What is pro audio? Btw in systems theory we have the concept of mathematical structure, which smoothes over incompatible regions. We have here the incompatible regions of music performed by alive humans , the dead matter of recording amplifying reproducing , and the alive human listener.
That cannot be mapped into/onto "information theory". because "alive" cannot.
Just two notes here.
One, the transistor should ideally be actually two transistors, a say MPSA 56/06, or BC 639/640, driving somthing like MJE 15030/15031. These are 50W, 8A devices, and can sink more than enough for anything but a power amplifier. The reason why you are making in effect a discrete Darlington transistor is bandwidth; standard ready made Darlingtons usually crap out at nominally 10 MHz, but in real life that's a lot less, and this combo will nominally hit 60 MHz no problemo.
Two, the ideal topology in my experience is, on the input side, a resistor limiting the current to the zener, a zener to the ground, reference voltage via another small value resistor (say, 47...100 Ohms) shunted to ground by a 220...330 uF capacitor which rids you of the zener noise if any, and via anothe small resistor (10...?) to the base of the driver transistor. That last one is there just in case that the stabilizer tries to osciallate, which you stop by gradually increasing the value of the resistor. Don't forget to throw in some caps on the output side as well.
If you, as I do, choose to use first a larger value electrolytic, I use 2,200 uF, in parallel with say 100 nF, you end up with what Technics calls a "virtual battery". The whole circuit effectively serves to charge that last 2,200 uF cap. If your current requirements are 1A or less, any sudden power draw of 1A or less will cause the output voltage to shift by less than 1 mV, which is negligible. The upshot here is that this last large cap defines the sound quality and needs to be a good quality one, while eveyrthing preceding it is of no apprecialble consequence.
Thanks Nigel and Andrew for the inputs,
I will re ask my question on the power thread to follow as my question don't come with the same level of experience and understanding+knowledge you have both (and the others here) in relation to a general discussion. This question follows some answers Andrew gave me already in the PS threads section.
I will re ask my question on the power thread to follow as my question don't come with the same level of experience and understanding+knowledge you have both (and the others here) in relation to a general discussion. This question follows some answers Andrew gave me already in the PS threads section.
I'm not sure I follow. There are optimum currents for Zeners and tricks for thermal compensation at specific voltages. However if your load doesn't care a lot about the supply voltage maybe there are easier solutions. A current source and a resistor can be a really good voltage reference if the current source is really stable. Usually that would be much more complex than a FET, often requiring either a Zener or band-gap somewhere to create the voltage reference for the feedback current control.
I must point out that there is no single best regulator and even "best reference" must be qualified for application.
First determine how your device interacts with the power supply. The uber-regulator I did was focused on the least dynamic of loads, one that is really sensitive to supply noise (simply because a crystal oscillator can be examined to such a high degree). If the load is free of dynamics (looks like a light bulb to the power supply) then what I did is a very good solution. It may not be good for a really dynamic load, like a DSP chip or headphone amp. A suitable metaphor would be tires. The best tires for a Ferrari would be a poor choice for a big passenger sedan over bad roads and really ridiculous on an earth mover. That's why there are a myriad of tires available. Same for power supply solutions.
Demian, I completely agree that it is not all the same, and there is no one-size-fits-all solution that I am aware of.
I maintain that really good power supplies will eat up as much time as the basic circuit to make them work without a hitch. I learnt that while developing shunt power supplies for my headphone amps. Looked easy, and in effect it is a relatively simple circuit, but making it wrok well really sweated me out. And I sort of chickened out by making them about 2.5 times as powerful as I expected from the worst case load, which I assumed to be just 8 Ohms. Such headphones were used in telephone exchanges in the 40ies. Realistically, I probably wouldn't need more than say 30-40 mA, but I made them work with 150 mA - just in case.
If anyone is interested, I can post the schematics here.
There is an excellent Design Note by TI, AN-1849, on power amp supplies I advise one and all to read, Clearly written, well explained, most worthwhile reading. I'd post it, but it exceeds the permitted file size.
Your point on PSU speed is well taken. Wanting as low an output impedance as possible, initially I used ready made Darlingtons. Despite their generally good specs, I soon found out that I would be better off by making my own discrete Darlingtons, using MPSA 56/06 as drivers for MJE 15030/15031 transistors. This sounded clearly better, and the actual cost difference was truly negligible., off hand less than 50 cents per PSU line. I use this for regualting the PSU lines of power amp IPS, VAS and predrivers. Works like a charm.
I maintain that really good power supplies will eat up as much time as the basic circuit to make them work without a hitch. I learnt that while developing shunt power supplies for my headphone amps. Looked easy, and in effect it is a relatively simple circuit, but making it wrok well really sweated me out. And I sort of chickened out by making them about 2.5 times as powerful as I expected from the worst case load, which I assumed to be just 8 Ohms. Such headphones were used in telephone exchanges in the 40ies. Realistically, I probably wouldn't need more than say 30-40 mA, but I made them work with 150 mA - just in case.
If anyone is interested, I can post the schematics here.
There is an excellent Design Note by TI, AN-1849, on power amp supplies I advise one and all to read, Clearly written, well explained, most worthwhile reading. I'd post it, but it exceeds the permitted file size.
Your point on PSU speed is well taken. Wanting as low an output impedance as possible, initially I used ready made Darlingtons. Despite their generally good specs, I soon found out that I would be better off by making my own discrete Darlingtons, using MPSA 56/06 as drivers for MJE 15030/15031 transistors. This sounded clearly better, and the actual cost difference was truly negligible., off hand less than 50 cents per PSU line. I use this for regualting the PSU lines of power amp IPS, VAS and predrivers. Works like a charm.
Which characteristic of the last, large value el. cap influence final sound quality ? ESR, ESL or famous brand? There are two approaches - standard cap
after rectifier and low ESR. With low ESR caps you get a high charging current and higher magnetic field. High quality, low ESR cap after the regulator is not a good thing according to many serious designers. Single high value cap or parallel combination? Snubber or not?
I see your point of view, and I agree, but what you probably don't know is that for the last 13 years I have been making power line filters and have sold them to 28 countries around the globe.
The point is that much of what you mention you have to deal with I don't because most problematic aspects of grid power lines never even make it to my transformer, let alone after that. I have no idea what "transformer hum" even means, although, truth be told, that's also because I like to overkill with transformer size, and I have mine custom made for me.
Too many question all rolled into one.
Basically, there are two approaches to rectifiers. One advocates use od ultra fast diodes and is currently very much in vogue, and the other one advocates block rectifiers which have paired diodes by deafult and tend to have exceptional current capabilities, like 25 and 35 A (eg. KBPC-400 35). I tried the ultra fast diode way and found nothing of particular interest there, so now I'm back to block rectifiers; other people do it other ways.
Snubbers - definite yes there. You may not actually hear the difference betwees yes and no, but they can't hurt and are not expensive.
On caps - there are two groups of caps. The smaller value ones, 10 uF and below, for which I like to use Wima, Siemens or Plessey caps, sometimes also styroflex (for RIAA circuits, for example). For the large caps, after all these years and trying out any I could lay my hands on, my choice are German made Fisher & Tausche capacitors. Not the best in the world, but in my experience they sound better than any Japanese cap of same properties. If price no object, I will use Siemens Sikorel caps, and to the ebst of my knowledge, they are the best in the world, but their price is not small and they have become very hard to find after the Siemens-Matsushita manufacturing mreger.
Personally, I always use parallelled caps, space permitting. Two 10,000 uF caps in parallel will always have a smaller output impedance than a single 22,000 uF cap. Sometimes, as with vintage products, space consideration will force me to use 22,000 uF caps, as I did while refreshing my Marantz amps. I used BC Components caps and must admit I am very happy with them.
I also liked Cornell-Dubilier caps. Elna, Nichicon, Nippon Chem, Sanyo, etc always left me with somewhat thinner sounding bass lines, however, I would not hesitate using them for smaller values, at least some of their lines.
I agree such effort with power line filtering is worthwhile. I just followed "keep it simple" rule and the "keep it cheap" rule, too. I am just thrilled by those moments in music when for ex. Keith Jarret's piano gently fades , resounding the memory of every keystroke, into silence until the audience wakes up from magicland. I am annoyed by chirps and jingles leaking in from the power line from whatever of hundreds of control signals.
For keep it simple, I agree to the point least complex but still capable of achieving the desired results - unfortunately, this general idea has been just as perverted as most other good ideas and taken to the absurd.
Same for keep it cheap. I'd love to, but if I need more than is usually used (by the industry) to achieve the set goals, I will do do, otherwise I'm wasting my time. If I want standard, I can go out and buy it wherever.
In my experience, really good requires more materials than the average, not terribly so, but definitely more. I could name a few pretty good sounding amps from the industry which will use say 8 transistors for the input stage and VAS, I will use depending on the topology say 14-20, but I will obtain better results and sound than the simpler amp. Krell might use 118 transistors for the same job.
It's all pretty relative, and it does depend on what goals you set.
Same for keep it cheap. I'd love to, but if I need more than is usually used (by the industry) to achieve the set goals, I will do do, otherwise I'm wasting my time. If I want standard, I can go out and buy it wherever.
In my experience, really good requires more materials than the average, not terribly so, but definitely more. I could name a few pretty good sounding amps from the industry which will use say 8 transistors for the input stage and VAS, I will use depending on the topology say 14-20, but I will obtain better results and sound than the simpler amp. Krell might use 118 transistors for the same job.
It's all pretty relative, and it does depend on what goals you set.
Please share where it goes. My goal was - 145 dB noise levels. Unrealistic but I got somewhere. Andrew is right to say I was unconcerned with the exact layout. For my friend I have chosen LD1084. He needs 1.5 amps and the LD1084 was the better device as that extra 1.3 V is heat I don't have to deal with. The Cfbp was to be LDO also.
If I have a moment I will try the capacitance multiplier and CCS plus resistor into a Cfbp. It should equal the LD1084 in all ways and be much faster.
Sinking current. The output cap should do a good job of that???
We talked some time ago about by choice using " low grade " caps with high grade decouplers to make a Pi filter. Dejan ( Dvv ) thinks 4700 uF a good choice and I suggest 22000 uF as a good low grade. Low grade should be high ripple current. In the old world the low grade would have been seen as the right grade. My Quad 303 is low grade. My friend is walking distance from Munsdorf and knows them very well. I could use them. I don't see the need. BBH Aerovox seem OK and very good price.The old Philips big sized ones that say 200 000 hours 55 C are my favourite. They were called BC and now vishay I think ? 23 years of service is about right. 114 years is about a humans maximum and is about 1 000 000 hours.
You slanderer!
I might litigate. Just name me one single time whan 4,700 uF was good enough for me? I keep rambling about two 10,000 uF in parallel per supply line, which makes 40,000 uF per side. and that only for the current gain stage.
I ALWAYS use regulated PSU lines for the IPS, VAS and predriver which is 5-7V above the current gain stage PSU lines. I have done so since 1974 when I read Otala and Lohstroh's text in IEEE.
In my view, I need as much capacitance as it takes to provide 2 Joules of energy per every 10W of dissipated power, no matter what's the working impdance I always assume the worst case. Since I expect my amps to work into 2 Ohms, that works out to quite a lot of capacitance.
Dejan. 4700 uF would be good and x 3 per side is your prefered why I think? Anyway it is a very good use of money. The 4700 uF should be > 105C and need not be very high ripple as that job has been done.
I will looks at capitance multipllier and some type of Vbe multiplier. I have a hunch that is very workable. The idea that a regulator should be active in the way that they usually are is not the best idea. It means it is an ampliyfing element albeit not asked to do much. It is not a resistor nor a battery.
I will looks at capitance multipllier and some type of Vbe multiplier. I have a hunch that is very workable. The idea that a regulator should be active in the way that they usually are is not the best idea. It means it is an ampliyfing element albeit not asked to do much. It is not a resistor nor a battery.
WEll, if you want to hog it (if you've got it, flaunt it), you could use just one pair of say 10,000 uF centrally, but add say 2,200 uF beside each and every output device. Then, the big caps can be just about anything, but you would need top quaity smaller ones for top results. That's an intelligent use of residual inductance for your benefit.
The output devices would need next to zero time to be supplied with up to 2A each, meaning you'd resort to the big 'uns only about 1% of the time, and even then only if you're a head banger.
The output devices would need next to zero time to be supplied with up to 2A each, meaning you'd resort to the big 'uns only about 1% of the time, and even then only if you're a head banger.
Max. ripple current is directly related to heating under load is directly related to ESR. For that reason, there are no low grade-high ripple current elco's.
- Status
- Not open for further replies.