Squeezebox 3 compares to which CDP/DAC? - Page 3

Yep, that sounds about right to me.

Keep in mind that I think the DAC1 is a terrible DAC.

Take a listen when you get the PSU in and let us know if you hear a difference.

I too am curious about the explanation for this. I'm willing to believe it, but I don't get how it happens.

As I see it, the Squeezebox takes bits in from its server, performs computations on them, and to the external DAC it outputs resultant bits plus the timing between the bits. The bits-to-bits processing should be completely deterministic and independent of PSU quality unless the power is so terrible that the CPU literally cannot compute properly. The timing between the bits would presumably be controlled by a crystal oscillator that ought to be pretty insensitive to PSU quality also. And the DAC's input ought, one would think, to be discriminating enough to respond only to the digital information in the signal, not to the analog properties of the PSU that drove the chips that output the signal, again assuming the power isn't so terrible that it utterly corrupts the ability of the driver chips to output a proper digital signal.

Clearly I'm missing something somewhere, but I don't know what. Can someone walk me through it?

Have a go at reading this article by Tangent, it's not only a good explanation of the different types of power supplies it also briefly touches on how power supply noise might affect a given circuit (the example given is a Cmoy-type amp). The problem in extrapolating this information to the SB3 is that its circuit may be more or less tolerant of power supply "noise".

With the stock power supply, the bits are the same but the 0s aren't quite as round as they could be and the 1s are too skinny and fragile, leading to a brittleness, especially at the back of the soundstage, behind the drummer and a bit to the left. Think about it like driving a car - sure you can drive on oval wheels and you'll still get there, but driving on round wheels clearly feels better.

let's try not to threadcrap this discussion, please.

The discussion got threadjacked when the suggestion was made that a PSU upgrade would help the digital coax out: the OP asked for comments on the SB3 as a DAC.

My response was simply a tongue-in-cheek attempt at pointing out the absurdity of the PSU suggestion. The PSU has no impact on the _timing_ of the stream of bits being fed out of the SB3. If the PSU had enough impact on the _quality of signal_ from the digital coax to the point that it impacted the downstream DAC, this would manifest itself as a garbled stream, not as a subtle difference requiring listening sessions to discern.

my mistake; please see my correction.

Not saying I disagree with you but I have experienced a strange phenomenon. I work in a College with a very old, very creaky and really overtaxed electrical system. I get a strange effect when the colors on my CRT wash in and out and the screen contracts , to my untrained eye it looks like the guns on the CRT are struggling, it is worse when there are more people here using the system but it is an effect observed by many of my colleagues, and not a little disconcerting. Of course we are talking about a CRT which has a rather more substantial set of electrical requirements than a CD or DAC or whatever the SB3 is.....

Based on that article you suggested, Nate, I'm going to have to take Max F's suggestion.

If I understood and extracted the information from that article correctly, essentially, the SQ difference will come from noise transmitted through the Squeezebox's coaxial output from it's power supply vs. what doesn't come through on my reg'd PSU. Seeing as I don't have the equipment, datasheets or expertise to properly interpret that kind of data anyway, I guess I have to rely on my ears and hope for the best...

I'll post results as soon as reasonable...

If you honestly think a coax digital out is compsed of zero's and ones you need to study up on basic electronics. A coax digital signal is an analog yes analog approximate squarewave which is higly sensitive to powersupply noise especially one with a rise impedance wrt to frequency. A digital cable carries an analog signal which is converted to digital only at the DAC chip.

I'm really trying to understand this.

If the integrity of the digital signal -- carried in analog form, sure, but that's true of every digital signal in every digital electronics device, and we've had those working reliably in the hundreds of megahertz for some time now -- is so susceptible to corruption that a change in PSU makes a difference in its legibility at the receiving end, then why would the audible difference between two PSU's be even slightly subtle? This is what I'm not getting (or one of the things I'm not getting). It seems as though we'd either have fully corrected errors or else a thoroughly unsubtle noise problem. Is the nature of the bitstream encoding such that loss of fidelity rises only slowly with the number of misinterpreted bits?

Thanks for the suggestion. I'll return the favour by suggesting an introduction to basic digital signal processing.

Whoa. This changes my whole perspective. To quote the Dude: New s@#$ has come to light, man.

<blank stare>

Let's discuss for a second how this "high sensitivity" might manifest itself. Given a digital signal coming down a wire in its highly-sensitive-to-crappy-PSUs analog form, there is a level of signal that is considered to be a 0 by the receiving processor and a level that is considered to be a 1. These are not discrete levels but ranges (since as you so kindly pointed out, the signal is not discrete/digital at this point) - basically all levels up to a point will get treated as 0 and all levels past that point are treated as 1s.

Now let's consider the level of signal representing a "reference" or perfect 1. There is a large threshold available at the receiver for interpreting this level. If you are say 3% below what is considered a 1, the signal processor does not consider that a 0.97, it's still a 1. If a PSU were messing up the signal level sufficiently for 1s to be so far below level as to be mistaken for 0s, it would affect not only the mystical digital coax out but also the digital processor at the heart of the SB3. Translation for those who are having trouble following along: The SB3 would be non functional.

Now let's take the thinking a step further: Let's say the stock PSU was sufficient to mess up the signal badly enough to accomplish the above and that through some miracle the SB3 itself was still processing all of it's digital bits correctly and only the coax out was affected. What level of signal degradation would we guess this would require? Maybe a 50% loss of the peaks of the wave would do it? But let's be generous and say that even a 25% degradation were enough. Now let's flip back to the analog outs. Care to guess what that level of degradation would do to the analog signal? Hint: We wouldn't be talking in phrases like "there is slightly less air around the notes" and "the soundstage is not quite as wide".

Now let's jump back to the digital domain and assume that even given all the above, we were still going to accept that the signal is getting sufficiently messed up for the receiver to be making mistakes interpreting it. And let's say we're playing a quiet passage so were getting a sample that says 0000000000001010 which is asking for an analog signal level of 10. Our PSU now messes this up and a couple of bits end up getting flipped by the time the receiver reads it, so now lets say we've got 0100001000001010... our DAC now spits out an analog signal level of 16906 for this sample. So we've mistaken a sample asking for a nearly 0 analog level to one asking for just over 16k out of a total of 64k possible levels. Good thing the highest order bit didn't get flipped or we'd be in the FS forum looking for new cans...

Is the absurdity of the PSU having subtle yet nonetheless discernible effects on the bitstream going out over the digital coax starting to sink in yet?

Your missing the issue, it is not the DAC mistaking a 1 level for a 0 level, it is the timing that gets messed up with a dirty power supply. Remember SPDIF includes a clock signal. It is easy to prove that a crappy SMPS still gives bit perfect data. Just hook up the SB3 to a HDCD decoding DAC, the DAC still lights up HDCD meaning bit perfect transmission. The issue is the timing (clock) gives more jitter with a dirty supply. Why do you think people build dedicated power supplies for their clock upgrades?

You quoted a reply of mine in which I stated that the PSU could affect signal quality but not signal timing by condescendingly telling me to study up on basic electronics because I must be oblivious to the fact that digital bitstreams are actually passed via analog signals. I proceeded to explain to you why the PSU affecting the analog signal quality will have no effect on the bitstream, so now you are lecturing me about jitter (ie: innaccuracy in signal timing). Supposedly I am missing the issue. Allow me to address the issue then:

The SB3 uses high precision crystal oscillators. At frequencies around 40KHz these are reliable to the 10s of picoseconds, independent of the supposed influence of a noisy PSU.

No peer-accepted electrical engineering authority has ever claimed that jitter below 200ps was audible and even this is conservative by several orders of magnitude. Through mathematical analysis we can prove that jitter below 100ps can't have any effect on 16bit/44.1KHz samples within the commonly accepted physical realities of our universe. No one that I know of has ever ABXed jitter in digital audio when the jitter was on the order of single-digit nanoseconds. Most commonly acceptable thresholds for jitter detection by human observers are on the orders of 10s to 100s of nanoseconds (not counting John Atkinson of course, who can detect single picosecond jitter without even connecting the transport to anything). Note that 10s to 100s of nanoseconds of jitter is 3 to 4 orders of magnitude worse than what the SB3 is producing.

Now please explain again how a better PSU is going to improve the SB3 as a digital transport, otherwise we can move right into discussing why anyone who bought a Transporter for use purely as a digital transport has wasted $1500, unless of course one is aiming for style points.

Because for some people hi-fi is a hobby, and if you're not tweaking something, it's no longer a hobby but just another household appliance.