PC Power supply?

[Triode]

Is this turning into an audio modding board?

The only thing I would say, is that if anyone is seriously interested in modding and building a linear psu, may want to consider taking it direct to the relavent rail(s) inside rather than just an external psu.

I'd be interested in any before and after measurements you do of the internal rails with your external psu Andrew.

[And I definately agree power supplies impact audio quality - I wouldn't have spent so much on exotic PSU capacitors if not, but that's another story...]

[Fabrice Rossi]

Patrick Dixon a écrit :
> I don't actually think the wine analogy is a very good one: tasting wine
> is something that's only done in very small quantities (for obvious
> reasons);

Au contraire (being French, I had to answer this ;-). The analogy is
perfect, because wine _drinking_ is not wine _tasting_. When you taste wine:

1) you spit
2) you wash your mouth with water and spit again

One of my friend do this for a living and she can taste more than one
hundred different wines in a couple of hours. I can taste around 30
different wines in the same amount of time without getting drunk (at
all) and without loosing my taste.

> listening to music is something that many people like to do
> for several hours at a time. My experience is that very small
> differences that you hear in an AB comparison, often translate into
> very significant differences over a longer period of time.

Absolutely the same thing for wines (remember, I'm designed by centuries
of evolution to like wines).

> If you find
> yourself wanting to listen to more music (and more varieties of music),
> then I think that's a pretty good indication that your music
> reproduction system is doing something pretty good.

And it's a pretty good indication your wine producer is doing something
pretty good...

> There are plenty of areas where scientific testing is playing catch-up
> rather than dictating the 'correct' methodology, and I happen to think
> music reproduction is one of them - because it's not just a 'simple'
> matter of electronics, acoustics and anatomy - the brain gets involved
> too.
>
> But then if you want to be sceptical, it's no skin off my nose - I just
> think if you're more open-minded you might have something to gain.

That's funny because I only complain about the lack of scientific
_observation_ not the lack of scientific _explanation_. A that's exactly
why I used the wine analogy and not a technical one. I'm not claiming
that fancy PSU will not improve the sound, I'm simply saying that if ABX
cannot prove it, then it does not.

Fabrice

[Fabrice Rossi]

Andrew L.Weekes a écrit :
> It only took one post though, better than normal

My pleasure ;-)

> The idea is presented as interesting, not as science. Why don't you try
> it yourself?

Well, the only ABX testing I can do right now is between compression
methods. I've not the specialized hardware needed to do proper ABX
comparison.

> I won't argue debate any further in the interests of others viewing,
> since these debates always end up as diametrically opposed views,
> neither side willing to change or acccept the opponents view. I have
> plenty opf statistical evidence for the audibility of PSU changes yet,
> like anyone else, can provide no *proof* since by definition
> subjectively assessed criteria cannot be proven, one can only provide
> statistical evidence.

Please, read my post again. As I said, I'm not claiming that PSU cannot
change the sound you hear, I'm saying that only ABX can prove it (in a
statistical sense).

<...>

> On the basis of a quick experiment earlier the comparison between the
> SB2's SMPS supply and an external linear, low noise, wide bandwidth
> supply is audible to *me*.

How did you test it?

> But hey, if you and others are totally happy where you are that's
> absolutely fine by me, I'm not an evangelist; I contributed to a thread
> in an area I have years of experience in, nothing more, nothing less and
> I shall contribute no more, unless asked!

Why not?

> In response to my previous post though, attached is a calibrated plot
> of the output noise of the SB2's SMPS supply, compared to a low noise
> discrete linear one. There's a sea of noise and discrete output spectra
> that is common to the non-linear behaviour of almost all SMPS PSU's.
>
> When you factor in the line rejection of the internal reg's (which
> isn't constant and degrades rapidly with frequency) and the PSRR of the
> internal IC's (again degrading rapidly with increasing frequency), the
> referred to input degradation which comes with the output op-amp if it
> has any gain, I very much doubt that the suppies as they stand get to
> full 24 bit performance levels across the entire audio band.

So it seems you have scientifical evidence, don't you? Thanks for these
information.

Fabrice

[Fabrice Rossi]

Phil Leigh a écrit :
> I was hoping (obviously in vain) that we could avoid the ABX/double
> blind bs on this forum...

Just for fun, could you explain why ABX/double blind is bullshit?

I've nothing, absolutely nothing, against _strange ways_ to improve the
sound reproduction. If it can be improve by putting a half empty crystal
glass of salty water on my loudspeakers, I'll do it. However, I'm always
puzzled by this kind of anti-scientific point of view. The scientific
method does not require scientific explanations. They come after,
sometimes. Fortunately for us, the drug industry is using a valid way of
assessing the performances of its product, while the way it produces new
drug is quite a mystery.

Fabrice

[Andrew L. Weekes]

Andy - please keep posting and please advise (via personal email if necessary) of a linear PSU design for the SB2.

I'll have more to say once I've completed the unit I'm currently building, but in response to the comments above re: the best results being obtainable inside the Squeezebox, the non-invasive method I've chosen currently can, in my view, acheive big benefits, just with a bit of lateral thinking.

The solution I'm currently working on is a two-box one, such that one contains the raw supply with the relastively noisy mains stuff in one box, followed by another box containing the high quality regulation, remote from the mains and as close as possible to the Squeezebox.

The idea being to get the regulator as physically close to the squeezebox as is humanly practical, without delving inside it. the regulator (which is a discrete circuit) then feeds the SB2 via as short a cable as is practical, little more than a couple of inches.

This keeps the loop area to a minimum and brings most benefit from the ultra-low output impedance of the regulator (which is at lest an order of magnitude below that of a three-terminal type, along with having lower noise and wider bandwidth) this would otherwise be lost, due to the cable impedance.

You could trial something similar, just using standard 3-terminal adjustable reg's, like the LM317 / LM1086 or, if you can get them, the higher current, but obsolete, LT range (LT1085 / 4 etc.).

In my case I have plenty of voltage head (due to the transformers I had around), so I have a pre-regulator in the mains PSU box, feeding the final regualtor.

I've just spent an hour listening to this, in comparison the SB2 supply is much harsher on the ears - subjectively the noise floor seems higher (in that less subtle detail is apparent in the mix) but most obvious is there is a harsh edge to the treble, that's tiring for me to listen to. The linear supply eliminates this, so much so I'm going to plug it into the main system later and see how close it gets to the CD player, 'cos it sounds rather enjoyable.

Regards,

Andy.

[Andrew L. Weekes]

I wanted to share some details about the itnernal power supply configuration in Squeebox2 which are really important to understand if you're trying to get better performance through modding.

BTW, I wanted to thank Sean for the incredible openness he's shown here relating details about the SB2's internals - it's unusually refreshing to see such stuff, given the normal secretive behaviour that goes with much of the audio world!

It should save a whole bunch of reverse engineering for those that want to delve inside - I'd strongly agree with his view that the 14V supply is likely the most critical one to address, sonically.

Thank you,

Andy.

[Andrew L. Weekes]

Just for fun, could you explain why ABX/double blind is bullshit?

I know I said I wouldn't get involved, but since you do seem genuinely open-minded (and I may have mis-interpreted your response) I'll make a comment, if I may ;-)

I wouldn't say it's bullshit myself, simply mis-applied to the job in question. The problem for me with ABX is it doesn't take into account the way we hear music, which is not through some short-term effect in the ear, but a longer term effect in the brain.

For most of us music is all about emotions; dancing, singing, happy, sad, angry, agressive, soft etc. That's the whole raison d'etre for it's existence and why musical artists do what they do. To break it down to simple measurement is like digitising a great painting and assessing the range of colours or the contrast etc Our ears know nothing about the emotional effect music has on us, that process happens in the brain.

Some of the best scientific work in this field comes from Hugo Fastl and the late Eberhard Zwicker - their publications and work are ammassed in Psycho-acoustics - Facts and Models, which isn't exactly light bedtime reading, but is fascinating nonetheless.

The ABX test forces us into trying to listen for a difference, usually in an inherently short-term fashion, by rapid switching. Thsi is not how we listen to music, which is a long term event by comparison.

The reason is that hearing takes place in the brain, not in the ears.

Hearing is related to data processing in the brain and makes use of enormous amounts of previously stored data and complicated adaptable data processing schemes. Therefore hearing fundamentally makes use of the long term memory were all those data sit.

Hearing is only partially related to the gathering of audio data in the ear and the primary storage of those data in sensory memory, which is short term.

Now in a rapid switching DBT you want to make an "objective" test by deleting the huge data processing part of the hearing equation which is related to long term memory, because the rapid switching prevents the long term memory from being activated. Take out this "subjective" part of hearing and what's left is the "objective" part related to short term memory, ain't that true?

No.

Take out the data processing part and nothing serious is left of
hearing: no dataprocessing in the brain means no audio image reaches our consciousness which means no conscious judgement can be made.

The curious thing is that in such a DBT you STILL want the test person to make a conscious *subjective* judgement (a judgement on personal experiences is by definition subjective).

So you want the test persons to give their subjective judgement on the data collecting process in the ear plus short term sensory memory storage process. But nobody can do that.

The subjective judgement canonly be made by our consiousness, and our consiousness has no knowledge of the early processes of hearing.

It's one of the reasons that level matching is deemed so critical, since that is one of the few things the early processes of hearing can determine - remove it and the tests differences often vanish.

To that end, I DO believe that double blind testing can work, but that it has to be inhernently long term - the comment Partick made earlier, about whether you are glued to the seat listening to music that engages you, or would rather be watching the TV, is the best measure by far in my view.

I think that if one allowed the listener to listen to each item for a long period (and here I'm thinking several days per item) then it's possible that a DBT, or an ABX could be made to function reliably. So I'm not anti science at all, but I do believe that the way it is often proposed is fundamentally flawed.

So why don't I try it out?

Simply because using the long term evaluation, the results become obvious; whether a particular change is better or not becomes obvious, simply owing to the amount of music I listen to - if i find myself still up at 2 in the morning when I have work the next day, it's usually a good sign

[NealG]

I count myself as one of those ‘open-minded DIY'ers’. I don’t have Andrew’s experience in PSU design but I am conversant in electronics and have a strong experimental side to audio. It was Andrew’s postings on PinkFish that made me experiment with a simple linear supply for the SB1, here the audible benefits are undoubted to my ears. I have also found that the SB2 benefits as well from a linear supply, I have no doubt that improving it would make the SB2 even better, perhaps with one of Andrew’s super regulators.

I think its possible to design a good SMPS, Linn do but it must come at a cost, something that the rather low cost off the shelf wall wart item could never emulate IMHO. From my experiments with tube amps where I have far more practical experience, the PSU is key to the overall sonic and measured performance of the amp, especially with SE design, you are listening to the PSU. I don’t see why the SB should be any different.

[cliveb]

I realise that I'm a bit of an outsider in this debate, but I'd like to make a comment. Let me begin by declaring my own personal position: I believe that ABX is a valid way of establishing whether there are any sonic differences between components, where by "sonic difference" I am referring only to the actual soundfield generated in the listening room.

Andrew makes some very valid points about ABX testing and the listening process. The fact that the listener's response is an emotional one and occurs over a long time period is certainly true. But I think he misses an important issue. The listener's emotional response is determined not only by the physical sound that is produced: other non-sonic factors influence things enormously, such as the listener's current state of health, their general mood, and the visual/tactile appeal of the equipment in use. My feeling is that, apart from transducers such as microphones/loudspeakers/etc, the current level of performance achieved by modern competently designed audio reproduction equipment is sufficiently high that these non-sonic factors have a greater influence on the listening experience than the actual soundfield.

To that end, I DO believe that double blind testing can work, but that it has to be inhernently long term - the comment Partick made earlier, about whether you are glued to the seat listening to music that engages you, or would rather be watching the TV, is the best measure by far in my view.

I think that if one allowed the listener to listen to each item for a long period (and here I'm thinking several days per item) then it's possible that a DBT, or an ABX could be made to function reliably. So I'm not anti science at all, but I do believe that the way it is often proposed is fundamentally flawed.

This method cannot work, because over the long term, the non-sonic factors will change radically. Your health will change from one day to the next; you may have had a row with your spouse/partner/children; and so on.

Where the ABX detractors make their mistake is in attributing their perception of a difference to purely sonic reasons. If the equipment sounds different, then a short-term ABX will show it. If you get a negative result in an ABX, then the differences in sound fields generated are probably beneath the threshold of audibility. But that doesn't stop you genuinely perceiving a difference over the long term, due to your emotional response to the equipment.

[NealG]

Clive, I have to disagree, the same reasons you give also apply to short term A / B listening. One's personal health and mental state will also affect the listeners perception, with short term listening the 'risk' or 'error' is greater for the listener. A 'snapshot' is not an accurate measure unless you can perform it over many samples and over a longer time period, thats why I agree with Andrew that long term listening is better, it will average out the 'errors'. This debate will never be resolved, it's been done to death many many times before, best to agree to disagree, IMHO.