Mark, you're talking about specific, personal preference here. It doesn't hold or there would be much greater market acceptance.
Its not about preference at all. Don't know why do I have to keep pointing that out. Its about discrimination. Its like another form of instrumentation. One what you don't read with your eyes, you learn to use your ears to assess the results by way of discrimination.
Look, you can hate the sound of the speakers. It doesn't matter any more than it matters what the fan sounds like on an AP 555, or if you like the way the video display looks. Ignore it. Learn to listen to what extremely low distortion speakers can tell you about the electronics.
@Markw4 this is a thread for measuring distortion on the cheap. You've made it plenty clear in other threads that you don't believe in distortion as a measure of audiophile goodness. That is your personal belief, which you are entitled to have. But please don't turn this thread into another of your subjective vs objective crusades.
Tom
Tom
Ah, but I was responding to talk about the cost of a 555x, which has nothing to do with low cost PSS distortion measurement. Happy to get back to the low cost topic.
Hi Mark,
I respect your beliefs, however you have to respect ours as well. Many of us have long careers in the audio business with a great deal of practical as well as theoretical experience. It isn't a belief from our experience, but learned lessons.
The price of admission is high, period. So be it. It is perfectly valid to compare expensive options to lower cost options to illustrate the trade-offs. The very high cost of all test equipment can be a barrier, and I have certainly sacrificed to get what is needed. If it weren't necessary, no way would I divert money from my family and living on a whim or for bragging rights. If a person doesn't have the ability to use the proper equipment, they haven't any way to see what is required and therefore can't comment, now can they?
Now, my comment on preference applies to your affinity for electrostatic speakers. That is your particular preference. I've heard all kinds of speakers, including panels and electrostats. Those also have problems, nothing is perfect. Most people prefer good dynamic speakers, voltage drive. That is the market, the best product wins.
I respect your beliefs, however you have to respect ours as well. Many of us have long careers in the audio business with a great deal of practical as well as theoretical experience. It isn't a belief from our experience, but learned lessons.
The price of admission is high, period. So be it. It is perfectly valid to compare expensive options to lower cost options to illustrate the trade-offs. The very high cost of all test equipment can be a barrier, and I have certainly sacrificed to get what is needed. If it weren't necessary, no way would I divert money from my family and living on a whim or for bragging rights. If a person doesn't have the ability to use the proper equipment, they haven't any way to see what is required and therefore can't comment, now can they?
Now, my comment on preference applies to your affinity for electrostatic speakers. That is your particular preference. I've heard all kinds of speakers, including panels and electrostats. Those also have problems, nothing is perfect. Most people prefer good dynamic speakers, voltage drive. That is the market, the best product wins.
I didnt read all the posts here, however it seems an AD620 may be a good front end to compare the source and output signals then amplify the output by 10 to 100 times.
The 620 may be able to do that at the same time without sacrificing differential cancelation. These are $20 but still cheap next to an actual instrument.
** Output is viewed on Scope.
The 620 may be able to do that at the same time without sacrificing differential cancelation. These are $20 but still cheap next to an actual instrument.
** Output is viewed on Scope.
Ahem, Beats headphones? (I know them very well, despite my involvement I'm not responsible for their tuning.) They fared poorly in the Toole headphone tests but did not slow the sales at all.
1) I replaced the fan in my AP sys 1 with a temperature controlled fan. Quiet and trouble free. Clearly out of warranty.
2) I would deal with this discussion as follows-
Fix what are known audio sensitivities- 1) Frequency response, 2) distortion, 3) any other basic stuff, clipping response, SNR in operating levels, stability into different loads, EMI RFI resistance etc.
Look for other less obvious issues. Settling time, thermak shifts changing response, I/O interface issues (e.g. input capacitance modulation) I'm sure there are others that could affect things.
All of the above have thresholds below which they cannot concievably affect the sound. However its irresponsible to not fix the obvious weaknesses.
I can't disagree Demian. I don't listen to headphones often, but still have my AKG K240 (upgraded from the head crusher Koss Pro4A). The vast majority of people use ear buds that have questionable quality listening to questionable source material. There is a high end market for headphone amps and headphones, I'll grant you that.
I'm waiting for warranty to expire to replace a few fans in new equipment. Never thought I would see the day when test gear was cheaply whacked together. I was use to using HP.
Absolutely! You nail what matters in order of importance. How far you go depends on the performance you want.
I'm waiting for warranty to expire to replace a few fans in new equipment. Never thought I would see the day when test gear was cheaply whacked together. I was use to using HP.
Absolutely! You nail what matters in order of importance. How far you go depends on the performance you want.
^Reads like a fundamental misunderstanding of the concept of thresholds in audio. Audibility thresholds are in fact estimates of the level at which 50% of the population cannot hear below the threshold and the other 50% of the population can still hear the effect below the threshold. IOW, a threshold is an estimate of a threshold for the average ear (think of the middle of a bell curve). Moreover, published thresholds, where they exist, cannot possibly be measured hard limits for every human on earth since it is impossible to measure billions of people. In addition, not all bell curves have gaussian distributions. In many systems it turns out in practice that bell curves are not gaussian, and that presumed rare events, or rare individuals, may be more common than have previously been estimated. https://www.edge.org/response-detail/11715
All the above is without consideration of the now well-known bias towards false-negative results in ABX blind tests. Modern perceptual science literature lists the false-negative bias of ABX as one of the cons that disfavor its use.
Last edited:
Perceptual biases unavoidably plaguing sighted testing far outweigh the possible false-negative results in ABx blind tests.
I could list many psychological effects that would apply here as well. Including the better-than-average effect. But that's a topic better severed in a different thread.
Tom
Tom
It could work only if there were no phase shift between the input and output, and there were perfect amplitude match throughout the full frequency band of interest. Better to suppress the fundamental (e.g. with a Twin-T filter) and monitor the residual.
Just a related question - do we know the power noise rejection of those devices and if we've discounted that as a limiting factor?
I got my DIY ADC down to -160dB on the power supply/clocking but the ADC chip itself only got to -124dB which was fine for my purposes.
I've also added a Arturia MiniFuse 1 for easier/general use. It's also a low cost interface.
I got my DIY ADC down to -160dB on the power supply/clocking but the ADC chip itself only got to -124dB which was fine for my purposes.
I've also added a Arturia MiniFuse 1 for easier/general use. It's also a low cost interface.
That's a nice question that I would like to expand with your permission. Regarding power supply PSRR, what would be the diminishing returns point for the digital converters and the opamp buffers.
I would not mix noise and PSRR. PSRR may be well under control. Noise has physical limits due to devices used.
The spell check error that I bolded actually might be accurate! 🙂 I like it!
(Like me, you're probably using a Mac, which does that sort of thing...)
Psychological effects should be contained in their own thread. I have noticed this topic is used to derail normal threads and create havoc. Everyone is well aware of this.
I think I will begin killing posts like this. Absolutely everyone knows where they lead, especially the usual suspects.
I think I will begin killing posts like this. Absolutely everyone knows where they lead, especially the usual suspects.
That is all certainly true.
But... PSRR may be vastly underrated as a problem.
Most linear devices used at audio frequencies like opamps, whether integrated or not, don't have great PSRR above the audio band. Since a lot of these circuits rely on the use of global feedback to improve PSRR, their limited open loop gain bandwidth means that they might not be so good above several tens of KHz.
Well, guess where digital artifacts from all the processing are situated. Guess where junk on the power line may reside. Guess where ringing in the basic transformer/rectifier circuit lies, if the resonant circuit isn't properly damped. These unwanted signals can easily mix back down into the audio band within the audio circuitry, which has lousy linearity at these high frequencies.
The digital circuits used for processing normally have almost no PSRR. That's not how they are designed.
PSRR in Logic Devices
I'd say the same is true of CMRR as well. At least my tests, albeit not performed with $100k of test gear, have indicated that. (Note: PMA has already demonstrated the effects of CMRR elsewhere. I'm just pointing this out for everybody not PMA.)
Of course, we tend to miss all these effects by testing system components in pristine ideal environments. Audio Precision, by being the best at ignoring these effects, also becomes the worst by ignoring these effects. The burden is up to the tester. If we want our audio gear to perform in real life as well as they measure in testing with an AP system, we need to make the rest of our audio system perform as well as the AP gear does.
Yes, I understand the hypothesis. But as always, I do measure my designs under real world conditions, with digital sources, unclean power etc. The measured spectra that I post show if I have troubles with mains related interference and digital artifacts or not. This is what I am interested in. Rather than dicussions about potential, however unquantified issues.
The problem with relying on audio band spectra is that some problems identified by well-known and very serious people do not show up well as distinct spectral lines. I don't know why the reasons for this are not more well understood.
For example, Bill Whitlock has repeatedly said that audio systems can have "veiled" and or "grainy" sound. What distinct spectral line shows that condition? None of them do!
So, another problem we have is people who believe all that matters will be revealed in spectral lines. Noise power (where noise is defined as any unwanted signal, not just random noise) can in some cases be distributed across a few or very many bins, yet still be audible. How? Its because we assume test signals and any kind of resulting spectra must all be steady state signals. We then assume humans can only hear steady state signals as represented by clear spectral lines. IOW, there is more than one false assumption involved. Signals that are not steady state during FFT acquisition will get smeared across bins. That's just the way the math works.
EDIT: There is also more that can be said about the effects of such types of noise (unwanted signals) and their possible appearance in the overall noise floor. Noise signal behavior, FFT windowing, etc., can make it difficult to reliably see effects in the noise floor. IOW, its hit or miss on that possibility. Also, spectral line noise skirts contain information about noise more strongly correlated with the audio signal. However because the skirts appear to be low level, they may be mistakenly ignored. Correlated noise effects, when not steady state, can be at full 0dBFS signal level and skirt height may only be an indicator of the time the full scale signal (distorted in some way) spends in a particular distorted state. IOW, spectral line skirt height is not a reliable measure of correlated noise peak amplitude. Again, this is just how the math works.
For example, Bill Whitlock has repeatedly said that audio systems can have "veiled" and or "grainy" sound. What distinct spectral line shows that condition? None of them do!
So, another problem we have is people who believe all that matters will be revealed in spectral lines. Noise power (where noise is defined as any unwanted signal, not just random noise) can in some cases be distributed across a few or very many bins, yet still be audible. How? Its because we assume test signals and any kind of resulting spectra must all be steady state signals. We then assume humans can only hear steady state signals as represented by clear spectral lines. IOW, there is more than one false assumption involved. Signals that are not steady state during FFT acquisition will get smeared across bins. That's just the way the math works.
EDIT: There is also more that can be said about the effects of such types of noise (unwanted signals) and their possible appearance in the overall noise floor. Noise signal behavior, FFT windowing, etc., can make it difficult to reliably see effects in the noise floor. IOW, its hit or miss on that possibility. Also, spectral line noise skirts contain information about noise more strongly correlated with the audio signal. However because the skirts appear to be low level, they may be mistakenly ignored. Correlated noise effects, when not steady state, can be at full 0dBFS signal level and skirt height may only be an indicator of the time the full scale signal (distorted in some way) spends in a particular distorted state. IOW, spectral line skirt height is not a reliable measure of correlated noise peak amplitude. Again, this is just how the math works.
Last edited: