Obviously the Fetlington resistor can be whatever you like. With my tastes I suspect it would never be asked to work. 0R22 not a bad choice and 0R33 OK. Although a rather obvious idea the switch is cake and eat it solution. Often sharing the work cools things down. My conjecture is the punch of this amp would be excellent and the sweetness also. Our ears not up to stating the difference when the dumpers kick in. If Bob carver is right this will work.That is commencement of switching in class BC or G, but not this amp type.
... which is similar to the reason why you'd mount narrow tires when crossing sand dunes.
Jan
Walking on sand hey?
Great Belgian desert story. Belgian guy in National Geographic circa 1962 is in the Sahara and he starts by saying it is raining on his tent as he writes which is highly unusual. The next day he is with the Tuareg on a trip. Soon they calculate water is short. He realizes it is Ramadan and says is collecting salt so as to avoid Ramadan ? Yes it is. Water is getting low and suddenly they find a water bottle someone has dropped and all is OK. They smile and say " you see God was with us". The Belgian reporter says he had discovered the desert travelers are not the experts we suppose and no more so than him. Experts Hey?
Great Belgian desert story. Belgian guy in National Geographic circa 1962 is in the Sahara and he starts by saying it is raining on his tent as he writes which is highly unusual. The next day he is with the Tuareg on a trip. Soon they calculate water is short. He realizes it is Ramadan and says is collecting salt so as to avoid Ramadan ? Yes it is. Water is getting low and suddenly they find a water bottle someone has dropped and all is OK. They smile and say " you see God was with us". The Belgian reporter says he had discovered the desert travelers are not the experts we suppose and no more so than him. Experts Hey?
Ain't no such animal as rock steady power, unless the supply is an absolute monster,
. That said, always buy the biggest model of conventional, say, Japanese receivers, as one can, absolutely nothing to do with the claimed power of them, but merely to get the biggest transformer and power supply, .Sorry, not true. It wasn't a brilliant comparison, but I was reminded of the incident, because the concept that a unit under the lowest stress gives the best service still hold true. There's no cutting in and out wildly, the more poweful unit does its job nonchalantly, takes it in its stride; the undersized unit grinds and grinds and grinds, has far more heat buildup, and excess wear and tear shortens its life.
Last edited:
Actually, the 6550 integrated has a single point supply, with just two 15.000 uF caps for both channels.
The newer 680 integrated my wife is using has a single app. 550 VA transformer, but it has separate secondary windings for each channel, thus two rectifiers and two sets of 8,200 uF caps, one for each channel.
Citation 24 is also like that, but it trades a classic transformer for a pretty big toroid, and the filter caps are larger, 2x10,000 uF for each channel.
All this proves is that for excellent load tolerance and prodigious amounts of current a massive power supply means little unless the circuit has been done properly. If it is capable, a reasonable PSU will suffice, although I admit I would have been happier with a bit more of capacitance.
Frank, there is enough and there is the absolute.
My only objection to a few threads here is only that they tend to be a little extreme in the sense that they search for the absolute. While it can be done (more or less), the kickback is the price, which most of those discussing either could not or would not pay.
I, on the other hand, like to use a rule of thumb which has thus far never failed me, simplistic as it is. Quite simply, look at what you want from the amp worst case, then multiply the last number in Watts by 1,33 and you have the number of VA you want from your trafo.
In my case, I want 100/200/400W into 8/4/2 Ohms, so worst case number is 400. So I need 400*1.33= 532 VA. This is a non-standard value locally, so I need to take the next nearest value up (always up, never down), which is 600VA per channel.
Admittedly, even this is extreme, given that I live in a small room and own relatively efficient speakers (92 dB SPL/2.83V/1m), but better safe than sorry.
My only objection to a few threads here is only that they tend to be a little extreme in the sense that they search for the absolute. While it can be done (more or less), the kickback is the price, which most of those discussing either could not or would not pay.
I, on the other hand, like to use a rule of thumb which has thus far never failed me, simplistic as it is. Quite simply, look at what you want from the amp worst case, then multiply the last number in Watts by 1,33 and you have the number of VA you want from your trafo.
In my case, I want 100/200/400W into 8/4/2 Ohms, so worst case number is 400. So I need 400*1.33= 532 VA. This is a non-standard value locally, so I need to take the next nearest value up (always up, never down), which is 600VA per channel.
Admittedly, even this is extreme, given that I live in a small room and own relatively efficient speakers (92 dB SPL/2.83V/1m), but better safe than sorry.
This is far too general, especially in the light of the fact that the xformer is disconnected from the supply some 80% of the time, even more for bigger transformers.
About the only thing you can generalize is that a bigger xformer will give shorter, higher and more harmonics-loaded charging pulses than a smaller one.
Whether that leads to better, or worse, sound depends on a lot of things outside the supply.
Jan
All right, Jan, then give me a proposal for a better way to work it all out. I'm always open to suggestions.
Anyway, everything is always far too general until you get down to the nitty gritty of a specific case at hand.
The reason why I will go for 2 600VA toroids is because I also own my old AR94 speakers, which are much less well behaved than my regulars. Also because I want it to maintain its character even at large power swings into very unforgiving speakers which I do not have, but may run into some day.
When I was refreshing my Marantz 170DC power amp, I first used 2 15,000uF caps. When pushing it hard into the AR speakers, I did notice a slight change in tonality. Then I tried again with two 18,000 uF and the change was smaller. In the end, it got 2 22,000uF caps and now there is no change in tonality that I can detect.
Also, I was determined not to change anything from the original. From the internatioal Marantz community, I learned that the replacement transistors for the power stage the preferred units were Motorola/ON Semi MJ 21195/21196, as they allegedly changed the sound the least (origial devices gone long ago). So just in case, I boutgh a bucket full of them from Germany, with an original certificate in writing. Now, the original devices are rated at 130W each, while the NJ devices are rated at 250W each, meaning that I could have used them and modified the protection circuitry to accommodate them and approached the problem that way. But in the end, the logic "don't fix it if it works" won.
Anyway, everything is always far too general until you get down to the nitty gritty of a specific case at hand.
The reason why I will go for 2 600VA toroids is because I also own my old AR94 speakers, which are much less well behaved than my regulars. Also because I want it to maintain its character even at large power swings into very unforgiving speakers which I do not have, but may run into some day.
When I was refreshing my Marantz 170DC power amp, I first used 2 15,000uF caps. When pushing it hard into the AR speakers, I did notice a slight change in tonality. Then I tried again with two 18,000 uF and the change was smaller. In the end, it got 2 22,000uF caps and now there is no change in tonality that I can detect.
Also, I was determined not to change anything from the original. From the internatioal Marantz community, I learned that the replacement transistors for the power stage the preferred units were Motorola/ON Semi MJ 21195/21196, as they allegedly changed the sound the least (origial devices gone long ago). So just in case, I boutgh a bucket full of them from Germany, with an original certificate in writing. Now, the original devices are rated at 130W each, while the NJ devices are rated at 250W each, meaning that I could have used them and modified the protection circuitry to accommodate them and approached the problem that way. But in the end, the logic "don't fix it if it works" won.
Dejan, there was a thread here that went into it, to an excrutiating level of detail: http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size.html. Ended up with a spreadsheet by gootee - key were the observations that a simple PS is anything but ...
of detail: http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size.html. Ended up with a spreadsheet by gootee - key were the observations that a simple PS is anything but ...I will look it over, Frank, but let me remind you that ever since the 1973 publication of the Otala/Lohstroh paper, I do NOT have simple PSUs.
Their logic of using higher voltage regulated PSUs for the IPS and VAS seemed so logical to me that ever since then, I have done nothing but.
I tried a lot of combinations, and to cut a long story short, I have settled on a simple class A regulator, in form of what is effectively a capacitance multiplier with an added zener reference voltage. Just like the one Krell later used on their KSA series of amps.
The only difference is that I have included the predriver of my Locanthi three stage output to the regulated simply because in my view that sounds better. Thus, only the driver and the output stage are fed off capacitor filtered power lines.
This changes the game, in my view, a hell of a lot. And it's all on my side. The only version I have not tried yet is to replace my trusty MJE 15030/15031 regulator transistors with say MOSFETs; that could be an interesting exercise.
Their logic of using higher voltage regulated PSUs for the IPS and VAS seemed so logical to me that ever since then, I have done nothing but.
I tried a lot of combinations, and to cut a long story short, I have settled on a simple class A regulator, in form of what is effectively a capacitance multiplier with an added zener reference voltage. Just like the one Krell later used on their KSA series of amps.
The only difference is that I have included the predriver of my Locanthi three stage output to the regulated simply because in my view that sounds better. Thus, only the driver and the output stage are fed off capacitor filtered power lines.
This changes the game, in my view, a hell of a lot. And it's all on my side. The only version I have not tried yet is to replace my trusty MJE 15030/15031 regulator transistors with say MOSFETs; that could be an interesting exercise.
How do we feel about SMPS? To my horror I find they do somethings very well. One even has the chance to regulate them as 340 V 1 amp is different to 100 V 3.4 amps or whatever.
I think I just said something. A 1.5 amp regulator with override. If my Dumpers came in to kick out the regulator we have a simple symbiosis. Class A type PSU with class B kick. To link the two is not difficult and possibly not needed. The FET's with their own rectifier and 2 x 10 0000 uF would be enough. The dumpers the same as they only get nasty for odd moments. The feedback of the amp should keep it tidy up to blood from ears level. That is because the driver is still clean.
Other Fetilington thought. In usual SF MOS FET amp the advised bias is 20 to 200 mA with 100 mA as a good choice. The game is fast and thin or lush and bloated. The Fetlington should change that and give a greater range as the MOS FET will divert to driving the Bipolar and do slightly less work for a given voltage. In the ideal world the triggering a bit more sophisticated and that is parallel working class G. I doubt it is a problem. It is mostly is to give a little bit more for no great cost. On a normal speaker they will never work. On a tough speaker it will sound better as nothing is worse than an amp using some form of protection to work albeit Ron. If a CCS was used to MOS FET's and RAW DC to dumpers the job is done. The CCS is under-driven when current not required. Thus it runs cool all the time. Naim NAP 250 had a variation of this. Had they really analyzed their design they could have done it cheaper and better. It was very rugged. Reviewers said no protection circuits used, Naim did not correct them. My Talon amp did this. It sounded very soft in the best possible way. If I am honest better than my stick in the mud Quad 303 and very similar although totally different design. Identical twins from different parents with no similar DNA. Everything up to the dumpers shunt regulated. I no longer believe in that although it did no harm. As far as I can tell the output cap of a series or shunt regulator does the important work. Dejan that's our 4700 uF decouplers. No ripple so we have better choices and 105C at that. As someone said if we have 4700uF we should deliberately used high ripple domestic quality caps with high series inductance as it is a free lunch and pudding. It is a RF CLC Pi filter.
Here is a variation.
45 Watt Class B Amplifier - RED - Page150
I think I just said something. A 1.5 amp regulator with override. If my Dumpers came in to kick out the regulator we have a simple symbiosis. Class A type PSU with class B kick. To link the two is not difficult and possibly not needed. The FET's with their own rectifier and 2 x 10 0000 uF would be enough. The dumpers the same as they only get nasty for odd moments. The feedback of the amp should keep it tidy up to blood from ears level. That is because the driver is still clean.
Other Fetilington thought. In usual SF MOS FET amp the advised bias is 20 to 200 mA with 100 mA as a good choice. The game is fast and thin or lush and bloated. The Fetlington should change that and give a greater range as the MOS FET will divert to driving the Bipolar and do slightly less work for a given voltage. In the ideal world the triggering a bit more sophisticated and that is parallel working class G. I doubt it is a problem. It is mostly is to give a little bit more for no great cost. On a normal speaker they will never work. On a tough speaker it will sound better as nothing is worse than an amp using some form of protection to work albeit Ron. If a CCS was used to MOS FET's and RAW DC to dumpers the job is done. The CCS is under-driven when current not required. Thus it runs cool all the time. Naim NAP 250 had a variation of this. Had they really analyzed their design they could have done it cheaper and better. It was very rugged. Reviewers said no protection circuits used, Naim did not correct them. My Talon amp did this. It sounded very soft in the best possible way. If I am honest better than my stick in the mud Quad 303 and very similar although totally different design. Identical twins from different parents with no similar DNA. Everything up to the dumpers shunt regulated. I no longer believe in that although it did no harm. As far as I can tell the output cap of a series or shunt regulator does the important work. Dejan that's our 4700 uF decouplers. No ripple so we have better choices and 105C at that. As someone said if we have 4700uF we should deliberately used high ripple domestic quality caps with high series inductance as it is a free lunch and pudding. It is a RF CLC Pi filter.
Here is a variation.
45 Watt Class B Amplifier - RED - Page150
Last edited:
Usually for regulators bipolar best. For valves MOS FET as customer prejudice wants them and they have small advantages . They do introduce second harmonic distortion which is good or bad depending on point of view. Bad is the truth of it.
N FET's is the only choice when high volts as the action is similar to Darlington with greater simplicity. You might need 4 x BU208 to do that. Cost of that FET< $1 and is often insulated. MJE 340/350 excellent in valve amps as it is class A, 6 MHz is plenty fast enough. I never use them because the customer says no. I use them to know what might be possible. At 10 mA they seldom need a heat sink as anode CCS at 300V. That would be driving a pentode to make it think the HT is 1000V.
N FET's is the only choice when high volts as the action is similar to Darlington with greater simplicity. You might need 4 x BU208 to do that. Cost of that FET< $1 and is often insulated. MJE 340/350 excellent in valve amps as it is class A, 6 MHz is plenty fast enough. I never use them because the customer says no. I use them to know what might be possible. At 10 mA they seldom need a heat sink as anode CCS at 300V. That would be driving a pentode to make it think the HT is 1000V.
Hi Banet, nice to see you here. Dejan this is the valve side of Beograd. We just for the fun of it concocted a 250 watt 813 PP amp with SE qualities using 3(4) valves and UL inter-stage transformer. Same as kids draw aircraft. Stupid thing is it should work. It uses 2 x EL 84 at 80 mA SE 300V to get the drive and transformer to have gain. Class A2 I would guess? Just for fun and not reality. It would get too hot. Being UL the gain distortion calculation is excellent. I wrote to Danbury in UK to consider making the inter-stage design as it could be useful. My own company could do it. Our engineer was taught by ex Partridge chief engineer. Partridge second hand are, name your price transformers. It is 100 pieces so not worth it.
Andrew T has 3 phase I think? Logically thinking it should be good. Pro audio uses it as they draw most current from the wave top. Having three wave tops helps. A nice Delta supply with it's own unique Star output is great. Then no shared neutral with others to cause neutral shift. Delta side is 11 000V from memory. 50kVA a small transformer these days( excellent).
What I like is the old 220V bi-phase. Belgium, Spain, Paris. The Paris has just been ripped out I hear. USA has it. Alas corrupted by 115 V side.
Off to get some metal bashed now. £35 an hour is the bit I like, £50 is usual and these guys look competent. They have older CNC which is fine and Colchester lathe for big stuff.
Andrew T has 3 phase I think? Logically thinking it should be good. Pro audio uses it as they draw most current from the wave top. Having three wave tops helps. A nice Delta supply with it's own unique Star output is great. Then no shared neutral with others to cause neutral shift. Delta side is 11 000V from memory. 50kVA a small transformer these days( excellent).
What I like is the old 220V bi-phase. Belgium, Spain, Paris. The Paris has just been ripped out I hear. USA has it. Alas corrupted by 115 V side.
Off to get some metal bashed now. £35 an hour is the bit I like, £50 is usual and these guys look competent. They have older CNC which is fine and Colchester lathe for big stuff.
Just do it. It is so cheap. Goldmund used a multiplier which gives options. If MOS FET big options. The ripple of a Cockcroft -Walton or whatever is nasty. When low current that can be completely removed using fast Bipolar transistors. You need capacitor input version if wanting it very cheap as did Goldmund. In a TV the TV tube was the 25 kV smoothing capacitor for the Cockcroft-Walton voltage multiplier ( Tripler).
Like Goldmund I favour an ultra simple series regulator and beefy cap. KISS( keep it simple stupid, the engineers watch words). It was 40 years ago I learned of those circuits. It was my best ever day at school/college in my life. As time goes by I realize first was also best. We now have the parts to make it really work and cheap.
Last edited:
OK, so your 100W/8ohm amp must once in while drive a burst into a bad behaving speaker that drops to 2 ohms. At what frequency is that? What is the duty cycle - 2%? 5% ?
Concluding that therefor you'd want a 400W xformer is grotesk overkill.
And if the high current burst appears between two mains peaks where the diodes are cut off, your transformer may as well be on the moon anyway.
I would start with sizing the reservoir caps and how much droop you want to accept, like 2V/ 5V?
Then your xformer should be able to top up the reservoir caps in a reasonable time, say a few milliseconds. One thing that figures here is the secondary ohmic resistance, and that the again points to the xformer power.
I don't have a surefire method, but isn't that where apps like PSUDII excell?
Real music reproduction requires only very little average power.
Jan
Last edited:
Richard, can I ask you, when you say 'polar types' do you mean polarized capacitors like (most) electrolytics? I understand that non-polarised film caps can be manufactured with 'polar' foil material but that this not mean that the cap is polarized in the sense that it has a + and - terminal.
Jan
- Status
- Not open for further replies.