Hi
Thank you for posting some really interesting findings! Just out of interest, would it be worth adding anything about differences between the Dynamik and it's predecessor? I'm not after any subjective stuff (as you know, I'm musically as deaf as a post) but it'd be really interesting to hear about some of the objectives and measured differences (it also falls in line with stuff I used to do as a hobby, so I'm particularly curious about it all). Of course, the key design principle behind the Dynamik might be subject to IP rights in that you don't want to give away any cool ideas to your competitors; I'll thus fully understand if it's not appropriate to respond to anything about the concepts behind it (I'm just being nosey again)!
Bri
(2010-11-18 11:26)murrays Wrote: [ -> ]We have done extensive measurements on power supply disturbance recently, and have compared results for both FLAC and WAV streaming. Our findings are as follows :
1. If we measure the power rail that feeds the main processor in the DS we can clearly see identifiable disturbance patterns due to audio decoding and network activity. These patterns do look different for WAV and FLAC - WAV shows more clearly defined peaks due to regular network activity and processing, while FLAC shows more broadband disturbance due to increased (but more random) processor activity.
2. If we measure the power rails that feed the audio clock and the DAC we see no evidence of any processor related disturbances. There is no measurable difference (down to a noise floor measured in micro-volts) between FLAC and WAV in any of the audio power rails.
3. Highly accurate measurements of clock jitter and audio distortion/noise also show no difference between WAV and FLAC.
The extensive filtering, multi-layered regulation, and careful circuit layout in the DS ensure that there is in excess of 60dB of attenuation across the audio band between the main digital supply, and the supplies that feed the DAC and the audio clock. Further, the audio components themselves add an additional degree of attenuation between their power supply and their output. Direct and indirect measurements confirm that there is no detectable interaction between processor load and audio performance.
Hi,
That are interesting results...
very positive that you did this testing (sorry for my english)
But why can I hear difference between flac en wav...
I had some blind tetst at my superb dealer in The Netherlands. I was not familiar to what I was listening to. he told me to sit down and listen.
I thought that he was ripping with different settings or sofware.
I could clearly hear differences..
After a while he told me that I was listening to flac and wav ripped files.
Therefore I know that there are differences between waf and flac files....
Many regards,
Ad.
@Briain
These tests were done using a Dynamik power supply, but I would expect similar results from a non-Dynamik supply. The bulk of the isolation comes from the design of the local power supplies on the main DS circuit board. The Dynamik advantage is more to do with reducing switching energy from the main supply. It achieves this by using a higher switching frequency (well outside the audio band) and spread-spectrum modulation (which I'm sure you're familiar with
).
@advr
Are you saying that you can reliably distinguish (in a blind test) between FLAC and WAV versions of the same audio data ?
(2010-11-18 13:58)murrays Wrote: [ -> ]@advr
Are you saying that you can reliably distinguish (in a blind test) between FLAC and WAV versions of the same audio data ?
YES, In a blind test I can hear differences. 4 of 5 times I prefered wav (but I didn't knew that when I was listening). All recordings were different between flac and wav.
Of course flac and wav ripped from the same cd, ripped with the same rip software and settngs.
System used: Audio Research Ref 3 pre, VT100 power, Klimax DS, Wilson Benesch Trinity's and Velodyne DD 18.
Dealer: AudioExcellence, Zevenaar. I can say that now. They know everything about ripping and streaming. Sometimes I am there guinea pig.
Ad.
(2010-07-29 20:28)ThomasOK Wrote: [ -> ] (2010-07-28 17:54)Tifany Wrote: [ -> ] (2010-07-28 13:18)Gibson Wrote: [ -> ]Try it before making any blind comment. Even the Linn dealers said there is difference between WAV and FLAC played by DS.
What do you mean by blind comment ? I just want to understand how a DS works, and from a scientific point of view, if the data are the same and the processes too, it should sound the same!
Only Linn engineers could give us useful information on these points.
A Linn engineer has weighed in. In a post a few months back Phillip Hobbs said that listening tests on the final KlimaxDS prototype found Linn engineers preferring FLAC over WAV. It was believed to have to do with which part of the DS was doing more work. Several others have also reported a preference for FLAC over WAV, if you search the forum you should find them.
I've a lot of respect for Phillip Hobbs, one of the best recording engineers around. I need a teashirt with "I like FLAC" on the front!
Thanks for the prompt reply, Murrays.
(2010-11-18 13:58)murrays Wrote: [ -> ]@advr
Are you saying that you can reliably distinguish (in a blind test) between FLAC and WAV versions of the same audio data ?
I have carried out double blind tests here and everyone has been able to distinguish between the two (I myself preferring the WAV). From an engineering viewpoint, I am unable to understand why this would be so, hence my curiosity when my brain has nothing better to chew on. Even Philbo (as mentioned in another post/thread has said there was a difference, but again there is no plausible reason as to why this is so.
(2010-11-18 11:26)murrays Wrote: [ -> ]1. If we measure the power rail that feeds the main processor in the DS we can clearly see identifiable disturbance patterns due to audio decoding and network activity. These patterns do look different for WAV and FLAC - WAV shows more clearly defined peaks due to regular network activity and processing, while FLAC shows more broadband disturbance due to increased (but more random) processor activity.
This is somewhat along the lines of what I would have expected.
(2010-11-18 11:26)murrays Wrote: [ -> ]2. If we measure the power rails that feed the audio clock and the DAC we see no evidence of any processor related disturbances. There is no measurable difference (down to a noise floor measured in micro-volts) between FLAC and WAV in any of the audio power rails.
3. Highly accurate measurements of clock jitter and audio distortion/noise also show no difference between WAV and FLAC.
This once again would indicate there ought to be no difference between one and t'other, as does.
(2010-11-18 11:26)murrays Wrote: [ -> ]The extensive filtering, multi-layered regulation, and careful circuit layout in the DS ensure that there is in excess of 60dB of attenuation across the audio band between the main digital supply, and the supplies that feed the DAC and the audio clock. Further, the audio components themselves add an additional degree of attenuation between their power supply and their output. Direct and indirect measurements confirm that there is no detectable interaction between processor load and audio performance.
Although the last is a bit suprising, although not unreasaonable if the processor still had enough overhead left. Which, once again, makes me wonder what the cause of these perceived differences is. It's no big deal as far as the enjoyment of the system goes; it's just that the puzzle is interesting to chew on.
Once again, Murrays, thanks for the response.
(2010-11-18 14:24)advr Wrote: [ -> ]YES, In a blind test I can hear differences. 4 of 5 times I prefered wav (but I didn't knew that when I was listening). All recordings were different between flac and wav.
Interesting! I did a little simulation. As in your experiment there are five trials. We consider a listener as being able to detect the difference between type A and type B material if he correctly identifies type A in at least 4 out of 5 trials. I wrote a script that randomly says type A or B in the five trials, with equal probability for A and B. This is a model for a listener who can't hear any difference. Hence, the choice is random. Next, I ran this simulation 10000 times and counted how often this virtual listener would be considered as hearing differences by the above criterion (at least 4 out 5 correct). One complication: the listener has to be considered as being sensitive not only when he has 5 hits, but also when he has 5 misses. This is because what you call A and B is arbitrary. In your experiment you didn't know what the two types of material were. What matters is that the listener consistently labels one type of material with a name.
Results: In 3746 cases the virtual listener had 4 or 5 hits, in the remaining 6254 cases he was considered as not hearing a difference. This means that in such an experiment there is a ~ 40% chance of declaring a listener sensitive when he actually isn't. From that I conclude that 5 trials give you too little data to derive a reliable statement.
Anyway, your experiment is the closest to a strong test that I heard about so far. I'll see that I find out how many trials are necessary.
(2010-11-19 19:02)tmalsburg Wrote: [ -> ]Interesting! I did a little simulation. As in your experiment there are five trials. We consider a listener as being able to detect the difference between type A and type B material if he correctly identifies type A in at least 4 out of 5 trials. I wrote a script that randomly says type A or B in the five trials, with equal probability for A and B. This is a model for a listener who can't hear any difference. Hence, the choice is random. Next, I ran this simulation 10000 times and counted how often this virtual listener would be considered as hearing differences by the above criterion (at least 4 out 5 correct). One complication: the listener has to be considered as being sensitive not only when he has 5 hits, but also when he has 5 misses. This is because what you call A and B is arbitrary. In your experiment you didn't know what the two types of material were. What matters is that the listener consistently labels one type of material with a name.
Results: In 3746 cases the virtual listener had 4 or 5 hits, in the remaining 6254 cases he was considered as not hearing a difference. This means that in such an experiment there is a ~ 40% chance of declaring a listener sensitive when he actually isn't. From that I conclude that 5 trials give you too little data to derive a reliable statement.
Anyway, your experiment is the closest to a strong test that I heard about so far. I'll see that I find out how many trials are necessary.
wow...
Do I have to sit 3746 times...??
This WAV/FLAC test was only to determine that there are differences between flac and wav. Not to indicated the a rip was a wav or a flac.
In my own surroundings I cannot hear this small differences.
Ad.
(2010-11-20 15:17)advr Wrote: [ -> ]Do I have to sit 3746 times...??
Haha! No, I guess a slightly increased number of trials should be enough, maybe 10 or 15. If you have only one miss in 10 trials you can be pretty sure that there is a true difference. The chance of that result being just a coincidence occurring while there is no true difference is only about 2%.
Titus
