Power Amplifier pole placement
My Hafler DH200 recently began making odd sounds, so while fixing it I've decided to modify it somewhat.
I came across some interesting threads which covered pole placement, such as "Hafler Dh 220 muffled channel" and "RC constants & power amps". The advice provided indicates that frequency of the input pole > frequency of the feedback pole > frequency of the power supply pole.
For the Hafler, I believe I can move the PS pole from 8 Hz (4 ohm load) to 3.6 Hz by increasing the filter caps to 22,000 uF.
Changing the feedback cap to 220 uF (from 470uF) should move the feedback pole to 7.2 Hz.
Finally, changing the input cap with 0.68 uF (from 10 uF) should move the input pole to 10.6 Hz.
Does all this sound about right?
Also, I have some Marantz Esotec amplifiers from the early 80's. They are DC coupled (there are separate AC coupled inputs). One has an opamp-based servo in the feedback loop; the other does not. How should the feedback pole be placed on these?
Thanks and regards,
Servo's are tricky, I leave them be. They are usually less than 0.1hz and can cause inverted output signals when the amplifier clips.
You will find that the amplifier 'sounds' different, even though it measures the same 20hz~20Khz.
Thanks for your reply, Dennis.
The DC-coupled amplifiers happen to have AC coupled inputs as well. The user can connect to either. In this case, does it make sense to move the pole associated with the input to be above the power supply pole?
Currently, the power supply pole would be at 3.6Hz, given the load.
In stock form, the input pole is at 0.3 - 0.5 Hz, with an input cap of 10 uF. Actually, this is a simple enough mod where I'll just try it and give it a listen.
Again, thanks for the information.
I would bump it up.
I always want the input pole to control the circuit.
Most speakers benefit from a cut-off on their low end anyway.
If you get nervous about the input pole becoming too high a frequency, convert it into an active filter. NAD does this on many of their amps, 10hz~15hz with a Q=1.
A quick follow-up Q
I had read somewhere that setting the input pole to a somewhat high frequency, say 14 Hz, results in phase distortion.
Is this something I should be concerned about?
yes, the phase effect goes many octaves above your chosen -3db cutoff frequency.
If your speakers cutoff about 2 octaves above your amp then IMO you should not be able to hear the phase effect, some may say you cannot hear it anyway.
I would prefer to lower the amp cutoff. I think it improves bass punch and power but only if your speakers are capable (modern low efficiency, small driver & long throw speakers cannot produce power but they can if well designed go low but only at low SPLs).
Lowering the input pole to a frequency lower than either the power supply pole of the feedback loop pole causes waveform envelope distortion.
The point of this excercise was to eliminate this.
Some notes from others:
Posted by Jon Risch ( B ) on April 18, 2004 at 15:24:15
In Reply to: OT - Can class H and some hollow-state lose close-spaced kick-drum pulses? posted by freddyi on April 18, 2004 at 12:54:38:
But rather the amp's AC coupling and how well it has been implemented.
If there are too many RC poles stacked up for LF roll-off (not that uncommon these days), then bass definition can suffer, and the bass notes wil run together at some frequency or rep rate.
It is basically about envelope distortion, something you have never seen tested in a retail consumer magazine (or very many pro publications either).
for more on this.
A simple cure is to use very large RC time constants, and stagger them by at least an octave apart.
Posted by djk ( M ) on April 18, 2004 at 20:06:39
In Reply to: Not the amp class or amp topology posted by Jon Risch on April 18, 2004 at 15:24:15:
"A simple cure is to use very large RC time constants, and stagger them by at least an octave apart."
I agree, and add:
the RC pole in the feedback loop must be a higher frequency than the RC pole formed by the load impedance and the power supply C, and the amplifier should have an input RC pole higher than the feedback pole.
"Most Class A DC coupled amps seem to be immune (Gee, I KNEW there was a reason that Class A sounded good!)"
Mainly because of the pseudo-regulation of the power supply that class A causes. Nelson Pass spoke to this once in a letter to Audio Amateur WRT the benefits of class A.
Operating at or near full power can cause the poles to 'beat' against each other (when not spaced properly) and cause the loudspeaker to bottom out from the self-generated infrasonic signals.
A JBL 2235 can handle 1KW clean program material, but bottom out with a 100W amplifier driven hard. While watching the woofer cone 'jerk', you could actually see the RC time constants in the supply and feedback loop EVEN THOUGH the input to the amplifier was high pass filtered at 20hz.
Posted by Jon Risch (B) on August 05, 2002 at 20:34:13
In Reply to: A question for Mr. John Curl posted by Jerry Parker on August 04, 2002 at 08:13:41:
Now, one could always come back and say something along the lines that an amplifier that distorts the waveform envelope is not properly designed. This would be true. It would also indicate an awful lot of "improper design" must be going on.
PRAT in Amplifiers
Just as with audio cables, many of the subtle things that might get lost in a lesser cable, can get through with a good one.
What does an amplifier DO to a signal?
It is not JUST passing it along, no, it is actually amplifying it, making it 'bigger and stronger' than it was.
It would seem that if we measured the THD and the FR, that that this would satify any requirement for linearity. Unfortunately, as many have discovered, these measurements, or even a dozen more traditional review measurements, do not fully characterize what the amp is doing with a dynamic, almost constantly changing musical signal.
An amp that preserves PRAT will not lose that almost indefinable sense of continuity to the music, a sense that the players are all together, and 'in the groove'. If you have never been a musician, then perhaps this will not make very much sense to you, the phrase "in the groove' will have very little meaning for you.
If we look at what exactly constitutes an amp that preserves PRAT, I think we will find that it has excellent dynamic behavior, it does not allow the waveform to be corrupted in the time domain.
How might this occur you ask? Well, rather than look at the square wave response leading edge, or to the amp's HF transient abilities, my personal feeling (hey, there ARE no meters that currently exist to measure PRAT, no agreed upon tests, no measurements from a textbook, just as there are no such measurements for imaging depth, or measurements for a sense of space around each instrument, etc.) is that it relates to the low frequency transient response of the amp, and how well behaved the unit is in terms of preserving the musical envelope.
If you examine an amplifier's LF transient response, you can do so in several ways.
One is to use a very low frequency square wave, say 1 Hz. This will show you what the time constants of all the coupling caps or servo loops is. Some amps have terrible response on this type of signal, the wavefrom exhibits a huge amount of overshoot and ringing at LF's, due to stacked coupling poles or poor servo design.
This might seem to be a clue, and indeed, it is for an amplifier's sense of bass impact and 'punch'. Where the signal crosses the zero line (translated into the FR domain), is generally where we will hear the amp as having solid bass response down to. I have seen some amps that have this occur at 40 or 50 Hz, even though the sine wave response -3 dB point is infrasonic. The amps tend to sound like they do not have good bass response down below this, you never seem to get a sense of bass 'pressure waves' on speaker systems that are capable of such a presentation. If there is a lot of ringing, this can cause a sense of bloated or boomy bass response, the amplifier equivalent of a poorly tuned vented cabinet.
Yet even this signal is NOT going to fully explain PRAT, as the signal is symmetrical.
If we look at the LF transient response in terms of a LF tone burst of moderate length, then we may begin to see some effects due to poor interstage operating point shift, and other related issues. The tone burst may end with a small tail or DC offset.
But again, this is basically a symmetrical signal.
BUT, if we look at the response to a medium length LF tone burst that has been offset so that the negative peaks are at the zero line, and examine the amps behavior with this kind of input, in many cases you would be shocked at what you see!
The amp may slowly 'center' the burst, and then, when it ends, there is a transient generated when the tone burst stops, AS IT IS NO LONGER AT THE ZERO POINT WHEN IT ENDS. This transient will USUALLY take the same general shape as the VLF square wave decay, but not always.
What is also very interesting, is that this effect may extend up into the midrange, even a 1 kHz offset toneburst will exhibit this effect! Yes, it is related to the DC coupling (and the stability of DC operating points within the amp), or lack thereof, but the fact remains, that the amplifier has now added some very serious waveform distortion that will not show up on a steady state THD measurement at all!
This alteration of the waveforms envelope is going to be continually adding spurious transients to the music, that are related initially to the dynamic nature of the music, BUT, also have the LF time constant/s signature of the amp superimposed over the music's envelope.
This directly affects the way we percieve the rhythm and timing of the music, as well as how much of this the speaker is exposed to, and ends up with the woofer offset from it's proper position, and the consequent added distortion as the woofer tries to reproduce it's upper range correctly.
Perfectly linear woofers will not suffer much, other than to loose effective dynamic range, but poor woofers will modulate the HF content due to displacement from nominal VC centering.
I once heard a system SO BAD on an offset tone burst, that an offset 1 kHz burst was clearly modulating the 1 kHz amplitude as each burst progressed from start to finish! The combination of the amp and speaker was clearly not able to handle this kind of signal properly, and it was intruding to a very large extent on the perception of the music.
There have been several AES papers on related matters, most notably, Lavardin's work, but these may only touch on the surface of what is going on with envelope distortions in audio systems.
Unfortunately, these kinds of papers have been looked upon as having been the work of 'fringe audio kooks' by the mainstream engineering establishment, and have not received the attention they deserve.
Some amps might seem to be inherently immune to such effects, such as a DC coupled amp. However, there are DC bias points and if these shift, they can also bring in secondary effects on the amplifier, some of which may affect mostly HF content as the bandwidth swings all over the place.
Literally, the rise time of the amp is modulated by the music's envelope content! Again, none of this shows up on traditional measurements usintg steady state sine waves.
Most Class A DC coupled amps seem to be immune (Gee, I KNEW there was a reason that Class A sounded good!), and so, if they do not commit any of the other numerous errors of design that can plague audio amplifiers, they tend to sound very good in terms of PRAT.
DC servo circuits are NOT a panacea in this regard, they too have time constants, and some poorly designed servo's will exhibit similar envelope distortions, or harbor such at somewhat lower frequencies than more conventional AC coupled amps.
So Yes, Virginia, there is amplifier PRAT, and some amps have it, and some don't.
Of course, all of the above is my personal take on all of this, and I am not saying that I have the whole answer, or that this is the only or primary problem, just that it illustrates one aspect of amplifier performance that is seldom, if ever, measured or looked at, sometimes even by the designer!
|All times are GMT. The time now is 12:39 AM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.
Copyright ©1999-2016 diyAudio