Group Delay in Amplifiers?

[Ron Resnick]

Which amplifier topology, solid-state or tube, has less group delay?

I'm sure the answer is: "it depends."

On what does it depend?

What tube circuit design might have less group delay than a transistor design?

What transistor circuit design might have less group delay than a tube design?

Do SET amplifiers have less or more group delay than complex, multi-output tube, push-pull amplifiers?

 

[Holmz]

Which amplifier topology, solid-state or tube, has less group delay?

I'm sure the answer is: "it depends."

On what does it depend?

What tube circuit design might have less group delay than a transistor design?

What transistor circuit design might have less group delay than a tube design?

Do SET amplifiers have less or more group delay than complex, multi-output tube, push-pull amplifiers?

They are all largely insignificant.
The speaker on the other hand can have group delay orders of magnitude higher, and speaker also have large phase errors, and often inverted phase.

Few are really that stressed about speaker phase, impulse response, and step function response.

Maybe for subwoofers people get concerned with group delay. And the terms “fast” and “tight” are likely referring to group delay with a sub

But for an amplifier it is a pretty linear device.

 

[DonH50]

Group delay is the (negative of the) change in phase divided by the change in frequency. Linear phase thus results in constant group delay (that is, group delay for a linear phase system is a single number, a constant). Group delay is usually related to overall bandwidth and signal coupling components. Overall group delay rarely matters for audio electronic components as there will always be some delay, but it is pretty tiny, and the difference in arrival time does not matter for a stereo system. The electrical path time from the needle (or laser, or streamer) to the speakers is normally swamped by the acoustic delay in sound. In HT systems, video processing is typically longer, so the audio is delayed to align with the picture. It is important that the delay to each speaker is adjusted so that the sound arrives at the listener at the same time to avoid corrupting the image.

With respect to electronics, tube amplifiers often have the greatest group delay due to to phase shift in the output transformer. Since that still pales in comparison to the contribution of the speakers I wouldn't worry about it at all. There are many other things that influence the sound much more than (circuit) group delay.

What perhaps matters more is the change in group delay over frequency, as that may be perceived, depending upon how much it changes over our effective resolution bandwidth (not our full-range bandwidth, but a narrower region over which we can distinguish, typically an octave or so IIRC). If you think of a wideband signal, then constant group delay means all frequencies are delayed equally, so a change in group delay is like moving the listening position without changing anything else (e.g. EQ, volume, etc.) If the group delay is not constant, then different frequencies will be delayed by different amounts, and that changes the sound we hear. The usual example is a square wave or pulse signal which has a very wide bandwidth. If group delay is constant, then the square at our ears matches the square wave from the source. If it varies with frequency, the square wave will not look like a normal square wave, as the harmonics that produce the square waveform will no longer be in phase to create a flat-top response. This is usually much easier to measure than to hear, however, and the largest culprit by far is usually the speakers themselves (mainly crossovers) and not the electronics.

There is a relevant article on ASR: https://www.audiosciencereview.com/forum/index.php?threads/group-delay-101.39333/

HTH - Don

 

[DasguteOhr]

For example, if you have an amplifier with a lot of group delay, you lose the ability to locate voices and instruments in the soundstage that your speakers create; it blurs. Your ear receives the sounds and your brain creates the illusion of the music. This is very sensitive to phase shifts. The correct phase ensures that you can correctly assign sounds to where they arise in the room. For example, a bad amplifier does not create a three-dimensional sound image; this is not due to channel separation (crosstalk). Another example of adjusting pickups is that they are often only set for the best crosstalk, which is completely wrong; the better way is to make the phase angle of both channels as equal as possible. Try it, you will be surprised.

There is no blanket statement that a transistor or a tube works better; it depends on the wiring of the amplifying elements and how many amplifier stages are involved.

 

[Holmz]

For example, if you have an amplifier with a lot of group delay,

Which amps do this?
I cannot think of a single one that has any effect above maybe 50 Hz.

… you lose the ability to locate voices and instruments in the soundstage that your speakers create; it blurs. Your ear receives the sounds and your brain creates the illusion of the music. This is very sensitive to phase shifts. The correct phase ensures that you can correctly assign sounds to where they arise in the room. ...

However the speakers dominate the phase error space.
Just look at the impulse response.
One can choose to tidy it up with say DIRAC or some other phase-EQ, but they cannot tidy it up with an amplifier.

 

[DonH50]

For example, if you have an amplifier with a lot of group delay, you lose the ability to locate voices and instruments in the soundstage that your speakers create; it blurs. Your ear receives the sounds and your brain creates the illusion of the music. This is very sensitive to phase shifts. The correct phase ensures that you can correctly assign sounds to where they arise in the room. For example, a bad amplifier does not create a three-dimensional sound image; this is not due to channel separation (crosstalk). Another example of adjusting pickups is that they are often only set for the best crosstalk, which is completely wrong; the better way is to make the phase angle of both channels as equal as possible. Try it, you will be surprised.

There is no blanket statement that a transistor or a tube works better; it depends on the wiring of the amplifying elements and how many amplifier stages are involved.

I do not believe it is the sheer magnitude of group delay that matters but rather how constant, or how nonlinear the phase response that leads to group delay not being constant over frequency. That shifts phase and delay differently for different frequencies and is what smears the image (soundstage, etc.) If the group delay is 0.1 or 1 but constant things will sound fine; if it is a function of frequency then we have (hear) problems.

Where I have seen the line blur is when an amplifier has limited frequency response (at either low or high frequency) that may create higher average group delay than other amplifiers but more importantly has delay that changes significantly over frequency, corrupting the signal. The amplifier's design may increase group delay, but the cause of hearing issues is that the delay is no longer constant across the audio (audible) band.

For example, adding a gain stage adds delay, but if bandwidth (etc.) is sufficiently high that the delay does not change across the audio band, there is no audible impact (from delay; there may be other impacts like noise and distortion). Adding ten stages is a much greater increase in delay, but audible problems are probably due to the phase changes from limited bandwidth (plus added noise and distortion).

Group delay is necessarily related to circuit design choices so is hard to discuss in isolation, or attribute audible degradation to it alone.

IME/IMO etc. etc. etc. - Don

 

[microstrip]

Except for the case of a few remarks on poorly designed amplifiers, the only relevant reference I remember on this subject is concerning the mechanical time alignment of the WAMM and XVX Wilson Audio speakers amplifiers - the advised standard positions were given for the time delay of D'Agostino solid state amplifiers, Wilson Audio would measure customer amplifiers group delay versus frequency and supply a corrected table. As far as I read in the article, differences were mainly relevant in the case of tubed amplifiers.

 

[DasguteOhr]

Which amps do this?
I cannot think of a single one that has any effect above maybe 50 Hz.

e.g. every tube amplifier with incorrectly measured size of the coupling capacitor and grid leak resistors between the stages. There are many of these today, believe me, to create a low lower limit frequency or a certain sound character. The charging and discharging times of the coupling capacitor mainly determine whether an amplifier is impulse fast or a lame crutch.
The time constant(tau )should not be greater than 20ms

However the speakers dominate the phase error space.
Just look at the impulse response.
One can choose to tidy it up with say DIRAC or some other phase-EQ, but they cannot tidy it up with an amplifier.

Yes, I completely agree. Active speakers actually have the advantage, but the steep filters make it difficult.

 

[DasguteOhr]

I do not believe it is the sheer magnitude of group delay that matters but rather how constant, or how nonlinear the phase response that leads to group delay not being constant over frequency. That shifts phase and delay differently for different frequencies and is what smears the image (soundstage, etc.) If the group delay is 0.1 or 1 but constant things will sound fine; if it is a function of frequency then we have (hear) problems.

Where I have seen the line blur is when an amplifier has limited frequency response (at either low or high frequency) that may create higher average group delay than other amplifiers but more importantly has delay that changes significantly over frequency, corrupting the signal. The amplifier's design may increase group delay, but the cause of hearing issues is that the delay is no longer constant across the audio (audible) band.

For example, adding a gain stage adds delay, but if bandwidth (etc.) is sufficiently high that the delay does not change across the audio band, there is no audible impact (from delay; there may be other impacts like noise and distortion). Adding ten stages is a much greater increase in delay, but audible problems are probably due to the phase changes from limited bandwidth (plus added noise and distortion).

Group delay is necessarily related to circuit design choices so is hard to discuss in isolation, or attribute audible degradation to it alone.

IME/IMO etc. etc. etc. - Don

I would like to agree almost completely and the last sentence sums it up. Sounds almost exactly like my last sentence in post#4
I borrowed a few pictures, a tube amplifier from a well-known manufacturer. The lower picture shows the original condition, the upper picture with the adjusted coupling capacitor. Sorry for the picture quality
10khz

1khz

 

[Atmasphere]

Which amplifier topology, solid-state or tube, has less group delay?

I'm sure the answer is: "it depends."

On what does it depend?

What tube circuit design might have less group delay than a transistor design?

What transistor circuit design might have less group delay than a tube design?

Do SET amplifiers have less or more group delay than complex, multi-output tube, push-pull amplifiers?

This may not be the right question to ask.

IME phase shift in general is the bigger issue, although its more complex than that. If the amplifier has no feedback, it cannot correct for phase shift caused by a low or high frequency rolloff. For that reason, zero feedback amps need lots of bandwidth to keep phase shift artifacts from being audible. The ear can't hear phase shift at a single frequency, but over a range of frequencies it tends to interpret it as a tonality. A rolloff in the highs as darkness, in the bass as leaness. Generally speaking...

Amps tend to behave as filters so filter theory applies. If there is no feedback the rolloff of the amp will likely be gentle at first, going to a 6dB slope. A 6dB filter will have phase shift to 1/10th its cutoff if a low pass (IOW if the rolloff is in the highs) and up to 10x the cutoff frequency if in the bass. This means if you want no phase shift in the audio band, you need 2Hz to 200KHz response which is challenging for a lot of amps. When you have feedback, the rolloff might be more profound than just 6dB /octave. The faster it rolls off the more the phase shift is bunched up near the cutoff frequency, to the point there may even be a peak just prior to the cutoff.

Now if you have enough feedback you can correct for phase shift since the feedback allows for rejection of that which is not the signal. This is why some class D amps (which typically don't have all that much bandwidth) can have low phase shift in the audio band (such an amp is usually of the self oscillating variety since they tend to have the most feedback of class D amps).

Before class D came along, you had to be really careful about feedback in amplifiers, due to something called 'phase margin'. Since the amp has phase shift, if you increase the frequency enough, at some point the phase shift is so profound that feedback becomes positive rather than negative. So if feedback occurs above that frequency, the amp can oscillate because the feedback is positive rather than negative. We've seem examples of that: Futterman OTLs that would oscillate with certain loads, solid state amps from the late 70s and early 80s that did the same thing (usually with highly capacitive loads). In those cases the phase margin of the amps was being exceeded. This is univerally due to poor feedback loop design. Norman Crowhusrt wrote about this issue in the 1950s. If you have wide bandwidth in the amp, you have to be careful that the feedback loop has a damping aspect at a higher frequency so as to shut out oscillation, regardless of the load.

In case you were wondering why we made OTLs when SETs were so much easier to make, phase shift is one of the reasons: OTLs can have excellent bandwidth (we made prototypes that went well into the MHz range; production amps were bandwidth limited); SETs struggle to get past 20KHz unless they are lower power examples (type 45 power tube or the like). By having almost zero phase shift in the audio band an amp can came off sounding faster and wider bandwidth.

Its pretty obvious Stewart Hegeman of Harmon Kardon fame (Citation 1 preamp and Citation 2 power amp) was all about wide bandwidth on this account.

SETs tend to have the most phase shift of any amplifiers made. Its one of their weak points, a symptom of poor bandwidth. Most of that is caused by the output transformer which is probably harder to design than one for push-pull. That is probably a topic for a different thread.

 

[morricab]

Which amps do this?
I cannot think of a single one that has any effect above maybe 50 Hz.



However the speakers dominate the phase error space.
Just look at the impulse response.
One can choose to tidy it up with say DIRAC or some other phase-EQ, but they cannot tidy it up with an amplifier.

You have not heard electronics that destroy soundstage depth?

 

[Holmz]

You have not heard electronics that destroy soundstage depth?

Not really…
I have more commonly heard poor speaker placement that destroyed it.

Maybe I have just used better amps than were required?

But I have not seem a lot of measurement on group delay in amps, probably because it is generally difficult to have group delay in amps?
We have still not seen much in the way of measurements that show group delay in amp, or phase behavior… at least none that I can recall.

 

[morricab]

Not really…
I have more commonly heard poor speaker placement that destroyed it.

Maybe I have just used better amps than were required?

But I have not seem a lot of measurement on group delay in amps, probably because it is generally difficult to have group delay in amps?
We have still not seen much in the way of measurements that show group delay in amp, or phase behavior… at least none that I can recall.

I have heard vast differences in soundstage depth and image 3D between electronics. So much so in fact that it is more important than room or speakers in most situations.

Whether this is due to group delay or phase shift or?? I cannot say for sure but it definitely has something to do with perception of high frequency and its relationship with the rest of the spectrum.

 

[Atmasphere]

Harmonic distortion can have an effect on the perception of depth in amplifiers due to something called 'super position'. For example a prodigious second harmonic can make the waveform have higher amplitude. The imaging information is usually higher frequencies so this can bring some of that out.

 

[Holmz]

I have heard vast differences in soundstage depth and image 3D between electronics. So much so in fact that it is more important than room or speakers in most situations.

Whether this is due to group delay or phase shift or?? I cannot say for sure but it definitely has something to do with perception of high frequency and its relationship with the rest of the spectrum.

It would be appropriate for this thread if we knew which amps, and if those amps had some group-delay or phase data versus frequency.

Or the harmonic distortion.

But at least we would know which (two) amplifiers to compare.

There must be something that is physically happening in those amps that comports to reality.
Then we could have some hypothesis as to what those things are.

Without something like ^that^, it makes it hard to “presort” the list of possible amps into a “short list”.

While I am unconvinced that they are amplifier group delay, we could at least see if those two amps have a lot of group-delay or a minimal group delay.
Without that, then we are left with hunting down amps in the wild and listening to each and every one.

I just want to be done with amplifiers, and put one in that just allows me to enjoy the music.
And, as @Atmasphere suggests, I did find one with a predominate 2nd harmonic.

 

[morricab]

It would be appropriate for this thread if we knew which amps, and if those amps had some group-delay or phase data versus frequency.

Or the harmonic distortion.

But at least we would know which (two) amplifiers to compare.

There must be something that is physically happening in those amps that comports to reality.
Then we could have some hypothesis as to what those things are.

Without something like ^that^, it makes it hard to “presort” the list of possible amps into a “short list”.

While I am unconvinced that they are amplifier group delay, we could at least see if those two amps have a lot of group-delay or a minimal group delay.
Without that, then we are left with hunting down amps in the wild and listening to each and every one.

I just want to be done with amplifiers, and put one in that just allows me to enjoy the music.
And, as @Atmasphere suggests, I did find one with a predominate 2nd harmonic.

What amp have you settled on?

I am not saying that group delay is the likely cause of what I hear in electronics regarding soundstage and imaging …probably more to do with distortion types…but who knows because group delay and phase shift data for electronics is not readily available.

 

[Holmz]

@morricab Nagra Classic - but some of that is FengShui (aka WAF).
Both for looks and she won’t have to flick any switches or knobs when watching the TV.

It is a combined 2channel and 5.x.x