Crossover distortion, unless extremely bad (both output devices cut off), is monotonic. Gm-doubling distortion is definitely monotonic.ThorstenL said:
My understanding of monotonic is that the output direction does not change sign, so for a non-inverting amp the output never goes more -ve when the input goes more +ve. Strictly monotonic is stronger, and also rules out stationary output.
What do you understand by 'monotonic'?
Rather , a caracteristics function , the transfer function being its laplace
transform.
You will notice that a transfer function can be non monotonic
with GNFB if ever the amp/device loop gain is above 1 for phase output
being at -180°.
As for a caracteristics function , as you point it ,it can hardly be non monotonic.
transform.
You will notice that a transfer function can be non monotonic
with GNFB if ever the amp/device loop gain is above 1 for phase output
being at -180°.
As for a caracteristics function , as you point it ,it can hardly be non monotonic.
Hi,
Monotonic is used in two ways known to me for amplifiers, one refers to monotonic ordering of distortion products, the other to the nature of the distortion, the second one implies that distortion changes in step with the signal, that is more signal more distortion, less signal less distortion without any sudden changes.
Example - the distortion of a class A amplifier is monotonic even if it is odd order harmonics dominant, yet the distortion of a class AB ampliifer is never (cannot be) monotonic if dominated by the output stage, EVEN if the decay of the distortion products is always even order dominant.
The implication of monotonic in any context is a smoothly increasing or decreasing sequence.
Ciao T
Monotonic is used in two ways known to me for amplifiers, one refers to monotonic ordering of distortion products, the other to the nature of the distortion, the second one implies that distortion changes in step with the signal, that is more signal more distortion, less signal less distortion without any sudden changes.
Example - the distortion of a class A amplifier is monotonic even if it is odd order harmonics dominant, yet the distortion of a class AB ampliifer is never (cannot be) monotonic if dominated by the output stage, EVEN if the decay of the distortion products is always even order dominant.
The implication of monotonic in any context is a smoothly increasing or decreasing sequence.
Ciao T
Maybe that gets to the heart of the disagreement. I thought it meant that distortion always decreases as signal level reduces.
[edit: ...or Wahab's definition. I think that may be more important]
I had a look at the schematic for the Rotel RA-840BX4 (couldn't find RA840). One interesting thing is it has a double EF output stage, and the biasing instructions say to set the voltage drop across the emitter resistors to 2.5mV.
This is ten times lower than suggested e.g. by Doug Self, so I would expect fairly high crossover distortion. OTOH, the topology is pretty much as Thorsten described, so (unlike e.g. the "Blameless") it probably doesn't have distortion that increases badly with increasing frequency, or artificial emphasis on high-order distortion products.
Perhaps this gives a clue as to what's important. I'm not saying crossover distortion is OK, just that it may not be top of the "bad stuff" list.
Wide open-loop bandwidth, modest GNFB, no miller compensation, multiple output pairs, and it sounds OK despite having a bit of crossover distortion. Well, assuming this is representative of the Rotels that sound OK.
Last edited:
Indeed , i agree that your meaning is more logical , and i use it
myself generaly.....
As for crossover distorsion , it s easy to get rid of it but often
there s designs that do not explicitly adress this issue.
One thing important is , as you also pointed it, in PP output stage for exemple ,
to keep a meaningful voltage difference between the two halfs
command bases, as it s mandatory that at least one device is conducting
at any given moment, as if both device stop simultaneously conducting ,
the amp s open loop gain can collapse to the point that it will
be lower in the up of the audio band than the amp s supposed
closed loop gain , thus GNFB will be absent and crossover distorsion
will be awfully high.
OK, here is my suggestion for a THD-type metric. The idea is that higher orders sound worse, and amps spend much of their time well below max output (same as GedLee).
1. Find/pick the max output level for the amp - perhaps dominant harmonic reaches 5%?
2. Drop down 50% in voltage (-6dB) and measure harmonics, then combine using the Shorter weighting (harmonic n gets multiplied by n<sup>2</sup>/4).
3. Drop to 10% (-20dB from max) and repeat.
4. Drop to 1% (-40dB from max) and repeat.
5. Combine these three numbers in quadrature, so the highest one dominates the result.
With modern test equipment this should be possible. For a 40W amp this means testing it at 10W, 100mW and 1mW. For smooth amps such as genuine Class A should be dominated by the first figure, whereas P-P Class B/AB may get worse at lower levels. Could also do this as a function of frequency.
I know THD is not everything, so something similar for IMD is needed too (and other things).
IMD test proposal:
Generate three tones as follows: 5000Hz, 5001Hz both at 25% of max voltage to provoke non-linearity at 1Hz, and 270Hz at 10%. Look for 1Hz sidebands on 270Hz tone. Repeat at lower signal levels, and at other difference frequencies. The idea is to look for both normal 3rd-order IM, and the effect of recycled 2nd-order IM. Could also look for sidebands at mains frequency and harmonics.
1. Find/pick the max output level for the amp - perhaps dominant harmonic reaches 5%?
2. Drop down 50% in voltage (-6dB) and measure harmonics, then combine using the Shorter weighting (harmonic n gets multiplied by n<sup>2</sup>/4).
3. Drop to 10% (-20dB from max) and repeat.
4. Drop to 1% (-40dB from max) and repeat.
5. Combine these three numbers in quadrature, so the highest one dominates the result.
With modern test equipment this should be possible. For a 40W amp this means testing it at 10W, 100mW and 1mW. For smooth amps such as genuine Class A should be dominated by the first figure, whereas P-P Class B/AB may get worse at lower levels. Could also do this as a function of frequency.
I know THD is not everything, so something similar for IMD is needed too (and other things).
IMD test proposal:
Generate three tones as follows: 5000Hz, 5001Hz both at 25% of max voltage to provoke non-linearity at 1Hz, and 270Hz at 10%. Look for 1Hz sidebands on 270Hz tone. Repeat at lower signal levels, and at other difference frequencies. The idea is to look for both normal 3rd-order IM, and the effect of recycled 2nd-order IM. Could also look for sidebands at mains frequency and harmonics.
Hi,
You can translate the "transfer function" of any Amp output stage into a "wingspread diagram".
If there is (for example) a smooth gain falloff (or rise) towards the extremes, we may call the system producing the graph monotonic, if the transfer function is not monotonic we can see discontinuities and up/down undulations... It may be worth looking at those for Class AB Amp's...
Ciao T
You can translate the "transfer function" of any Amp output stage into a "wingspread diagram".
If there is (for example) a smooth gain falloff (or rise) towards the extremes, we may call the system producing the graph monotonic, if the transfer function is not monotonic we can see discontinuities and up/down undulations... It may be worth looking at those for Class AB Amp's...
Ciao T
DF96. Your proposal in post 610 is an interesting one and certainly should go a long way toward a decent SQ metric. My audio mentor, Mr. Hiraga, often pointed out that good amps not only have a favorable harmonic structure, but can maintain it a different power levels and into complex loads.
It does not take much measuring to see that the distortion spectrum changes with power levels. It takes a bit more to see that the harmonic fingerprint also changes with frequency and into complex loads. I've seen amps where one harmonic changes depending on load while the the other harmonics don't. Certainly something like that is important.
As some others have stated in this thread, it would be nice to boil down SQ into a few easy numbers. Getting there won't be easy, tho.
It does not take much measuring to see that the distortion spectrum changes with power levels. It takes a bit more to see that the harmonic fingerprint also changes with frequency and into complex loads. I've seen amps where one harmonic changes depending on load while the the other harmonics don't. Certainly something like that is important.
As some others have stated in this thread, it would be nice to boil down SQ into a few easy numbers. Getting there won't be easy, tho.
A lot of discussion on thermal modification of transistors, has anyone considered the various issues in caps? They are variable with respect to capacitance, DF, esr and leakage due to voltage, temperature, and humidity. Various film types are more stable than say, z5u or x7r multilayer monolithic, but none the less very imperfect devices. If I remember, inductance is pretty stable.
It was actually Walt Jung's Audio article in the 70's that led me to solve many issues back in my lab days. Of course the seminal paper on z5u's was done in the FA lab at Goddard after a shuttle launch failure that tied the earlier Japanese humidity problem together with the low voltage carbon track issue. Combined with the mechanical termination problems of axial glass packaging, we assumed over 5% defective caps installed on our digital boards. With logic voltages dropping from 5 down to 1.5, I wonder how the low voltage carbon migration was solved?
Back when I was current, polystyrene was the hot ticket for audio, but they were big, expensive and fragile, so we settled for mylar. There were debates over the self healing features of deposited film vs foil and the termination issues. Manufacturing improvements march on, so I don.t know the current thinking.
It was actually Walt Jung's Audio article in the 70's that led me to solve many issues back in my lab days. Of course the seminal paper on z5u's was done in the FA lab at Goddard after a shuttle launch failure that tied the earlier Japanese humidity problem together with the low voltage carbon track issue. Combined with the mechanical termination problems of axial glass packaging, we assumed over 5% defective caps installed on our digital boards. With logic voltages dropping from 5 down to 1.5, I wonder how the low voltage carbon migration was solved?
Back when I was current, polystyrene was the hot ticket for audio, but they were big, expensive and fragile, so we settled for mylar. There were debates over the self healing features of deposited film vs foil and the termination issues. Manufacturing improvements march on, so I don.t know the current thinking.
If it seems to be preferable to have less global feedback and more local stage, then it would make some sense why the chip-amp club is quite limited as this is fixed by the original designers giving the application integrator no leeway. Where the opportunity in a chip to use many more transistors and means of compensation to produce amazing new op-amps, unless they choose to change their design concept the chip amps are limited. Reasonable extraction?
audio power chip amps don't have many external design degrees of freedom, and not too many options - discrete or hybrid design certainly still makes sense
for small signal applications you can't buy, effectively apply discrete Q of the quality/speed of the last decade's advanced analog IC semi fab capabilities used in recent op amps
the internal design of op amps has advanced by a discontinuous step up from the huge investment by big companies with already deep institutional knowledge in the competition for medical imaging, telcom A/DSL markets demanding low distortion at 100 kHz thru low MHz - in circuits using global feedback
Curl's favorite examples, "proving" global feedback is "bad" use comparisons with 30 year old monolithic op amps - right up to Ron Quan's 2010 paper - he didn't even test a TL07x - he used the low power TL06x part instead in order to (be able to) measure phase modulation errors
for home audio small signal applications only phono preamps require lower noise parts than are available in monolithic op amps
for small signal applications you can't buy, effectively apply discrete Q of the quality/speed of the last decade's advanced analog IC semi fab capabilities used in recent op amps
the internal design of op amps has advanced by a discontinuous step up from the huge investment by big companies with already deep institutional knowledge in the competition for medical imaging, telcom A/DSL markets demanding low distortion at 100 kHz thru low MHz - in circuits using global feedback
Curl's favorite examples, "proving" global feedback is "bad" use comparisons with 30 year old monolithic op amps - right up to Ron Quan's 2010 paper - he didn't even test a TL07x - he used the low power TL06x part instead in order to (be able to) measure phase modulation errors
for home audio small signal applications only phono preamps require lower noise parts than are available in monolithic op amps
Last edited:
This is unfortunate. While it is true that my 'examples' were derived from past work from as long as 35 years ago, and that Ron Quan's 'examples' were rather dated as well, it should be understood that we both test modern IC's as well. It is just that they are not such OBVIOUS 'examples' of the problems with op amp design and use. When we design, or even write a paper, often, REASONABLE WORST CASE is what we normally use, because that is what other engineers will probably try to get away with, for cost and convenience reasons, if they are not shown the problems generated by doing so.
As far as 'listening' and trying IC's, we get many of the latest products and invariably hope that we find the 'perfect' solution to our needs. Sometimes, they work well enough, like in my latest JC-3 phonostage, so that I get the 'phonostage of the year' in the Jan 1012 TAS, and I breathe a sigh of relief. Sometimes they are initially designed into my products and have to be 'ripped out' or replaced, after reviewers reject my design effort.
Yet, in both cases, it would be difficult to MEASURE the reason for the acceptance or rejection of the IC in its location, in my designs.
Therefore I hope to promote all-out efforts in audio design to show what CAN be possible, rather than just, 'good enough' and I have a customer base that supports it. Then, later, the successful approaches can be made in more 'cost effective' form for the rest of us.
It is a little like Honda making racing engines first and then putting the technology learned into a Honda Civic, for example.
As far as 'listening' and trying IC's, we get many of the latest products and invariably hope that we find the 'perfect' solution to our needs. Sometimes, they work well enough, like in my latest JC-3 phonostage, so that I get the 'phonostage of the year' in the Jan 1012 TAS, and I breathe a sigh of relief. Sometimes they are initially designed into my products and have to be 'ripped out' or replaced, after reviewers reject my design effort.
Yet, in both cases, it would be difficult to MEASURE the reason for the acceptance or rejection of the IC in its location, in my designs.
Therefore I hope to promote all-out efforts in audio design to show what CAN be possible, rather than just, 'good enough' and I have a customer base that supports it. Then, later, the successful approaches can be made in more 'cost effective' form for the rest of us.
It is a little like Honda making racing engines first and then putting the technology learned into a Honda Civic, for example.
John, you got some props today. I stopped by about the only stereo store left within a 100 miles after getting my wallet cleaned out at the woodworking show. Anyway, this shop still has a nitty-gritty machine in the window, so that should be a hint. I said I was looking for a small integrated with remote for my guest room. He said, "well, if you need Phono, you want one designed by John Curl."
Then he went on trying to convince me tubes don't generate distortion and I needed some $5000 KT88 unit. I just smiled and nodded my head, checked on what else they carried, thanked him and went on my merry way.
My parts to refurb my DH-20 came, so off the to the workbench!
Then he went on trying to convince me tubes don't generate distortion and I needed some $5000 KT88 unit. I just smiled and nodded my head, checked on what else they carried, thanked him and went on my merry way.
My parts to refurb my DH-20 came, so off the to the workbench!
- Status
- Not open for further replies.