Head-Fi.org › Forums › Equipment Forums › Computer Audio › New theory on signal/noise versus sound quality.
New Posts  All Forums:Forum Nav:

New theory on signal/noise versus sound quality.

post #1 of 8
Thread Starter 

Why when I bypass the power supply caps with better caps does the sound improve dramatically but S/N ratio suffer? Is S/N ratio anything that we really need to worry about or are we putting too much emphasis on this one spec & losing actuall resolution as a result?

 

In my experiments into obtaining higher perceived resolution I have found that when I achieve my goal that S/N ratio suffers some as a result. My SACD player was a perfect example as after my mods the player sounded much better than stock but actual noise when no signal was present was high enough that you could easily hear the DAC's idle tones when volume was set to a normal lestening level. The S/N ratio according to the manufacturer was suposed to be 106db which would have made these tones inaudible though in this case they did have muting transistors at the output to block the sound when there was no signal to the DAC's & i removed these transistors so this may be in part why it increased though in part I believe that the difference is that the bypass caps I used in the power supply to the analog section & the DAC allowed me to hear the true nature of the idle tones from this DAC. There was no off setting negative phase DAC output to be able to cancel these idle tones as there is in my soundcard DAC & the sound of the idle tone were what I would describe as motor boating & were definately coming from the DAC, not anywhere else as these tones were present even when all other activity was stopped.

 

On my sound card the DAC has balanced outputs & as such the idle tones are cancelled quite well & remain inaudible even after the cap bypass mods , however the S/N spec suffered here as well though interestingly even though the distortion spec went up also the distotion level & spectrum distribution remained Identical only the noise part of the spec went up by about 10db. Part of this may be the noisy computer environment but I believe that the actual reason the noise portion but not the actual distortion went up is there is an improvement in the actual resolution of the DAC at very low levels & analog section that it is able to actually resolve the nature of the noise better than before the power supply cap bypass mod.

 

In my experiments with amplifiers Electrolytic caps used to control the gain of a high gain amplifier going to ground from the negative feedback loop were notoriously lossy in terms of sound quality & when either wire or a metalized film cap was sustituted the sound was much better in every way.

 

Power supplies for transistor gear must use electrolytics in order to have sufficient storage capacity to function properly however  by themselves appear not to be a totally adaquate solution & there remains some loss of resolution at the low levels as a result & of coarse that is also where noise resides which is at the lowest levels. So what appears to be happening is that the low level noise goes up in proportion to the gain in resolution but actual distortion remains at the same level as before. At 108db S/N ratio the noise is still inaudible so I'm not worried that noise went up because the actuall distortion reamains very low as in identical to stock except for the noise componant of the distortion spec, no change in the level of the distortion spectral componants according to RMMA tests Nor is there a change in frwequency response according to RMMA tests. Interestingly the distortion spectrum that I get actually looks better than the ones that Robscix got in his tests.  My 2nd & 3rd hamonic were identical in level to his on his review of the X-fi Titanium HD card but all componants bove the third harmonic were much lower. 4th through 6th were virtually nonexistant & higher harmonics were present but at a much lower in level than Robscix's tests. Noise was 10db higher accross the board in this test. So much for the theory that my mods increase distortion that some have expressed.

 

The fact that the noise is increased across the board lends creadance to the fact that it is not so much computer generated noise as noise coming fron the DAC or ADC input circuits. If the the noise was due mostly to EMI interferance I believe the noise would be more focused in one or two areas as opposed to being across the board.


Edited by germanium - 7/21/12 at 12:46am
post #2 of 8
Quote:
Originally Posted by germanium View Post

Why when I bypass the power supply caps with better caps does the sound improve dramatically but S/N ratio suffer? Is S/N ratio anything that we really need to worry about or are we putting too much emphasis on this one spec & losing actuall resolution as a result?

I stopped reading here because I agree with the premise of this - yes, we should stop worrying about single specs and hoping they explain the entirety of the world! Seriously. biggrin.gif

As to why S/N suffered, hard to say unless you have measurements that validate it. The manufacturer's claims are usually inflated or just taken from the part-spec (so they get a chip from TI that specs to 106, and they claim the entire system is now 106, when in reality it may not be due to whatever they've added up/downstream) - you'd have to be able to measure initial and final to see if the S/N really changed and if so by how much. The rule of thumb I've always heard for audible noise in a circuit is around -40 dB or higher, and once you break down below around -80 it should be inaudible and anything else is gravy. SNR doesn't always mean DNR; just like field- and frame-rate (which people also seem to misunderstand, something which advertisers are glad to perpetuate).

Also +1 on breaking up the EFI/RFI bogeyman nonsense - if there were really as much "evil noise" inside a computer as many audiophiles believe, you'd have more issues than "slightly grainy audio" or whatever other complaints.
post #3 of 8
Quote:
Originally Posted by germanium View Post

Why when I bypass the power supply caps with better caps does the sound improve dramatically but S/N ratio suffer?

 

Most likely because of expectation bias.

post #4 of 8
Thread Starter 
Quote:
Originally Posted by stv014 View Post

 

Most likely because of expectation bias.

Not expectation bias, true resolution increase.

post #5 of 8
Thread Starter 

Moved to a different slot on mother board, now noise is lower but distortion higher, still well below audible levels though. Strangely mixer responds differently to RMAA test program when I moved the soundcard to different slot.

 

Before I moved the card higher order distortions were almost nonexistant but noise was 8db higher than after I moved the card. Distortion now coincides with the distortion spectrum that Robscix got though.

 

After running these tests I updated to latest version of RMAA test which is 6.2.5 & interestingly high frequency noise spikes mostly vanished from noise plot. Noise floor more consistant now.Harmonic distortion is still higher than Robscix's tests but not much, about 5db higher on the 3rd harmonic. noise only 1- 2db now over spec. Intermodulation plot looks way cleaner on my tests than Robscix's though. Intermodulation is a more impotant spec than harmonic distortion as it would be more audible than harmonic distortion.

 

looks as though in order to really get bonafide test results I would have to get it tested by an outide electronics firm that has the proper gear to test this.

 

I guess though that I was mistaken on my earlier theory as noise can be lowered to stock levels but harmonic ditortion is indeed slightly higher, no where near adible levels though. The sound still sound way better to me than stock.


Edited by germanium - 7/21/12 at 9:05pm
post #6 of 8
Quote:
Originally Posted by germanium View Post

Not expectation bias, true resolution increase.

 

How do you know for sure ? It is easy to imagine that modding improves sound quality, especially since the comparison usually relies on long term auditory memory (which starts to fade already after less than a second); "op amp rolling" appears to improve the sound mainly for the same reason. Did you try any of the following ?

- level matched blind test between two identical cards, one modded, one unmodded (you may also want to do the test first with both unmodded, to confirm that the sound is really identical without the mods)

- blind test between loopback recordings of the same music, with the same mixer etc. settings before and after the mods

- audio differencing (null test) similarly to above; mods that affect the ADC and/or DAC clocking can make this invalid, however

post #7 of 8
Quote:

Originally Posted by germanium View Post

 

Harmonic distortion is still higher than Robscix's tests but not much, about 5db higher on the 3rd harmonic. noise only 1- 2db now over spec. Intermodulation plot looks way cleaner on my tests than Robscix's though.

 

You need to compare against your own card without the mods, under exactly the same conditions (installed in the same slot in the same machine, same mixer settings, same sample rate and resolution, same loopback cable, and so on). Robscix's results might be different due to random manufacturing variations, or any differences between the test setups.

post #8 of 8
Thread Starter 
Quote:
Originally Posted by stv014 View Post

 

You need to compare against your own card without the mods, under exactly the same conditions (installed in the same slot in the same machine, same mixer settings, same sample rate and resolution, same loopback cable, and so on). Robscix's results might be different due to random manufacturing variations, or any differences between the test setups.

 This is true though they are within range to be attributed to manufacturing variation anyway so I suppose it really doesnt matter. The mods sound great & the stock card to me didn't sound great. Good but not great.

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Computer Audio
Head-Fi.org › Forums › Equipment Forums › Computer Audio › New theory on signal/noise versus sound quality.