-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Let's develop an ASR inter-sample test procedure for DACs!
- Thread starter splitting_ears
- Start date
Should not you ask him first if he is interested in such kind of testing? 
Topping DX5 clipping:
Clipping disappears when digital volume control knob is set to -2dBFS.
But, when oversampling (48kHz==>96Hz) is used to the original signal, then the digital attenuation does not help.
Topping DX5 clipping:
Clipping disappears when digital volume control knob is set to -2dBFS.
But, when oversampling (48kHz==>96Hz) is used to the original signal, then the digital attenuation does not help.
Last edited:
Software resampling should be used with software volume control. This means the 2dB (if enough) reduction should also be done with software before sending to the DAC instead of using the hardware volume control on the DAC.
For example the TI PCM1794 chip on my soundcard does not support hardware volume control, I always do it within the playback software to prevent clipping.
This is certainly the "gold standard", however, not everybody uses a PC for music playback. I try to avoid it if possible, as I find it much more convenient to use hardware volume control than to use a volume slider in jRiver MC...
The software volume control can be assigned to a physical device (e.g. a physical knob) if you don't like to use a mouse to drag a slider:
When software resampling is used without also using suitable amount of software attenuation, the signal is already clipped in software domain before entering external hardware. This article explained it.
SRC and peak levels
It is often assumed that resampling (SRC) or dither are safe operations that do not change peak levels of the file and cannot cause clipping. If the file clips, people are surprised and suspect there is something wrong with the SRC algorithm. In fact, both SRC and dithering can…
izotope-rx.livejournal.com
Here are some illustrations showing software domain clipping on the software resampler output using integer output format:
Digital filter "game"
Some of you probably have seen a lot of filter plots like these on DAC datasheets: How about making these things without using specialized software like Matlab, Octave and such? They are not particularly easy to use. So many people just "see" them rather than try to figure out the meaning...
www.audiosciencereview.com
Last edited:
@Kal Rubinson: could you provide more background infos about the volume knobs you are using with jRiver MC? Like manufacturer, part number. Is any specific software required to get it working with jRiver Mac? See previous post.
Thank you in advance!
Edit: Just saw that you are using the Microsoft Surface Dial, which can be configured to send keyboard shortcuts as required by jRiver MC. Will consider adding such a device to my system.
Edit2: In another context the Contour Multimedia Controller Pro v2 was mentioned (used by Bob Katz for controlling Acourate Convolver). There is also a smaller version, the Contour Multimedia Controller XpressMusic. Both use USB cable connection.
Thank you in advance!
Edit: Just saw that you are using the Microsoft Surface Dial, which can be configured to send keyboard shortcuts as required by jRiver MC. Will consider adding such a device to my system.
Edit2: In another context the Contour Multimedia Controller Pro v2 was mentioned (used by Bob Katz for controlling Acourate Convolver). There is also a smaller version, the Contour Multimedia Controller XpressMusic. Both use USB cable connection.
Last edited:
little-endian
Member
- Joined
- Apr 16, 2023
- Messages
- 92
- Likes
- 98
Some additional thought about this topic beforehand:
Intersample peaks do not only occur in the audio domain, but actually should be something to be considered also for video, which at least theoretically may be described as the same phenomenon, just in two dimensions.
In the same way, reconstructed audio may have a higher level then it's "surrounding" sample points, so may pixels (which are the equivalent) call for e.g. brighter levels than their digital values suggest. Especially when images are scaled, transcoded, etc. I could imagine it to happen and I better don't ask if that is regularity accounted for, fearing to already know the answer.
For some strange reason, what is entirely understood and practically handled such as filtering for aliasing, isn't followed when it comes to images or video. How many tools have crappy downscaling algorithms full of aliasing? Yes!
* A devil's advocate argument which comes to mind could be that arbitrarily high intersample peaks may be "artificially" constructed raising the question whether any amount of headroom will ever be enough. In that sense, there can be no "perfect" DAC.
Technically, a DAC doesn't "fix" such masterings but instead doesn't make them even worse. Again, one could argue here that in many cases the source it already so bad that it doesn't matter anymore anyway.
But even under the premise that it is a non-issue - so are SINAD differences in the region of dB-fractions which are also lively discussed here. At the end, I find a DAC's behavior to those intersample peaks even more interesting than a few dB of noise or distortion here and there.
Without intending to sound provoking or too cocky, but out of interest: why didn't Benchmark already implement such headroom in the DAC1 already?
My impression is that this whole topic was even a more niche one back in the days of the DAC1 and "discovered" by the scene later on and now of course also for Benchmark is a welcome marketing point.
Mind you, the DVD audio to me looks and sounds even better. If you take the 5.1 and create a downmix to stereo, you will have an even more dynamic result without intersample clipping, showing that as usual, despite the otherwise very good production quality, the official CD release has been unnecessarily limited compared to the multi-channel DVD.
It's just inconvenient yet as it's something new and it's very human to question it at first, but once a testing routine has been established, it should cause much more effort than testing for the usual SINAD fuss, we all got used to by now and also for sure has been something exotic back then.
Same goes for testing against the deemphasis support, which I've raised very recently.
Intersample peaks do not only occur in the audio domain, but actually should be something to be considered also for video, which at least theoretically may be described as the same phenomenon, just in two dimensions.
In the same way, reconstructed audio may have a higher level then it's "surrounding" sample points, so may pixels (which are the equivalent) call for e.g. brighter levels than their digital values suggest. Especially when images are scaled, transcoded, etc. I could imagine it to happen and I better don't ask if that is regularity accounted for, fearing to already know the answer.
For some strange reason, what is entirely understood and practically handled such as filtering for aliasing, isn't followed when it comes to images or video. How many tools have crappy downscaling algorithms full of aliasing? Yes!
Like Benchmark's most previous baby "DAC1" back in the days.
I would formulate it as "ISP overs are unnecessarily provoked by questionable mixing/masterings" (but should be rendered correctly nonetheless).
Are they? My understanding is that one - at least in test scenarios - may have samples at 0dBFS with peaks in between, calling for analog reconstruction at levels above the 0dBFS equivalent, without using any oversampling as well. One could argue that it's the reconstruction of that analog signal calling for the additional headroom*.
* A devil's advocate argument which comes to mind could be that arbitrarily high intersample peaks may be "artificially" constructed raising the question whether any amount of headroom will ever be enough. In that sense, there can be no "perfect" DAC.
I think that can be righteously argued from both sides. On one hand, especially nowadays' mixing and mastering on average sadly is as bad as virtually never before, where the irony of technical capabilities and actual use and interest from the average music joe is staggering to say the least. On the other, one could imagine a perfectly fine highly dynamic master still containing intersample peaks for whatever reason. Since it is technical a valid input for a DAC, it should be rendered correctly.
Technically, a DAC doesn't "fix" such masterings but instead doesn't make them even worse. Again, one could argue here that in many cases the source it already so bad that it doesn't matter anymore anyway.
That's indeed a matter of standardization and definition. One pro argument to allow intersample peaks however would be that this way, one makes use of the full SNR a certain bit depth allows without the need to have it lower by constraining yourself in the "digital domain".
However, doesn't the SNR of most DACs nowadays exceed the one of the sources anyway?
Excellent point. What is practically required has been long exceeded even in this forum, nominally not being exactly voodoo-affine.
What is really likeable about RME is that they care about the details, such as dynamic loudness or manually activatable deemphasis.
That point may be seen from two different angles and neither I am sure which one to be preferred. One can argue that such inputs containing intersample peaks still contain valid data and of course they do as nominally, all bit combinations are allowed.
But even under the premise that it is a non-issue - so are SINAD differences in the region of dB-fractions which are also lively discussed here. At the end, I find a DAC's behavior to those intersample peaks even more interesting than a few dB of noise or distortion here and there.
On the other hand, even if one reserves 12dB of headroom, the SNR of most products would only go from "totally insane" to "still insane". I mean, most figures are below what we will ever be able to hear anyway so how much damage is caused by turning the wheel academically even further? One might sassily ask what is worse - having mostly inaudible intersample clipping or mostly inaudible noise?
Correct me if I'm wrong, but neither the Benchmark DAC1 had such an overhead implemented and since it's also over/undersampling (to something like ~ 110 kHz if I halfway remember that correctly), it should run into similar issues.
Without intending to sound provoking or too cocky, but out of interest: why didn't Benchmark already implement such headroom in the DAC1 already?
My impression is that this whole topic was even a more niche one back in the days of the DAC1 and "discovered" by the scene later on and now of course also for Benchmark is a welcome marketing point.
Yep, that one seems to be yours/Benchmark's "favorite".
Mind you, the DVD audio to me looks and sounds even better. If you take the 5.1 and create a downmix to stereo, you will have an even more dynamic result without intersample clipping, showing that as usual, despite the otherwise very good production quality, the official CD release has been unnecessarily limited compared to the multi-channel DVD.
Maybe impertinent to ask for your opinion on a competitior's product, but from a technical/scientifical standpoint, given your enviable technical expertise, how would you judge RME's implementation of volume control in their ADI DACs, especially in regard to the prevention of intersample clipping?
This for me is the core argument why the behavior of intersample clipping is something which should be tested for.
It's just inconvenient yet as it's something new and it's very human to question it at first, but once a testing routine has been established, it should cause much more effort than testing for the usual SINAD fuss, we all got used to by now and also for sure has been something exotic back then.
Same goes for testing against the deemphasis support, which I've raised very recently.
And yet ... the DAC1 also got away with(out) it, critically acclaimed nonetheless.
Interesting measurements of Benchmark DAC3
goldensound.audio
Archimago on IS overs:
archimago.blogspot.com
Benchmark DAC3 B Measurements - GoldenSound
The benchmark DAC3B offers some mixed objective results, with some areas going beyond any other manufacturer and some falling behind
goldensound.audio
Archimago on IS overs:
MUSINGS / MEASUREMENTS: A look at Intersample Peaks and Overload Tolerance... (RME ADI-2 Pro FS, TEAC UD-501, Oppo UDP-205)
A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.
archimago.blogspot.com
little-endian
Member
- Joined
- Apr 16, 2023
- Messages
- 92
- Likes
- 98
Yes, especially Archimago seems to be a decent and "down to earth" source whereas GoldenSound for me is a somewhat confusing case and mixed bag.
On one hand, that guy on Youtube with his very pleasant voice and eloquence seems to know his field, having tremendous knowledge on the subject, including the intersample peak niche, one the other, his videos are also full of nothing but voodoo nonesense where he seriously compares the sound of DACs (which is questionable to begin with as in most cases only possible in borderline cases while crunching your teeth during comparison, if at all) and best of all, presented in principle here through the viewer's DAC while watching it, which doesn't make any sense to even begin with and is pure satire.
Guess the latter is a simple result of "bumping the click rate" as - I have to admit it - that voodoo stuff is somewhat entertaining despite knowing better.
On one hand, that guy on Youtube with his very pleasant voice and eloquence seems to know his field, having tremendous knowledge on the subject, including the intersample peak niche, one the other, his videos are also full of nothing but voodoo nonesense where he seriously compares the sound of DACs (which is questionable to begin with as in most cases only possible in borderline cases while crunching your teeth during comparison, if at all) and best of all, presented in principle here through the viewer's DAC while watching it, which doesn't make any sense to even begin with and is pure satire.
Guess the latter is a simple result of "bumping the click rate" as - I have to admit it - that voodoo stuff is somewhat entertaining despite knowing better.
AnalogSteph
Major Contributor
Which is where they must have learned the hard way about these things - that model also featured an ASRC, and boy do those hate overs. Fortunately this fad dissipated rather quickly after the early 2000s.
I think GoldenSound has more or less stopped doing subjective reviews of DACs now. He seems to have had a change of heart on the matter and now considers measurements to be king.
I am not trying to judge him and it is really not my intention. In the specific article of him that I have linked above I can see good technical approach and that is the reason why I posted the link. I do not prefer black and white approach.
little-endian
Member
- Joined
- Apr 16, 2023
- Messages
- 92
- Likes
- 98
Neither do I still fully understand why Benchmark opted for using a potentiometer (prone to wear and eventually noise or even distortion) for the analog volume control in the DAC1. Even at that time, there were already DSP- and IC-based ones available - as commonly used by AVRs. If I remember correctly, they didn't want to use ICs due to concerns about their quality (fair enough), but neither DSPs due to their SNR reduction (academically comprehensible but worth being discussed).
I am very well aware of John Siau's explanation which goes like "analog volume reduction preserves the SNR while a DSP based one before a DAC chip by principle doesn't". While that is technically true, I still fail to see the practical relevance as I as the listener will lose that part of SNR either way as my ear by definition also will only achieve the full dynamic range when utilizing the threshold of unpleasantness (which probably isn't healthy anyway). As soon as I lower the volume, I'm also left with only my fraction of dynamic range and SNR so I don't see why the technical loss should matter that much. I think RME argued in a similar way once, seeing DSP-based volume controls despite the technically reduced SNR as the best solution of all.
Well, meanwhile Benchmark also uses a DSP-based solution for the volume or actually "gain" control (although in a bit more traditional way with a fixed max and min and no "rotary" control as Amir put it), at least for digital sources and as mentioned in the article, besides taking any mechanical wear out of the equation, they also have the advantage to elegantly circumvent the intersample peak issue alongside as one rarely will use the DAC at its full volume setting but way further down than -3dBFS so the headroom for intersample peaks will be huge in practise.
Still, it would be interesting to know how big the headroom is at the 0dB setting of today's DACs (when used in conjunction with an amplifier and not the maybe built in headphone output) and also how they implement any SRC which I guess is the crucial part if happening before* the volume control.
* if I understand it correctly, usually a sample rate conversation is done before the volume control where e.g. RME offers an optional fixed headroom of 3dB. What in general speaks against doing such SRCs after the volume control? Guess nominally a loss in precision for the math, but how relevant is that in practise if properly dithered in between, I wonder.
I am very well aware of John Siau's explanation which goes like "analog volume reduction preserves the SNR while a DSP based one before a DAC chip by principle doesn't". While that is technically true, I still fail to see the practical relevance as I as the listener will lose that part of SNR either way as my ear by definition also will only achieve the full dynamic range when utilizing the threshold of unpleasantness (which probably isn't healthy anyway). As soon as I lower the volume, I'm also left with only my fraction of dynamic range and SNR so I don't see why the technical loss should matter that much. I think RME argued in a similar way once, seeing DSP-based volume controls despite the technically reduced SNR as the best solution of all.
Well, meanwhile Benchmark also uses a DSP-based solution for the volume or actually "gain" control (although in a bit more traditional way with a fixed max and min and no "rotary" control as Amir put it), at least for digital sources and as mentioned in the article, besides taking any mechanical wear out of the equation, they also have the advantage to elegantly circumvent the intersample peak issue alongside as one rarely will use the DAC at its full volume setting but way further down than -3dBFS so the headroom for intersample peaks will be huge in practise.
Still, it would be interesting to know how big the headroom is at the 0dB setting of today's DACs (when used in conjunction with an amplifier and not the maybe built in headphone output) and also how they implement any SRC which I guess is the crucial part if happening before* the volume control.
* if I understand it correctly, usually a sample rate conversation is done before the volume control where e.g. RME offers an optional fixed headroom of 3dB. What in general speaks against doing such SRCs after the volume control? Guess nominally a loss in precision for the math, but how relevant is that in practise if properly dithered in between, I wonder.
little-endian
Member
- Joined
- Apr 16, 2023
- Messages
- 92
- Likes
- 98
restorer-john
Grand Contributor
Benchmark is in the business of selling D/A converters and other audio products- don't forget that.
It's not unexpected that any of their commentary, "discoveries", educational white papers and design choices would support whatever direction they are taking at this point in time. Of course, all that is subject to change going forward.
The construction quality of their products is outstanding in my opinion. Totally fit for purpose. Not inexpensive, but not unreasonable either.
Just don't believe everything they say is gospel.
restorer-john
Grand Contributor
So the original .wav files they sent you were 16/44 or something else (16/48)? Just wondering if it was the effects of a SRC at the mastering for CD/disc manufacturing end.
little-endian
Member
- Joined
- Apr 16, 2023
- Messages
- 92
- Likes
- 98
Just to make sure - in my comment about the "mixed feelings" I was entirely referring to GoldenSound, not Benchmark.
restorer-john
Grand Contributor
I get that. Benchmark were of course on message, prosecuting their narrative in the comments/responses to GS's review linked.
They have a pattern. It probably works for them and likely for many potential customers of theirs.
Similar threads
- Poll
