From my experiments with multi-tone testing, I saw that multiple tones indeed beget more harmonics, but statistically, they do in general not fall at the same frequency where they could reinforce each other or cancel out. Which is no surprise of course. But maybe it would manifest itself as an increase in perceived noise level?
jan
There are limits on “how many tones” we can distinguish, ~640 from my Psychoacoustics book
but human perception is more limited in its resolution of multicomponent signals – only 24 critical bands are used in explaining masking phenomena
Chris Montgomery – originator of Ogg Vorbis has posted here that perceptual lossy compression uses only ~6-7 bits per critical band
over the years I have posted a few comments on distortion, perception, measurement with references, links
AP High Performance Audio Analyzer & Audio Test Instruments : Downloads
presumably modern op amp design teams have at least someone familiar with:
Distortion Analysis of Analog Integrated Circuits (The Springer International Series in Engineering and Computer Science): Piet Wambacq,Willy M.C. Sansen: 9780792381860: Amazon.com: Books
but human perception is more limited in its resolution of multicomponent signals – only 24 critical bands are used in explaining masking phenomena
Chris Montgomery – originator of Ogg Vorbis has posted here that perceptual lossy compression uses only ~6-7 bits per critical band
over the years I have posted a few comments on distortion, perception, measurement with references, links
AP High Performance Audio Analyzer & Audio Test Instruments : Downloads
presumably modern op amp design teams have at least someone familiar with:
Distortion Analysis of Analog Integrated Circuits (The Springer International Series in Engineering and Computer Science): Piet Wambacq,Willy M.C. Sansen: 9780792381860: Amazon.com: Books
I bought that when it came out, not a bad book. I've seen OOP technical go MUCH higher than that.
The other category that gets pricey is books officially in print but of which no one has stock. And then to compound matters the data entry is sometimes done incorrectly and the misspellings make copies hard to find. Van der Ziel's books are often found with the author's name as Zeil, for example.
I have a copy of Robert Langlands' The Zeta Functions of Picard Modular Surfaces. If one searches including the author's name in the fields, nothing appears. But by title only, one outfit on Amazon has a copy of the title by "Janglands"
Its condition is described as "used, acceptable". The price: a mere $1979.95, not including shipping
Now how long this will sit in their inventory is another story.
As you add more tones the power in each must be reduced or the crest factor will cause clipping. At 20k each tone would be very small and any one combination of two tones would give vanishingly small distortions (far less than .01%).
EDIT Sorry PMA already said that but let me add that in doing these multitone tests for years the results were always well behaved and for the most part could be predicted by simple IMD testing at an appropriate level.
Last edited:
Springer has a booth at every ISSCC and offer a lot of IN print books on topics related to highly specialized interests for $200-$300 or even more.
Ah for the gravy days when I would buy anything when I saw it that even remotely might be of use. I think I've bought maybe three tech titles this year so far, the most recent one Basso's latest on switchmode power supply loop compensation, which ships next month according to theory.
Since I am also running out of room, even with two 10 x 30 storage spaces, it's fairly easy to resist acquisitions that aren't very pertinent.
The more I read of these discussions, the more it seems that the world needs good clipping indicators on things like preamps and power amps. I intend to include them on every output of a fancy mastering EQ box I'm designing, even though it will sport stupendous overload margins.
True enough.... it can just increase the percieved noise floor and mask fine structure... detail if anything. But, I usually look at the whole chain view.... the harmonics of each amplifier would add to the next harmonic generator/amp stage in the system. Then there are phase effects on harmonics at each end of spectrum that shifts harmonics for more or less additive (esp. with tube/transformer circuits).
Then there can be the subjective effect of all those harmonics masking and/or acting as noise in some cases. Until rigorous testing has been done on the whole of it, we can only get a few markers or clues. Clearly a single tone thru an amplifier with low harmonic levels isnt a clue to the sound of mulit-frequencies/music going thru an amp or system with many more harmonics present. Maybe 640 tones is the worst case number to use. Doesnt matter -- just need to standardize on a reasonable number (withOut the MP3 like codec justifications, pls) and then see if that correlates better to what people say they hear in the amp or system. It would certainly be a more realistic testing which might give greater insights. Thx - RNM
Last edited:
Would it bring enough gain to keep Id of the JFets unmodulated?
But agreed, let's elaborate on this in bcarso's special thread.
Definitely also interested in this.
..and I have to apologize to RNMarsh (and other DC contributors) for my rough words earlier, I agree that DC and thermal stability must not be ignored - just had the feeling that we had put to much efforts on that, while not discussing the AC topics!
Since reading the last pages I feel more on an audiophile track again. Thanks to all for that.
I made new my schem.
This amp has good performance characteristics and rail to rail input and output.
amp works from +/-7v
Amp delivers 3 Vrms of out with thd lower then 0,001%
Pot R1 - dc offset nulling
Resistor R3 change quiescent current of the output stage.
schem:
This amp has good performance characteristics and rail to rail input and output.
amp works from +/-7v
Amp delivers 3 Vrms of out with thd lower then 0,001%
Pot R1 - dc offset nulling
Resistor R3 change quiescent current of the output stage.
schem:
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
No problem-o Thx for saying so. Fortunately I have pretty thick skin. Now we are friends again
-RNMarsh
Choco, the bias of the folded cacode could also be referenced to the input stage on top of the transdiode. I would then make the voltgae over the output Fet higher and a string of Leds could do that. Sorry, this may be off totic but i have not seen that bcarso has opened that road until now.
I would love to be more informed on distortion and masking etal. BUT does that mean we cant do anything except the same single tone or two tone Im test? Are we back to that so quickly? A lot has been studied and tested and written about. yet we keep doing single tone tests. If multi-tones aint it...Can we get suggestions for a more complete set of tests that are meaningful for our SOTA opamp being designed here? There's designing it and there is testing it. How will we know it is really better than a simpler circuit or something else? -RNmarsh
Last edited:
I'll try to get to this soon, but don't wait for me --- I will be pleased if anyone jumps in and starts it. I am easily distracted
I will say that after a cursory review of two patents I was kindly sent, I think that at least one aspect of linuxguru's basic approach, namely the nearly constant operating point of the main device, both current and voltage, is not covered per se.
friends again
...now I have to think again - difficult after a day that drove my brain in sort of a synaptic collaps.
No need to be Manichean. Lots of good test signals out there.
The fundamental problem to me is that no-one has agreed on an actual, specific design goal. Vague target means that the conversation wanders all over the place and nothing definitive will ever come out of it.
I use dynamic measurements with very high resolution all the time. I had a lot of hope to find a correlation to audible quality. Curious enough the designs that had very low harmonic distortion measured the conventional way with single tone up to 10kHz also scored best on the dynamic tests. For example an AD797 scores great with gain up to x10 challenging my equipment. Can i build a better sounding line stage with gain of x10 compared to an AD797 with the same gain ? You bet, but why ? Added distortion that makes the sound "fuller" or more "natural" ? I do not think so, i can not hear better then -60dB, low order, and my line stage is much better at around -100dB THD.
See an old post from 2009 on my MPP thread. That was the time i measured much but i have given up.
See an old post from 2009 on my MPP thread. That was the time i measured much but i have given up.
Attachments
Harmonics will not be seen. The idea is that noise will increase. But how this
noise will be measured in the presense of such a big population of main tones?
But new relevant methods are already proposed
:http://www.listeninc.com/PDFs/paper_121AES_non_coherent_distortion_Listen_Inc.pdf
In any case, for an amateur
, the two or three tone tests conducted at various levels and with a real load connected at the output are still OK.The resulting new frequencies all fall within the audio band, most of them within the ear’s most sensitive band (1-3Khz)
Two tones: The 19 kHz and 20 kHz equal-level sine waves test
Three tones: The 20.00 kHz, 10.05 kHz, and 9.00 kHz equal-level sine waves test
From Bob Cordell book, CHAPTER 22 “Distortion and Its Measurement”
George