Previously you said this
" In order so as not to run out of memory or incoming data, transport clock jitter occuring at less than 1 Hz is allowed to come through, but transport clocks and cable jitter have very little jitter activity below 1 Hz, and it is too low frequency to affect the sound of music, therefore it does not create any problems. It is did cause problems, they could just lower it to 0.1 Hz or whatever, but 1 Hz is good enough."
"Starts attenuating at around 0.1Hz" is a quote from where?
You think that close-in FM sidebands decreasing in amplitude as one goes from the fundamental tone to 1Hz either side of that tone, has no perceptual effect?
I don't know for a fact. In the Benchmark DAC-3 manual here: https://cdn.shopify.com/s/files/1/0321/7609/files/DAC3_Series_Manual_Rev_B.pdf?9982830537634228604 on the bottom of page 64 they list some jitter specifications, also in the middle of page 63. There is also a jitter FFT graph on page 61, where they state, "Note that none of the 31 FFTs show any signs of jitter-induced sidebands."
In short, it looks to me like jitter is not the primary limiting factor with this DAC. Distortion and noise appear worse than jitter.
Therefore, without some research into measuring what changes when SRC is disabled and a clean transport clock is used instead, and also blind testing of some kind with skilled listeners, I am going to withhold forming any conclusions about to what extent the SRC may be audible. It seems doubtful there is a large effect or problem with SRC as currently implemented, but doubt is not a conclusion.
Very little hope of seeing close-in jitter on those plots as the x-axis is so gross - depending on the the bandwidth plotted such close-in jitter/phase noise shows as a widening of the base of fundamental tones or in the case of Bruno's plots, individual spurs are seen (due to the x-axis bandwidth)
When such issues aren't being investigated, plots such as on page 61 are shown & such statements therein are made.
I think you may have seen from Bruno's plots (if I interpret them correctly) a significant difference in close in side spurs based on the clock frequency feeding of the incoming datastream, no? So this gives you a handle on the varying jitter/FM modulation occurring as the jitter on the incoming datastream varies.
Another good starting point might be to research if FM modulation of 1Hz & lower is perceptible with complex signals but obviously we hit the barrier of research funding & interests & these types of questions are typically not directly answered unless they happen to be part of another, more active research area.
But does this not boil down to how sensitive we are to varying pitch (not absolute pitch) in music?
There are some hints out there related to this question but listening experience is by far the quickest way to answer this for oneself
It's because the low pass filter corner frequency in Bruno's plots is higher than the corner frequency used in the Benchmark DAC. Like he said, if you get the bandwidth right, it's not a problem, and we see it is not a problem in DACs today, even though it was a problem with the early version Bruno was testing.
Please don't deny what is obvious - close-in jitter requires x-axis to be in much higher resolution than Benchmark show on their plots - each x-axis gridline on their FFT is at 2KHz - Bruno's plots have gridlines @ 400Hz.
How do you know that the Bruno's ESS SRC is using a higher corner freq than current ESS ? Have you got some reference for this?
Come on - you are better than that.
These two plots of the Benchmark done by Stereophile clearly show the close-in jitter widening at the base of the plot (the x-axis resolution is finer than the plots shown in the Benchmark manual). The base widening is the same in both plots but some side spurs (& higher noise floor - random jitter?) are also evident when USB is used as input - hmmm, thought the Benchmark marketing copy told us it eliminated all jitter??
Last edited:
In addition to my last post; to raise the probability of doing a sound experiment one should (has to):
-) state clearly a research hypothesis/question/objective before starting to work out the experimental conditions
-) use utmost care to raise the probability that only the independent variable
(aka the difference between the DUTs) has impact on the result, that includes usage of negative control(s)
-) include good measurements, as we (usually) are only interested in differences that are not already known to be perceptable (beside using as positive controls)
Although i agree to your list, your "disaster assertion" about the subjective evaluation raises some concerns.
Within a (hopefully) scientific experiment we are still using the subjective evaluation although we are trying to "measure" it in an objective way.
As we know that humans are still prone to bias effects - even when used as detectors in this said experiment - and taken into consideration that we most likely will not be able to control anything, we have imo to conclude that our detectors must nevertheless be able to control the impact of those bias effects (at least up to a certain degree) to get useful results.
We know from other fields too that even the participation in an experiment can/will have some impact, therefore the more and more accepted conclusion that for example observative studies should accompany the "gold standard" laboratory experiments.
In more general words, it seems to be a good idea to use not only quantitive methods (like "blind tests" of the kind we are discussing in this thread) but qualitative methods as well.
To get back to the "disaster assertion"; it might be true, but the true problem exists because we don´t know if the subjective evaluation results are correct or incorrect. We aren´t able to evaluate if the subjective evaluation was objective, valid and reliable; otoh we can´t say that it wasn´t too, because we don´t know that either.
-) state clearly a research hypothesis/question/objective before starting to work out the experimental conditions
-) use utmost care to raise the probability that only the independent variable
(aka the difference between the DUTs) has impact on the result, that includes usage of negative control(s)
-) include good measurements, as we (usually) are only interested in differences that are not already known to be perceptable (beside using as positive controls)
Although i agree to your list, your "disaster assertion" about the subjective evaluation raises some concerns.
Within a (hopefully) scientific experiment we are still using the subjective evaluation although we are trying to "measure" it in an objective way.
As we know that humans are still prone to bias effects - even when used as detectors in this said experiment - and taken into consideration that we most likely will not be able to control anything, we have imo to conclude that our detectors must nevertheless be able to control the impact of those bias effects (at least up to a certain degree) to get useful results.
We know from other fields too that even the participation in an experiment can/will have some impact, therefore the more and more accepted conclusion that for example observative studies should accompany the "gold standard" laboratory experiments.
In more general words, it seems to be a good idea to use not only quantitive methods (like "blind tests" of the kind we are discussing in this thread) but qualitative methods as well.
To get back to the "disaster assertion"; it might be true, but the true problem exists because we don´t know if the subjective evaluation results are correct or incorrect. We aren´t able to evaluate if the subjective evaluation was objective, valid and reliable; otoh we can´t say that it wasn´t too, because we don´t know that either.
Everybody can try a very easy test to control about questioning ABX listening tests. Let listen carefully first time any type of possibly acoustic music, which is not very well known for you, from the beginning to the end of the track.
Than listen this track one more time, without any changing of gear, light, seat, etc.
Secondly the feeling of the whole music feeling will be different. You will hear more details, more empathy.
Now let thinking about any type of comparing test, are they really correct?
Even worse, when during listening change any type of circumstance, you hear an other fragment of the track, and the empathy of the music is absolutely off. So when you want to compare two gear or cable, etc. you must hear first and second time the same gear, then change what you want and listen only once. Don't switch during tracks, it may be deceptive. ABX test are useless exept sceptic dumbass, who must listening whole life entry level HT amps with sour cream cup speakers...
Than listen this track one more time, without any changing of gear, light, seat, etc.
Secondly the feeling of the whole music feeling will be different. You will hear more details, more empathy.
Now let thinking about any type of comparing test, are they really correct?
Even worse, when during listening change any type of circumstance, you hear an other fragment of the track, and the empathy of the music is absolutely off. So when you want to compare two gear or cable, etc. you must hear first and second time the same gear, then change what you want and listen only once. Don't switch during tracks, it may be deceptive. ABX test are useless exept sceptic dumbass, who must listening whole life entry level HT amps with sour cream cup speakers...
Indeed.
I want to add to this the fact that many musiclovers under normal circumstances never reach conditions where these subtle differences become audible. Then the reasons for buying expensive electronix are not obvious. In the end you want to immerse yourself in music?
There are a number of useful points made in this thread & one of them is that this 'subtle differences become audible' is only needed to pass such ABX tests - it's a spot the difference test.
Normally one isn't conscious of such subtle differences but they may well affect our emotional engagement with playback without us being specifically conscious of the specific differences in the sound - the nature of auditory perception is like this.
Last edited:
The first things you write are methods to amplify the difference between the dacs so you can prove that the differences between them are bigger than the thereshold of human hearing. If that is not so important you don't have to care about overpriced gear and can kick back with some good music.
For some Pride Of Ownership is important. It feels good to own things you can show your friends. That can be a reason to buy the expensive stuff.
The reason I am on this forum is the guys who can go through the garbage and make high class equipment with its contents. That impresses me and gives me ideas for how to do similar myself. If i succeed with such it gives me great feel good when playing music and the greatest Pride Of Ownership.
Do you mean that our emotional engagement is not in place in such difference tests?
I come from the academic side of the street, but I dont know very much about hearing and perception. One thing I do know however is the our ear-brain apparatus have a very short memory. This have been proved time and time again. Based on that fact it is a no brainer that spot tests are the only way to go to find differences between audio equipment and how can you choose between two pieces of gear if you're not conscious?
The topic of this thread is a spot the difference test between dacs and I will try to discuss this primarily.
Not usually in spot-the-difference ABX style blind testing, no. It usually requires a conscious identification of some specific aspect of the sound which is identified as different between the two tracks A & B so that X can be stated as being either A or B - so something like the hit of a snare, the rasp of the brass, the ringing of the cymbals, the tail of the bass notes, etc - something specific, in a specific part of the sound that can be identified as different between A & B
You are talking about echoic memory (the part of working memory that is used as short term sensory memory). Yes, good for the type of aspects of the sound above that I mentioned are used in spot-the-difference but are these the ONLY differences possible? What gives us more emotional engagement with one device in the playback chain Vs another?
Ah, OK, from your questions I thought you were actually interested in whether there may be differences between DACs when using our auditory perception as the yardstick, not when using some limited features of it in an unnatural & undefined listening test which ignores how to do auditory perceptual tests??
Last edited:
The discussion about the audibility of "subtle differences" started back in the ~1960s (maybe even earlier) and the argumentation was always nearly the same, means some measurements show that the differences are below the (normal) thresholds of hearing therefore the difference can´t be percepted.
The two interesting questions are:
-) are the differences really that "subtle" (at the beginning in the ~1960s they were surely les "subtle" than today)
-) do we know about the normal circumstances
the latter because your argument represents the opposite in concluding from rather "unnormal" circumstances to the "normal" circumstances.
Might be true due to the mentioned "unnormal" circumstances; as said already earlier, emotional response sometimes need more time than available during short music samples and the artificial situation might even be distractive enough to prevent emotional involvment.
Without evaluation it´s hard to say.
Which part of the ear-brain apparatus do you mean with the "very short memory" assertion?
Usually it´s only the socalled echoic memory part associated with a very short memory time span (according to the literature in dependence from variables it is between 150 ms - 5s), in addition there exists the socalled working memory and of course the long term storage. Time spans for the working memory range from minutes to days while long time storage might last livelong (though usually no part of human memory works flawlessly)
According to your premise (of the very short memory) wouldn´t it be more a "no brainer" to do spot tests (or any tests at all) as you wouldn´t remember any difference after a couple of seconds?
Assumed that we interested about something relevant to the normal listening circumstances.
The problem with resampling inside the DAC is that you are stuck with the sound and quality of that oscillator. Cannot improve on it much. This is precisely why I don't put a reclocker inside my DAC.
The other thing I don't agree with is that incoming S/PDIF jitter to 99% of DACs is unimportant. Every customer of mine who changes S/PDIF cables hears a difference. This proves that these DACs are affected by jitter. They have all kinds of expensive and inexpensive DACs.
Steve N.
Empirical Audio
It would be nice if jitter was actually rejected. It usually isn't. Simple tests like changing S/PDIF cables prove this.
I don't like the method of measuring jitter through a DAC either. I like to measure jitter directly on the S/PDIF cable. I have found this to correlate better to sound quality of sources.
See some plots:
Can S/PDIF cable jitter be measured?
Jitter from Sonos and how much the Synchro-Mesh reduces it
Steve N.
Empirical Audio
I don't see any problem with putting an extremely low jitter clock in the DAC and leave it at that. Most cable jitter can be reduced to unmeasurable levels after high quality reclocking.
Here is perhaps the best DAC in world, if you believe Bruno Putzeys: https://www.mola-mola.nl/readmore/read-more-tambaqui.PDF It uses an interpolator to resample. Rejects jitter down to extremely low levels.
Benchmark DAC-3, according to Stereophile has state of the art measured performance and using the AES jitter test, FFT analysis shows no measurable jitter.
It is definitely possible to reject jitter quite well. Costs too much to do it with barebones implementations it would seem. Deliberately not providing jitter immunity for an expensive DAC and playing around with cables to get some kind of euphonic jitter distortion sounds like the type of thing often attributed to audiophiles, and exactly not what mastering and recording engineers want to do their work with. Some people want different things for different reasons. That's fine so long as we are honest about what is actually being done and not pretending added euphonic distortion represents increased accuracy.
Last edited:
Bruno's DAC is not a good representative of Delta Sigma DACs (99% of DACs in the marketplace) as it's based on different technology - it's PWM - pulse width modulation technology which may make it less jitter sensitive than PDM based DS DACs
Funny enough, the Stereophile jitter plots I posted are of Benchmark DAC3 HGC & it clearly shows both close-in jitter & with USB input, there are jitter spuriae & a higher noise floor (due to random jitter?).
Stereophile don't mention close-in jitter (in any of their measurements for any DACs?) but do say this
"However, with USB data a pair of sidebands appeared at ±1518Hz (fig.12), though it is fair to note that these are at a very low level."
Read more at Benchmark DAC3 HGC D/A preamplifier-headphone amplifier Measurements | Stereophile.com
Last edited:
I would agree it's not perfect, and I would rather not see the sidebands that showed up with USB. Highly questionable as to how audible they are, but most likely the DAC has much less jitter than DACs that make no attempt at jitter rejection at all and rather rely on playing around with cable jitter to arrive at a sound one finds to one's liking.
Look, I'm not saying any DAC is perfect or completely transparent. I am saying that some DACs measure quite well and sound very good too. Some DACs are also a lot better at jitter rejection than others.
Personally, what I happen to want among other things is a DAC that measures well, sounds good, and rejects jitter as well as possible. I don't want a DAC that requires playing around with cables to get an acceptable sound. Other people may want something different. That's fine. If you have one you think is really good, why not try to get Stereophile to review it and let's see how it comes out. If it comes out better than Benchmark, great. If it's cheap, even better. Maybe I'll get one. But, I'm not holding my breath.
Mark is describing a typical FIFO and buffering setup where the S/PDIF is simply transferring the data losslessly to a completely independent memory/clock/DAC. Suggesting binary data has memory of its jitter "history" is simply nonsense.