most often a full transparent face shield when I first put power to a newly built supply. I do not recommend 63V caps in a 55V supply; should there be a power line surge, they would have no extra capacity to deal with it, and could be damaged and even explode. Bear in mind that there is no circuitry in your supply between the secondary of the transformer and the caps that would protect them. Supply surges happen all the time; you must design your supply so that it will tolerate them. Yet another reason to use a regulator; you won't have problems affording the caps that have sufficient voltage to handle surges. Let me put it this way: a 10000uF cap is probably equivalent to a respectable portion of a hand grenade. Think about it. I would fuse the input to the transformer from the mains on the hot side, if you're running 230VAC (my guess since you seem to be from Britain) then at 2A, slo-blo of course, and fuse the output of the transformer to the rectifier bridges on the hot side at 5A, standard fuse; the LM339 has internal current-limiting circuitry that will protect the load from drawing more than 5A continuously. Make certain that you heatsink the rectifier bridges and the regulator extensively; consider a fan. If you really want to protect your amp, you can add a crowbar circuit to the output of the LM338, which will disconnect the 338 and short the DC line to the amp. I own a 10A 12V bench supply with a crowbar circuit, and can make a schematic available. The inrush preventer is a good idea. It may preserve your fuses, and will reduce the stress on the entire system, transformer, rectifiers, caps, regulators, and amplifier, as well as your speakers. That giant thump you hear when you turn it on is not a Good Thing. Safety ground is safety ground; I do not recommend putting anything but wire between the earth connection at the mains input to the facility (by "ground" or "earth," I mean the third wire that goes to the local hard earth ground where the mains enter your house- this is the safety ground- a green wire with a yellow stripe in the harmonized standard, and green in the old British standard) and the chassis; I also do not recommend tying the chassis to the neutral mains connection :bigeyes: of course. If possible, a GFCI (you call them "RCDs-" "residual current devices" in Britain) would be a good idea. But remember: RCD/GFCI only protects against current that is lost to a local earth ground; phase-to-neutral (or phase-to-phase, if you happen to be using 3-phase) can still kill you and the RCD/GFCI won't do a dang thing about it, so don't get across a mains circuit. 230VAC can ruin your whole freakin' day, kill you very quickly, and even if it doesn't the burn scars will last a lifetime. As a general practice, when designing a power supply, ensure that the components have at least 25% headroom for conditions that can cause any sort of problems if they go on for very long, 50% for conditions that will cause damage to the equipment if they go on for more than a second or so, and 100% for conditions that will cause immediate or instant failure of the equipment, or immediate or instant death or injury. The transformer can handle an overage for a little while before it becomes critical; but the rectifier diodes may be damaged and potentially cause damage elsewhere if they are exposed to something that's wrong for more than a second or so. Capacitors, to finish the lesson, can EXPLODE. Watch out. AndrewT Hi, a good regulated power supply for the output stages of a power amp is just as complex a design job as the power amp itself. To cascade the reg and power amp will require some might debugging powers that will stretch any good designer. Using a chip reg is a cop out that is unlikely to produce the goods. I would (do) rely on brute force unregulated PSUs and they allow an amp to perform simply. jackinnj quote: Originally posted by Shredly I would regulate the power supply Unless the amplifier has known poor PSRR, regulation isn't necessarily a good idea. If the output devices fail, the regulator devices can be blown up, off the shelf regulators need to be heatsinked, when you regulate you put in another dimension of non-linearity into the equation, and you turn "music power" into wasted heat in the process. wrt exploding capacitors -- check the ripple current rating on the manufacturer's datasheet -- this one went off like a shotgun shell: Shredly quote: Originally posted by AndrewT Using a chip reg is a cop out that is unlikely to produce the goods. I disagree- what is essential is that a good integrated regulator be used, and that it be adequately bypassed for stability. Early regulator designs suffered from instability and had trouble dealing with ripple when used in adjustable configurations. Up until the mid- to late-'90s, they generally had current limits of an ampere or two. However, as competitive pressure from the switched-mode designs then becoming popular increased, the linear IC manufacturers saw their sales fall off, and decided they'd better get some designs capable of more. The LM339 was a result of this effort. It handles 5A minimum continuous, typical is 8A, and 7 to 12A in half-millisecond pulses. It pushes ripple down by 75dB. It handles voltage input-to-output differentials of up to 35V, and the dropout is 3V or less in all supported temperature, current, and voltage regimes. Error voltage developed by bleed current from the adjust pin is minimized, at a maximum of 100uA. It is packaged in the venerable, reliable, well-supported and performant TO-3 package, capable of dissipating at 35C/W to the ambient without heatsinking, and with a very low thermal resistance of 1C/W to the case. (For those not familiar with thermal resistance calculations, with a maximum temperature of 125C, ambient at 20C, this is a difference of 105C, implying that the TO-3 case without heatsinking can dissipate 3W to the ambient. With a 1C/W case resistance, the device could dissipate 105W into an ideal ambient heatsink; with a well-cooled and properly thermally coupled real-world heatsink it should be capable of dissipating 50W easily, 80W with difficulty. Because there are a million different heatsinks for the TO-3, you can get about anything you can imagine.) The use of a minimal number of relatively small and inexpensive capacitors, two diodes, and two resistors, gives a highly serviceable, small, cheap, quiet, stable, mid-current, low- to mid-voltage fully regulated supply. The IC is internally protected from thermal overload, output short circuit, and the pass transistor is safe-area protected. If bypassed properly with protective diodes it is immune to almost anything that might happen in either the line or the load, with the exception of kilojoule-range power hits from lightning strikes and the like; and if you're running your amps without power protection, you're looking for trouble. I don't run anything more complicated than a can-opener without a power bar with thyristors in it. With +/- 40V rails, 5A on each rail gives a maximum supply of 400W; rated conservatively, this will run a 200W RMS audio amplifier indefinitely. At 50V, you get 500W of supply and 250W RMS conservatively. I actually intend to implement this design for my own use, and as a guitar player who plays regularly, you may be sure I will be pushing my equipment to the max on a regular basis. Any defects in this design will present themselves quite unmistakably, as I am very picky about my sound. I'll be using this supply with a Lin amp that I discussed elsewhere; the only reason it's not already implemented is because the toroids are on backorder at Digikey. I'm putting up with the power hum from the full-wave rectifier that's making it through the 6800uF caps, bypassed with .1uF ceramics, that Ampeg used in their brute-force supply design, and quite frankly it's driving me nuts, and always has. I'd never wish that hum on anyone. I just never got up the gumption to do anything about it; but since I've had to replace the output transistors, I figure the heck with it, it's already frankensteined, let's fix EVERYTHING. quote: Originally posted by AndrewT I would (do) rely on brute force unregulated PSUs and they allow an amp to perform simply. Well, I gotta ask- can you hear whether the amp's on or off without sending any signal through it? Have you measured the power hum? 6800uF is a heckuvalot of cap, and I got one on each rail, and I sure can hear it. Theoretically, that's 0.88V of ripple. About 2%. That means those two 6800uF caps are decreasing the ripple about 17dB or so- ridiculous for the price. 10,000uF would increase that... to a whopping 18dB. Gimme a 75dB regulator any day. Shredly Worth mentioning, while I was researching that last: http://sound.westhost.com/power-supplies.htm. Nice work. jackinnj LM338 max to ground potential is 40V. Shredly quote: Originally posted by jackinnj LM338 max to ground potential is 40V. No, max from input to output. From the data sheet: "Since the regulator is “floating” and sees only the input-to-output differential voltage, supplies of several hundred volts can be regulated as long as the maximum input to output differential is not exceeded, i.e., do not short-circuit output to ground." Perhaps it might be wise to add some protection on the output for a supply over 40V. It's a relatively simple circuit. richie00boy quote: Originally posted by Shredly Well, I gotta ask- can you hear whether the amp's on or off without sending any signal through it? Have you measured the power hum? 6800uF is a heckuvalot of cap, and I got one on each rail, and I sure can hear it. Theoretically, that's 0.88V of ripple. About 2%. That means those two 6800uF caps are decreasing the ripple about 17dB or so- ridiculous for the price. 10,000uF would increase that... to a whopping 18dB. Gimme a 75dB regulator any day. The unregulated caps may be a bit bigger, but nothing compared to the regulators heatsinks which are expensive, big and hot. quote: Originally posted by Shredly No, max from input to output. From the data sheet: "Since the regulator is “floating” and sees only the input-to-output differential voltage, supplies of several hundred volts can be regulated as long as the maximum input to output differential is not exceeded, i.e., do not short-circuit output to ground." Which is exactly what will happen upon power-up when the output stability caps are discharged and present a very low impedance to ground. AndrewT Hi, Shredly is entitled to his opinion. Newcomers should stick to unregulated supplies until they have sufficient experience to advance. The knowledgable will already have the skills to go regulated or not. BTW Ipk for 140W into an 8ohm speaker is about 5.92Apk. +-6m8F (puny certainly not brute force) of smoothing on each channel will sound pretty poor, particularly in the bass and/or sub-bass. I would recommend a minimum of 2mF/Apk requiring +-12mF/channel & I have a leaning towards +-3mF/Apk/channel for bass duty. My overiding smoothing capacitor selection procedure is to adopt an RC time constant for the PSU ~=160mS to 200mS requiring +-20mF to +-25mF for 8ohm speakers and +-40mF to +-50mF for 4ohm speakers. A well designed power amplifier will not become unstable nor hum with an appropriate quantity of ripple on the supply rails. jackinnj quote: Originally posted by Shredly No, max from input to output. From the data sheet: "Since the regulator is “floating” and sees only the input-to-output differential voltage, supplies of several hundred volts can be regulated as long as the maximum input to output differential is not exceeded, i.e., do not short-circuit output to ground." Perhaps it might be wise to add some protection on the output for a supply over 40V. It's a relatively simple circuit. I think you have misread the data sheet. If you want the regulator to "float" it has to float on something -- a zener diode for instance. Here's a floating variant that I use in a tube amp: I did a group buy for a hundred boards of a variant which was used in the Jung and Hollander Marantz 7 and Last PAS modifications. Here's a writeup: http://www.tech-diy.com/DIY_PWR_LastPass.htm Shredly Errrmmm, well, jackinnj, it does float on something: the LM338 requires both the 120 ohm resistor between output and adj, and a resistor from adj to ground, which adjusts the output voltage. See the data sheet's application hints section for more information. It's worth noting that the ST datasheet is considerably less informative than, and seems to have a few translation problems from, the National datasheet. As far as the regulator heatsinks, richie00boy, if one is careful to adjust the output voltage as close as reasonable given line fluctuations to the regulator brownout voltage, then one finds that the dissipation in the pass transistor (and therefore by far the bulk of the dissipation in the regulator) at 5A is in the close neighborhood of 15-20W. Use of a fingered, on-board heatsink some 3" (80mm) tall, with a fan and good attention to airflow, is sufficient, and such heatsinks are available for $1 to $2 US each, including hardware and insulation pads. The footprint is the size of perhaps a 1mF cap, but the height is considerably less than a 3300uF. It might be worth mentioning at this point that there is now an alternative to stinky, nasty, messy silicone heatsink grease; and I've now tested it in a 20A Cuk converter and my guitar amp's power amp's output stage and it works very, very nicely: it's a paraffin compound distributed by Digikey. It's a bit more expensive than the nasty silicone stuff, but it's non-toxic, non-smelly, doesn't outgas, and the application is as a stick of the material encased in a plastic holder; you just rub it onto the insulation pad and the heatsink, hit it with a heat gun for a minute, put some on the device's can's contact surface, and mount the device while the heatsink assembly is still hot but not too hot to handle. I was careful to not overstress the output transistors for the first hour or so of use, until they had been hot long enough that I felt good mechanical compliance was unquestionable, and I'd recommend this practice in general. It's one of the reasons manufacturers do burn-in before they sell a unit. That said, richie00boy, thanks for pointing out the current outrush potential problem at startup; I'll look into this before I hook it up. I think that makes the current limit circuit at the output mandatory. AndrewT, since there are two regulators each capable of delivering 5A continuous and a minimum of 7A for up to half a millisecond, with a typical value of 12A over the same period, I think that 5.92Apk should be supportable by the regulator. Note that the current limit on the regulator, bolstered by a limiting circuit at the output of the regulator circuit, should handle this smoothly, probably introducing clipping if the signal is beyond its load limits; the current limit of the device, according to the datasheet, smoothly decreases with time, rather than forcing hard shutdown. However, it is a valid point, and I'll look into it carefully, so thanks for bringing it up. I suspect this will limit my amp to below 180W total output; output amp clipping is EVIL where speakers are concerned. In addition, AndrewT, it is apparent you have great experience with power amps; I have looked over some of your other posts and you seem to consistently have excellent advice. If I could reasonably ask it of you, I'd very much appreciate a little guidance on how I might address at least some of this hum problem on my power amp; the schematic is on another thread, so instead of spamming this one with a diagram that's off-topic, I'll provide a link to the thread. jaycee indicated that the problem might potentially be coming through the Zener diode used in the biasing circuit, but didn't indicate quite how. Any advice you can provide would be most helpful and very much appreciated. Shredly OK, I'm just about done analyzing a 2-rail supply with full protection, using LM338s. It's a modification of the National "floating regulator" application note, with a 5.1V zener to hold the input-to-output voltage across the regulator to under that value, a 2N3055 to block the input voltage, and a 2N3904 to do foldback limiting on the base of the '3055. I've used a pair of 1-ohm 10W resistors, Radio Shack $1.95 each, for the current sensors. The '3055 dissipates about 20W max; the zener and '3904 are protected by a 10k resistor, limiting current to about 4.7mA total between them. The only thing I'm having trouble with is there might not be enough current to give the output I want; with a DC Beta of about 35, I might have to Darlington the '3055 to get enough current through it. I'm lookin into it. I'll post when I'm done. Keep yer eyes open. Shredly OK, here's two of 'em. The second one is a pair of LM338s with appropriate protective circuitry. When I was done with the design, I realized I could get about the same performance and a smaller size with a discrete supply, and since most of the work to do the discrete supply was already done protecting the '338s, I did that too. The one in this post is the discrete design. I invite analysis and commentary. I intend to breadboard the '338 design; the discrete design should be breadboarded carefully and operated with caution until its operating parameters are clear. I'll include my analyses following these two posts, and I'll post breadboard results when I have them. The transformer has arrived, but today is band practice so it might be a few days. Shredly ...and here's the '338-based design. Shredly OK, now here's how they work. First, the discrete one- it makes the LM338-based one easier to explain. Q1 and Q2 pass transistors are a darlinton pair of MJE3055/2N3055, basically the same transistor but a little less beefy and a little more gainy in the MJE line- it's a TO-220 instead of a TO-3 like the 2N3055. The darlington pair configuration guarantees enough gain to make sure that there is solid current flow in the main circuit. The 5.1V zener was chosen because in this value range, zeners have a very low temperature coefficient, and inherently low bulk resistance. This means that the 5.1V breakdown voltage varies almost not at all with either current or temperature. Q3 is the feedback amplifier, and R7, R8, and R9 are its voltage divider. Because this is closed-loop voltage gain, we do not need to consider the DC Betas of the transistors until we calculate to be certain there is sufficient current, and that the 2N3904s won't burn; the voltage gain is A = 1/B Where, A is the closed loop gain, and B is the feedback fraction. The feedback fraction is just B = (R9 + part of R8)/(R9 + R8 + R7) For reference, this should be calculated for R9+R8/(R9+R8+R7) and for R9/(R9 + R8 + R7) to show the limits of the adjustment. But we won't use this equation; we just need to get the closed loop gain, so let's solve for that. Rearranging, A = ((R7 + part of R8)/(R9 + the other part of R8)) + 1 And as above, this should be done for R7/(R8+ R9) and (R7 + R8)/R9, to get the limits. This gives us a lower limit of 1.132258064516129032258064516129 and an upper limit of 5.85. Now, the output voltage is Vo = (Vz + Vbe)/B Where, Vo is output voltage, Vz is the zener diode breakdown threshold voltage, 5.1V, and Vbe is the base-to-emitter voltage of Q3, about 0.65V according to the data sheet, with a max value of 0.85V. But there is an easier way to do this: Vo = A(Vz + Vbe) So we have to calculate for A = 1.13 and A = 5.85, and Vbe = 0.65V and 0.85V. The four answers are: 6.5104838709677419354838709677419V @ Vbe = 0.65V and the pot turned all the way up, 6.7369354838709677419354838709677V @ Vbe = 0.85V and the pot turned all the way up, 33.6375V @ Vbe = 0.65V and the pot turned all the way down, and 34.8075V @ Vbe = 0.85V and the pot turned all the way down. Right off the bat, I got a problem, and I know where it came from. When I did the original calculations, I used Vbe = 1.5V + 1.8V, automatically using the darlington drop and the max Vbes from the darlington pair. So R9 is too big to get high enough. We'll have to bump it up. We'll shoot for A of 8.7, to get us just over 50V; with Vin at 47V, we can't ever actually reach this, of course, but it ensures that we can get as high as possible. There is a warning to be aware of here, though; if you push the regulator too far up, you'll brown out, just like 3-terminal regulators do. I'll show you how to calculate the brownout voltage shortly. This means that (R7+R8)/R9 should be as close as possible to 7.7 (remember you have to add one!). The 5k pot is pretty standard, I use the little ones with high stability. So if we use a 1.3k for R7, and an 820 for R9, that will give us 49.93V at the top, and 7.03V at the bottom. And another warning: if you run this supply at 7.03V, don't be expecting to get 5A out of it- that would be 40V across the pass transistor, which would get you 200W dissipation out of a transistor that only can handle 115W, and that's assuming a perfect heatsink! Clearly, for the voltage levels desired by the original poster, a yet higher value for R7 and perhaps a somewhat lower value for R9 will be needed; with the demonstration above, it should be little trouble to get what you need using this regulator. If you want me to check your results, post them, and be detailed: show me your work. If you're confident enough to press on on your own, more power to you (so to speak. :D ) OK, now let's talk about another important feature of this circuit. It uses foldback current limiting, implemented using R5 and R6 as the sense resistor, R3 and R4 as the voltage divider, and Q4 to control the base voltage of Q1 and implement the foldback. The current limit in this scheme (this is standard for foldback limiting, so pay attention) is Isl = Vbe/K(R5||R6) Where, Isl is the current into a shorted load, Vbe is the base-to-emitter voltage of our 2N3904, 0.65 to 0.85 volts, K is the feedback fraction of R3 and R4, that is, R4/(R3+R4), and (R5||R6) is the value of the parallel pair of R5 and R6, that is, 0.5 ohms since they are both 1 ohm resistors. Note carefully that these are 10W resistors, for a total dissipation of 20W between them. If our current limit is to be 5A, these resistors will drop 2.5V; this means that they will dissipate 12.5W, or 6.25W each. The maximum current limit we can implement with them is therefore (20W/0.5ohm)^.5, or just over 6A. For this circuit, Isl = (.065V/(((47k/(47k + 3.6k))x(0.5ohm))) = 1.399A. You'll notice you've come out very close to my targeted value of 1.4A. At this value, with input of 47V and output of 40V, the pass transistor will dissipate 7V x 1.4A = 9.8W, eminently reasonable. Now under normal operating conditions, the total current will be: Imax = Isl + (Vo(1 - K)/K(R5||R6) Where: Imax is the maximum load current. Thus we see that we can get Imax = 1.4A + (40V(1-(47k/(47k + 3.6k))))/((47k/(47k + 3.6k))0.5ohm) = 6.13A. This is why this type of limiting is called "foldback;" the current limit decreases as the load increases beyond the programmed current limit. In this case, the programmed current limit is 6.13A, and if we are dropping 7V across the pass transistor, then it is dissipating 42.89W. This is a heavy load, but it is well within the capabilities of the 70V, 15A, 115W TO-3 cased 2N3055, if it is properly heatsinked; most likely, if this amount of current is to be drawn for any length of time, or on a regular basis, it would be best to use a fan to cool the heatsink and transistor. But at 4A, we dissipate a much more reasonable 28W, which we can expect to dissipate without a fan and with a reasonably sized (and priced!) heatsink. Now for the remainder of the supply's parts. R1 limits the current through the zener to 20.95mA, eminently reasonable for this 1W device, which dissipates 107mW under this load. R2 limits the current to Q3 to the same level, resulting in an even lighter load under normal conditions, since the transistor operates in the active region, not at the rail. And even at the rail, 107mW is by no means excessive for this 250mW device. R2 also provides current to the base of Q1, forward biasing the BE junction and taking this transistor into the active region; with a minimum gain of some 25-30 under its light loading, and with an equal minimum gain of 25-30 in Q2, we see that as little as 1mA results in the flow of a minimum of some 625mA, and with the more substantial current levels available of 10mA, we can send 6.25A out; this is at our current limit, so there will be no problem maintaining the setpoint voltage, provided we don't brown the regulator out. The functions of the filter caps are obvious enough that I won't bother unless asked. The 1N4004 diodes are to clamp the voltages positive; this prevents oscillations at startup from resulting in a blown supply, ensuring stability. The negative half, of course, uses the complements of the '3904, MJE3055, and 2N3055, and works in precisely the complementary way. Because the output voltage is controlled by the closed-loop gain, which is not determined by the DC betas of the transistors, but by the ratio of R16/(R16+R17+R18) and so forth, we do not need to worry about the fact that the 2N2955 has a rather higher DC beta than the 2N3055; it just means that the MJE2955 won't have to work quite as hard as the MJE3055 does (and the MJE3055 isn't working very hard). Comments welcome. ETA: Oh yeah, brownout. Well, you got two Vbe drops across the darlington pair, and your drop across the sense resistors; like I said above, the max drops are 1.5V and 1.8V, so that's 3.3V, and at 6.13A you'll be dropping 3.065V across the sense resistors, so that's 6.3065V maximum brownout, and more like 1.4V minimum. Our 7V looks like it leaves us a nice margin. Shredly OK, now the '338-based one. The foldback limiting works the same way, and you'll note it uses the same resistors and transistor as well. The 2N3055 stands off the voltage in excess of 5.1V above the output voltage. Note that this means that the current capacity of the transistor is compromised if the output is a lot below the input- just as in the discrete regulator. The zener diode sets the voltage from the output of the regulator chip to the base of the MJE3055; the MJE3055 drops 0.6 to 1.8V, and the 2N3055 drops a further 0.7 to 1.5V, so between 1.3 and 3.3V of the 5.1V the zener makes gets dropped across them, leaving 3.8 to 1.8V; this is below the regulator's brownout, but note that the higher Vbe(on) occurs when Vce is high- these are the worst conditions. Under more normal conditions, like the constant intended operating conditions of this regulator, this should not be a problem, however, it must be carefully breadboarded and load tested to ensure that there is sufficient headroom. If there is not, a zener rated at 1 to 1.5V more should do the trick. The foldback limiter circuit is required to protect the 2N3055 from an output short, which would otherwise overload and quickly destroy it. The current limit on the chip would not limit enough to prevent this. The 2N3055 protects the LM338 from higher voltages than it is rated for, and the transistor can handle 60V, and pass 15A, and dissipate 115W. The darlington pair again increases the current amplification of the transistor to allow sufficient current to pass. The protection diodes protect the LM338 from cischarges of C5 and C6 as in the standard circuit. The main advantage of this circuit over the discrete regulator is that it will be much quieter; perhaps as much as 15dB, or even more. I'll get into the ripple calculations, and show a few other things this circuit does that the discrete one does not, later. AndrewT Hi, can you move your foldback limiting to upstream of the regulator? Otherwise the 3V (@Iout=6Apk) drop across the sense resistors ruins the regulation. i.e. the output voltage modulates with load demand. What is the purpose of the polarity switch S2? Cap C2 is X rated and cap C1, usually two of them, is Y rated jackinnj if you are going to go to all this work -- why not use a decent error amp and reference instead of the quite (very) inadequate LM338 -- take a look at the impedance plot of this device -- Shredly quote: Originally posted by AndrewT Hi, can you move your foldback limiting to upstream of the regulator? Otherwise the 3V (@Iout=6Apk) drop across the sense resistors ruins the regulation. i.e. the output voltage modulates with load demand. ...and that's why I put it out for review. Nice suggestion, Andrew. I'll do that. There's no reason not to, and good reason to do it. quote: Originally posted by AndrewT What is the purpose of the polarity switch S2? Since I intend to use this in my amp, I included their power wiring; I didn't question it other than to briefly mentally note what I'm about to say, but since you ask, my assumption based on what I see it doing is that it's intended to cancel noise on hot or neutral, or else decouple ground from either if it's off. I've never been able to hear a difference no matter how it's set, and as a result I generally leave it off in the center position. I suspect this is as close to a ground lift as they could safely come. I should have mentioned that everything left of the transformer is how the amp is wired from the factory. Thanks for pointing this out. I will repost without this shortly. Since it seems odd to you as well, perhaps you begin to see why I question every detail of their design, with particular attention to the way they have dealt with grounding. quote: Originally posted by AndrewT Cap C2 is X rated and cap C1, usually two of them, is Y rated Correct, and I should have noted it. For those not familiar with this rating system, all capacitors on the power line side of a transformer, that is, for those who use the British terminology, the "mains" side, must be specially rated to handle power disturbances, and specially constructed so as not to expose the equipment or through it the user to dangerous line voltages. For this application, Class X2 should normally be sufficient for C2, although X1 would be more effective, and in fact X2 is installed; since I use an Isobar surge suppressor, I didn't see a need to increase to the higher surge rating of 4kV that X1 provides. Technically, X1 is considered necessary by the IEC in all applications where the capacitor is upstream (on the line or mains side) of the fuse, so again, X2 is what is required here. There is only one C1 because they provide a switch to use it for hot, neutral, or neither. It would, as Andrew points out, be more normal to provide two of them, one to hot and one to neutral. Shredly quote: Originally posted by jackinnj if you are going to go to all this work -- why not use a decent error amp and reference instead of the quite (very) inadequate LM338 -- take a look at the impedance plot of this device -- The longer I look at it, the more convinced I become that you're right. That's partly why I did the discrete design, in order to have a starting point for investigation of such options. I may well take your advice, although I have to say that I'm not sure it's worth putting a $200 power supply into my $5 amplifier (I exaggerate, but you get the idea, I'm sure). I also have to point out that as the design becomes more complex, it becomes more and more out of the reach of a beginner, which is at least partially what this thread is about; that's why I did the detailed analysis of the discrete regulator, in addition to the fact that the design as given doesn't give quite the performance that the OP indicated they need; this last because the values are appropriate to my application rather than the OP's. But of course, what's really interesting is how it will behave when I put it to the test. The 'scope tells all (at least if you know how to make it do so). :D I'll also point out that in an amplifier that's designed to handle at least some noise, as all Lin amps appear to me to be by use of a diff amp at their front end, and when both cost and construction difficulty are considerations, bootstrapping a design from a usable if not necessarily ideally designed component might perhaps be better than attempting a fully featured and therefore highly complex design, and in addition, in this application, you must admit that ANY (competent) regulation is better than none. To put it another way, while it may not be perfect, it should be more than adequate and it certainly is better than inadequate simple capacitive filtering and smoothing alone. You yourself pointed out that it wasn't short-load protected; so fine, now it's short-load protected. I'll point out again that at this juncture, due to the complexity of the design necessary to protect it, the value and complexity of power supply may now be approaching that of the amplifier. While I'm sure that this seems appropriate where the absolute acme of performance is desired, I'll point out that if an output transistor goes short, what happens to the $1.98 voltage regulator may well be a matter of little concern to the user, and I'll also point out that someone suggested using a 6A fastblow fuse and this may well be seeming a more and more attractive as the design becomes more and more complex. While the $1.98 regulator may blow and protect the $0.50 fuse in this case, it will be of little consequence compared to the destruction of the $10 worth of output transistors that caused it. I'm kind of waiting for the OP to respond and let me know whether s/he was able to make any sense of it. If so, I'm prepared to respond as appropriate to various levels of knowledge; I find that not only is it more polite to assume a reasonable level of skill, and let them tell me if I'm going over their heads, but it's also more helpful in terms of folks actually learning some ideas they can use later, rather than just another cookbook design. Speak up. Is this useful, or are you completely lost at sea? Don't be embarrassed, I spent a long time and a lot of money learning this stuff, and an even longer time (although the money generally went the other way :D ) using it, and if you're a beginner, you're doing well to even ask the question in the first place. AndrewT Hi Shredly, your value judement is right. So back to my earlier point. The hum is almost certainly due to an earthing/grounding loop/wiring error. Inserting a regulator is unlikely to be the solution that is required. Any decent amplifier should have sufficient ripple rejection to give a quiet output. Just compare all the other ClassAB threads running at the moment that complement the quietness of their amps. richie00boy I agree with Andrew. Regs might help a little, but you will be masking the underlying problem. jackinnj quote: Originally posted by richie00boy I agree with Andrew. Regs might help a little, but you will be masking the underlying problem. and as I said at the outset -- you burn watts that could otherwise be turned into music -- two identical amplifiers -- two identical setups -- ceteris paribus the one which is tuned 0.1 dB louder will sound "better" and speaking of $5 amps -- nothing warms the heart more than an inexpensive amp which sounds good -- Shredly quote: Originally posted by AndrewT Hi Shredly, The hum is almost certainly due to an earthing/grounding loop/wiring error. Inserting a regulator is unlikely to be the solution that is required. Well, you've seen the schematic, and according to your lights, the ground is wired the way you say it should be. It's a circuit board, it's kind of hard to wire it incorrectly. I've described the grounding in detail. As I told you in the other thread, there's no question of it being anything outside the power amp; it hums when shorting plugs are plugged into its input, and for that matter when the inputs are shorted on the board, to any of the several available grounds, including the one that goes to the jacks. There isn't any place except on the circuit board itself for the hum to come from; and if there's a flaw on there, it's a design flaw. Yet, it's done with a "star ground" just the way you say it should be. So fine, where's the "ground loop" or "wiring error" or "grounding error?" According to your description of how the grounds are supposed to be arranged, there isn't any. You still haven't answered how it can be a ground loop in something outside the amp board when the inputs are shorted. Let me repeat that yet again: there is no difference in the hum at the output when the inputs are shorted, and no difference when all the other boards in the amplifier are disconnected completely. There's nothing left inside the chassis except the amplifier board and the transformer and 120V wiring. The inputs are shorted. And still it hums. You have the schematic in front of you. You nor anyone else can point to a flaw in the design. Yet still it hums. The grounds are arranged precisely as you claim they should be, going separately to a star grounding point. Yet still it hums. And if it's some faulty component in the power amp, the exact same thing is wrong with both of 'em- remember, this is stereo, there are two identical amps here. And that's how it's been since the day I bought it. And I'll tell you something else: most guitar amps made back then hummed just like this one does. It's not excessive, by the standards of the day. But it's there, and it's objectionable to me. quote: Originally posted by AndrewT Any decent amplifier should have sufficient ripple rejection to give a quiet output. Just compare all the other ClassAB threads running at the moment that complement the quietness of their amps. See, this is frustrating. The schematic is right in front of you, and the grounding is done just the way you say it should be. So where's the problem? I know where the problem is, I can see it with my 'scope. It's on the rails. They have 5V of ripple on them. They didn't have to be any better than that back then to be no worse than the competition. The designers hoped to beat the competition on features, and IMO they did so. It has a really nice sounding chorus, and it has a really nice reverb, and it has a dirty and a clean preamp channel each with a 3-band EQ, and the clean channel has a treble boost. You put two 4-ohm 2x12" cabinets on this puppy and crank the input up, and it will blow the windows out of your house; it's loud enough to be painful standing in front of it in the open air within 10 feet, plenty for any gig you don't need a kW PA system for. The S/N ratio at that point is probably 90dB, and perhaps better than that, and that's pretty quiet. That's not the point. The point is, it hums when it's turned on. I'm gonna fix it. If you think there's some kind of design problem with the amplifier, point it out; everything you need is right in front of you. Want me to take some measurements? Sure, name them. Got a suggestion? I'm willing to listen. But at this point, what I hear is a mantra, not careful consideration of what you're looking at and a direction for troubleshooting. I have a direction for troubleshooting, and I'm gonna pursue it, unless I get a suggestion that makes some sense. Shredly quote: Originally posted by richie00boy I agree with Andrew. Regs might help a little, but you will be masking the underlying problem. Again, WHAT "underlying problem?" The schematic is right there, posted on this board. A complete description of the grounding arrangement has already been made. Nobody's pointing out any problems; all I keep hearing is, "it's wrong," without any indication of WHAT is wrong. Someone suggested there might be a problem with the zener diode transmitting noise into the current source for the diff amp. I bypassed the zener diode three different ways, and the noise was unchanged. It's not the zener diode. I can see what's wrong; its using an unregulated supply because the manufacturer was too cheap to put in a regulated one. And I'm gonna prove it. What this looks like to me is a means of covering your *** when it works. Again: if there's something wrong with the design of the amplifier, POINT TO IT. The schematic is available on another thread on this board, and a complete description of the grounding as well. I'll point out yet again, because we seem to be losing sight of the fact, that this is a commercial product; that it has always behaved as it does now; that the hum happens when the inputs are shorted to ground on the board; and that we're not talking about excessive hum, but about objectionable hum. IMO, power amplifiers should be seen and not heard when there is no input. And one more time for the stragglers, Naim seem to think that regulated power supplies are just fine for their amps. I agree with them. Shredly quote: Originally posted by jackinnj and as I said at the outset -- you burn watts that could otherwise be turned into music -- So what? It ain't like I'm runnin it off a battery, man. quote: Originally posted by jackinnj two identical amplifiers -- two identical setups -- ceteris paribus the one which is tuned 0.1 dB louder will sound "better" Ummm, IIRC you can get 15A out of a wall plug. At 120V. That's 1800W. We're nowhere near that, either in my amp or in the one the OP described; we're talking about 200W at the outside. So what was the limitation you had in mind again? quote: Originally posted by jackinnj and speaking of $5 amps -- nothing warms the heart more than an inexpensive amp which sounds good -- And nothing is worse than an otherwise excellent amplifier that hums when you turn it on. richie00boy Sorry to see this is causing you some distress. Sometimes it's only when you have the amp in front of you that you can click what the problem is. Unfortunately the logistics of that mean preclude much further comment on that basis from myself. Forgive me for not reading the thread again, but did you find whether the hum is 60Hz or 120Hz? I think you've already got it now, but zener diodes or other inherently 'noisy' components don't hum, they hiss. Hum comes from coupling or grounding issues. Have you 'scoped the power rails? Shredly quote: Originally posted by richie00boy Forgive me for not reading the thread again, but did you find whether the hum is 60Hz or 120Hz? It's 120Hz, just below B on the second fret of the A string on a concert-tuned guitar, plus several harmonics , the most prominent of which is at 240Hz, according to my spectrum analyzer. Most musicians erroneously call that "60Hz hum," because they don't bother to know what frequencies their instruments make. It's exactly the same hum I've ever heard from any piece of equipment that has a lousy power supply in it, or uses one of those cruddy wall warts, or when the ground is broken in an instrument cable. Exactly the same hum I've fixed scores of times by putting in a regulator, or replacing the wall wart with a proper regulated power supply; bigger caps always make it quieter, but a regulator makes it GO AWAY, at least as much as one can in a world full of electric lights. quote: Originally posted by richie00boy I think you've already got it now, but zener diodes or other inherently 'noisy' components don't hum, they hiss. Hum comes from coupling or grounding issues. And from cheap power supplies. quote: Originally posted by richie00boy Have you 'scoped the power rails? Yes. They have a few V of ripple on them, just as the theoretical calculations say a 40V full-wave rectifier with a 6700uF smoothing cap should have with a demand around what the amp is drawing. richie00boy OK I was just wondering about the possibility of a failing PSU cap. With 120Hz hum and strong 2nd harmonic it's not likely to be induced hum (would be 60Hz) from wiring. AndrewT Hi, can you confirm that you have 5Vpp of ripple on the supply rails when the input is shorted and speaker/load connected? I would expect only a few tens of mV in this condition. AndrewT Hi, which post is the schematic on. I have just gone back and cannot find it. A photo of both sides of the circuit board would be helpful, particularly the triple grounding trace. freakyone quote: Hopefully you won't mind having a quick look at the grounding picture/modified circuit I will post a bit later to make sure I have understood all of your advice correctly. Hello Andy Im interested to see what your modified unregulated circuit looks like, Any chance of posting this? Cheers Dan smithy666 In my opinion, you don't have large enough caps in the power supply - the old rule of the bigger the better. I built the Stochino fast amps (100W into 8R) and use 100,000uF per channel. I'm currently working on new anodized alumiunium cases for these amps and am going to upgrade to Mundorf caps, increasing to about 130,000uF per channel. Definitely increases dynamics and bass punch. Look at any high end amp and see what they use - usually big power supplies even for 100-200W amps. AndrewT Hi, Shredly is of the opinion that regulated supplies can improve the performance of an amplifier by design rather than the brute force low technology route that many of us follow. He is possibly right, for some excellent regulated throughout amps have impecable performance with the advantages that you just listed, probably due to the very low PSU output impedance that is achieveable with a regulated supply. However, the purpose of this thread is to find and cure a HUM problem. We cannot seem to agree what is causing the hum and until we have sufficient data to analyse, then it is going to be difficult to suggest workable solutions. richie00boy The simple fact is that some amps with 'mere' 10,000uF supplies don't hum, so there is no need IMO to go to such crazy levels of capacitance as 10x that, for this particular reason at least. MikeBettinger To clear things up for me, the hum problem that is being discussed is in the guitar amp whose schematic is posted in the other thread and not related to any of the amp regulators that are posted in this thread, correct? If this is true, is there a diagram of how the grounds are actually implemented in the guitar amp? The schematic I'm examining doesn't say much as to how all of these circuit and input/output grounds are actually implemented. It's tough to say lift the ground at the circuit imput and be doing anything other than guessing. The sequencing of the grounds and the loop layout can make all of the difference in the world. mike. AndrewT Hi Mike, you hit it on the head. We have not seen the details yet. All we have is a reverse engineered schematic that I have lost contact with. But, our poster has got himself bogged down with regulated supplies. MikeBettinger quote: Originally posted by AndrewT All we have is a reverse engineered schematic that I have lost contact with. Here's the schematic. Mike. Shredly OK, well, I did it, and it worked like I thought it would, except for a couple problems I'll get into shortly. The hum is definitely coming from the power supply, no question about it. The only hum I can hear now is coming from the reverb-- which is unshielded, another cost-cutting measure by the manufacturer, I think-- I replaced the cheapie cables to and from the reverb coil, and it helped, but there's still just enough to be irritating. Coil (aka "spring") reverbs, BTW, are infamous for picking up hum, both because they usually don't have a good faraday cage around the coils, and because you have to amplify the signal coming back from the coil a lot. I'll put a cage around it and that should take care of that. So you know, I put a florescent light about six inches above the top of the amp case, to see if the amp itself is well shielded, and properly grounded, and I could only just barely hear the difference with the light off or on, which is pretty darn good; with the original power supply in there, the hum it already made was quite a bit too loud to hear the difference, so that'll give you an idea how much hum it was. However, it appears that I don't have enough capacitance on the input- it's browning out at the top end. IOW, if I keep the gain and output level on the preamps down, it sounds awesome- but if I crank it up, it drops out if I hammer a power chord, and comes back when the chord quiets down just a bit, about 300 or 400ms into the attack; interestingly, this seems more prevalent at the top end, whereas one would expect the bass to demand more real power, being lower frequency and therefore staying on one side of the pushmepullu longer. It also has startup trouble- it won't come up properly if the speakers are connected when I turn it on, but if I disconnect the speakers, power it up, and then connect them, it works just fine. I think 5A isn't enough for this amp. I could gang two regulators to get 10A to fix the startup problem, you can either use a pair of resistors, or, if you're finicky, you can use an op amp to get them to track, and I could probably take care of the brownout problem by that and increasing the input capacitance, but if I'm gonna use two regulators I can get almost as much power (8 or 9A) in less space using a pair of 2N3055-based discrete regulators with foldback current limiting; I'll use the circuit I posted a couple pages back, more or less. Two TO-3 cans instead of four; the rest of the components except for the smoothing caps are all small, the largest other than the pass transistors will be the MJE3055s in TO-220 cases that I'll darlington with the 2N3055s to boost their beta high enough to give me the output I want. If it were just one problem or the other, I might try to fix it with the LM338s, but both problems together make the discrete regulator the right choice. Now, I'd say that means that you prolly want to avoid using the LM338s with a Lin amp above about 130 or 140W; you might get away with a monobloc 150W, but for stereo I'd stay below 75W/channel. It looks to me like the startup demands of a dual 100W amp (200W total) with two 8-ohm speakers on it are just too great for these regulators to handle. I tested the 6700uF caps from the original supply to see if they were leaky, and they were within spec according to my scope; certainly they're not having any trouble below 1kHz, and for a power supply running at 120Hz, that should be fine. They have a bit more ESR than I'm entirely happy with, but hey, these were from like 1985 or something, and this manufacturer was kinda cheap anyway, so I'm not astonished or anything. I agree that two 6700uF caps is probably not really enough for this amp. I'll probably want something a bit heftier in there, and in addition, I'm running the transformer at 46V and the caps are rated at 50V, which is kinda hinky- I'd be happier with 100V caps in there, so I'll prolly get some 10000uF 100V caps on order today. I'm kinda tired of doing everything with a face shield and heavy shirt on- it's hot down in the garage, and I'd be more comfortable with a long-sleeve T-shirt and shooting glasses. :D Oh, it's worth mentioning that I had to add two turns to the transformer to get up out of the brownout zone at 42V output- which is the lowest margin I'm comfortable with. I suspect I might be able to take one or both of them off with the discrete regulator, but I'll find that out when I bench test it. So if someone uses the LM338-based design with that transformer, with a 60W or 70W design, you'll definitely want to use a toroid rated for 30VAC RMS, in case you need to add a few turns to it. I was able to get what I needed with 14-gauge automotive primary wire; I don't have any magnet wire on hand; if it were fifteen or twenty turns, it might be worth getting some, but for two turns stranded wire oughta be fine. It's my considered opinion that you can get quieter power for the money with regulation than with massive smoothing caps, but I have to admit that it's a much easier design to just put giant caps in there. If you want serious power, you'd better be prepared to test your regulator design out before you count on it working. I certainly have the knowledge and experience to implement this type of design, but I'm now on my second iteration, and I have a backup plan if this doesn't work, and anyone who does this had better have one too. It might be a bit beyond the abilities of even an experienced amateur who isn't really ready to do power supply designs, and it's well beyond a beginner. So the OP on this thread, who is building around 100W/channel stereo would be better off with just big smoothing caps; I take it back, you guys were right, regulator design for high power amps isn't for beginners. But stick around, because I'm gonna do a design, and it'll be tested, so you might learn something, and at worst you'll have some designs to add to your cookbooks, and try out yourselves. And you'll know their limitations, as well. Shredly This is the design I'll start with. I'll probably adjust R3 and 4 and 12 and 13 to get a higher current limit, and I may go with 10A or even 25A rectifiers; I've already mentioned my thoughts on the smoothing caps. Hey, thanks for digging up the other schematic; I've been busy building or I'd have taken care of Andrew's request earlier. :D Shredly Errata: 1. The 1.2k and 820 resistors in the feedback amp divider were reversed. The 1.2k goes to + out and the 820 to - out. 2. The decoupling caps I have are 250V, not 100V. This makes little difference, but it is worth noting; I recommend 100V minimum for all caps in this design. 3. My output clamping diodes are 1N4002, not 1N4004; the 4002 is a 100V diode, 1A just like the 4004, so this makes little practical difference, but I'd recommend the 4002 as the minimum, since the 4001 is only good for 50PIV. Again, it makes little difference, but 100V minimum is the recommendation. 4. I added a 0.1uF decoupling cap to the base of the feedback amp transistor, Q3/Q7, to the negative rail. This is to provide dominant bypass, a required stability feature for this type of regulator. My 10000uF 100V smoothing caps will arrive near the end of the week. I hope to be far enough along to know whether they are really worthwhile by then. They will also serve as a backup plan in case my regulation schemes fail utterly (unlikely, but the wise engineer makes contingency plans for everything. Remember the seven Ps). Testing results: Testing showed that the original current limit of 7.52-7.95A (at .65V to .85V Vbe for Q4/Q8) is insufficient for startup. I get motorboat oscillations in the power amplifier, showing that the output impedance is insufficient. I have upgraded to 10.12A, with a short load current of 1.44A, by increasing R3/R13 to 5.1k. I also discovered high frequency oscillation of the current limit transistor, Q4/Q8, which I avoided by bypassing from emitter to collector with another 0.1uF 250V cap, and I'm delaying startup of the current limit by adding a 1uF cap to the negative rail to its base; this cap will also decouple even quite low-frequency oscillations in the sensitive base of this transistor. I have not yet determined whether another 0.1uF decoupling cap should be added to the base of Q1/Q5; I don't expect so, but I am prepared if it is necessary. This (1uF across R4/R12) is permissible to permit charging of capacitances on the amp board; the time constant is on the close order of 32ms. Over this time it should be impossible to draw sufficient current to destroy the 2N3055 pass transistors, even into a direct short to ground. The resistance of R5||R6/R14||R15 will prevent it, since there is build time for the transformer's magnetic field, and charging time constant across the input smoothing caps combined with the transformer internal resistance and the rectifier voltage drop and stray capacitance to consider as well. Testing at 10A current limit, with the listed bypass and slow-limit caps in place, will commence this evening after work. It is also worth noting that the preamp and chorus boards are taking their power from the positive rail of the power amp power; since a matched pair of 7815/7915 regulators is in use on the chorus board (where this power comes to them), it is my intent to take the input to this regulator pair from the positive input rail to the new regulator, rather than from its output, to ensure good balance between the two regulators, and to free the power amp board from having to deal with the consequences of events on these other boards other than their output. The draw is no more than 2A (limited by the capacities of the 7815 and 7915), and this should make little difference at the output, since line regulation is one of the strengths of this design. ETA: I will also probably go get a pair of 25A rectifiers this evening; 8A is a little under what I really need, and the next they have is 25A. Shredly Well, it worked, but when I was all done it dissipated too much power inside the chassis. No more dropouts, though, and no more hum. But while I was poking around, I came across the capacitance multiplier- a much simpler circuit than the full-on regulator- which reduces ripple to managable (and even infinitesimal) levels without overstressing the pass transistors. The basic principle behind the circuit is to use the DC beta of a transistor to multiply the capacitance in a capacitive-input filter. From the point of view of a computer, the regulation would be unacceptable, but for a power amplifier, you frankly want as much of the voltage from your transformer and rectifiers as you can get, to reduce dissipation in the power supply, with the minimum ripple possible. You're not worried about either line or load regulation, beyond that which the transformer can provide; the regulation provided by the transformer is more than sufficient for an audio power amplifier. What you get from this circuit is lots and lots of amperes, without excessive dissipation in the pass transistor, and with a minimal component count. The cost is that you don't have any more load or line regulation than the transformer can provide, but because that's more than enough regulation for a power amp, what you're doing is throwing away the parts you don't need. I had a look at this, but I don't think I need that much sophistication in this design; I intend to go with something much more like Ron Elliot's design. Mr. Evil's design is, however, very interesting for the extreme efficiency it allows. I'll design this circuit today and implement it; I expect I'll get noise specs about as good as from my regulator, and considerably less dissipation. When I have settled on a design, I'll publish it here. I think this might be just what the OP was really looking for: excellent filtering, good-enough regulation, low parts count and therefore ease of understanding and low cost, as well as low possibility of failure either in the design or the implementation. BTW, Andrew, you asked Mr. Evil how to foldback limit his design in the second thread on it, and I think I see an approach that should work; Mr. Evil might like to see it too. If one can put a normal current limit on it, one should be able to put a foldback limit on too, at the cost of higher parts count, and potential feedback effects that might need to be damped. Let me know if you'd like me to develop it a bit further. richie00boy I'll be interested to see how you get on with the Cap Mulitiplier, when speaking with Rod Elliot in some depth about it, he recommended against it's use in anything other than class-a amplifiers due to the wildly varying current demands making it pretty much as wasteful as a regulator. Shredly Here's a nice circuit diagram. The minimum apparent capacitance in the filter is about 0.25 farad; at 10A, the minimum beta of the 2N3055 is 5, and at that same 10A and beta, it's drawing about 2A from the TIP31 or 32; at 2A, the beta of the TIP31 or 32 will be 15 minimum; that's a total beta of 75. Under normal operation, with both betas in the range of 50 to 70, it should be more like about 8 farads. This should be more than sufficient to calm my ripple problems. I've eschewed the dual RC filter of Mr. Elliot's original design; if it turns out to be an issue, I'll obtain some 470uF 100V caps and set up an RC filter. This is fast, cheap, and highly performant. If closely enough matched transistors are available, the 6.8k and 5k resistors can be left out. Any handy power transistors, provided they are relatively well matched-- and I don't mean anything like as well matched as they should be for a pushmipullu, same maker and part number should be fine-- should work fine for this circuit. The betas of power transistors are necessarily limited, so a Darlington pair is recommended; remember that beta often decreases with collector current, so be sure that the first transistor in the pair can handle all the current that will be required at the absolute minimum beta of the second transistor, at the maximum expected current. For the 2N3055, this absolute minimum beta is about 5, and the maximum current they can handle is 15A. I wouldn't put 15A through a 2N3055, YMMV but be careful you don't let the smoke out. I think 10A is probably pushing it, and if I expected that much load on a continuous basis, I'd be either using MOSFETs or parallelling the 2N3055s with base resistors at about 0R1, and at 5A or a bit more, like I expect, I'll be heatsinking them and making sure they're right in the main air path. Note that putting MOSFETs into this circuit just about requires Mr. Evil's circuit in the thread I referenced earlier for reasonable performance. At an absolute minimum beta of 5, I'll have to pump 1.666...A into the bases of the '3055s to get 10A out of their emitters (and pull 8.333...A into their collectors at the same time). Thus, the front transistors (TIP31/32s, MJE3055/2955s, whatever you happen to have on hand) must be able to handle this current. Be sure to check that the dissipation in the resistors that connect the bases of the front transistors can handle the current to the bases of the '3055s divided by the front transistors' minimum betas, too- in this case, that's a minimum beta of around 15, so you'll need 5W resistors (huh!?!? hope you were paying attention!) to be absolutely certain you can handle whatever this thing dishes out- it'll be drawing 111mA, and 220R dissipates 2.7W with that current. If you don't want to use 5W resistors, you can skimp by getting 150R 2W, 75R 1W, though I wouldn't go further down than that, with the predictable reduction in performance. If you're not used to thinking about transistors, you might not see why you can use the NPN 2N3055 on the negative rail, but your Darlington front has to be a PNP. The reason is that for an NPN, Vbe must be positive to turn it on, whereas for a PNP, it must be negative. The relatively obvious substitution of an NPN for the front transistor won't work, because it will never turn on; if it ever did, it would work, but it can't start because there's no source. I've breadboarded the negative circuit and tested it at 13.6V 1.36A (my test load was four 10R 25W resistors in series-parallel, for 10R and 100W, overkill perhaps but I don't really like 150C components floating around on my bench, plus they tend to scorch the test leads); I didn't want to take my amp apart again to get the 6700uF caps out, so I used an extra 3300uF cap for the input filter, C4. For the breadboard at this voltage and current, a 100R 1W resistor was fine for R2. I left out D2, since the load is static and totally resistive. I noted small defects in the output waveform coinciding with the top of the charging current waveform; they were about 0.2V. That's about 1.4% ripple, whereas with the same load directly connected to the 3300uF caps ripple is a hefty 20% or more. I suspect that the defects are due to the Vbes of the transistors. I'll bring it up to full voltage and run with nine of my test load resistors, three parallel groups of three in series, 10R 225W; at 4A 40V, dissipation will be 160W, and I have a bench fan to keep them relatively cool. Shredly quote: Originally posted by richie00boy I'll be interested to see how you get on with the Cap Mulitiplier, when speaking with Rod Elliot in some depth about it, he recommended against it's use in anything other than class-a amplifiers due to the wildly varying current demands making it pretty much as wasteful as a regulator. I'm not so sure about that. He correctly points out in the article that you have to allow for both line and load variations with a regulator, which means that the voltage drop across the pass transistor can vary by 20% or more; this makes what otherwise would be a relatively simple circuit quite complex and increases the power handling requirements for the pass transistors enormously. That's the problem I ran into; if I left enough headroom to ensure that I would never brown out, I was dropping 6 to 8V across them or even more, and at 10A that's 80W or more apiece, and remember you have to double that. That's a lot of heat. That was the problem I ran into. With just a cap multiplier, the voltage drop across the transistors is pretty much constant, and mostly dependent upon the Vbes of the transistors unless you fool with the bias point with R3/R5 (or R4/R6, depending on which side of the circuit you're looking at). This means that the most I'll ever drop will be about 3V; at 10A, that's only 30W, which a TO-3 can handle nicely with a small heatsink considering I'm using a fan, and considering I already need the fan to keep the drivers on the Lin amp cool. From the point of view of bang-for-the-buck, those 150,000uF caps you guys are on about go for $50US or so each; 2N3055s are $1, TIP31/32s are $0.50US, the 220Rs are $2US, and the 6800uF 100V caps are the enormous sum of $5US. I'm figuring I can put this together for under $30 total, plus my time, instead of the $100 or more it would cost to get something that would perform worse and introduce not only the hazard of an exploding 150,000uF cap but a component that has a limited lifetime and costs like the dickens. From a space point of view, I expect from my breadboard to have space left over, whereas I simply don't have enough room for two 150,000uF 100V caps, if I could even source them. But where the rubber meets the road is once I get it tooled up and put it in there. Will the hum be reasonable, or perhaps even gone? Only time will tell. If it works for mine, it should work for anybody's. I may wind up with a switcher yet. I hope not, but considering the time I've put in so far, it's looking more and more attractive. :P Shredly In case I wasn't clear earlier, a big thank you to Mr. Evil for both the nice MOSFET design and the commentary, both of which were most useful in evaluating this idea at the design phase (when it doesn't cost anything but skull sweat). richie00boy The cap multiplier doesn't offer any extra energy reserve, merely the same sort of (low) ripple that the much larger caps would have. In your case your amp seems very sensitive to sub-200Hz ripple (must be a shocking design IMO) so the cap multiplier might be what you're after. Please keep us updated. And whatever did happen to Mr Evil? Shredly quote: Originally posted by richie00boy The cap multiplier doesn't offer any extra energy reserve, An important point that I'd like to emphasize for the OP of this thread; you're entirely correct, richie00boy. This was why my first attempt had dropouts at high volume; I didn't have enough capacitor at the frontend of the power supply. No matter what you put in to handle ripple, you have to have enough capacitor at the frontend to supply all the current that the power amp will demand- you can't count on it being available from the transformer/rectifier at any given moment, because the wave might be down at zero. If you run out of electrons, your amp will drop out, and you'll get interesting dropouts that might even happen on, say, every third peak, causing various weird distortions of the waveform that you'll have a heck of a time diagnosing. My goal has never been to eliminate the capacitors-- just not require insanely huge and expensive ones. quote: Originally posted by richie00boy merely the same sort of (low) ripple that the much larger caps would have. In your case your amp seems very sensitive to sub-200Hz ripple (must be a shocking design IMO) so the cap multiplier might be what you're after. The design goal was to offer competition to the Roland Jazz Chorus 120. They designed it to give 140W, just a bit more than the JC120; it has a bucket-brigade chorus, like the JC120, but with depth (that's how many repeats you get- technically speaking, this is actually a clean/effect mix control) and rate (that's how often the chorus sweeps per second) for each channel which the JC120 doesn't have; it has reverb on both channels which the JC120 doesn't have; and it costs less. You get what you pay for, and in this case, they took it out of the power supply and (clearly) the power amp design, as well as not providing vibrato. The power amp therefore isn't as good at rejecting noise as it might be, and although it does give really nice high end (which for a guitar is up in the 6-10kHz area), it's a little hissy and not quite as straight-ahead clean. At the time I bought it, considering the price/performance and the options the amp offered that the JC120 just didn't have, it was a good deal; and looking back, I'd buy it again, even knowing everything I know now. But if I can significantly improve it (and I already have done so), for reasonable effort and at a reasonable cost, then it's worth my time, and besides, I enjoy fooling around with stuff like this. If I got right down to it, and spent weeks and weeks on it, I bet I could find just what's wrong with the Lin amp, but you know what? Even after all the effort I've spent, I'm still ahead of the game. And even if I wind up with a switcher, I'll still be- there's no guarantee once I found the problem in the amp that there would be any good way to fix it; and if I did, it would probably be Frankensteined, with resistors and **** hanging out all over the place. Knowing as I do now that the bulk of the problem is the ripple, I can address it without taking that risk. My design goal, therefore, is to a) remove the noise, b) without generating excessive heat, c) for a reasonable cost, d) in the available space, e) without unreasonable effort. So far, so good. I've shown that by reducing the ripple, I can eliminate the noise, and I've also "taste tested" it with a ripple-free power supply and shown that all else (heat, dropouts) aside, it really does sound significantly better in all volume regimes with reduced ripple. My last design even eliminated the dropouts, but at the expense of more heat than I'm happy with. As a final backup position before doing a switcher, I can always double-gang a pair of LM338s on each rail. But if I don't have to regulate it, why regulate it at the cost of the heat, and the extra component cost? If I can eliminate the ripple without regulating it, why not just do the minimum necessary? I think that's the question that Mr. Elliott (apologies for the earlier mispeling) was answering. quote: Originally posted by richie00boy Please keep us updated. You bet. AndrewT Hi, the linked schematic by Mrevil is a full regulator, not an improved multiplier. The multiplier can and does reduce the ripple but the transformer regulation will still allow the voltage to dip, but that's OK since it gives some SOAR protection to the output transistors. The basic multiplier you posted shows 220r and 3.3mF which will reduce the ripple significantly, However, you can reduce the RC time constant slightly and put in a cascaded RCRC smoother feeding the pass transistor with much improved ripple attenuation. 100r feeding 3.3mF then another 100r feeding a second C will respond a little quicker than 300r. The problem you should try to avoid is the pass transistor becoming fully open (saturated) and the voltage falling on the supply side. The ripple dips then pass through the pass device and you have ripple on the output. A very stiff and/or slow base feed runs into this problem when feeding large changes in output current. It is the large changes in ClassAB that causes the ripple dip feedthrough. This does not happen in substantial or fully ClassA amps. The reason ClassA benefits from the multiplier is that the PSU is supplying substantial current to feed the large output Iq when zero or tiny input signal is present and this leads to ripple on the amplifier output or very large capacitors in the PSU smoothing section (roughly two to three times that usually used for ClassAB). Bringing me, neatly, to your reference to 150mF 100V smoothing caps (where did that recommendation come from?). The recommendation I put forward, gleaned from a number of knowledgable sources, is 2mF to 3mF per Apk of amplifier output. Your 50W amplifier does not need 150mF which would imply an output capability of 50Apk to 75Apk into your normal load (impossible). and 100V caps are not required either. What is the open circuit voltage of the amplifier PSU when mains supply voltage is at maximum? A 50W amplifer might just be able to work, safely, with 50V caps but I suspect they could go over voltage on worst case conditions. Therefore almost everyone will use 63V caps. These are cheap and small. I have 10mF and 15mF 63V and they are only 35 diam by 60 long, probably not much different from the 6m8F fitted as standard. Finally, many that have fitted multiplier PSUs say they do not sound as nice as the standard PSU. That one is down to listening and your priorities. Shredly quote: Originally posted by AndrewT Hi, the linked schematic by Mrevil is a full regulator, not an improved multiplier. I'm not so sure- a regulator by definition regulates the voltage, which that circuit doesn't really do- IOW, if the line drops, the output drops, and if the load goes up, then the output from the transformer will cause the line to drop. So it really has no better line or load regulation than the transformer already did. However, it should and according to Mr. Evil's testing does drastically reduce ripple. OTOH, it also doesn't really appear to be a capacitive multiplier any more either. I think I'd let Mr. Evil dub it. quote: Originally posted by AndrewT The multiplier can and does reduce the ripple but the transformer regulation will still allow the voltage to dip, but that's OK since it gives some SOAR protection to the output transistors. That's my thinking, too. And I'll go one step farther: since the transformer's regulation when added to the existing 6.7mF smoothing caps appears to be sufficient in the designers' minds, and since I see no serious defect in the amp at high output as it stands (the hum is pretty much constant, and when the amp is putting out a lot it's overwhelmed), I have every confidence that it will still be sufficient after I am done. quote: Originally posted by AndrewT The basic multiplier you posted shows 220r and 3.3mF which will reduce the ripple significantly, However, you can reduce the RC time constant slightly and put in a cascaded RCRC smoother feeding the pass transistor with much improved ripple attenuation. 100r feeding 3.3mF then another 100r feeding a second C will respond a little quicker than 300r. I am in fact running my breadboard at the moment with a 100R 2W, having been too lazy and busy to get out and get the 220R 5W I should really be using. Now it appears that I shouldn't really be using that; how serendipitous. Glad I didn't spend any money on