Post by Javier on Jun 7, 2013 13:39:09 GMT
I was playing with my PK DAC and made a very weird discovery for which I have no explanation.
After having recently recapped the DAC replacing the original 1uF film DAC IC's decoupling caps with HQ X7R bypassed with another 100nF X7R, output analogue filter film Wima's with same value NP0/C0Gs (4,700pF)and the XO's original 47uF lytic with a combination of 10nF X7R plus 10uF lytic, I had been measuring the performance of the DAC using RMAA (Right Mark Audio Analyzer) and was quite disappointed with the figures I was getting for both SNR (Signal to Noise Ratio) and DNR (Dynamic Range A-Weighted). On a bad day I was getting 89dB SNR and on a good one it went to 94dB top. This is WAY below the specc'ed performance of the ES9023 chip of 112dB. On the other hand distortion was very good matching the data sheet numbers and even slightly bettering them.
A few days ago I wanted to do some testing of Signalyst's HQPlayer so I booted my PC from another HDD were I a have a fresh W7 installaion I ocasionally use for this kind of tests. I have a disk image backup that has just the very basic but all Windows updates, patches and fixes are included but no apps whatsoever, it allows me to format this spare HDD, restore the image to it and start a new in a matter of minutes. Thin is, when I got bored of playing with the said media player I fired up RMAA again and was impressed, not to say shcocked, with the results. SNR jumped to a staggering 110dB!!!! that is a whooping 17 to 21dB increase compared to what I was getting when booting from my day to day W7 HDD.
Since I found this, I have done these tests over and over again in different days and at different times. On each occasion I'd boot from one drive measure, restart from the other and measure again only to find there is a solid consistency in the outcome. The "good" Windows always has at least 15dB better numbers than the "bad" one.
The puzzling thing is in both cases hardware is EXACTLY the same. Both disks, the one conatining the "good" Windows and the one with the "bad" one were always connected at the same time which rules either drive as the cause of the noise differences. RMAA was the same as well so the only difference is the OS and installed apps. I went as far as shutting down all not strictly needed services and closing resident programs so the OS was as similar as possible with identical services running and no apps except RMAA, still same result.
How come one Windows has a higher noise floor than the other? It can't be due to electrical noise as I stated before hardware is identical thus it'd afect both the same. Both OSs are updated to the latest released fix and have the same services and apps running with the exception of the presence of a running anti virus on the "bad" Windows which shouldn't affect audio in any way, at least not increasing the noise floor as it is doing nothing when running RMAA (0% CPU usage).
EDIT:
Just made a quick test to rule out Anti virus or network apps as cause of noise difference. I have run RMAA 6 consecutive times, odd tests (1,3 and 5) have Anti Virus running, for even (2,4 and 6) tests Anti Virus was stopped:
From the above pic it seems clear that Anti Virus makes no significant difference to the resulting numbers the small variations being way to small and inconsistent, for example if we compar 1 Vs 2, 3 Vs 4 and 5 Vs 6 SNR is a tad better when AV is running but cross talk is a bit worse.
After having recently recapped the DAC replacing the original 1uF film DAC IC's decoupling caps with HQ X7R bypassed with another 100nF X7R, output analogue filter film Wima's with same value NP0/C0Gs (4,700pF)and the XO's original 47uF lytic with a combination of 10nF X7R plus 10uF lytic, I had been measuring the performance of the DAC using RMAA (Right Mark Audio Analyzer) and was quite disappointed with the figures I was getting for both SNR (Signal to Noise Ratio) and DNR (Dynamic Range A-Weighted). On a bad day I was getting 89dB SNR and on a good one it went to 94dB top. This is WAY below the specc'ed performance of the ES9023 chip of 112dB. On the other hand distortion was very good matching the data sheet numbers and even slightly bettering them.
A few days ago I wanted to do some testing of Signalyst's HQPlayer so I booted my PC from another HDD were I a have a fresh W7 installaion I ocasionally use for this kind of tests. I have a disk image backup that has just the very basic but all Windows updates, patches and fixes are included but no apps whatsoever, it allows me to format this spare HDD, restore the image to it and start a new in a matter of minutes. Thin is, when I got bored of playing with the said media player I fired up RMAA again and was impressed, not to say shcocked, with the results. SNR jumped to a staggering 110dB!!!! that is a whooping 17 to 21dB increase compared to what I was getting when booting from my day to day W7 HDD.
Since I found this, I have done these tests over and over again in different days and at different times. On each occasion I'd boot from one drive measure, restart from the other and measure again only to find there is a solid consistency in the outcome. The "good" Windows always has at least 15dB better numbers than the "bad" one.
The puzzling thing is in both cases hardware is EXACTLY the same. Both disks, the one conatining the "good" Windows and the one with the "bad" one were always connected at the same time which rules either drive as the cause of the noise differences. RMAA was the same as well so the only difference is the OS and installed apps. I went as far as shutting down all not strictly needed services and closing resident programs so the OS was as similar as possible with identical services running and no apps except RMAA, still same result.
How come one Windows has a higher noise floor than the other? It can't be due to electrical noise as I stated before hardware is identical thus it'd afect both the same. Both OSs are updated to the latest released fix and have the same services and apps running with the exception of the presence of a running anti virus on the "bad" Windows which shouldn't affect audio in any way, at least not increasing the noise floor as it is doing nothing when running RMAA (0% CPU usage).
EDIT:
Just made a quick test to rule out Anti virus or network apps as cause of noise difference. I have run RMAA 6 consecutive times, odd tests (1,3 and 5) have Anti Virus running, for even (2,4 and 6) tests Anti Virus was stopped:
From the above pic it seems clear that Anti Virus makes no significant difference to the resulting numbers the small variations being way to small and inconsistent, for example if we compar 1 Vs 2, 3 Vs 4 and 5 Vs 6 SNR is a tad better when AV is running but cross talk is a bit worse.