Hi,
Amplify the residual by (say) 60dB and listen to it.
I find often that once I heard the residual in isolation I can afterwards identify it clearly in listening, if it sounds objectionable...
Ciao T
Yes, my tests were before I learned about Bill's DiffMaker. Interpreting the results is still no easy task, tho. So I get a difference - how good or bad is it? I mean to the ear.
Amplify the residual by (say) 60dB and listen to it.
I find often that once I heard the residual in isolation I can afterwards identify it clearly in listening, if it sounds objectionable...
Ciao T
I do it with headphones. I put the music on one channel and put the residual on the other channel.
Ivigone and VK, I commend both of you, especially, because you have shown a strong sensitivity to what I have found in my own design work. I consider the effect of just about everything, from power supplies, wiring, circuit thru-path, etc., etc. It is the only way to succeed in making a 'world class' product. The real competition out there, does the same thing.
For everyone else, be not too sure about your complete understanding of phase distortion. There is more evidence out there from Bell Labs, etc., than is commonly published.
Last years I have got a strong feeling, that standard vision of distortions, and of their role in audio, that described and discussed by many professionals, can only lead to a mid-fi level designs, that something important, is not explicitely shared with wide circles of DIYers. I do not blame anybody, it is difficult to judge who is open minded, and who is not.
At least, John Curl, I believe, can not be blamed in being not opened to others.
Yes, it is really Richard Heyser's fault. He started all this when people believed that the ear was 'phase deaf' to monophonic aberrations.
Back in the day, the 1960's, scientists believed that the only factor important in speaker design was amplitude, and as long as you held the effective path differences between the drivers to less than the delay that an 'echo' could be detected, you were OK. That included then, the K-horn and many other speakers of the past.
When he gave his first paper on this, he got protests from the audience at the AES, listening to his lecture. And he only stated, that in some circumstances, perhaps a 2 foot path difference might be detected.
Back in the day, the 1960's, scientists believed that the only factor important in speaker design was amplitude, and as long as you held the effective path differences between the drivers to less than the delay that an 'echo' could be detected, you were OK. That included then, the K-horn and many other speakers of the past.
When he gave his first paper on this, he got protests from the audience at the AES, listening to his lecture. And he only stated, that in some circumstances, perhaps a 2 foot path difference might be detected.
Yes, you are correct, VK. That is why I have tried to keep this thread running for so long and through a lot of adversity. We need to be reminded that there is a difference between mid fi and hi fi (or hi end as we often find it). In mid fi, they work to a spec, with cost and build-ability in mind. Many of us make mid fi, even me, on occasion, when I am asked to make something cheap, but reasonably good sounding. Right now, I am playing with a 'cost effective' headphone amplifier, and I bet that it will sound pretty good, but NOT AS GOOD, as my direct drive triode STAX power amp, which uses all discrete devices. However, I will give it my best shot.
One must presume that ones listening experience, especially if proven reliable in the past, is real, and not just the product of external influences entirely, if one is to move forward to produce something more ideal than something already mass produced. After all, there are very good designers out there, perfectly happy with their design, and the 'proof' that was generated in a double blind test, that virtually anything they made cannot be discerned from anything else of comparable measurement.
One must presume that ones listening experience, especially if proven reliable in the past, is real, and not just the product of external influences entirely, if one is to move forward to produce something more ideal than something already mass produced. After all, there are very good designers out there, perfectly happy with their design, and the 'proof' that was generated in a double blind test, that virtually anything they made cannot be discerned from anything else of comparable measurement.
Speaking about single-arbitrary-shape pulse test, we should remember, that, being expressed as a sum of harmonics, the pulse will be converted in tooo long sum of them (infinite-like). And the harmonics are quite definitely aligned in the time (phase) domain.
I doubt that situation in this case is close to the multy-tone (even hundred tone) IMD tests.
Would be interesting to know a good metrics, how to estimate numerically the difference, between the input and output pulses.
I doubt that situation in this case is close to the multy-tone (even hundred tone) IMD tests.
Would be interesting to know a good metrics, how to estimate numerically the difference, between the input and output pulses.
Chris, you are starting off well, but you WILL run into a dead end. However, IF you can get signal averaging, it is much easier to use the ST equipment at normal working levels for the device under test. You can also improve the front ends of these ST analyzers by replacing the input IC's with something much quieter.
You used AD797's didn't you? I need to get a few pretty soon; supply seems to be drying up, at least from Digikey. The LME49710 might suit though - another project! Gee, thanks. I think.
Not that there is much wrong with the equipment that you can afford, but because it will not test for dynamic time errors, do we have a problem. This seems to be the direction we need to go in. Of course, it is difficult to convince any manufacturer to make something that shows dynamic time shift with signal level and frequency, but that seems to be where the problem lies.
I've been following your postings about this pretty closely, but haven't thunk up any great thoughts yet. The absolute values are so small that it's going to require something with RCH sized timing itself, IOW very high sampling rate with very low jitter. Beyond my modest sound-card rig, fersure.
Much thanks, again,
Chris
Speaking about single-arbitrary-shape pulse test, we should remember, that, being expressed as a sum of harmonics, the pulse will be converted in tooo long sum of them (infinite-like). And the harmonics are quite definitely aligned in the time (phase) domain.
Vladimir,
has noted by Joachim, for the reasons you speak about, the pulse I used is of the raised cosine type, limited to typical audio bandwidth. It hasn't been selected to get an absolute measure of the fidelity in their reproduction, but only for the comparison of the transient response of the DUT with the reference.
I doubt that situation in this case is close to the multy-tone (even hundred tone) IMD tests.
Would be interesting to know a good metrics, how to estimate numerically the difference, between the input and output pulses.
Multi-tone test is...another kind of test: it uses simmetrical and periodic signals that are not so effective in displaying what I was looking for: transient response. Molti-tone test is a frequency domain test: I want to get information on the transient "distortion" in time domain.
Hi Luigi (lvigone),
Your findings with your shaped pulse are fascinating me.
If I got it correctly, you can know in a advance how an amplifier will sound with the aid of that shaped pulse. If this is the case, your finding is a real breakthrough.
If I got it correctly, all you have so far is comparison to reference amp, not anything you can learn from that shape pulsed in itself. If this is the case, that's the major drawback of your finding.
Anyhow, it looks to me that your findings deserve true scientific study and development. Hopefully, someone, some time, will do it.
It looks to me very interesting, that your findings about the importance and the effects of power supply, wiring and parasitic components seems to be in accord with what John Curl, the late Allen Wright and other SOTA audio manufacturers were saying for years about the importance of those elements on the sound quality of audio amps - while their effects are usually not shown be conventional measurements.
So, there are some people who say: "If I cannot measure it (by the measurements I know off) - it doesn't exist. When people report about effects on sound quality, as long as cannot measure it, those who report about it must be deluded, they only imagine it, it's only a Placebo Effect". Others try to find ways to measure those reported effects on sound quality, or at least, try to find a way that can positively and consistently tell the sound quality of an amp by means of measuring equipment.
Your findings with your shaped pulse are fascinating me.
If I got it correctly, you can know in a advance how an amplifier will sound with the aid of that shaped pulse. If this is the case, your finding is a real breakthrough.
If I got it correctly, all you have so far is comparison to reference amp, not anything you can learn from that shape pulsed in itself. If this is the case, that's the major drawback of your finding.
Anyhow, it looks to me that your findings deserve true scientific study and development. Hopefully, someone, some time, will do it.
It looks to me very interesting, that your findings about the importance and the effects of power supply, wiring and parasitic components seems to be in accord with what John Curl, the late Allen Wright and other SOTA audio manufacturers were saying for years about the importance of those elements on the sound quality of audio amps - while their effects are usually not shown be conventional measurements.
So, there are some people who say: "If I cannot measure it (by the measurements I know off) - it doesn't exist. When people report about effects on sound quality, as long as cannot measure it, those who report about it must be deluded, they only imagine it, it's only a Placebo Effect". Others try to find ways to measure those reported effects on sound quality, or at least, try to find a way that can positively and consistently tell the sound quality of an amp by means of measuring equipment.
Sooner or later, we will have to have some sort of precision pulse test that we will have to pass through our audio playback equipment in order to consider it high fidelity. The pulse itself may actually change somewhat, but the ESSENCE of the pulse, that which our ears find important, will have to be conserved. The ear hears some phase or time shift easily, and ignores other phase shift, such as what is created by a phase shifter, that can completely change the 'look' of the waveform, but not its essence.
Sooner or later, we will have to have some sort of precision pulse test that we will have to pass through our audio playback equipment in order to consider it high fidelity. The pulse itself may actually change somewhat, but the ESSENCE of the pulse, that which our ears find important, will have to be conserved. The ear hears some phase or time shift easily, and ignores other phase shift, such as what is created by a phase shifter, that can completely change the 'look' of the waveform, but not its essence.
Wouldn't that imply we should add terms to the transfer function of amps, because everything that is in there we can measure right now.
We then could add an inverse filter enforcing the devine transfer function on any system that is linear enough. Sorry JC, i am out of my mind.
I may have missed additional comments on this topic, but a critical aspect of multi-tone testing is whether the tones are phase correlated or not. Once upon a mass tape replication system ago I used a multi-tone signal to check frequency response, and when the tones were digitally synthesized and locked in phase to each other the actual peak levels were far lower (20dB or more in some cases IIRC) than when the test signal was generated by summing 12 multiple free-running analog oscillators. In order to more accurately model compression we used the analog-generated version.
Multiple free running oscillators can't work, you have a non-stationary crest factor and the measurements are generally meaningless.
The AP software has the options to set the phase of the tones as you wish. In phase, random or whatever angle you want. Not sure what the means to the measurements, tho.
We need more than this. What we need is a time arrival difference, with level measurement. Settling time 'might' be related to this.
Multi-tone Testing
And I guess that was exactly my point Scott: that type of signal is more similar in complexion to music than is a phase-correlated multi-tone series. If you are actually using free-running oscillators at the very time of making measurements, then indeed, the crest factor is pseudo-random and therefore meaningless.
However, and this is the critical difference: If you are using a recording of free-running oscillators, the test signal becomes pseudo-random, because you can repeat it. It's crests become predictable in timing. This signal can be used in a controlled test, indeed it was a test signal I used for proofing dynamic bias modulators (Dolby HX).
It is this signal's extreme crest and slew maxima which make it a good signal for exciting dynamic tracking issues in a circuit/record system.
Perhaps a standard recording of this type could prove useful in amplifier testing as well. This is not to say that a series of other test couldn't scare the same data from an amp, but this one tells you a lot by itself.
Not a panacea fore sure, but another tool in an engineer's arsenel of problem-finding signals.
Howie
Howard Hoyt
CE - WXYC-FM 89.3
UNC Chapel Hill, NC
www.wxyc.org
1st on the internet
Multiple free running oscillators can't work, you have a non-stationary crest factor and the measurements are generally meaningless.
And I guess that was exactly my point Scott: that type of signal is more similar in complexion to music than is a phase-correlated multi-tone series. If you are actually using free-running oscillators at the very time of making measurements, then indeed, the crest factor is pseudo-random and therefore meaningless.
However, and this is the critical difference: If you are using a recording of free-running oscillators, the test signal becomes pseudo-random, because you can repeat it. It's crests become predictable in timing. This signal can be used in a controlled test, indeed it was a test signal I used for proofing dynamic bias modulators (Dolby HX).
It is this signal's extreme crest and slew maxima which make it a good signal for exciting dynamic tracking issues in a circuit/record system.
Perhaps a standard recording of this type could prove useful in amplifier testing as well. This is not to say that a series of other test couldn't scare the same data from an amp, but this one tells you a lot by itself.
Not a panacea fore sure, but another tool in an engineer's arsenel of problem-finding signals.
Howie
Howard Hoyt
CE - WXYC-FM 89.3
UNC Chapel Hill, NC
www.wxyc.org
1st on the internet
Speaking about the transfer function, we must not forget about limits of validity and the assumptions made, while the math has been elaborated. It is applicable to periodic functions. If a signal looks like a pulse, one should immediately think about possible limitations of MEASURABLE (not theoretical) transfer function. Can we measure it at uV and GHz?
Devil of misunderstanding is always sitting in tiny details.
Roots of listening impressions, IMHO, frequently relate to micro-level signals and their time shifts. If the last takes place (micro-signals are distorted and time shifted), in particular due to NFB action, that can not distinguish them from noise, and due to electro-magnetic interference, the micro-detailes become "ironed-out", a typical sound of mid-fi equipment.
Step to high-end is made due mainly to hardly explicable tricks, from scientific point of view, sometimes on the level of know-how.
Devil of misunderstanding is always sitting in tiny details.
Roots of listening impressions, IMHO, frequently relate to micro-level signals and their time shifts. If the last takes place (micro-signals are distorted and time shifted), in particular due to NFB action, that can not distinguish them from noise, and due to electro-magnetic interference, the micro-detailes become "ironed-out", a typical sound of mid-fi equipment.
Step to high-end is made due mainly to hardly explicable tricks, from scientific point of view, sometimes on the level of know-how.
Last edited:
- Status
- Not open for further replies.
- Home
- Member Areas
- The Lounge
- John Curl's Blowtorch preamplifier part II