MikeB said:
Hi Mike,
Phase shift in professional world has other drawbacks also, because hi frequency phase-shift can cause "smearing" effect on the sound coming from multiple hi frequency drivers located in non-linear axis and also located at different angular locuses, when driven by different amplifiers with different phase shifts..
I agree with you amp having just flat ferq response in open loop,
But do you agree that phase shift at high frequency is certainly a type of Distorted version of input signal in terms of output phase changes.
regards,
Kanwar
If I'm not forgotten, the value is under 2deg at 20khz. Mr. Maynard achieve this with JLH topology, and standard 3stages topology cannot reach this.
Maybe this is one factor why JLH topology sounds different (better?) than ordinary 3 stages topology.
Kanwar, you just give a good example 😀 What happened if one "sound system" consist of amps that have different phase shift, eg one has 5deg at 20khz, one has 50deg at 20khz, etc. Wouldn't the sound just become strange?
Hi workhorse !
Oops, using different amps with different phasehifts ?
That must cause problems...
But for normal stereo at home these should be harmless, as long
as the phasehift is only a constant timedelay.
lumanauw, your 2° at 20khz, are these openloop or closedloop ?
Well, here is one point why flat openloopfreqresponse is important:
If it is not, you change the amount of feedback with frequency,
also changing the amount of phaseshiftcompensation with rising
frequency, giving dynamic phaseshift dependent on the signal...
Mike
Oops, using different amps with different phasehifts ?
That must cause problems...
But for normal stereo at home these should be harmless, as long
as the phasehift is only a constant timedelay.
lumanauw, your 2° at 20khz, are these openloop or closedloop ?
Well, here is one point why flat openloopfreqresponse is important:
If it is not, you change the amount of feedback with frequency,
also changing the amount of phaseshiftcompensation with rising
frequency, giving dynamic phaseshift dependent on the signal...
Mike
How do JLH and 3 stage topologies compare at lower frequencies?
For me, due to age, anything over 12kHz becomes academic real fast.
On top of that, extended (3 years) casual observing spectral content with an RTA patched in between preamp and amp indicates there is very little content over 8kHz. There is even less over 10kHz. This includes symphonic, jazz, rock, movies, CD, SACD, DVD-A's, DTS, 2 channel, multi-channel, etc., etc.
What's happening at 20kHz is nearly irrelavant unless it relates to stability problems or is an indicator of something going on at lower frequencies.
Hi sam9, i share your thoughts that it is irrelevant what goes on
at 20khz, no normal human hears 20khz...
But observing an amp at 20khz gives a very good idea about the
performance of an amp, means optimizing an amp at 20khz
ensures good highfrequency response.
Yes, music contains not many frequencies above 8khz, but these
are important. Try filtering out all frequencies above 8khz,
the result is a muffled sound...
My experience with constructing amps is that they easily have big
troubles with trebles, small wrong details immediately gives badly
distorted trebles, beeing a pain to the ear.
Listening to commercial amplifiers reveals that most of them have
audible defects in trebles, they are heavily distorted or simply not
present. Since i optimize my constructions for 20khz i have much
less problems with trebles.
Mike
at 20khz, no normal human hears 20khz...
But observing an amp at 20khz gives a very good idea about the
performance of an amp, means optimizing an amp at 20khz
ensures good highfrequency response.
Yes, music contains not many frequencies above 8khz, but these
are important. Try filtering out all frequencies above 8khz,
the result is a muffled sound...
My experience with constructing amps is that they easily have big
troubles with trebles, small wrong details immediately gives badly
distorted trebles, beeing a pain to the ear.
Listening to commercial amplifiers reveals that most of them have
audible defects in trebles, they are heavily distorted or simply not
present. Since i optimize my constructions for 20khz i have much
less problems with trebles.
Mike
MikeB said:
We optimize our commercial amplifier designs to get trouble free performance upto 50KHZ using N-channel Mosfets, so their is no event left for harsh trebles or distorted effects.😉😉
regards,
Kanwar
Sorry, had no intention to offend you...
With commercial i meant cheap stuff like sony and co !
These amps driving us into DIY...
Mike
With commercial i meant cheap stuff like sony and co !
These amps driving us into DIY...
Mike
"Mr. Graham Maynard mentioned the importance of minimum phase shift (at 20khz being benchmark) for good sounding power amp"
I trust by this he means phase nonlinearity, quite seperately from group delay which could be weeks as long as it's both channels!
Looking at the graph of a simple RC pole at 160KHz (R=1K, c=1nF typical of my input AND feedback filters compounding to produce a second order 100KHz -3dB filter).
From the graph it can be seen that the NONLINEARITY of phase is -1.6 degrees at 20Khz which would be the same for both channels and absolutely inaudible - equivalent to mm of head movement from the sweet spot.
It has nothing at all to do with open loop gain.
I trust by this he means phase nonlinearity, quite seperately from group delay which could be weeks as long as it's both channels!
Looking at the graph of a simple RC pole at 160KHz (R=1K, c=1nF typical of my input AND feedback filters compounding to produce a second order 100KHz -3dB filter).
From the graph it can be seen that the NONLINEARITY of phase is -1.6 degrees at 20Khz which would be the same for both channels and absolutely inaudible - equivalent to mm of head movement from the sweet spot.
It has nothing at all to do with open loop gain.
Attachments
amplifierguru said:
By this I assume you mean the deviation from an ideal linear phase characteristic that has the same zero-frequency group delay as the simulated circuit? This graph only appears to show the absolute phase shift rather than the deviation from linear phase.
Also, I believe Graham actually meant the phase shift itself, rather than the deviation from phase linearity. He uses this "first cycle distortion" metric, where he does an FFT on the first cycle of a tone burst response to a sine (not cosine) input, and computes its distortion. His posts on this caused a big stir a while back, and jcx pointed out that a simple linear first-order low-pass filter has "first cycle distortion". In another post here, I talked about why the first cycle distortion goes to zero as the bandwidth gets large for a first-order LPF. The phenomenon is related to phase shift, rather than phase linearity. I don't really agree with his use of this metric, I'm just providing this info as it gives some context to Graham's statements.
Hi andy c
OK it wasn't obvious - but the 'characteristic' frequency of a simple 1st order LP filter or pole is it's 3dB/45deg point, so, in the example graph this ocurred at 160KHz, so phase shift at 20KHz should have been 45/8 = 5.6 deg for linear phase but it showed as 7.2 deg. A difference of -1.6 deg.
My amps have 2 such lags (input and feedback) so -3.2 deg of identical phase nonlinearity ON BOTH CHANNELS. And I wont be changing a thing because it's a necessary part of my 'infinite slew factor' design.
As for the thread/post references thanks I'll take them on board and get back. But IMO sinewaves are for testing steady state not some boundary effect. Smacks of a witch-hunt.
OK it wasn't obvious - but the 'characteristic' frequency of a simple 1st order LP filter or pole is it's 3dB/45deg point, so, in the example graph this ocurred at 160KHz, so phase shift at 20KHz should have been 45/8 = 5.6 deg for linear phase but it showed as 7.2 deg. A difference of -1.6 deg.
My amps have 2 such lags (input and feedback) so -3.2 deg of identical phase nonlinearity ON BOTH CHANNELS. And I wont be changing a thing because it's a necessary part of my 'infinite slew factor' design.
As for the thread/post references thanks I'll take them on board and get back. But IMO sinewaves are for testing steady state not some boundary effect. Smacks of a witch-hunt.
I wonder, how much phase shift does 1 transistor make? For example, in VAS, how much shift does the input signal in the base than the output signal at the collector?
JLH is known for 2 stages, does it's phase accuracy is made by this 2 stages (compared to 3 stages that cannot achieve this?)
JLH topology have something 😀 Even NP is now making one variation of it.
JLH is known for 2 stages, does it's phase accuracy is made by this 2 stages (compared to 3 stages that cannot achieve this?)
JLH topology have something 😀 Even NP is now making one variation of it.
could anyone link me to Graham's articles or posts?
I thought phase shift is important in open loop, because the 'correction' goes several degrees later distorting the signal even more. Than the more distorted signal is going to be again corrected and so and so.. not only in terms of TIM.
I am very interested in others' opinions.
regards
I thought phase shift is important in open loop, because the 'correction' goes several degrees later distorting the signal even more. Than the more distorted signal is going to be again corrected and so and so.. not only in terms of TIM.
I am very interested in others' opinions.
regards
Hi lumanauw,
I have stayed out of this discussion thus far, and note that your valid observations.
I abstain from discussing the illustrated external NFB control because I am not sure that it can be related to driving back EMF generating dynamic loudspeakers.
Generally it is amplifier output current driving inabilities when driving dynamic loudspeaker loads that cause NFB loop induced error, and their effects can only be exacerbated with increased external NFB.
Hi Sam9,
I too cannot hear those higher audio frequencies in isolation, but I do hear differences in reproduction that relate only to their reproduced presence.
So it is not their spectral content that I detect, but their phase related energy contribution to the reproduced composite waveform.
This is why some audiophiles add supertweeters to their systems, like the Townshend one that does not come in until 12-15kHz !!! Thus what happens to a composite waveform at 20kHz cannot be irrelevant, or this supertweeter would not sell at its £800 per pair !!!
Hi amplifierguru,
Yes of course it must be phase non-linearity, but I am sure you already understand that amplifier 'group' propagation delay can have a bearing when global NFB is applied.
MikeB is checking for minimal open loop phase non-linearity before applying NFB. The amplifier itself muct have a much greater NFB input bandwidth than the signal input bandwidth for it to remain dynamically clean under the influence of loudspeaker system stored energy and resultant back EMF that is not directly related to input. Sometimes a passive input filter is used to 'protect' a NFB amplifier from that which it cannot alone dynamically cope with, and whether a -3dB 100kHz input filter is audible can relate more to input/NFB topology than steady state sinusoidal amplitude/phase measurements can reveal on the test bench.
Your 3.2 degree performance looks excellent.
Hi Mr Evil,
I cannot put my hand on my heart and state authoritively that there is nothing wrong with voltage drive to a loudspeaker, but it is my opinion that the only problem with voltage drive is related to voltage sensing for the NFB loop and the amplifier's consequentially propagation delayed current modifications that modulate the on-going voltage output and fail to instantaneously control the immediate voltage error.
Thus the output voltage gains waveform that is not original, and something that is effectively added and which by its nature is of higher (tweeter) frequency is likely to be much more distracting.
Hi jcx,
I note the 10uS interauaral discrimination figure and its 70 degrees equivalent at 20kHz.
This is a figure taken in passive sinusiodal isolation.
I could also say that an angular 70 deg at 20kHz represents -12dB in relative amplitude wrt 1kHz, which is the summation of what happens when you turn your head, if not the stereo resultant.
Harmonic energy and thus instrumental/vocal characteristics would be perceived quite differently when shifted almost into quadrature.
Anone can work out for themselves how audio is degraded by hf phase shift by drawing a 1khz sinewave (or squarewave for simplicity) and one of its higher high harmonics. Draw the simple summation, then add the harmonic with a phase shift or say 10uS delay. The dynamic differences in spl. will be quite different no matter what the system or loudspeaker/phones or listening environment.
Hi Andy,
I use both forward and reverse first cycle induced distortion measurement as an assessment method, in exactly the same way that square wave testing is used. Both are equally useful, but both are equally incorrect, especially from a waveform viewpoint.
It is my experience that an amplifier that copes well with a first cycle (low propagation delay and low open loop phase shift) sounds better than one that might produce a much superior steady sinewave thd figure on the test bench.
An input filter can mask poor first cycle performance and still allow good thd figures to be quoted.
The original four transistor JLH amplifier behaves almost as if it does not have NFB when first cycle reverse examined due to its low propagation delay and low open loop phase shift, and I am not aware of any other investigation method that reveals this.
Cheers ........... Graham.
I have stayed out of this discussion thus far, and note that your valid observations.
I abstain from discussing the illustrated external NFB control because I am not sure that it can be related to driving back EMF generating dynamic loudspeakers.
Generally it is amplifier output current driving inabilities when driving dynamic loudspeaker loads that cause NFB loop induced error, and their effects can only be exacerbated with increased external NFB.
Hi Sam9,
I too cannot hear those higher audio frequencies in isolation, but I do hear differences in reproduction that relate only to their reproduced presence.
So it is not their spectral content that I detect, but their phase related energy contribution to the reproduced composite waveform.
This is why some audiophiles add supertweeters to their systems, like the Townshend one that does not come in until 12-15kHz !!! Thus what happens to a composite waveform at 20kHz cannot be irrelevant, or this supertweeter would not sell at its £800 per pair !!!
Hi amplifierguru,
Yes of course it must be phase non-linearity, but I am sure you already understand that amplifier 'group' propagation delay can have a bearing when global NFB is applied.
MikeB is checking for minimal open loop phase non-linearity before applying NFB. The amplifier itself muct have a much greater NFB input bandwidth than the signal input bandwidth for it to remain dynamically clean under the influence of loudspeaker system stored energy and resultant back EMF that is not directly related to input. Sometimes a passive input filter is used to 'protect' a NFB amplifier from that which it cannot alone dynamically cope with, and whether a -3dB 100kHz input filter is audible can relate more to input/NFB topology than steady state sinusoidal amplitude/phase measurements can reveal on the test bench.
Your 3.2 degree performance looks excellent.
Hi Mr Evil,
I cannot put my hand on my heart and state authoritively that there is nothing wrong with voltage drive to a loudspeaker, but it is my opinion that the only problem with voltage drive is related to voltage sensing for the NFB loop and the amplifier's consequentially propagation delayed current modifications that modulate the on-going voltage output and fail to instantaneously control the immediate voltage error.
Thus the output voltage gains waveform that is not original, and something that is effectively added and which by its nature is of higher (tweeter) frequency is likely to be much more distracting.
Hi jcx,
I note the 10uS interauaral discrimination figure and its 70 degrees equivalent at 20kHz.
This is a figure taken in passive sinusiodal isolation.
I could also say that an angular 70 deg at 20kHz represents -12dB in relative amplitude wrt 1kHz, which is the summation of what happens when you turn your head, if not the stereo resultant.
Harmonic energy and thus instrumental/vocal characteristics would be perceived quite differently when shifted almost into quadrature.
Anone can work out for themselves how audio is degraded by hf phase shift by drawing a 1khz sinewave (or squarewave for simplicity) and one of its higher high harmonics. Draw the simple summation, then add the harmonic with a phase shift or say 10uS delay. The dynamic differences in spl. will be quite different no matter what the system or loudspeaker/phones or listening environment.
Hi Andy,
I use both forward and reverse first cycle induced distortion measurement as an assessment method, in exactly the same way that square wave testing is used. Both are equally useful, but both are equally incorrect, especially from a waveform viewpoint.
It is my experience that an amplifier that copes well with a first cycle (low propagation delay and low open loop phase shift) sounds better than one that might produce a much superior steady sinewave thd figure on the test bench.
An input filter can mask poor first cycle performance and still allow good thd figures to be quoted.
The original four transistor JLH amplifier behaves almost as if it does not have NFB when first cycle reverse examined due to its low propagation delay and low open loop phase shift, and I am not aware of any other investigation method that reveals this.
Cheers ........... Graham.
Is it that cct in post #5 is doing different thing than JLH topology (while they both can produce very small phase shift?)
So there is no way that standard 3 stages can perform like JLH?
I wanted to have JLH performance with cold (AB) amp, is it possible or just dreaming?
Hi lumanauw !
I think it's possible, of course you can't beat the JLH, it is still ClassA,
at least with my latest amp i reached this quality, according voices/
clarity/details. Of course with much more power and high damping
factor giving very tight and strong bass and very intense dynamics.
Mike
I think it's possible, of course you can't beat the JLH, it is still ClassA,
at least with my latest amp i reached this quality, according voices/
clarity/details. Of course with much more power and high damping
factor giving very tight and strong bass and very intense dynamics.
Mike
Hi, Mike,
Thanks for your little secret 😀 Does it's openloop response really flat (not descending like opamp's? Flat until what Khz?)
So, highgain design sounds good too, as long as it's open loop response is flat.
Hi, Darkfenriz,
Are we twins or what? 😀 Can think the same thing.
Besides Xover distortion, I think this is the one who resposible for the "transistor sound", especially in feedback designs.
Thanks for your little secret 😀 Does it's openloop response really flat (not descending like opamp's? Flat until what Khz?)
So, highgain design sounds good too, as long as it's open loop response is flat.
Hi, Darkfenriz,
Are we twins or what? 😀 Can think the same thing.
Besides Xover distortion, I think this is the one who resposible for the "transistor sound", especially in feedback designs.
lumanauw said:
If this is responsible then the sound coming out from the transistorised feedback designs is totally unnatural and much deviated from the orignal version...............
You're right😀 I have a dream to make audio power amp that in blind test 100% will think it is coming from live reproduction and wouldn't think even a slight if is from reproduction.
And should work with anykind of music (classical, disco, reggae, etc)
- Status
- Not open for further replies.