Bob Cordell's Power amplifier book

[Bob Cordell]

Originally Posted by Edmond Stuart

Hi Bob,

There is some truth in Walter's comment. The supply rails of the PGP front-end, for example, were filtered by means of a 100uF cap and a 10 Ohms metal film resistor, rated at 1/4W. (see: Front End PCB Schematic R1 & C2, in the upper-right corner and R72 & C29 in the lower-right corner)

After powering up a few times, these resistors were blown. Admittedly, they were 10 Ohms instead of 1 Ohms. Nevertheless, I think we should not underestimate the peak power during startup, even in case of very small resistors, like 1 Ohm.

Cheers,
E.

Hi Edmond,

Understood, but who is talking about using 1/4 watt resistors in such a location?

Anyway, it is a good point to look at the peak power dissipation of the resistor during turn-on, and this is something that can be SPICEd.

Of course, resistors are rated for continuous power dissipation. Most can handle several times that in peak dissipation and suffer no reduction in reliability.

As pointed out earlier, it also depends a lot on how fast the power supply comes up.

Consider some back-of-the-envelope numbers.
Suppose at turn-on we have 100 Amps through the rectifier, maximum. To simplify and make things more worst-case, assume this 100A persists over the full time it takes to charge the reservoir capacitor to 60V.

Assume we have a 50,000 uF reservoir capacitor.

The voltage slew rate on the rail will be I/C = 100A/0.05F = 2000 V/s.

The rail will come up in 30ms.

Assume that the rail voltage slew rate (SR) is applied directly to a 100 uF filter capacitor (ignore the effect of the series resistor to simplify and be more worst-case).

The capacitor current will be SR * C = 2000V/s * 0.0001F = 0.2A.

If we run this through a 10-ohm resistor, the peak voltage drop will be 2V and the peak dissipation will be 2V * 0.2A = 0.4W.

This seems too small even for me to believe. What have I done wrong here? The math? The assumptions?

Cheers,
Bob

[Edmond Stuart]

Originally Posted by Bob Cordell

Hi Edmond,
[...]
If we run this through a 10-ohm resistor, the peak voltage drop will be 2V and the peak dissipation will be 2V * 0.2A = 0.4W.

This seems too small even for me to believe. What have I done wrong here? The math? The assumptions?

Cheers,
Bob

Hi Bob,

I also was puzzled. We both made a wrong assumption by thinking that the front-end PSU was directly tied to the main PSU via 10 Ohms resistors. That's not the case. After rereading syn08's description of the PSU, it appears that the front-end supply rails are switched on by means of relays. (see the final note on this page: Power Supply ). No wonder that the resistors were blown.
My apologies for the confusion.

BTW, did you try the VDMOS sub-threshold enhancements as mentioned on page 2304?

Cheers,
E.

[PB2]

Hi Bob,

I've not seen much discussion around here of the Quad 405 amplifier. I do remember the articles in the AES and Wireless World years ago but I don't think any of them showed a full schematic. I recently found the service manual and I have to say that it is clever and fairly simple. Here's the service manual in case you've not seen it. The N1 and N2 modules are output current limiting and are often removed since they tend to trip too early:
http://www.keith-snook.info/Schemati...ice%20Data.pdf

I see things that I'd do differently but it is an interesting design.

Here is an AES paper that discusses the unique output stage:
http://quad405.com/jaes.pdf

And this page has reviews, mods, etc:
Quad 405 Information Page

Wondering if you studied it in detail and what you think of it.

[jcx]

I've read up on Black's Feedforward, Quad, Vanderkooy (…and Annison, Danyuk, Sandman, Stochino) paper's circuits, simmed a few alternatives

after thinking about it I come to the question that if you can measure the error to amplify and add in to cancel the main amp distortion - why don't you just use more feedback?

for audio power amps we have the ability to use plenty negative feedback loop gain over an extended definition of audio frequency range if allowed to use RET or MOSFET output Q to push unity loop gain intercept up and use higher order compensation, or nested feedback (including Bob's, Hawksford's EC)
feedback error disappears into noise for really high loop gain feedback amps except for the last few octaves of audio

the feedforward schemes can knock down the >20 kHz errors ~ 20-30 dB – but at some cost in parts, design complexity for arguably inaudible “improvement”

and the cancellation is very sensitive to gain/power coupling network tolerances

remember also that the audible IMD products are reduced by the feedback at the product frequency – so any high frequency difference products folding down into audio are reduced by the high loop gain feedback

feedforward may just reduce Bob's THD 20 kHz metric without adding to/giving any of the implied "goodness" at actually audible frequencies that we hope the THD20 correlates with


Alternative topologies


High loop Gain Composite Op Amp Circuits

[PB2]

Originally Posted by jcx

I've read up on Black's Feedforward, Quad, Vanderkooy (…and Annison, Danyuk, Sandman, Stochino) paper's circuits, simmed a few alternatives

after thinking about it I come to the question that if you can measure the error to amplify and add in to cancel the main amp distortion - why don't you just use more feedback?

for audio power amps we have the ability to use plenty negative feedback loop gain over an extended definition of audio frequency range if allowed to use RET or MOSFET output Q to push unity loop gain intercept up and use higher order compensation, or nested feedback (including Bob's, Hawksford's EC)
feedback error disappears into noise for really high loop gain feedback amps except for the last few octaves of audio

the feedforward schemes can knock down the >20 kHz errors ~ 20-30 dB – but at some cost in parts, design complexity for arguably inaudible “improvement”

and the cancellation is very sensitive to gain/power coupling network tolerances

remember also that the audible IMD products are reduced by the feedback at the product frequency – so any high frequency difference products folding down into audio are reduced by the high loop gain feedback

feedforward may just reduce Bob's THD 20 kHz metric without adding to/giving any of the implied "goodness" at actually audible frequencies that we hope the THD20 correlates with


Alternative topologies


High loop Gain Composite Op Amp Circuits

Thanks jcx for the interesting comments and links.

You mention the use of better output devices - sure it is clear that we can do better with better devices but this design is not much more complex than others and seems to do very well with inexpensive, slow devices. You say that cancellation is very sensitive to gain/power coupling network tolerances which I'm sure is true if your goal is perfect cancellation but why not just accept a 20 dB improvement and say good enough. And really, 1% resistors are not very expensive. I also like the advantage of there being no thermal tracking issue with this design.

I have a rough idea of where the audible limits of distortion are and I believe that many designs here are well past those limits. But to have a low distortion class B design like the Quad with no thermal adjustment or tracking issues is an interesting alternative.

Interesting that this thread just started on the Quad 909:
Quad 909 Clone

[jcx]

I don't see the justification for /attraction of the "Class C" dumper - its not like thermal runaway in Class AB is intractable or even expensive in component count to control - unless you're Doug Self trying to nail his "Optimum Class B" bias point

with any bias at all the feedforwad has lots less work to do, the mismatch isn't so critical as when you're trying to fix a 1.2 V deadzone at zero crossing

Thorsten’s fixing up a power chip amp with Stochino’s Feedforward circuit may be interesting if you want to avoid power circuit design – the power chip amps are so slow you can’t do that much for them with added (nested) negative feedback

[PB2]

Originally Posted by jcx

I don't see the justification for /attraction of the "Class C" dumper - its not like thermal runaway in Class AB is intractable or even expensive in component count to control - unless you're Doug Self trying to nail his "Optimum Class B" bias point

with any bias at all the feedforwad has lots less work to do, the mismatch isn't so critical as when you're trying to fix a 1.2 V deadzone at zero crossing

Thorsten’s fixing up a power chip amp with Stochino’s Feedforward circuit may be interesting if you want to avoid power circuit design – the power chip amps are so slow you can’t do that much for them with added (nested) negative feedback

We obviously have different perspectives based on this and our past discussions. There's no reason why I wouldn't bias it up a bit more say to a .2V dead zone, it is an obvious modification. Quad did eventually add one diode to reduce the deadzone to .6V. There's been a lot of discussion about thermal tracking, output devices with built in diodes and thus some are concerned about it - I like the fact that all that goes away. I don't have any issue with power circuit design I just like to review all the options with a level headed view of the advantages/disadvantages.

[richiem]

I'd like to interject here that there are many advantages to a combination of Class A and Class B operation (NOT AB) such as used by McIntosh in tube amp design, where a relatively low-power Class A driver stage drives the load through the crossover region, and where the Class B output stage then drives the rest of the signal swing.

This design was successfully transferred to solid-state by Mattes in the "Sharma" circuit many years ago, and something similar was done also by Dynaco. Negative feedback then sorts it all out for low levels of THD. I think of this as a feedback form of current dumping. I haven't done a web search for this design and don't know if it has any adherents these days, but it should....

[jcx]

the Vanderkooy, Lipshitz (quad/jaes.pdf) paper spells out the requirement for “true feedforward”

if you're talking "bootstrapping" a heavy bias floating supply Class A with the Class B high V stage I think the analysis shows you have to consider it as a feedback amplifier - possibly with nested loops although there is some isolation by the load's own impedance - I think it is at best equivalent to their fig 4a

"true" feedforward requires non-interaction of the 2 amps - one shouldn't affect the stability of the other - with fixed impedance transmission line loads Black proposed "biconjugate network" coupling

in SE audio amps the practical solution seems to be essentially a power XO with the correction amp controlling the load at high frequency

I certainly enjoy the intellectual exercise studying Black's Feedforward, knowing its historical relation to negative feedback
I just don't yet see it as anything but a curiosity in practical audio power amps if you aren't limited by other considerations from "maxing out" negative feedback, open loop linearity

but at this late date I think most audio power amp design can be characterized as "conceptual art" - so whatever floats your boat is fine


"con·cep·tu·al art

Noun:

Art in which the idea presented by the artist is considered more important than the finished product, if any exists

[PB2]

Some questioned if it actually worked:
http://quad405.com/lips.pdf