There looks to be some theoretical reasons to prefer a Common Emitter Output Section.
Dr Ed Cherry has advocated this for many years, in contradiction to Self's rather poor opinion of the idea, and usual practice.
I have already reinvented some of Cherry's work, which reassures me a bit but is not a very productive use of time.
I would prefer not to reinvent any more, so does anyone have any references to any other analysis of the CE OPS?
Not sure yet if there is some sort of optimum bias point, similar to the Oliver bias point for an emitter follower.
David
Dr Ed Cherry has advocated this for many years, in contradiction to Self's rather poor opinion of the idea, and usual practice.
I have already reinvented some of Cherry's work, which reassures me a bit but is not a very productive use of time.
I would prefer not to reinvent any more, so does anyone have any references to any other analysis of the CE OPS?
Not sure yet if there is some sort of optimum bias point, similar to the Oliver bias point for an emitter follower.
David
I am only aware of one major advantage of CE output stages: They are able to run extremely close to the rails (even more so with MOSFETs), which is why they tend to be standard fare in rail/rail opamps. Literature on those may prove enlightening.
Other than that, there's plenty of reasons why people have traditionally preferred followers. The output stage tends to be the slowest component as-is, and going CE makes it even slower and more prone to capacitive loading. (Keep in mind an emitter follower may have a bandwidth of >fT if driven hard.) I don't think there is any optimum biasing per se, it generally is the more the merrier. It stays a current source output (with output impedance then knocked down by feedback) until way into Class A territory.
Other than that, there's plenty of reasons why people have traditionally preferred followers. The output stage tends to be the slowest component as-is, and going CE makes it even slower and more prone to capacitive loading. (Keep in mind an emitter follower may have a bandwidth of >fT if driven hard.) I don't think there is any optimum biasing per se, it generally is the more the merrier. It stays a current source output (with output impedance then knocked down by feedback) until way into Class A territory.
Dave is correct of course. It suboptimal to wrap gnfb around the local nfb introduced by cc/cd ops. But the practice remains the mainstay of amplifier designs for many of the reasons mentioned above. Plus scaling up a design by simply adding more followers is substantially preferred to redesigning the amp for larger output power. Nelson Pass has been using them in his designs for decades and so assertions that they do not work well in practice are misguided.
My principle reason is for feedback optimization, but the rail-to-rail is a nice bonus, as you say, especially for mosfets.
I hoped that someone would have some references to the op-amp literature, manufacturers don't seem to release as much information as they used to.
Simple theory says the increased global feedback should compensate for the reduced local feedback.
So either the popular conception is incorrect or my analysis is too simplistic.
Hence my curiousity.
Not that I know of, appreciate a reference if it exists.
That's my main motivation.
I suspect Nelson does this partly to keep the rails volts down, makes it more DIY friendly, and partly just for the interest in an unconventional solution.
He seems reasonably unconcerned about the lower distortion possible and I haven't seen any analysis from him.
Best wishes
David
You might also be interested in this thread... http://www.diyaudio.com/forums/solid-state/284744-atc-active-style-amps.html. These long tried and tested amps employ common source outputs stages similar to a more recent design of John Vanderkooy published in Linear Audio I believe (?).
Most interesting in the ATC implementation is the provision of adjustable local feedback around the output stage that nulls distortion by balancing the transconductances of the output devices. It is a very effective technique that I would recommend trying...
Also of note is that the ATC implementation adopts a low voltage (op amp) driver stage and necessarily separate power supplies for each of the three output stages (for their 3-way active speakers). IMHO this circuit topology has a number of advantages and deserves more attention than conventionality (if that is a word) affords.
Most interesting in the ATC implementation is the provision of adjustable local feedback around the output stage that nulls distortion by balancing the transconductances of the output devices. It is a very effective technique that I would recommend trying...
Also of note is that the ATC implementation adopts a low voltage (op amp) driver stage and necessarily separate power supplies for each of the three output stages (for their 3-way active speakers). IMHO this circuit topology has a number of advantages and deserves more attention than conventionality (if that is a word) affords.
The concept (floated power supply CE) seems similar to the QSC power amps but I haven't looked too closely at either yet so that's just a first impression.
It would be nice to see the circuits, can you post the ATC's in that thread if they are public domain?
Cherry discusses some of these options in an article in Wireless World (or whatever it was called at the time).
I would expect that power supply stray capacitance would load the amp, which seems undesirable, but haven't checked if it's a problem, or even if it happens.
Exactly the sort of issues that I want to explore here, so thanks.
Best wishes
David
A correction. The QSC in that thread is indeed a "Grounded Collector".
So it's an emitter follower, except with the collector grounded and the supplies float.
I had a recollection that QSC did some kind of Common Emitter circuit in some amps.
A "Grounded Source" would make sense for the FETs that have the Source connected to the case (some laterals I think).
Or possibly a CFP OPS if the Drain/Collector is connected to the case.
It seems QSC was more interested in the practicality benefits of Grounded case rather than the CE circuit as such.
Many of the picture links in earlier discussions in DIY have disappeared, any QSC experts are welcome to be pedantic here.
David
OK, I reread the referenced thread and the distinction between the QSC and the ATC is now clear.
The ATC is a particularly neat implementation if you have cases connected to Source/ Emitter.
Most BJT aren't, of course.
What is the problem that requires attention to the MOSFET pin inductances?
I suspect this is the flip side of my concern over the power supply float capacitance.
Best wishes
David
The lead inductance issue is that they need be somewhat longer than normal due to a 90 degree bend. The 90 degree band I suspect also has an influence here. But with MOSFETs particularly operating above about 30V drain-source all manner of parasitic oscillations are possible - and probable if not careful. A higher than recommended gate resistor helps, as can well placed RC snubber networks.
Nevertheless I believe the use of MOSFETs in this application where source resistors can be omitted is preferential to BJTs. But please check out the local feedback nulling arrangement that ATC apply if you get the chance - IMHO it ranks as one of the most simple and effective fixes available...
As to power supply capacitance, ATC appear to use a large E-I transformer specified to medical equipment grade - that is inter-winding screening. It does get warm, however. But then in a single channel amp case, a separate output stage transformer would likely not pose much of an overall cost issue?
The floated supply makes the amp much more vulnerable to noise coupled from the mains, so presumably the transformer is screened to reduce this.
Hi Arthur
The amp is in an article in "Proc. IREE Australia, vol. 39, pp. 1-8, Jan/Feb. 1978" E.M. Cherry.
Not public domain but I will ask Dr Cherry what I can do.
No simulations yet, I don't plan to simulate the entire amp anyway, I mainly wanted to see what his OPS looked like.
The rest is fairly conventional LTPs for input and second section but with his favorite "nested feedback" scheme, that I am less keen to follow.
More later.
Best wishes
David
I haven't studied all the links in this thread but a quick glance seems to indicate the commercial amps are all more (or less) sophisticated versions of Vanderkooy & Krauel
AES E-Library A Simple Reliable Power Amplifier with Minimal Component Count
If I was actually designing an amp for production this Millenium (instead of doing the 1ppm THD20k wank) I would use
____________________
I've spent a lot of time dreaming up new fangled PA topologies but have always come back to refine the basic Lin/Dinsdale circuit. No pedantry about why the modern topology isn't Lin/Dinsdale bla bla please.
In my book (and also according to Self & others) Lin/Dinsdale is a 3-stage amplifier
JLH & Cordell treat the Transconductance + VAS as one block for compensation.
Cherry treats the VAS + 'EF' output as a single block for compensation
I spurn the bastardised TMC
versions (hope I got dem TLAs rite)
_____________________
In fact Vanderkooy's circuit isn't 'new' either. Circa 1980, Great Guru Baxandall proposed something just like it to instruct me in the 'similarities' between CE & CC .. but as a thought experiment.
The real problem in trying to abandon Lin/Dinsdale, is getting good Class B. I would like to run 50mA or less through each pair of devices. I know the performance above is possible with the 'standard' EF2 circuit (without Self's extra EF feeding the VAS) with 'modern' devices .. with low crossover distortion.
The technology which made this possible is pure Cherry compensation which I've done to death in other threads.
What pure Cherry allows is
________________________
On CFPs for the 'EF output stage, apart from the Thermal issues I highlight in Bob's thread ...
Self shows in his 4th (and IIRC earlier) edition Fig 5.40 that though THD at full power may be better than EF2, at 1W & below, his EF2 is a LOT better than his CFP output stage.
He has no explanation for this. Anyone with the latest edition of Self and can tell us if he has investigated this further?
________________________
I can't comment on Cherry's IREE, Oz paper, not having seen it.
And the rail2rail OPA crowd must have something useful on this topic.
________________________
Dave, I tried to send you a PM but your mailbox is full
AES E-Library A Simple Reliable Power Amplifier with Minimal Component Count
If I was actually designing an amp for production this Millenium (instead of doing the 1ppm THD20k wank) I would use
- something like this
- attempt to get THD20k etc below 0.01%
- ensure good overload &
- unconditional stability with load as I've pontificated at length in several threads on this forum.
____________________
I've spent a lot of time dreaming up new fangled PA topologies but have always come back to refine the basic Lin/Dinsdale circuit. No pedantry about why the modern topology isn't Lin/Dinsdale bla bla please.
In my book (and also according to Self & others) Lin/Dinsdale is a 3-stage amplifier
- Transconductance i/p
- VAS
- 'EF' output
JLH & Cordell treat the Transconductance + VAS as one block for compensation.
Cherry treats the VAS + 'EF' output as a single block for compensation
I spurn the bastardised TMC
versions (hope I got dem TLAs rite)_____________________
In fact Vanderkooy's circuit isn't 'new' either. Circa 1980, Great Guru Baxandall proposed something just like it to instruct me in the 'similarities' between CE & CC .. but as a thought experiment.
The real problem in trying to abandon Lin/Dinsdale, is getting good Class B. I would like to run 50mA or less through each pair of devices. I know the performance above is possible with the 'standard' EF2 circuit (without Self's extra EF feeding the VAS) with 'modern' devices .. with low crossover distortion.
The technology which made this possible is pure Cherry compensation which I've done to death in other threads.
What pure Cherry allows is
- 'current driving' the EF OP stage so they behave like CE
- maximum feedback around the output stage which is always the biggest THD source
________________________
On CFPs for the 'EF output stage, apart from the Thermal issues I highlight in Bob's thread ...
Self shows in his 4th (and IIRC earlier) edition Fig 5.40 that though THD at full power may be better than EF2, at 1W & below, his EF2 is a LOT better than his CFP output stage.
He has no explanation for this. Anyone with the latest edition of Self and can tell us if he has investigated this further?
________________________
I can't comment on Cherry's IREE, Oz paper, not having seen it.
And the rail2rail OPA crowd must have something useful on this topic.
________________________
Dave, I tried to send you a PM but your mailbox is full
Last edited:
The ATC amplifier has been in use since the mid 1980s and was based (I believe) on a PA amplifier that was manufactured well before that. The Vanderkooy implementation is very similar!
Please see me earlier comments about the local feedback that is unique to the ATC design (I believe). It has the ability to null the transconductance differences between the two halves and renders significantly lower distortion with no compromise in stability.
Also to add is that in this topology MOSFETs behave particularly well and when high power and reliability are concerned, any on-paper advantages of BJTs are best left there...
This is the core of my enquiry but I am not sure if it's a problem.
CE OPS with same size emitter resistors as a typical EF seems like it should work, yes ?
I actually have some nice results on this issue since we last swapped ideas.
NOT what I expected intuitively, will discuss this more, this is just a quick reply to some of your points.
It is perfectly predictable from his own plots of the transfer functions of CFP versus EF.
"Sub-optimal", to put it politely.
I will scan it and at least put the circuit up for you and Arthur.
After almost 40 years I think this qualifies as "fair use"
That's what I expected too but damned if I can find much.
They seem more proprietary than they used to be.
I used to donate and had more mail space.
But donations are in US$ so it has become more expensive and the mail space shrunk when my donation expired.
Need to clean it out.
Best wishes
David
I am not convinced about the wisdom of a floated supply in amp optimized for really low distortion.
Makes much sense for Public Address reliability and is probably adequate for Hi-Fi in practice (i.e. not detectable in DBLT).
But the use of a CE OPS makes sense to me.
Drive it with an AD797, how simple is that?
And even ultra low distortion <0.001 @ 20 kHz should be doable.
Best wishes
David
Last edited:
My expectation was that it could couple into the feedback loop but I haven't checked this.
I did read an advocate of floated supplies, who admitted that mains noise transmission was one of the downsides so it seemed plausible.
Do you have a link to study?
Best wishes
David
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.