No, because the forward path characteristics (singularities) before and after the loop is closed are not the same.
Re: Phase Intermodulation and FM distortion
Mr Cordell,
What you described here sound to me like the most obvious kind of non linearity: deviation of the gain vs amplitude function from a straight line.
This also causes harmonic distortion, intermodulation, and certainly much more.
RF guys have been measuring this for a while to describe amplifiers, front ends... with the dual tone intermodulation test, and the 1dB compression point. Maybe we (analog, low freq, audiophiles...) should also use this kind of tests more often.
(although I'm not sure if the 1dB compression point can also take into account your example of the collector-base capacitance because it's a dynamical behaviour and the -1dB thing is a static representation)
Bob Cordell said:
Mr Cordell,
What you described here sound to me like the most obvious kind of non linearity: deviation of the gain vs amplitude function from a straight line.
This also causes harmonic distortion, intermodulation, and certainly much more.
RF guys have been measuring this for a while to describe amplifiers, front ends... with the dual tone intermodulation test, and the 1dB compression point. Maybe we (analog, low freq, audiophiles...) should also use this kind of tests more often.
(although I'm not sure if the 1dB compression point can also take into account your example of the collector-base capacitance because it's a dynamical behaviour and the -1dB thing is a static representation)
Traderbam,
You are right; one normally shouldn't take a design intended for open-loop operation and then just wrap a feedback loop around it. However, in most cases in audio amplifiers, the change in design has mainly to do with proper feedback compenastion and open-loop gain. Those aspects of linearizing the open loop amplifier would often be pretty much the same, in that you do the best you can.
One caveat here: some manufacturers, knowing that they will be able to reduce distortions by the subsequent application of negative feedback, "optimize" their open loop amplifier economically by cutting corners and skimping on open-loop linearity. Such designers, who use NFB as a crutch to allow cheap open loop design, are the ones who generally have given NFB an undeserved bad rap.
Bob Cordell
You are right; one normally shouldn't take a design intended for open-loop operation and then just wrap a feedback loop around it. However, in most cases in audio amplifiers, the change in design has mainly to do with proper feedback compenastion and open-loop gain. Those aspects of linearizing the open loop amplifier would often be pretty much the same, in that you do the best you can.
One caveat here: some manufacturers, knowing that they will be able to reduce distortions by the subsequent application of negative feedback, "optimize" their open loop amplifier economically by cutting corners and skimping on open-loop linearity. Such designers, who use NFB as a crutch to allow cheap open loop design, are the ones who generally have given NFB an undeserved bad rap.
Bob Cordell
mikeks said:
OK, so is there another way to keep the same basic non-global feedback 30dB amp and somehow wrap gain & feedback around it so we can compare with and sans feedback with the same basic amp? I'm sure you can come up with an ingeneous scheme!
Jan Didden
If one matched two issues on this topic, than might find a conclusion: Higher PIM amplifiers, which usually have higher gain in stages preceding compensation network are more prone to RFI applied to differential stage.
This way unhearable PIM might be a symptom of something potentially hearable. Sorry for generalizations.
regards
Adam
This way unhearable PIM might be a symptom of something potentially hearable. Sorry for generalizations.
regards
Adam
Comparing NFB and NoFB Amplifiers
Michael, you make a good point. There is probably no ideal, totally fair way to compare the two in general. To me, the debate is on whether NFB is a good thing or a bad thing in an end product. For this reason, the comparison I think is relevent is to comapre two end-product-like amplifiers built with the same care and expense, not worrying that one or the other needs certain cost-neutral accomodations (like FB compensation) for its NFB or NoFB architecture. Equal-cost things that are important are same type and speed of output transistors, same type and amount of Class AB biasing, etc. However, even within the constraints I have mentioned, I admit there are many different ways to do the comparison. As a starting point, a comparison of PIM between an Ayre amplifier and a Boulder amplifier might be very interesting.
Bob Cordell
Ultima Thule said:
Michael, you make a good point. There is probably no ideal, totally fair way to compare the two in general. To me, the debate is on whether NFB is a good thing or a bad thing in an end product. For this reason, the comparison I think is relevent is to comapre two end-product-like amplifiers built with the same care and expense, not worrying that one or the other needs certain cost-neutral accomodations (like FB compensation) for its NFB or NoFB architecture. Equal-cost things that are important are same type and speed of output transistors, same type and amount of Class AB biasing, etc. However, even within the constraints I have mentioned, I admit there are many different ways to do the comparison. As a starting point, a comparison of PIM between an Ayre amplifier and a Boulder amplifier might be very interesting.
Bob Cordell
janneman said:
If we don't want to penalize either amplifier, both implementations probably must look quite different.
The closest thing I could think of would be to make a lot of equal
gain blocks, each with local feedback only and a gain of 3 or 6 dB.
Then we could use perhaps 10 times 3 dB to make the nonglobal FB
30 dB amp. We could assume that distortion tends to be generated
in the final stages mostly. We would use Class A to make sure that the
first stages aren't punished by crossover distortion.
For the global feedback version, we could add a few stages in the front to get some excess gain.
Still we would need a blameless subtractor to close the loop and the o.l. bandwidth for the global loop version
would take a small hit.
These amplifiers would at least use the same mechanism for gain.
Maybe we could use some extra stages all the time and make up for them with a front attenuator in the o.l. case.
That should be easy to simulate with subcircuits.
Gerhard
Few years ago, I made a measurement of the amp attached, measurement here:
http://web.telecom.cz/macura/pma1_measure_en.html
and this amplifier:
http://web.telecom.cz/macura/pma2_en.html
They differed only in class A output stage outside or inside feedback loop. Both measured very low distortion, sounded pretty different.
http://web.telecom.cz/macura/pma1_measure_en.html
and this amplifier:
http://web.telecom.cz/macura/pma2_en.html
They differed only in class A output stage outside or inside feedback loop. Both measured very low distortion, sounded pretty different.
Attachments
PMA said:
Quite interesting! Which one sounded best to you? Is it possible to describe the differences?
Jan Didden
lumanauw said:
Lumanauw,
I guess I'm a little confused at what I'm interpreting to be two different observations that you seem to have interrelated. I'll explain my take on what you've presented so far and then explain my last post (briefly, correct me if I am not getting the essence).
In your earlier series of post you described a experiment where you applied a 1khz signal to your amp, monitored the output with your distortion analyzer and inspected the residual output on your scope. With no load the output was primarily distortion. Changing nothing in the setup you connected an 4 ohm load and observed spikes on the residual. This I do not know what to make of but it would be interesting if I could duplicate it to see what had the most influence on the spikes presence. What knob to turn so to speak. I did not respond to this due to time constraints and my lack of ability to duplicate your test at the time.
The second observation I see as an evolution of the first, sort of a left field approach to determining an amps sensitivity to the noise emitted by a SMPS. I interpreted your post as suggesting that by placing a noise source where it can couple to the speaker cable and monitoring the amplifier output, one could determine an amps susceptibility to noise pickup. I liked the suggestion because it mimics what might happen in the real world; only worse.
This made sense to me, in fact I had a slightly worried sense that the test might prove a weak point in the layout and my understanding might be sent back to the drawing board.
To perform the test I set up the system with the PicoScope monitoring the amplifier output, displaying an FFT view with a 160 kHz bandwidth; no signal, just the noise floor with the preamp volume set at a normal volume setting. I used the power supply body as a probe moving along the speaker wires, over the chassis, power supply and amp circuitry. The results were as I posted with the actually input stage and voltage amp (in very close proximity to each other) being the only susceptible parts of both the amp and preamp.
I actually think that with a bit of cleaning up this approach would be a good DIY approach to determining the effectiveness of a layout/grounding scheme.
Regards, Mike
feedback - no feedback test
With regards to the feedback / no feedback PIM test, the following can be done.
Assume the amplifier under test – no feedback – has a transfer function A(s) with a DC gain A(0) = Ao. Now we put in front an ideal stage of gain Aa (we assume it independent of frequency in the range of interest) and close the feedback loop with a factor B. System gain is now:
G(s) = A(s).Aa/(1 +A(s).Aa.B)
We want to make G(0) = Ao so system gain matches the original bare amplifier gain. Given an arbitrary selected value for Aa, then we can solve for B:
B = (Aa –1 )/Ao.Aa
And substituting:
G(s) = A(s).Aa/(1+(A(s)/Ao ).(Aa-1))
It may be easily verified that G(0) = A(0) as required.
In this conditions, the output level will be exactly the same for equal input levels, both for the bare amplifier and for the compound amplifier with feedback, implying internal signal levels for the amplifier under test will also be the same, the only difference being its actual input signal is not the original test signal alone but the composite of the test signal and a correction term.
Due to feedback, system frequency response will change as poles are moved around as expected, and transfer error will be attenuated by an amount approximately equal to A(s).Aa.B as noted in the previous post .
But the amplifier will be operating essentially under the same stress as if alone, and the distortion mechanisms will be the same.
Rodolfo
With regards to the feedback / no feedback PIM test, the following can be done.
Assume the amplifier under test – no feedback – has a transfer function A(s) with a DC gain A(0) = Ao. Now we put in front an ideal stage of gain Aa (we assume it independent of frequency in the range of interest) and close the feedback loop with a factor B. System gain is now:
G(s) = A(s).Aa/(1 +A(s).Aa.B)
We want to make G(0) = Ao so system gain matches the original bare amplifier gain. Given an arbitrary selected value for Aa, then we can solve for B:
B = (Aa –1 )/Ao.Aa
And substituting:
G(s) = A(s).Aa/(1+(A(s)/Ao ).(Aa-1))
It may be easily verified that G(0) = A(0) as required.
In this conditions, the output level will be exactly the same for equal input levels, both for the bare amplifier and for the compound amplifier with feedback, implying internal signal levels for the amplifier under test will also be the same, the only difference being its actual input signal is not the original test signal alone but the composite of the test signal and a correction term.
Due to feedback, system frequency response will change as poles are moved around as expected, and transfer error will be attenuated by an amount approximately equal to A(s).Aa.B as noted in the previous post .
But the amplifier will be operating essentially under the same stress as if alone, and the distortion mechanisms will be the same.
Rodolfo
Distortion measurement
Bricolo,
Yes, we can borrow some distortion measuring techniques from the RF guys, although I'm not sure 1-dB compression would be that useful. Different distortion measurements are often just means of looking at different syptoms of the same underlying nonlinearity. However, some distortion tests are better at stimulating a given nonlinearity than others. The other half of the problem is how the results of the nonlinearity make their way to the output of the device being tested. For example, doing an end-to-end THD-20 test on a compact disc, from recording to playback, would typically be fruitless, due to the brick-wall bandlimiting of the anti-alias and reconstruction filters. At the same time, a multitone intermodulation test would expose the nonlinearity, both for odd order and even order, with in-band IM products.
Bob Cordell
Bricolo,
Yes, we can borrow some distortion measuring techniques from the RF guys, although I'm not sure 1-dB compression would be that useful. Different distortion measurements are often just means of looking at different syptoms of the same underlying nonlinearity. However, some distortion tests are better at stimulating a given nonlinearity than others. The other half of the problem is how the results of the nonlinearity make their way to the output of the device being tested. For example, doing an end-to-end THD-20 test on a compact disc, from recording to playback, would typically be fruitless, due to the brick-wall bandlimiting of the anti-alias and reconstruction filters. At the same time, a multitone intermodulation test would expose the nonlinearity, both for odd order and even order, with in-band IM products.
Bob Cordell
Hi, Mike,
Reading your post, you are right. There are 2 things, one is my experiment, the other thing is "suggestion" to use SMPS to inject disturbance to NFB loop.
Reading your post, you are right. There are 2 things, one is my experiment, the other thing is "suggestion" to use SMPS to inject disturbance to NFB loop.
With no signal (and no load), I don't know if your experiment and my experiment is the same thing? I think the fundamental wave is important in this experiment. This fundamental wave(+load) will be forcing the amplifier to "work", then we can see what "leftovers" from this "work". As you can see, the spikes only appear if there is load (dummy or real speaker). Without load, the spikes don't appear.
Re: feedback - no feedback test
Rodolfo,
This was what I had in mind, although you explain it much more concise than my feeble attempts. I presently am experimenting with a non-global feedback amp with 30dB gain, so one of the building blocks I have available.
Something to try out for me!
Jan Didden
ingrast said:
Rodolfo,
This was what I had in mind, although you explain it much more concise than my feeble attempts. I presently am experimenting with a non-global feedback amp with 30dB gain, so one of the building blocks I have available.
Something to try out for me!
Jan Didden
janneman said:
Jan,
I made a number of A-B listening tests with these 2 modifications of the same amplifier. I invited about 10 visitors, always 1 person for 1 test. Everybody took his favorite CDs to listen to with him. All of the listeners were able to find differences in A-B testing and to estimate which of the amp was playing (A or B). Preferences were about 60% to 40% for the non-global FB version. Preferences depended on kind of music the listeners had preferred. The people who listened mostly to club jazz and pop music preferred the non-global FB version. Those who liked classical orchestral music preferred the global FB version (me too). There is no universal truth in listener's taste 😉 .
Regards,
Pavel
PMA said:
Pavel,
I can partly relate to this in that in my experience, club / jazz / solo music is better to bring out differences in amps than massive orchestral works. This may be because WE can better identify this if the musical structures are simpler, rather than that the amps sound different withy different types of music, though.
Jan Didden
janneman said:
Jan,
for me, as a test material, I feel it different. For me the classical music, philharmonic orchestra, music like Beethoven, is the best test material. It discovers amp's possibilities and performance, it is difficult to play it clean with high resolution, and non-agressive.
On the other hand, I am very suspicious when I read about "great air, space" etc. In my experience, "air" and "space" are often another words for distortion of different kind. When this "Ayry" 😀 amplifier tries to play complex classical music, the result is often poor resolution and great mismatch.
Regards,
Pavel
Re: Re: feedback - no feedback test
Hi Jan,
This could be an interesting experiment. The "ideal" front amplifier could be an OPA 134 or LM4562 in an Aa=40 dB gain configuration for example, to have a meaningful increase in gain yet not introducing too much additional auxiliar amplifier distortion. Layout should be tricky with regards to stability etc.
As results I place my bets in a lower overall distortion level with a different spectral structure because of the added correction term, in comparison with the test amplifier alone.
Rodolfo
janneman said:
Hi Jan,
This could be an interesting experiment. The "ideal" front amplifier could be an OPA 134 or LM4562 in an Aa=40 dB gain configuration for example, to have a meaningful increase in gain yet not introducing too much additional auxiliar amplifier distortion. Layout should be tricky with regards to stability etc.
As results I place my bets in a lower overall distortion level with a different spectral structure because of the added correction term, in comparison with the test amplifier alone.
Rodolfo
traderbam said:
I think I said that in my post. I am better able to hear differences in amps if i listen to "simple" music. So, I thought, maybe that is because we can better do this, rather than that the amp would perform differently with different music.
I know that more frequency components in the signal will lead to more distortion (IM etc) components, so you could say, complex music leads to worse sound quality in a given amp. I'm not sure that is true because from a freq content point of view 'simple' music already has a lot of freq components too. Complex music also has more masking of distortion components.
I realize this is a point of view, not a proven fact. But you got to start somewhere 😉 .
Jan Didden