CHANGSTAR: Audiophile Headphone Reviews and Early 90s Style BBS
Lobby => Amp and DAC Measurements => Topic started by: Marvey on May 13, 2015, 05:09:42 AM
-
Take with a grain of salt. Using Focusrite POS. Input is 16/96 (with gain shoved up), so might be some aliasing. Low noise floor, but quite a bit of crap in the treble. Ignore the uV - I didn't calibrate. However, it does give us some idea what is going on.
16 and 24 bit data -90db undithered 1kHz sine wave
-
I will never know how the first three hours of the yggdrasil sounded as I let the DAC warm up for an hour and proceeded to make technical measurements. If the measurements presented are familiar territory, then you should be as impressed as I am. If such graphs are not familiar, then please note that these graphs are extended, some down to -160 dB for example. Extreme. Much farther than most other graphs you will see elsewhere on other DACs on the internet or in brochures. So be careful when making comparisons. Watch the X and Y axes in other data. The graphs presented here are not marketing graphs. They have not been prettied up as is often the case in mainstream consumer goods.
Schiit has specifications for these measurements published on their website and in their operating instructions. In my measurements, my yggdrasil measured better than Schiit's conservative specifications in every single measurement. Every single one. Well done Schiit!
Commentary added to the end of this post.
The data presented were collected as follows:
1. PrismSound dScope III, picoscope 2205, FLIR E63900
2. DAC balanced output used for measurements except oscilloscope square wave response which used unbalanced output
3. 100 Kohm load used for measurements
4. 44 KHz sample rate, 24 bit depth
5. USB input - cables to go USB cable
6. Balanced cables Tecnec Canare starquad Neutrik XLR cables
7. Unbalance cable BlueJeans Belden RCA
8. Vaunix Lab Brick USB hub
Commentary:
By now, some of you probably have noted the THD+N graph may not be the lowest ever seen.
The THD+N graph measures artificially high due to a driver / OS interaction on my computer. An update will be forthcoming.
Designers are often faced with trade-offs for various reasons. I believe the Schiit design team has made trade-offs that result in best auditory experience over best measurements possible on all fronts. I have a colleague who gives a "13 dB miracle" demonstration showing that SNR isn't the best predictor of auditory experience. So THD+N may not be that influencial on the final sound while jitter may have a bigger effect. Those of you who will reference the 1992 and 1998 AES papers on the topic of jitter may note that no trained subjects were used to determine the masking thresholds. With many more listeners now well experienced in digital audio listening we might see much lower thresholds if the experiments were repeated today and trained listeners employed.
-
Sorry to spread this over multiple posts. More measurements from 3 hr on time.
-
Yggdrasil measured at 408 hrs on time.
Test setup is same as described for the measurements taken at 3 hrs on time with the following exception: A PicoScope 5243B was used in place of the 2205. The 5243B has 15 bit 2 ch ADC capability and higher sample rates than the 8 bit ADC used on the 2205.
I will try to include a few extra screenshots though this posting system only allows 8 per post and I only reserved two posts for these measurements. Most measurements were so close to the original to be considered of no consequential change. Except one spectacular change. If the 3 hr on time jitter spectrum was impressive, take a look at the 408 hr on time jitter spectrum. OMG! My congratulations to Mike Moffat. That is a truly impressive achievement!
The THD+N graph measures artificially high due to a driver / OS interaction on my computer. An update will be forthcoming.
-
Holy shit, perfect 3 steps and no ringing due to gibb's phenomenon. The closed form filter works.
-
Yggdrasil measurements at 408 hrs on time, part B.
I have included a 1KHz -70 dBFS time domain signal to demonstrate that the blib seen in the -90 dBFS version is pretty insignificant.
-
On the 1 kHz, what are those "blips" at the 0 transitions?
-
Crossover distortion.
-
Makes sense. I can see it in Marv's plots too, but the Focusrite may be masking it a bit due to noise at the interface. The 2i2 I have has some line in gain knobs that can alleviate the problem if properly adjusted. But that's a pain in the butt and even then it may still be noisy.
Holy shit, perfect 3 steps and no ringing due to gibb's phenomenon. The closed form filter works.
Based on the time domain ringing period it's about 22 kHz BW (in agreement with the 44 kHz sampling rate settings). Would be interesting to see the FR all the way to say 96 kHz or more. It is not a perfect sinc but perhaps something else that alleviates ringing with a slower/smoother FR roll off.
-
Crossover distortion.
Indeed, not very severe though.
All 'normal' ladders have it in some severity.
Only with signed magnitude it isn't obvious as it isn't in the 0V line, instead it has 2 of them at the -6dBFS points.
It is only possible to 'see' it when you zoom in on that point. Nobody is going to.
It may be less audible in SM because in softer passages as there is no 0 crossing and the signals would be well below -6dB or even -12dB.
impressive jitter figures...
-
impressive jitter figures...
My sentiments exactly! Made my jaw drop when I measured it. Agreed about the crossover distortion inaudibility. Will be looking again at approx 400 hrs. Would have liked to catch 200 hrs but that darn career thing keeps soaking up my time such that when I do get a free moment, I'm more inclined to listen to music than measure the system.
-
Yeah really nice clock indeed .... well done mr. Moffat !
Those crossover blibs are just at -104dB and at a frequency of around 17kHz (for this test tone).
Judging from the waveform I figure it may not be possible to improve this using forward correction, perhaps using signed magnitude.
What would interest me is the differences in jitter between 'cold', 'several hours' and 'mucho hours' as well as seeing if the 'blib' gets smaller when warmed up.
After all the accuracy of most components is temperature dependent.
I don't think any other aspects/measurements will change over time, unless they are related to the 'blibs', which I think is visible in the form of the relatively high frequency 'poles' in the upper part of the spectrum.
But why scrutinise measurable thresholds that may not be that audible ? Ah yes ... because we do want to know.
If it is temp dependency of one specific component, sticking a tiny regulated heater (resistor) on it may greatly shorten the long 'wait before use/magic apppears period' .
Wouldn't it be great if you could switch it on and in 5 mins would do 'the magic' (IF caused by actual warming up of certain components)
I totally get the 'available time' thing vs curiosity.
-
If it is temp dependency of one specific component, sticking a tiny regulated heater (resistor) on it may greatly shorten the long 'wait before use/magic apppears period' .
Wouldn't it be great if you could switch it on and in 5 mins would do 'the magic' (IF caused by actual warming up of certain components)
After all there are oven controlled DACs out there.
-
By now, some of you probably have noted the THD+N graph may not be the lowest ever seen. Designers are often faced with trade-offs for various reasons. I believe the Schiit design team has made trade-offs that result in best auditory experience over best measurements possible on all fronts. I have a colleague who gives a "13 dB miracle" demonstration showing that SNR isn't the best predictor of auditory experience. So THD+N may not be that influencial on the final sound while jitter may have a bigger effect.
Try measuring at -3db and -6db, ladder DACs tend to higher THD as you approach 0dBFS.
-
That's what the Lavry Gold does, among others.
-
I dunno about the Prism, but according to the legend, THD+N was measured at -0.1 dBFS where 0 dBFS maps to 14.76 dBu (according to the script log). This is close to 4 Vrms which matches Schiit's balanced drive specs. Atomicbomb would have to confirm...
I'm a bit puzzled about the "Soundcard Outputs" control. It seems it's using WDM Microsoft Sound Mapper. Where these (https://schiit.com/drivers/)drivers used?
-
After all there are oven controlled DACs out there.
Yes, but I assume these are temp controlled clockgenerators only.
-
Wait, there are actual oven dacs? I thought people were joking...
-
Clockgenerators with a built-in 'oven' are quite common if a stable frequency (does not mean low jitter) is needed.
They are called OCXO (Oven Controled X'tal Oscilator)
Quite costly ($ 50 to $ 400) and about 10 to 100x more stable than 'normal' XO's
Here is an example: http://www.farnell.com/datasheets/1883682.pdf (http://www.farnell.com/datasheets/1883682.pdf)
A bit overkill for a DAC
(should be in DAC guts stuff perhaps)
-
Oh that makes more sense. I was picturing little dac saunas for a second there.
-
Oh that makes more sense. I was picturing little dac saunas for a second there.
I was picturing something more like diesel-engine glow plugs. And I'm a bit surprised at the pic showing the Yggy's temperature. In this climate that is a problem
(Or to put it another way, at last I have a reason, apart from the money, not to buy one!)
-
Yes, but I assume these are temp controlled clockgenerators only.
Usually, but sometimes not only the clocks. E.g. Lavry's Gold:
"The PCM DAC is constructed of custom made laser trimmed thin film resistor networks; yet any resistor is subject to short term drift due to temperature variations and long term drift due to component aging. The resistor networks are kept at a constant temperature by a linearly controlled heating element. (A bang-bang controller, such as a home thermostat is undesirable because it produces turn on and turn off surges-- thus audible kicks.) Keeping the resistors at constant temperature overcomes resistor dependency on environmental temperature variations."
If all else fails, why not sell this as an Yggy accessory case:
(http://i.walmartimages.com/i/p/00/05/08/75/80/0005087580716_500X500.jpg)
-
Those will be fine..
these won't
(http://mef.com.my/webshaper/pcm/pictures/Panasonic/Microwave/NN-SM330M_l.jpg) :P
-
I don't think either of those would be big enough... but maybe a tabletop pizza oven?
-
Indeed, not very severe though.
All 'normal' ladders have it in some severity.
Only with signed magnitude it isn't obvious as it isn't in the 0V line, instead it has 2 of them at the -6dBFS points.
It is only possible to 'see' it when you zoom in on that point. Nobody is going to.
It may be less audible in SM because in softer passages as there is no 0 crossing and the signals would be well below -6dB or even -12dB.
impressive jitter figures...
I'm surprised by the cross over distortion. I don't see it when measuring the outputs of my M7 (using Rigol DSO)... Will post a SS when I'm not posting from iPhone.
I'm not experienced on the matter, and the FFT measurements show all other signals other than fundamental are near noise floor. But I wouldn't of expected it in a device where fidelity is meant to be the main goal...
-
Those crossover blibs are just at -104dB and at a frequency of around 17kHz (for this test tone).
Judging from the waveform I figure it may not be possible to improve this using forward correction, perhaps using signed magnitude.
I'm surprised by the cross over distortion. I don't see it when measuring the outputs of my M7 (using Rigol DSO)... Will post a SS when I'm not posting from iPhone.
I'm not experienced on the matter, and the FFT measurements show all other signals other than fundamental are near noise floor. But I wouldn't of expected it in a device where fidelity is meant to be the main goal...
You won't see cross over distortion from your M7. Delta-Sigma recovery filters assure there won't be any such artifact. Part of their "charm". Look at Solderdude's comments again. This crossover distortion is in the last two bits of a 20 bit DAC at the zero-crossing only. It is above 17KHz in nature. It is far less audible (if at all) than jitter spectrum on many high-end DACs. I was hesitant to post this particular graph due to exactly this potential for mis-interpretation. I also have linearity graphs but am really hesitant to publish them for the same reason.
-
Ahh that makes a bit of sense to me - and I'm glad you posted that information so we could discuss it and so I could learn something new :P As I said I think practically it makes no difference, your explanation re bit accuracy further justifies it (earlier I was trying to consolidate the 20/21 bit accuracy rating of the DAC with the time domain distortion but I think you've covered it but I'll have to do some more reading I think).
The M7 is an R2R DAC though btw...
-
Bob, what other parameters/measurements (besides jitter) you feel are important to look at that correlate well with perceived performance?
Also, any reason to use WDM over ASIO for the measurements? (I'm not familiar with Yggy driver support nor the Prism, and that's what the screen captures seem to display)
Where the measurements done at -0.1 dBFS (4.something Vrms)?
Any higher rate measurements?
Thanks! :)
-
The M7 is an R2R DAC though btw...
Ok, brain fart. I saw and wrote M7 but was thinking M51. Mental dyslexia kind of thing.
-
Bob, what other parameters/measurements (besides jitter) you feel are important to look at that correlate well with perceived performance?
Also, any reason to use WDM over ASIO for the measurements? (I'm not familiar with Yggy driver support nor the Prism, and that's what the screen captures seem to display)
Where the measurements done at -0.1 dBFS (4.something Vrms)?
Any higher rate measurements?
Thanks! :)
I'm not sure I'm the one to answer questions relating measured performance against perceptual performance. One of the reasons I obtained a dScope and several other rather expensive test bench items such as a Klippel analyzer and Keyence laser displacement meter. I am trying to figure this out for myself as part of my acoustic research.
I'm using the WDM drivers only because the dScope is finicky during USB sound card setup when using ASIO drivers and it takes a bit of fiddling to make it all play reliably. Using either AES/EBU or SPDIF is a breeze by comparison. But everyone wants to know about USB performance so that is what I am focused on initially.
There may be higher rate measurements later on. Setup, measurement and screen capture all take some time for each test. I haven't reached retirement or independent wealth, so my hours are filled performing acoustic research for a corporate entity. I'm hoping to catch any changes after 400 hrs of constant power on time this weekend. I missed the 200 hr mark as my corporate gig needed attention at that time.
-
LOL! I more than understand Bob. Thanks!
-
LOL! I more than understand Bob. Thanks!
BTW, I've probably watched A Night in Casablanca more times than anyone else alive. It comes to life heard through a WA7 and T90 from an OPPO 103
-
What kind of work do you do that gives you access to a dscope? Or do you just have one for home/personal use?
-
What kind of work do you do that gives you access to a dscope? Or do you just have one for home/personal use?
That is a very good question. I found my passion in life and have developed it for both professional and personal benefit. It is a long story I should document sometime. Suffiice it to say that I am responsible for voice prompting systems used by Emergency Medical Services devices in 40 languages, in almost every country in the world. Robust systems that are critical to be heard in noisy environments. Psycho-acoustics and the human auditory system are quite important to me. Building my personal acoustic lab has been an obsession, rather than owning a boat, sports car, motorcycle, helicopter, etc. My personal lab is far more sophisticated than my corporate lab, though budget is slowly being made available to bring parity at sometime in the future. I intend to keep at my acoustic research as long as I am able to function.
-
So, if you ever need to hire an engineer... :)p8
-
I'm surprised by the cross over distortion. I don't see it when measuring the outputs of my M7 (using Rigol DSO)... Will post a SS when I'm not posting from iCrap.
I'm not experienced on the matter, and the FFT measurements show all other signals other than fundamental are near noise floor. But I wouldn't of expected it in a device where fidelity is meant to be the main goal...
The M7 has a sign magnitude DAC (PCM1704) that is actually intended for audio.
Because it is sign magnitude the MSB doesn't 'topple' at the '0' Volt point but it does at the -6dBFS point instead where it is 'masked' by the large signal level.
The chip (AD5791) used in this DAC is not designed for audio in a sense that it needs another physical input signal (non standard audio related format) and post processing of the analog signal (deglitching) compared to chips designed for audio.
Because of its wide operating temperature range and stability it has been approved/designed for military and aerospace.
Not because it is more accurate or 'better'. :o
The somewhat confusing part here is that Schiit uses 2 of these chips per channel and claim 21 bit resolution so it stands to reason mr Moffat made it signed magnitude as well.
As the smallest 'step' of the PCM1704 is 23 bit (accuracy seems decent so who knows how small the steps can be) and the AD5791 is 'merely' 20 bit the accuracy of the first steps of the 1704 are likely to be better as in more accurate as theoretical the smallest possible step is 8x smaller in the PCM1704.
The large amount of spikes that run fairly high in level compared to most other DAC's around are caused by the LSB step size most likely or a very, very small DC offset in the DAC output signal which would only be a problem if used in signed magnitude configuration.
Again, that chip isn't designed for this specifically.
It stands out even more because the noise floor of the analog section is very good as well.
In practise it is too bad that extremely low noise floor is decreased by say 20 to 30dB as when music is played there will be lots more 'poles' all over the freq band creating a higher noise floor.
I wouldn't care about this because I would like to see a recording that actually has a noise floor below -100dB anyway.
However, like with everything else it's not just the DAC chip that matters but the implication.
The digital and analog signal handling, PCB design, power supply and other used parts are of greater importance.
I seriously doubt if any 'garbage' below say 120dB would be audible as you would have to play extremely loud for it to reach audible levels.
At least that's what I think.
Aside from that, Schiit uses magic in their designs where AGD just uses chips. 8)
It seems to me mr. Moffat knows his Schiit and if its good enough for the goose .....
And ... did I tell you I am impressed with the amount of jitter (even though I could not care less about this)
-
I wouldn't care about this because I would like to see a recording that actually has a noise floor below -100dB anyway.
Not going to happen in a live situation due to the room background noise. Might be possible in a studio that has attended to HVAC, Low Frequency isolation, etc. Not many studios manage both along with good room acoustics. Very expensive. Then again, we have the microphone self noise with which to contend. The greater the number of mics in use, the higher the residual noise.
I seriously doubt if any 'garbage' below say 120dB would be audible as you would have to play extremely loud for it to reach audible levels.
The sound level necessary would only be listenable during a very very quiet passage of a classical recording or dramatic pause in anything else. The rest of the recording would be distorting the electro-acoustic transducers due to the SPL being produced at such a high gain.
And ... did I tell you I am impressed with the amount of jitter (even though I could not care less about this)
Very impressive jitter performance, yes. I used to be of the same opinion that jitter below a certain point didn't matter. But rereading the 1992 and 1998 AES papers on the audibility of jitter one notes that no listener training was performed. The subjects didn't know what they were supposed to detect. When we have all kinds of comments about digital hash, glare, something not quite right etc. one has to wonder if jitter is possibly a contributor along with the recovery filter characteristics.
-
Yes, measurments on AES/EBU input would be nice...given all the attention USB gets, I'm wondering if it by now measures better (e.g. my paranoia).
First post for me, didn't do intro post yet.
-
I'm surprised by the cross over distortion. I don't see it when measuring the outputs of my M7 (using Rigol DSO)... Will post a SS when I'm not posting from iPhone.
I'm not experienced on the matter, and the FFT measurements show all other signals other than fundamental are near noise floor. But I wouldn't of expected it in a device where fidelity is meant to be the main goal...
Did you make some measurements with your DSO? Which model?
I'm guessing it was hard to find anything to observe, as even Rigol 2000 series and above have only 500uV/div resolution. The yggdrasil crossover distortion measured on the dScope is at approximately 43 uVpp so it would be very hard to see on most DSOs.
-
Oh I have the piccies on my iPhone but need to get onto a PC to post them, will update this post with them when I get home.
I've done some more reading to try and understand all this. The PCM1704 chips implement signed magnitude so I wouldn't expect to see any cross over distortion. But my waveform SSs also aren't zoomed in to the scale you mentioned so it may not be evident (if it exists). One thing I still don't understand is reconciliation of the rated distortion figures of the AD chips vs the distortion measurements - is what you've measured within spec of +/- 1 LSB? (Probably because I don't understand the DBFS scale lol.... Next thing to read up on).
I'm not sure how Schiit have implemented the chips per channel but I don't think it signed magnitude - prolly one chip per balanced hot signal.
My DSO is a Rigol DS1054Z (software upgraded).
For context, I took these measurements to understand the full scale output of the DAC so I could fix gain of my amp accordingly (did these a few weeks ago) - I didn't trust AGD website's ratings... So they may not serve our purpose here...
(http://www.overclockers.com.au/pix/image.php?id=kpddb&f=1')
Click to view full size! (http://'http://www.overclockers.com.au/pix/index.php?page=image&id=kpddb')
-
I've done some more reading to try and understand all this. The PCM1704 chips implement signed magnitude so I wouldn't expect to see any cross over distortion. But my waveform SSs also aren't zoomed in to the scale you mentioned so it may not be evident (if it exists). One thing I still don't understand is reconciliation of the rated distortion figures of the AD chips vs the distortion measurements - is what you've measured within spec of +/- 1 LSB? (Probably because I don't understand the DBFS scale lol.... Next thing to read up on).
I'm not sure how Schiit have implemented the chips per channel but I don't think it signed magnitude - prolly one chip per balanced hot signal.
My DSO is a Rigol DS1054Z (software upgraded).
Very good, thank-you. The DS1054Z has an 8 bit ADC for input with an ability to SW oversample to an equivalent of approx 12 bits. Maximum sensitivity is 1mV/div so it will be very hard to observe phenomena below 100uV. I have a picoscope 5243B which has a 16 bit ADC and same SW oversample feature to 20 bits, yet only has 2mV/div resolution. We both need a 20dB low noise probe amplifier to see this type of zero crossing distortion on our oscilloscopes. One of the reasons the dScope is sooo expensive. The front end resolution it can attain.
-
I think scopes in general can do much higher sampling rates. Audio Analyzers don't go that high in sampling rate. Proly different ADCs.
-
Here is a link to an ECN article which discusses the oversampling feature in an oscilloscope:
http://www.ecnmag.com/articles/2013/01/how-get-more-8-bits-your-8-bit-scope
-
Yes. I can see that working out. May also depend on SNR.
-
Wowzors, -90 DBFS is a seriously tiny signal (0.00003162 of full scale voltage). Puts the mid-scale/zero crossing glitch amplitude in perspective...
Just out of curiosity, how does the time-domain vertical scale work in your graph? It's not consistent increments in DBFS between tick-marks? Not familiar with the Dscope meter.
And following what Solderdude said earlier, I'd expect these glitches to appear as second-harmonics (2 x Fo)? Which in this case is 2 kHz not at 17 kHz? I had this confirmed in a paper I read here - Page 5 of http://www.analog.com/media/en/training-seminars/tutorials/MT-013.pdf
-
Indeed! Why neither Solderdude nor I am concerned about the zero crossing distortion. My graph is extremely magnified in the Y axis to see it. dScope information can be found here:
http://www.prismsound.com/test_measure/products_subs/dscope/dscope_home.php
The frequency is being estimated by looking at the x-axis glitch period and inverting.
May 17 my yggdrasil will hit 400 hrs. I hope to have some time free to repeat the measurements at the beginning of this thread and fill in those reserved posts.
-
I should clarify that the artfifact shown (the crossover blib) is not a glitch as described in the linked article.
The glitches (which the used DAC chip indeed has) have already been removed.
The 0 crossing thing you see is (most likely) caused by DAC chip '1' taking over the positive half of the sign wave to the negative half of DAC chip '2'.
I suspect (no evidence here, just a hunch, I guess mr. Moffat or Jason can correct me) that the chips are in signed magnitude config (hence the 21 bit resolution with 2x 20 bit chips) and perhaps the 'blib' is caused by an extremely small DC offset between the '0Voutputs' of both signal halves where they are 'combined' to create the complete signal.
The 'size' of the blib may actually differ per Iggy as well.
In all cases ... the blib may be quite measurable, because the noise floor of the electronic itself is very low, but is so low in amplitude it will be inaudible.
The practical noise floor, with an actual audio signal on it, will be higher than better measuring/performing DAC's for sure but as mentioned before will still be much lower than any real world recording out there so will be 'masked'.
Indeed as atomicbob already answerred the 'fundamental frequency' of the blib is estimated by looking at the length of a known period (the actual sinewave of 1kHz) in millimeters on the screen and the periodlength of blib and simply dividing those numbers and multiply that result x 1kHz (the used sinewave).
As that 'blib' has 'sharp edges' it will also have harmonics but those are far above the audible range.
The amplitude of the blib is estimated in the same way by measuring (in peak mm's on the screen) a known amplitude (-90dB) and the amplitude of the blib and crunching some numbers.
There is no higher math, assumptions, or theories involved other than quite basic stuff.
-
408 hour on time measurements now added. Now I'm going to listen to my damn schiit. If I don't come up for air, it is because I have lost track of time again. This DAC is truly that good!
Maybe next weekend I will get to the AES and s/pdif input measurements. I'm out of energy for today.
-
No changes in measurements after that many hours ?
Jitter still the same ?
-
The jitter spectrum improved significantly. Have a look at the graphs. They are now in the posts previously reserved on the first page of this thread.
-
The jitter spectrum improved significantly. Have a look at the graphs. They are now in the posts previously reserved on the first page of this thread.
This is very interesting. Perhaps the biggest improvement of yggy's sound comes from jitter rejection?
-
Thanks..
Jitter improved indeed even more .
Should you ever have time to make some 'cold' measurements again the jitter part and crossover parts would be most interesting.
Unless I calculated incorrectedly the ENOB is about 18 bits, which isn't bad at all.
-
So, either subjective accounts of improved audio performance correlate to the improved jitter performance over time, or we are still measuring the wrong thing and the data/impressions relationship is just coincidental. What was the supposed proclaimed level where jitter became inaudible and everything sounded the same again?
-
So, either subjective accounts of improved audio performance correlate to the improved jitter performance over time, or we are still measuring the wrong thing and the data/impressions relationship is just coincidental. What was the supposed proclaimed level where jitter became inaudible and everything sounded the same again?
Jitter isn't the problem in the first place.... is what people at hydrogen audio say.
-
Jitter isn't the problem in the first place.... is what people at hydrogen audio say.
I don't care what fucktards say obviously, that's why I ask here.
-
I don't care what fucktards say obviously, that's why I ask here.
Yes, that's why we do not have any defined answer to jitter problem because people think such problem does not exist in the first place.
From my very limited and flawed readings from the papers, it is supposed to be the length of jitter matters, not the level. When it becomes a certain ns short enough, jitter is supposed to be inaudible...
Sigh, here my lack of expertise is shown. Someone who knows better should answer this question.
-
Maybe a measurement or plot artifact, but the 400+hour square wave closeup looks clipped on the first overshoot (both pre and post ringing), whereas the same measurement at 3h looks clean.
The squarewave plots are listed as being taken single ended. Is the squarewave full scale?
-
Ok, about jitter, I found an outstanding article from Amir of WTB forum.
http://www.madronadigital.com/Library/AudibilityofSmallDistortions.html
Using Hawksford and Dunn research for example, a DAC which generates less than 20 picoseconds would be transparent to its source (assuming jitter is the only distortion we are worried about).
So I guess it's 20 picoseconds.
-
... from Amir of WTB forum.
Oh my god. No thanks
I do spend time at Hydrogen Audio, and have would prefer to avoid Amir for ever.
-
So, either subjective accounts of improved audio performance correlate to the improved jitter performance over time, or we are still measuring the wrong thing and the data/impressions relationship is just coincidental.
Good question.. the answer will depend on who you ask.
Is it jitter ?
What kind of jitter ?
Is it the jitter frequency or frequency spectrum ?
Is it caused by being R2R or other type of conversion ?
Is it caused by filtering ?
Is it a combination of all of the above ? And in what 'mix' ?
If it's only the jitter that alters during the warm up time then obviously when we can trust the ears on here the treshold of 'magic' is below 3ps and with unknown spectrum.
Obviously 18 bits resolution is enough for 'magic' and as many report red book also shows magic than the 'simple conclusion' could be:
18 bits/44.1khz, upsampling while retaining bit perfect reproduction and jitter below 3ps should do it.
Regardless of transfer type ? USB vs other connection methods.
What was the supposed proclaimed level where jitter became inaudible and everything sounded the same again?
That too will depend on who you ask.
I am curious about those numbers as well.
Maybe someone should spend time on this researching it while using audiophile/trained ears and capable equipment/recordings (that obviously needs to have low jitter in the ADC stage as well)
Also would be interesting to 'test' people plucked of the street and determine the gap between trained and untrained.
Who is going to fund the research ?
I can't participate in the listening tests though as I am (fortunately for me) DAC-deaf
-
Oh my god. No thanks
I do spend time at Hydrogen Audio, and have would prefer to avoid Amir for ever.
While he is as stubborn as folks at Hydrogen Audio, it is not wise to dismiss his knowledge on digital audio completely. There are many stuffs to learn from his debates with HA people.
-
So I guess it's 20 picoseconds.
That figure is pretty much full of shit. http://www.head-fi.org/t/668878/jitter-correlation-to-audibility/30#post_9544922
-
While he is as stubborn as folks at Hydrogen Audio, it is not wise to dismiss his knowledge on digital audio completely. There are many stuffs to learn from his debates with HA people.
You may be right, but I still think I'll stick with some other information sources.
Hydrogen audio has a dogma which might not suit many. I find it a useful balance (although I've given up some of fantasy-based audiophoolery stuff I used to hang around) but, personally, I can't live life in the the lab, I want to live it chatting over a drink too, so I like my objective with a dose of subjective and vice versa.
Jitter is something that has always bothered me because I suspect that, in audiophile-land, it gets used often by a lot of people who have no clue what it means or sounds like or what, technically, it is. I'm not talking Changstar: I doubt that there is anyone on this page who doesn't understand jitter far better than I do. But for the ordinary, non-engineer audiophile (like me!) it to to easy to think, wow, flutter, rumble, jitter and start supposing stuff that may, or may not, be true.
There are some sample files on hydrogenaud.io, with various levels of applied jitter. Whether one uses DBT with them or not (I didn't) it is interesting to get, at last, actual samples of actual jitter. I know that I had to go some way up the various intensities before I could hear it, but I do not make any claims other than personal preference, based on my own hearing, which is getting, according to the audiograms, quite bad.
I also understand that training and experience can enable people to detect digital artifacts: a necessary part of the skills of those working with digital audio, lossy file formats, etc.
-
Thanks..
Jitter improved indeed even more .
Should you ever have time to make some 'cold' measurements again the jitter part and crossover parts would be most interesting.
Unless I calculated incorrectedly the ENOB is about 18 bits, which isn't bad at all.
Yes, starting over with the yggdrasil having cooled for at least 24 hrs is in the plan. It may not occur for a few months as I have a short but intense set of tasks in my queue at the corporate gig near term.
-
Even if talking about any of Stereophile measurements, these jitter measurements are really the best I've seen from usb ever. For the imaging test, it would be nice to superimpose a 19khz 44.1ksps tone measurement on top of the wideband response.
-
Been reading around the internets lately trying to learn about DACs, filters, up sampling, etc and came across an exact interpolation process. I.e. time domain -> FFT -> zero pad FFT -> inverse FFT resulting in up sampled time domain signal (exact interpolation values and keeping of original samples) as opposed to estimated interpolation samples based on accuracy of FIR filters.
Key advantages are the process:
- Keeps original samples
- Exact interpolation (bit perfect?)
Sound familiar?
I ran some basic/simple code to confirm the maths and it seemed to work well based on random number generated waveforms. I think Schiit managed to figure out how to implement this process in a DSP, batched process maybe?
Additional info here:
http://www.dspguru.com/dsp/howtos/how-to-interpolate-in-time-domain-by-zero-padding-in-frequency-domain
http://dsp.stackexchange.com/questions/14919/upsample-data-using-ffts-how-is-this-exactly-done
Anyone seen this implemented in audio DACs before? I don't have a Yggy myself (yet) and am thinking to upsample with this process on the PC and run my DAC in NOS (although I'm not sure if the M7 runs true NOS based on my previous waveform observations)...