Bob Cordell Interview: Negative Feedback

[Fanuc]

I agree. This kind of signal shows time response of the system, though the system can be perfectly linear. Every sudden change, even if it is straight line of zero level suddenly changed to sine starting at zero, not maximum, needs infinite bandwith to be transferred without linear distortion. It is all about understanding of linear and non-linear distortions and time response to any signal.

Yep. I've seen Graham mention this arguement before etc and it seems to lack common sense (irrational). Loudspeakers do not really have back EMF - nothing serious anyway.

 

[jan.didden]

REGISTER ENTRY FOR GB2424137

Form 1 Application No GB0505024.0 filing date 11.03.2005

Title AMPLIFIER

Applicant/Proprietor
AUDIO PARTNERSHIP PLC, Incorporated in the United Kingdom, Gallery Court,
Hankey Place, LONDON, SE1 4BB, United Kingdom [ADP No. 07018526002]

Inventor
DOUGLAS SELF, AUDIO PARTNERSHIP PLC, Gallery Court, Hankey Place, LONDON,
SE1 4BB, United Kingdom [ADP No. 09246794001]

Classified to
H3W H3T
H03F

Address for Service
MAGUIRE BOSS, 5 Crown Street, ST IVES, Cambridgeshire, PE27 5EB, United
Kingdom [ADP No. 07188725001]

Publication No GB2424137 dated 13.09.2006

Examination requested 11.03.2005

Grant of Patent (Notification under Section 18(4)) 16.01.2007
Publication of notice in the Patents and Designs Journal (Section 25(1))
14.02.2007
Title of Granted Patent AMPLIFIER

________________________________________________________________________________

15.06.2006 Notification of change of Address For Service address of
MAGUIRE BOSS, 5 Crown Street, ST IVES, Cambridgeshire, PE27 5EB,
United Kingdom [ADP No. 07188725001]
to
MAGUIRE BOSS, 24 East Street, ST IVES, Cambridge, PE27 5PD, United
Kingdom [ADP No. 07188725001]
dated 08.06.2006. Written notification filed on GB2373569

**** END OF REGISTER ENTRY ****

________________________________________________________________________________

RENEWALS DATA

Date Filed 11.03.2005

Date Not in Force

Date of Last Renewal

Year of Last Renewal

Next Renewal Date 11.03.2009

Status PATENT IN FORCE

 

[Graham Maynard]

Hi PMA,

10kHz is approx 60mV/uS/V, input or output, nothing approaching infinite about that.
For an amplifier to be accurate to the limits of an audio source it must of course have a much greater bandwidth and short propagation delay if it is to be expected to damp real world loudspeaker back-EMFs at such frequencies - as can be generated by loudspeaker system crossovers.
This is why it is essential to preserve output stage damping linearity throughout the AF range for class-AB operation, or to use some form of class-A augmentation.

If a class-AB amplifier tests out better with a suddenly starting 10kHz sine (no matter whether this might be thought unrealistic) then it is also likely to be capable of establishing a more stable stereo image of an instrument like a cymball or violin due to either or both amplifiers being less disturbed by back-EMF.

 

[Bob Cordell]

Hi Bob and Nelson,

Re Douglas' Patent. I surmise he advanced upon earlier designs by claiming novelty for an *active* voltage controlled current source which he uses to displace the class-AB crossover beyond the range of loudspeaker generated back-EMF, such that his 'XD' amplifiers run in class-A at normal listening levels.

Re back-EMF induced distortion, see the attached illustration from my 2004 EW article.

This shows class-AB amplifier output simulations of a generic type similar to 'Blameless'
All outputs are 15V sine, with each trace *fundamental nulled* to examine individual distortion residuals.

The black trace illustrates the -40dB level of the Blue output.
Blue with choke and Miller C.dom into R load
Red with choke and Miller into 'Ariel' equivalent.
Mauve with choke but no Miller C.dom into Ariel
Green no choke but with Miller C.dom into Ariel
Yellow no choke and no Miller C.dom into Ariel (same architecture)

Yellow circuit not constructed, so might not be stable.

Those who state it is wrong to study how an amplifier responds to a suddenly starting wave, may follow the repeated error residuals from 150uS onwards.
Note that individual amplifier component variations overlaid upon the same semiconductor layout have an influence upon group delay, and thus crossover timing within each voltage output waveform.
Note how 'loudspeaker' loading exacerbates emitter follower output stage crossover distortion when a Miller C.dom is used at the VAS within a closed NFB loop.

Thanks, Graham. Can we get a copy of your EW article? I'd also like to see the schematics of the amplifiers you refer to in your simulation so I can better understand what's being done.

In regard to Self displacing the crossover point away from where back EMF has an influence, would that perhaps not be the same as running the amplifier in Class AAB, where it is in Class A up to between 1 and 10 watts, or so? Or is Self trying to achieve the same effect without the consequent increased idle bias power dissipation?

Cheers,
Bob

 

[jan.didden]

[snip]In regard to Self displacing the crossover point away from where back EMF has an influence, would that perhaps not be the same as running the amplifier in Class AAB, where it is in Class A up to between 1 and 10 watts, or so? Or is Self trying to achieve the same effect without the consequent increased idle bias power dissipation?

Cheers,
Bob

Bob,

That's part of it. In the whitepaper I refer to above, there is a table comparing efficiencies of A, AB, and XD in several incarnations.

http://www.cambridgeaudio.com/assets/documents/840Awhitepaper8-2-06web.pdf


Jan Didden

 

[Bob Cordell]

Hi PMA,

10kHz is approx 60mV/uS/V, input or output, nothing approaching infinite about that.
For an amplifier to be accurate to the limits of an audio source it must of course have a much greater bandwidth and short propagation delay if it is to be expected to damp real world loudspeaker back-EMFs at such frequencies - as can be generated by loudspeaker system crossovers.
This is why it is essential to preserve output stage damping linearity throughout the AF range for class-AB operation, or to use some form of class-A augmentation.

If a class-AB amplifier tests out better with a suddenly starting 10kHz sine (no matter whether this might be thought unrealistic) then it is also likely to be capable of establishing a more stable stereo image of an instrument like a cymball or violin due to either or both amplifiers being less disturbed by back-EMF.

I can't argue with this, but I think it is not much different than saying that the small-signal bandwidth must be greater than 100 kHz or so, and that the amplifier have adequate slew rate, which, in the extreme case here, might be enough slew rate to handle a full-power 100 khz sinewave, in very rough terms, if the 10 kHz sudden tone burst were bandlimited to 100 kHz. I think that only comes out to about 24 V/us for a 100 watt amplifier.

Neither of these requirements is asking a lot for a decent solid state amplifier.

Bob

 

[PMA]

Hi PMA,

10kHz is approx 60mV/uS/V, input or output, nothing approaching infinite about that.

Only in steady state (settled). When turned on, from zero line, even at at phi = 0 deg, the corner (transition) from line to sine has infinite frequency spectrum. Every RC lopass, even 100MHz, will introduce some delay (1.59 nanoseconds in this case) and will "round" transition between line and sine.

 

[Graham Maynard]

Hi Bob,

In his '90s texts Doug showed how there would be a transconductance change going from class-A to AB (not recommendable).
He also achieved highly linear circuits when driving resistors, but there were comments about distortion when driving real world (back-EMF generating) loudspeakers.
His class-A XD arrangement allows him to use his established highly linear class-AB circuits for driving loudspeakers because the back-EMF induced current crossover non-linearities are shifted away from normal listening voltage outputs.

 

[Graham Maynard]

Hi PMA.
Yes, every circuit introduces a delay.
The same delay that can fail to prevent a NFB loop controlled class-AB amplifier output stage from damping loudspeaker back-EMF due to an inability for a VAS to reverse output drive.
Output stages store charges and these must be actively removed, but the control response is finite, thus the development of 'crossover distortion'.

 

[PMA]

So called "back EMF" issue can be completely described by speaker electrical equivalent circuit. "Back EMF" is just exchange of energy between amplifier and parallel resonant circuit, representing acoustical impedance of speaker in the enclosure transferred to speaker terminals by electro-acoustic transformation.

 

[Nelson Pass]

This is a displacer circuit offered by Douglas Self in the Nov2006 EW.

Ah yes, thanks for reminding me I read the paper, but I'm
obviously getting old.

The novelty of tracking the current source against output
voltage is apparent.

It seems to me that it ought to work well into resistance, but
less well into a reactive load, as the current is what we ideally
want to track.

 

[Bob Cordell]

Hi Bob,

In his '90s texts Doug showed how there would be a transconductance change going from class-A to AB (not recommendable).
He also achieved highly linear circuits when driving resistors, but there were comments about distortion when driving real world (back-EMF generating) loudspeakers.
His class-A XD arrangement allows him to use his established highly linear class-AB circuits for driving loudspeakers because the back-EMF induced current crossover non-linearities are shifted away from normal listening voltage outputs.

Yes, in another thread here we discussed what Self described as gm doubling.

I read the white paper and let you know what I think.

Bob

 

[Bob Cordell]

So called "back EMF" issue can be completely described by speaker electrical equivalent circuit. "Back EMF" is just exchange of energy between amplifier and parallel resonant circuit, representing acoustical impedance of speaker in the enclosure transferred to speaker terminals by electro-acoustic transformation.

I agree completely. It is, nevertheless, fair game to look closely at unanticipated effects this exchange of energy might cause. For example, there have certainly been cases in the past when protection circuits went off due to these kinds of currents.

Bob

 

[Graham Maynard]

Hi Nelson,

Old enough to know what's important though !

Hi Bob,

Music is an endlessly changing complex waveform comprising transients and simultaneously acting complex components, so we can have established back-EMFs interacting with new drive. Again Otala, Self etc have covered the instantaneous current requirements that can arise, but back-EMF can also reverse drive an amplifier when its music input changes.

Another way of looking at this is when an amplifier has an inductive output impedance then different propagation delays can arise at some frequencies between amplifier input and loudspeaker terminals at moments when the leading loudspeaker current is reverse driving the amplifier compared to when the amplifier is powering the loudspeaker. Obviously I could be informed that any distortion components arising would be way above audible frequencies, but they would not have been original, and passband limiting an amplifier's input will not protect against their generation; this could be another way of studying fractional image error.

The attached image is of simulated group delay at the output of a perfect amplifier with series 6uH choke (//8.2R) driving an Ariel Loudspeaker equivalent. Someone once challenged me that this shows the signal arriving at the loudspeaker before it leaves the amplifier, but of course this is for established sine excitation, and it shows how waveform coherence might become even more disturbed due to highly variable music throughput.

 

[ingrast]

....Music is an endlessly changing complex waveform comprising transients and simultaneously acting complex components, so we can have established back-EMFs interacting with new drive. ......

With due respect Graham, and far from simply dismissing what is your long involvement with audio and most certainly satisfying achievements, I cannot help but strongly disagree with your line of reasoning.

In a way, and aware of coincidences and differences, this reminds me of the pre Newton controversy regarding what is the trajectory of a cannonball. There were then scores of intuitive and observation based theories from ludicrous to pretty clever, yet wrong in the end.
Then comes a deceivingly simple theoretical framework which predicts with an accuracy far more deadly than one should like it to be.

The back EMF issue and its relation with negative feedback is a phenomenom fairly well described even by first order approximations in terms of linear systems with reactive components. Once the appropriate complex (i.e, AC) characterization is determined to the desired accuracy, including real world reactive loading, the interactions you describe intuitively are perfectly defined to the same degree of accuracy, and design considerations tailored to comform to a performance objective can be safely made.

Of course the listening experience is the final judge, but what I want to stress is that definitely the listening cannot possible unveil hitherto unknown phenomena not apparent in the design model, as long as it is accurate as said above.

What you say regarding NFB delay is embedded in system bandwidth. What you stress regarding back EMF is reactive loading (and it can be scary in circumstances). What you say regarding sudden waveform application implies infinite input signal bandwidth which is not the case.

Thanks anyway for stirring the cauldron, it allways makes for tasty soup.

Rodolfo

 

[Tube_Dude]

This is a displacer circuit offered by Douglas Self in the Nov2006 EW.

The attachment that Graham give at post # 398 , is precisely the VAS stage of a NAD 3020 amplifier...

 

[PMA]

I agree completely. It is, nevertheless, fair game to look closely at unanticipated effects this exchange of energy might cause. For example, there have certainly been cases in the past when protection circuits went off due to these kinds of currents.

Bob

I also agree completely. Reactive load issues have to be taken into account and the reactive load model should be close to "real world" impedance, to assess peak currents and instant peak power in the output stage, and to investigate non-linearities influence as well.

 

[AKSA]

The electrical inductance of the speakers is probably trivial at 1-2mH, and would not influence proceedings until 30KHz and beyond. But the back emf is created by inertia of the cone reverse energising the voice coil; this can and does affect sonics because the damping factor of the amp is not infinite. Propagation delay through the amp pales into insignificance due to this effect, which is measured in the milliseconds.

This might mandate more care in crossover design of speakers than any special attention to the amp. Of concern would be controlling phase shift at crossover and resonance. I have noticed small anomalies in passive crossover design can create marked phase shift around the crossover point, typically high in the midrange on a two way, and this is unquestionably audible.

Cheers,

Hugh

 

[Graham Maynard]

Hi Jorge,

Though this one is a free standing attachment to the output stage and dissipates a fair amount of heat.

 

[PMA]

The electrical inductance of the speakers is probably trivial at 1-2mH, and would not influence proceedings until 30KHz and beyond. But the back emf is created by inertia of the cone reverse energising the voice coil;

The equivalent speaker circuit connected to output stage can be seen in the following image: