Jacco shared a link to a company not too long ago which was doing something like that in at least one of their designs. I think it was these guys - Array Technologies, high-end amplifiers from the Netherlands
Front row centre isn't close enough to make out what's being said clearly. How about 'on stage' ? 😀
Why .?
Just use proper output devices and my special case is not so special most Sota amplifiers will meet this criteria , even Class -d Hypex claim this, its the dinky toy special amp so loved by many is the problem ...
My question that started this was can we do a low-z amp with low ppm , its not unusual to find speakers with low-z effective loads, look at wilson and Magico as 2 examples ...
Well look around , must be alot of amplfiers suffering from poor PSRR, 4 pr is the most common of output stage, the PSU design has to take this into concern..
Yep I agree, there indeed must be oodles of amps suffering from poor PSRR.
Otherwise if not, why do amps sound different from each other in the bass?
Otherwise if not, why do amps sound different from each other in the bass?
Your favorite brand , will require mucho mojo to have good bass ...
B&W 800 Diamond loudspeaker Measurements | Stereophile.com
They sound differently everywhere , midrange and treble changes too, is this from poor PSRR , more class -a bias changes the game again , its really a balancing act , look at how many amps have more distortion at Mw levels vs 10 watt, signal launch especially on classical music is in the Mw range , yet distortion rises at that level ...
B&W 800 Diamond loudspeaker Measurements | Stereophile.com
They sound differently everywhere , midrange and treble changes too, is this from poor PSRR , more class -a bias changes the game again , its really a balancing act , look at how many amps have more distortion at Mw levels vs 10 watt, signal launch especially on classical music is in the Mw range , yet distortion rises at that level ...
Well my hypothesis is its PSRR, what's yours?
More classA bias reduces PSU noise so makes PSRR less relevant. Distortion on single tones are you talking about?
More classA bias reduces PSU noise so makes PSRR less relevant. Distortion on single tones are you talking about?
Yes thats what i meant , more bias lower PSRR , so it changes , but if this holds true then it's a simple fix , err not ! Look at the other extreme , ClassD , does class D suffers from too much PSRR..?
Damping factor and output impedance affects the bass too, again its a sliding scale lots of variables
Damping factor and output impedance affects the bass too, again its a sliding scale lots of variables
ClassD has much worse PSU noise I reckon, but I haven't studied it closely. Certainly I have yet to see classD amps with oodles of caps on their supplies like the pics Frank sometimes shows of amps.
Sorry can't understand your last question - why would any amp have too much PSRR? ClassD amps tend to have very low PSRR as the output stage generally has none (OK then I'll allow 6dB 😉) before application of global feedback. At least an EF OPS has some by virtue of its local FB before GFB is applied.
Sorry can't understand your last question - why would any amp have too much PSRR? ClassD amps tend to have very low PSRR as the output stage generally has none (OK then I'll allow 6dB 😉) before application of global feedback. At least an EF OPS has some by virtue of its local FB before GFB is applied.
That's a plot of single tone THD+N for an amp? If so then only the part below 10W is of any relevance when playing music. Looks noise limited there.
Yes , but isnt the noise floor part of the issue, doesnt PSRR affects the noise floor , at the end of the day we are really listening to the PSU...
Transformer type , toroids are notoriously noisy ...
Transformer type , toroids are notoriously noisy ...
Yeah you're right - its not so helpful to call the 'noise floor' the 'floor' because it moves. I prefer 'noise elevator' in such circumstances.
At 10W out the PSU isn't particularly stressed out but it might be possible to work out how much of the THD+N residual comes from the supply (i.e. poor PSRR). If one had FFT plots of the residual at differing power levels that is.
At the end of the day we are indeed listening to the PSU, hence for amps to sound different they must be PSRR limited. Assuming distortion is under control which it generally is - partly because distortion is easily measured. PSRR is not so easy to quantify.
Toroids don't add more PSU noise - probably less in fact by virtue of having better regulation than other types. They add CM noise which is an entirely separate issue.
At 10W out the PSU isn't particularly stressed out but it might be possible to work out how much of the THD+N residual comes from the supply (i.e. poor PSRR). If one had FFT plots of the residual at differing power levels that is.
At the end of the day we are indeed listening to the PSU, hence for amps to sound different they must be PSRR limited. Assuming distortion is under control which it generally is - partly because distortion is easily measured. PSRR is not so easy to quantify.
Toroids don't add more PSU noise - probably less in fact by virtue of having better regulation than other types. They add CM noise which is an entirely separate issue.
According to my simulatoion, 7 pairs would be working, admittedly under worst case conditions, very near their absolute limits if asked to deliver full nominal power into 1 Ohm. This takes into account ASSUMED heating up, which is probably pessimistic, given the whale sized heat sinks Wayne has (I had to assume, as I don't know their specs, and when in doubt, I assume the worst).
In other words, I assumed a STEADY SТАТЕ delivery of 20 Amps. I realize this is probably theoretical, Wayne might hit that in short term peaks, but I feel it's good to be prapered for even the worst imaginable case. This thing will probably cost like €1,000 to build, God forbid that something should happen to burn it out, or some such.
THAT'S why I said i would be happier with 8 pairs. This offloads the power stage back into SOAR as it should be even under worst case conditions.
Of course, if one is willing to put up with some fan noise for forced air cooling to keep the heat sink temperature down, even 6 pairs could pull it off. This is another unknown - in Belgrade, where I live, winter temp is easily -10 deg C (app. 0 deg F) and summer temps will go above 40 deg C (app 104 deg F) outside. I have it easy as my air con keeps a constant temp of 24 deg C, but not everyone does. Don't know if Wayne does, so I assumed he doesn't, and he's in Florida (right, Wayne?). The only thing I don't have to think about is that a croc might wander in and slobber over the amp. 😀
And the logic is simple, the more points of heat generation I have for the same power, the less each point has to bear, and the better use I make of the heat sink.
Rick, I don't think anyone mised anything, we just didn't talk about it enough from all angles I need to consider.
And Wayne is a Storm Bringer, isn't he? 😀
And, on the subject of PSRR, I'd like to say two things.
One, please do remember that the amp is most vulnurable to these problems at its input stage, as poor PSRR results become embedded into the signal and amplified by the VAS, although one should not scoff at the power stage either. Anyway, in my case, the power supplies feeding the input stage, the VAS and the predriver are both heavily filtered and regulated. Cosequently, I should by default have a much lesser problem than normally (re. industry).
Two, my general feeling is that a lot of the sound of the amp is or is not resolved in the input stage. Make mistakes there and you might as well not bother.
Some months ago, somebody posted a message here with a schematic of 10 or 12 types of the CCS for the input stage, with figures for the PSRR factor. Briefly, it varied from 22 dB to 82 dB, a difference of merely 1,000:1. At the time, I asked that person if I could send over a few, in my view exceptionally well done CCSs, so he could put them up against what he had, but I got no answer.
So I did some studying. One of the worst results that man got was with the popular two transistor circuit, lower transistor's base to emitter of the upper transistor. If memory serves, that got the worst mark. Then I looked at some real world solutions for this circuit, which is a must with say H/K. Ah yes, but they connect it to a separate regulated and filtered line with relatively low voltage, which allows them to use small signal, low noise transistors. And so forth.
I've been doing some work on Wayne's case, The Current Factory, and I've had some very welcome help from Thorsten. As ever, some of his suggestions made me ask myself how the hell did I not see that myself, but some were quite original for me. So it changed, some BJTs were replaced by FETs, some values were very much changed, and overall I got even better results (thank you, Thorsten, much appreciated!).
The input stage CCS was 100% dumped and 100% replaced precisely because of the PSRR factor. The objective was to push it down as far as I posibly can, not just to -82 dB, but even lower if at all possible. Ideally, to eliminate it in practical terms, because when it sinks below noise and distorion produced by the circuit itself, it practically vanishes as a significant player.
Oh, this one will make me really sweat.
One, please do remember that the amp is most vulnurable to these problems at its input stage, as poor PSRR results become embedded into the signal and amplified by the VAS, although one should not scoff at the power stage either. Anyway, in my case, the power supplies feeding the input stage, the VAS and the predriver are both heavily filtered and regulated. Cosequently, I should by default have a much lesser problem than normally (re. industry).
Two, my general feeling is that a lot of the sound of the amp is or is not resolved in the input stage. Make mistakes there and you might as well not bother.
Some months ago, somebody posted a message here with a schematic of 10 or 12 types of the CCS for the input stage, with figures for the PSRR factor. Briefly, it varied from 22 dB to 82 dB, a difference of merely 1,000:1. At the time, I asked that person if I could send over a few, in my view exceptionally well done CCSs, so he could put them up against what he had, but I got no answer.
So I did some studying. One of the worst results that man got was with the popular two transistor circuit, lower transistor's base to emitter of the upper transistor. If memory serves, that got the worst mark. Then I looked at some real world solutions for this circuit, which is a must with say H/K. Ah yes, but they connect it to a separate regulated and filtered line with relatively low voltage, which allows them to use small signal, low noise transistors. And so forth.
I've been doing some work on Wayne's case, The Current Factory, and I've had some very welcome help from Thorsten. As ever, some of his suggestions made me ask myself how the hell did I not see that myself, but some were quite original for me. So it changed, some BJTs were replaced by FETs, some values were very much changed, and overall I got even better results (thank you, Thorsten, much appreciated!).
The input stage CCS was 100% dumped and 100% replaced precisely because of the PSRR factor. The objective was to push it down as far as I posibly can, not just to -82 dB, but even lower if at all possible. Ideally, to eliminate it in practical terms, because when it sinks below noise and distorion produced by the circuit itself, it practically vanishes as a significant player.
Oh, this one will make me really sweat.
If the input stage is the PSRR limitation then it can certainly be solved with extreme regulation - after all its running in classA and hence regulator output impedance isn't an issue. The output stage though is another kettle of fish because its going to be running in AB and hence you've got all those nasty haversine currents to deal with which can't be ameliorated with any kind of regulation (at least none I'm aware of).
Hence it seems to me that the OPS (and hence by extension the PSU) is where the major engineering challenges are. Unless I'm missing something vital?
Hence it seems to me that the OPS (and hence by extension the PSU) is where the major engineering challenges are. Unless I'm missing something vital?
The 164 volts is merely rail-to-rail voltage, from +82V to -82V, brochure words.
Single protection the amp has is that the front end can't swing 160 Vpp under normal circumstances.
Plenty of Perreaux amp output stages that blew up thanks to the high rail voltages, some with giant holes in the TO-3 hats.
That's a very poor design by current standards, for a great many aspects, from the input stage to the driverless output.
A decent chassis and 4 sets of well-matching lateral MOSFET triplets is all it's good for.
- Status
- Not open for further replies.