Whats wrong with large filter caps for Gainclone?

[planet10]

Originally posted by JoeBob
Adding an L would smooth things out, but are there any obvious downsides I don't know about?

On a SS amp with low voltage & high current you have to worry about the voltage drop across the choke (and the power it will have to dissipate).

dave

[millwood]

Originally posted by JOE DIRT®
try the test out with a random signal (pink noise)

I don't know what additional insight one gains by doing this. Simulations, by definition, aren't real and 100% reflective of reality. But as long as it captures the key "themes" of the issues one is interested to study, they have done their jobs.

And in this case, I think pedja's simulation so far has answered the questions vs. cap size. Using pink or purple or white noises does not foundamentally or directionally change the conclusions we have so far.

It seems to me, after having gone through all the charts, that it is clear that unless there are some particular designs in the lm chip, it is unlikely that it will benefit from small PS caps.

[DrewP]

AFAIK, Pi filters are really good for class A amps where the current draw from the psu is a constant value. The whole point with a choke is to not allow sudden rushes of current through.

In an amp with a class AB output stage such as the LM3875 (although all stages up until the output stage appear to be current sourced class A stages) the need to pull a quick slug of current from the power tranny would be compromised by the presence of a choke.

Yes a choke would cut HF PSU noise but I think it would kill dynamics by destroying the fast recovery attributes of the PSU as well. (assuming that one believes in the benefits of a fast recovery PSU)

Drew

[JoeBob]

How about the use of a very small choke? Just to filter out high-frequency noise?

[janneman]

Originally posted by DrewP
AFAIK, Pi filters are really good for class A amps where the current draw from the psu is a constant value. The whole point with a choke is to not allow sudden rushes of current through.

In an amp with a class AB output stage such as the LM3875 (although all stages up until the output stage appear to be current sourced class A stages) the need to pull a quick slug of current from the power tranny would be compromised by the presence of a choke.

Yes a choke would cut HF PSU noise but I think it would kill dynamics by destroying the fast recovery attributes of the PSU as well. (assuming that one believes in the benefits of a fast recovery PSU)

Drew

Drew,

With the choke filter, it's the last cap that furnishes the "slug of current". Since the diodes only conduct 20% or less of the time, the xformer couldn't do it for 80% of the time anyway, even if there is no choke.

Jan Didden

[IanHarvey]

It might be interesting to repeat the simulations with a low-value resistor (say 1 or 2 ohms) between the diodes and the filter caps.

This will lower the mean voltage on the caps (and so the charging pulses should be longer and less spiky), but without the voltage ripple of the small capacitor version.

Cheers
IH

[Joe Rasmussen]

Originally posted by ALW


You should do an FFT of your raw PSU if you think it's just 100Hz

Andy.

I have! Or just connect a little speaker with a suitable DC blocking cap across your reservoir cap and listen to it. Then ponder that you must be listening to that minus x dB down.

Have you ever done it with a choke input PSU ? Very instructive! Nice enough to hum along to.



But this whole discussion about reservoir caps can only be resolved by doing the obvious.... try it. Even better vary its value as we often vary other parameters and then listen to it. Why should these be any different?

Personally I suspect the whole thing come down to dynamic behaviours that are very difficult to predict and even more difficult to model. The way it affects the ability to portray pitch and pace in the mid-bass area alone seems to point to a possible resonant interactions between the dynamic behaviour of the circuit and the ability of a frequency based LF PSU to hang in there. The PSU does not have a constant source impedance (even with regulators difficult to achieve) and the constantly changing load imposed on it... and a merry dance indeed!

But don't bother figuring it out when the solution is: Just experiment - it will be better for your sanity.

Batteries anyone? How about setting up a trickle charged batteries where the averaged trickle is set above the averaged out (over time) current requirement? (Then allow for suitable peak current ability and test PSU Z on peak load) I have done it to Turntable power supplies (TT using DC motors). Could work a treat?

Certainly the noise floor would be dramatically lower, that should have other benefits like reduced haze and greyness this potentially adds to the sonics of most amps.

I make some interesting tube amps that use an infinitely high AC Z, an inverse bandwidth power supply. The amp circuits rides the power supply (and hence is isolated) and it is the power supply itself that is left to sort it out. There are many ways to tackle same or different problems. Maybe that is why we like our chosen toys ?

Joe R.

[DrewP]

With the choke filter, it's the last cap that furnishes the "slug of current". Since the diodes only conduct 20% or less of the time, the xformer couldn't do it for 80% of the time anyway, even if there is no choke.

Yes, the cap provides the slug to the amplifier chip but then how quickly is it refilled?


On our 50hz (Australian) supply, if as stated (and I have every faith that this is true) the diodes conduct for 20% of each half wave (which occurs 100 times a second). So, you're trying to top up the second PSU reservior (the one after the choke) with a short term current pulse that lasts 2ms (20% of 10ms).

So once the final reservior is depleted of its charge, how quickly can it be refilled with a choke in the way?

You folks out there who love doing simulations and drawing beautiful curves and graphs for us, please simulate a 100hz square wave with 2ms on pulses and 8ms off pulses trying to flow through a series choke being fed by a low impedance transformer and show me how compromised and fubarred the output to the final psu cap after the choke is.

My prediction is that the result will not be fast refill of the reservior.

Drew

[RichardN]

Would probably work best if the amp was amplifying. Listening to the steady state noise probably isn't that instructive.

The original Gain has the transformer and diodes in a separate box. I wonder how much the characterisics of this connection scheme (impedence and inductance of the wiring etc.) have been used to alter the characteristics of the high frequency burst as seen by the power opamp.

[millwood]

Originally posted by Joe Rasmussen
But this whole discussion about reservoir caps can only be resolved by doing the obvious.... try it. Even better vary its value as we often vary other parameters and then listen to it. Why should these be any different?

there is no question an actual prototyping will give the ultimate answer but if simulation can get us 80% of it using 20% of the resources, it is a pretty good starting point.

the proliferation of simulation in recent years is a good supporting evidence of that.

Originally posted by Joe Rasmussen
Batteries anyone? How about setting up a trickle charged batteries where the averaged trickle is set above the averaged out (over time) current requirement? (Then allow for suitable peak current ability and test PSU Z on peak load) I have done it to Turntable power supplies (TT using DC motors). Could work a treat?

Joe R.

maybe the opposite is more useful in an amp. batteries have high output resistence. so one way to go might be to use a battery in parrallel with a conventional PS (isolated with a diode) that outputs slighly below or at the battery's voltage. Under slight load, the battery is the only one contributing. Great.

Under heavy load, the battery's voltage drops, and the conventional PS, with its low impendence, starts to kick in, providing the neccessary "headroom".

Originally posted by DrewP
My prediction is that the result will not be fast refill of the reservior.

Drew

I agree with your intuition. a pi filter is a low-pass filter. and it will not react quickly with a high-frequency current drain.

I will do some simulation later today.