PDA

View Full Version : Digital Output



Mike Reeve
2005-03-03, 09:21
Hi

Regarding problems with lock on external DACS ...

About a year ago it was found that the Squeezebox digital output was
somewhat off 44.10KHz BUT ...
.... Sean quickly [as usual!] fixed this in the firmware.

Fortunately I have always routed Squeezeboxes via a
device that buffers the input in a FIFO
and then clocks it out using its own high-precision, lower jitter clock,
so any such inacurracy has never been a problem for me.

However, out of curiousity I yesterday connected a Squeezebox
to an Apogee Big Ben and it registers the Squeezebox at 44.13KHz :-O

[Another long-term issue has been the long settling time of the clock
when switching into 'PCM mode' so I left the Squeezebox on repeat
all night so that this was not an issue ...]

So, I'm wondering why the 44.13KHz:
(a) Has the firmware changed?; (b) Is this 'manufacturing variance'? :-O;
(c) ...

Clearly 44.13KHz is way out of the range of some 'high-end' DACs
(e.g. AudioSynthesis DACs have a capture window of only +/- 100ppm
- approx. 4Hz),
so this could be the cause of some peoples' lock problems.

I know other people who read (or at least read) this list
(such as Michel, Harry & Victor from the TacTHackers group)
have both Squeezeboxes and Big Bens
- what do you guys (or anyone else having the two) see as frequency
these days ...?

Regards

Mike

seanadams
2005-03-03, 10:54
Mike,

This clock imprecision is indeed a known and documented issue.
Squeezebox1 uses a "black box" DSP processor from a third-party
supplier. It is a great mp3 decoder and has a lot of good features, but
it also has a couple of irritating limitations like this that power
users might encounter. While this off-the-shelf DSP was immensely
helpful in getting us to market, unfortunately we have limited access
to its internal workings, and next-to-zero support from the supplier,
so this s/pdif issue has remained open for a long time.

Ideally what I'd like to have is separate, dedicated, high-precision
crystal clocks to generate the DAC's clocks and the s/pdif carriers
directly (no PLL), plus full software control of the s/pdif data and
framing. This would yield VERY precise s/pdif clocks, eliminate the
possibility of certain nearly-impossible-to-fix bugs, and ensure that
we can reliably deliver bit-perfect streams for DTS and such.

Sean


On Mar 3, 2005, at 8:21 AM, Mike Reeve wrote:

>
> Hi
>
> Regarding problems with lock on external DACS ...
>
> About a year ago it was found that the Squeezebox digital output was
> somewhat off 44.10KHz BUT ...
> ... Sean quickly [as usual!] fixed this in the firmware.
>
> Fortunately I have always routed Squeezeboxes via a
> device that buffers the input in a FIFO
> and then clocks it out using its own high-precision, lower jitter
> clock,
> so any such inacurracy has never been a problem for me.
>
> However, out of curiousity I yesterday connected a Squeezebox
> to an Apogee Big Ben and it registers the Squeezebox at 44.13KHz :-O
>
> [Another long-term issue has been the long settling time of the clock
> when switching into 'PCM mode' so I left the Squeezebox on repeat
> all night so that this was not an issue ...]
>
> So, I'm wondering why the 44.13KHz:
> (a) Has the firmware changed?; (b) Is this 'manufacturing variance'?
> :-O;
> (c) ...
>
> Clearly 44.13KHz is way out of the range of some 'high-end' DACs
> (e.g. AudioSynthesis DACs have a capture window of only +/- 100ppm
> - approx. 4Hz),
> so this could be the cause of some peoples' lock problems.
>
> I know other people who read (or at least read) this list
> (such as Michel, Harry & Victor from the TacTHackers group)
> have both Squeezeboxes and Big Bens
> - what do you guys (or anyone else having the two) see as frequency
> these days ...?
>
> Regards
>
> Mike
>
>
>
>

Marc Sherman
2005-03-03, 11:22
Sean Adams wrote:
>
> Mike,
>
> This clock imprecision is indeed a known and documented issue.
> Squeezebox1 uses a "black box" DSP processor from a third-party
> supplier. It is a great mp3 decoder and has a lot of good features, but
> it also has a couple of irritating limitations like this that power
> users might encounter. While this off-the-shelf DSP was immensely
> helpful in getting us to market, unfortunately we have limited access to
> its internal workings, and next-to-zero support from the supplier, so
> this s/pdif issue has remained open for a long time.

Oooh, there's quite an interesting sideband in that message... :)

- Marc

Michel Fombellida
2005-03-03, 11:23
Hi Sean,

FYI, a tweaker I know (in France) has developped a word clock input for the SB.
He has a working prototype and told me that he will soon be able to start a
small serie production. I hope to get one soon.

PS: Mike, I haven't checked the frequency with my BigBen recently.

Michel

Triode
2005-03-03, 11:42
Mike,

I tinkered with a patch which attempted to add fine grain adjustment to the squeezebox frequency, but didn't have the test gear to
prove whether it was any good. [And the one person who tested suggested it didn't work, but the frequency missmatch you are
recording is much larger than I expected!]

If you are interested in following this up give me a mail off list.

Adrian
----- Original Message -----
From: "Mike Reeve" <reeve_mike (AT) hotmail (DOT) com>
To: <discuss (AT) lists (DOT) slimdevices.com>
Sent: Thursday, March 03, 2005 4:21 PM
Subject: [slim] Digital Output


>
> Hi
>
> Regarding problems with lock on external DACS ...
>
> About a year ago it was found that the Squeezebox digital output was
> somewhat off 44.10KHz BUT ...
> .. Sean quickly [as usual!] fixed this in the firmware.
>
> Fortunately I have always routed Squeezeboxes via a
> device that buffers the input in a FIFO
> and then clocks it out using its own high-precision, lower jitter clock,
> so any such inacurracy has never been a problem for me.
>
> However, out of curiousity I yesterday connected a Squeezebox
> to an Apogee Big Ben and it registers the Squeezebox at 44.13KHz :-O
>
> [Another long-term issue has been the long settling time of the clock
> when switching into 'PCM mode' so I left the Squeezebox on repeat
> all night so that this was not an issue ...]
>
> So, I'm wondering why the 44.13KHz:
> (a) Has the firmware changed?; (b) Is this 'manufacturing variance'? :-O;
> (c) ...
>
> Clearly 44.13KHz is way out of the range of some 'high-end' DACs
> (e.g. AudioSynthesis DACs have a capture window of only +/- 100ppm
> - approx. 4Hz),
> so this could be the cause of some peoples' lock problems.
>
> I know other people who read (or at least read) this list
> (such as Michel, Harry & Victor from the TacTHackers group)
> have both Squeezeboxes and Big Bens
> - what do you guys (or anyone else having the two) see as frequency
> these days ...?
>
> Regards
>
> Mike
>
>
>
>

seanadams
2005-03-03, 12:06
> FYI, a tweaker I know (in France) has developped a word clock input
> for the SB.
> He has a working prototype and told me that he will soon be able to
> start a
> small serie production. I hope to get one soon.

I suppose this is possible if you can configure the MAS3539 to be a
slave for the i2s clocks, but I never tried it so I can't say how well
it would work. You'd need to also supply master clock and bit clock
though, and I don't know if this is possible except in PCM mode.

BTW the 74AC125 is used to reverse the direction of the of the bit
clock between the ip2k and the DSP. When playing mp3, the ip2k is the
master and the interface is simply 2-wire SPI (no word clock). When
playing PCM, the DSP is the clock master and the format is i2s.

I'd be happy to take a (quick) look at his schematic if it's of any
help.

Mike Reeve
2005-03-03, 12:14
Triode <triode1@...> writes:

> I tinkered with a patch which attempted to add fine grain adjustment
> to the squeezebox frequency, but didn't have the test gear
> to prove whether it was any good. [And the one person who
> tested suggested it didn't work, but the frequency missmatch you are
> recording is much larger than I expected!]
>
> If you are interested in following this up give me a mail off list.

Adrian

As I mentioned in my initial post, I buffer the Squeezebox via a FIFO
(whose output clock is a tightly controlled 44.1KHz),
so the inaccuracy is not an issue for me.

However, some of those who have been raising
the external DAC lock issue recently
may well be very interested in your patch ...

On the other hand, in the spirit community, I could give it try,
given that I see its effect (albeit with limited precision)
on the Big Ben - drop me a line off list.

[Does the patch work with 5.4.1 - I haven't upgraded to 6 yet
as I am travelling a lot at the moment and don't want to leave
the rest of the household with a 'bleeding edge' SlimServer.]

Regards

Mike

Michel Fombellida
2005-03-03, 13:00
Hi Sean,

He indeed had to find a way to reconfigure the DSP. Perhaps it only works in
PCM mode (I don't know) but of course this is essentially a tweak for us the
PCM lunatics :-) I'll tell him that you could check his schematic.

Michel

Mike Reeve
2005-03-03, 19:15
Mike Reeve <reeve_mike@...> writes:

>
> Triode <triode1 <at> ...> writes:
>
> > I tinkered with a patch which attempted to add fine grain adjustment
> > to the squeezebox frequency

For those interested in the results of my testing Adrian's patch ...

= GOOD news! And, also, BAD news! :-O
=
= The GOOD news: 20134 is the magic number :-)
=
= 100% = 44.10KHz
=
= 99 = .10
= 98 = .10
= 97 = .11
= 96 = .11
= 95 = .11
= 94 = .11
= 93 = .11
= 92 = .12
=
= 101 = .10
= 102 = .09
= 103 = .09
= 104 = .09
= 105 = .09
= 106 = .09
= 107 = .08
=
= The BAD news: Entering 'variable pitch mode' appears to be setting
= the "Fs" bits in the Channel Status Block to 48KHz :-(
=
= The unit that I usually have downstream of the Squeezebox
= implements these properly so it re-clocks its output to 48KHz.
=
= Some DACs may also 'object' to this erroneous sample-rate indication
= (although my guess is that very few consumer electronics companies
= bother too much with these details ...).
=
= However this begs two questions:
= (1) Why/where are the "Fs" bits being changed?
= (2) When the frequency errors were raised about a year ago,
= Sean tweaked the PLL to get it right and measured 44.10KHz.
= However, I am definitely seeing 44.13,

[Ken: FYI, as I noted in my original post,
this is after leaving it playing all night.]

= so is there really such a variance between units
= that it needs customised tuning?
= And if so, why?

Mike

Mike Reeve
2005-03-03, 22:01
Hi

Some interesting observations ... [? :-O]

Those who recall the history of the digital-out-channel-swap bug
will remember that about a year or so ago I spent several hours
doing (amongst other things) successive resets of a Squeezebox
to try to get a handle on the bug ... :-O

[BTW for those who were not on the list at the time,
Sean's conclusion was that the channel-swap bug is unavoidable
(with the current DSP)
- even if, like me, you only play PCM files!
The DSP powers up in 'MP3 mode' so it will always have to
switch (if only once) from 'MP3 mode' to 'PCM mode',
and whether the channels swap or not on that switch
is complete chance (although things do, fortunately,
seem to be biased towards not swapping ... :-)]

I just did the same thing (successive resets) again this evening
but focusing this time on the frequency of the digital out
(as measured, after a reasonable settling time, with an Apogee Big Ben):
most times the Squeezebox I was using would
settle at 44.13KHz but occasionally it would actually be at 44.10KHz.

So is the frequency, like the channel-swap bug, a matter of chance
(within certain limits) ...?

[As an aside I think I noticed, but I was not keeping detailed records of it,
that most times it settled at 44.10 the channels were swapped ... :-O]

I guess all I have done is confirm what we all knew ...
.... i.e. that the current DSP is not at its best in 'PCM mode'.

Mike

Mike Reeve
2005-03-04, 10:23
Mike Reeve <reeve_mike@...> writes:

> = The BAD news: Entering 'variable pitch mode' appears to be setting
> = the "Fs" bits in the Channel Status Block to 48KHz

BTW I am only assuming that this is the case from the behaviour of
the downstream box - I could be wrong ...
.... if anyone has an SPDIF analyzer could they take a look.

Mike