[sdiy] Taking a Step towards Digital Synthesis?....

[Jay Schwichtenberg]

Colin,

I'll but some comments inline.

First though I'll state my arrogant opinions (don't laugh, they really are)
on digital audio and do some ranting. Most converters and the support for
them (power supply and analog) are designed for the masses. Remember up to a
point saving that dime here and there is what matters and not the quality of
audio. If you look at the trend for audio it's going to compressed formats
and to soft distribution. Most people that buy music are in the 14-35 year
range and are probably listening on a portable device, DVD or computer. The
quality of most audio on these devices is adequate. Even taking a good CD
player and putting your own or higher end converters on it will probably
improve the sound. In my opinion 24 bit 48 kHz either raw PCM or with loss
less compression would be the ultimate.

The reality is that most of the 24 bit/96 kHz stuff is marketing garbage.
The limit of converters making real audio is about 20-21 bits. After that
those 3-4 bits are down in the noise and aren't really there. But there is a
major difference in sound quality between 16 bit and 24 bit converters. The
difference between 44.1 kHz and 96 kHz is very subtle. You will need to have
a good system to notice the difference. Well you go "I'll get a 96 kHz
system to record on and then down convert it latter". Unless you can afford
a real good down converter (either in hardware or software) then don't down
convert. Most of the ones built into standard recording apps (Cakewalk,
Cubase....) are not any good and you are better off recording at 44.1 kHz.
If you do have real good converters then playing at 96 and recording at 44.1
is a better solution.

Enough of this rant, I'm starting to sound like Harry on BBDs!

Frozen in in Portland.
Jay

> -----Original Message-----
> From: owner-synth-diy at dropmix.xs4all.nl
> [mailto:owner-synth-diy at dropmix.xs4all.nl]On Behalf Of Colin Hinz
> Sent: Tuesday, January 06, 2004 11:33 PM
> To: synth-diy at dropmix.xs4all.nl
> Subject: RE: [sdiy] Taking a Step towards Digital Synthesis?....
>
>
> On Tue, 6 Jan 2004, Jay Schwichtenberg wrote:
>
> > Jim,
> >
> > Here are a few suggestions on picking a DAC and working with
> digital audio.
> > Stuff I figured out after doing 4 cards. The major players in
> the field are:
> > Crystal, AKM, Burr Brown/TI and Analog Devices.
>
> How do the Philips parts compare to the "big players" out there? I ask
> as the UDA1355H sure seems like a neato part -- not just analog CODEC
> functions, but S/PDIF I/O as well. Lots of strange internal routing
> possibilities, too, for those with a microcontroller handy to fly
> the thing.

Looked at these and they aren't impressive. Typically when I here codec I go
multi-media or portable. Codecs usually don't have the quality of seperate
converters.

>
> > 1) Use multibit converters. These are more tolerent of clocking
> issues and
> > sound better.
> > 2) PC layout is super critical. Crosstalk, power supply and
> grounding are
> > all issues. Read all the app notes on this from all the venders.
> > 3) Power supply is critical. Most chips run off of 5 and 3.3 volts. Make
> > sure you have a clean and stable power supply.
> > 4) Clocking is less critical with multibit converters up to a
> point. Try to
> > get them as good as you can. Use a clock that is an exact
> multiple of your
> > sample freq or one of the digital audio clock chips.
>
> How do the codecs with integrated clock synthesis shape up? Again, I
> refer to the UDA1355H, which specifies a fixed 12.288 MHz crystal
> regardless of the analog sampling frequency. Yes, there's an internal
> PLL, but they're pretty pervasive these days.....

Depends on the clock freq, dividers, PLL is and the type of converters. I
wasn't able to find any tracking or jitter figures in the spec sheet. They
clam: "The PLL recovers the clock from the SPDIF or WSI signal and removes
jitter to produce a stable system clock (see Fig.4)." but unless they have
some digital magic in there the PLL will have to track and there will be
some jitter. With single bit converters you are 'suppose' to keep the clock
jitter below 100 pico secs. Above that supposedly you can there things. As
mentioned above multibit ones are more tolerant. It has to do with the
internal integrators used for the conversion. I haven't looked into a lot of
what's new nowdays. I was interested in a clock chip that Burr Brown makes.

>
> > 5) Use a 256X or 512X master clock. Some converters don't
> support 384X, but
> > all support 256X and most 512X. This will give you more options
> if you want
> > to change down the road.
> > 6) Do the math on the output filters yourself. Don't rely on
> the data sheets
> > and app notes. Some of them are in error.
>
> Rhetorical question: the datasheet claims that "on-chip filters take
> care of it all" are not to be trusted, then?
>

I wasn't able to tell much from the data sheet for this chip and I didn't
see any app notes.

With a typical DAC there is a brick wall digital filter in them to limit
bandwidth. Most DACs have very low current output, the signals are biased at
about V+/2, some are have differential and some single ended outputs. So you
add a analog filter to get rid of 'out of band' noise and to act as a
buffer. Check out Crystals AN48. I would use 2 dual op-amps on a channel to
get to the output stage. Even though most new op-amps have a high common
mode rejection I didn't use same packages for different channels to avoid
crosstalk. First chip have filter and gain, second chip would be a
differentual driver. Real anal designers will use different power supplies
on each channel too.

> > 7) Some AKM converters have a nasty thump on power up when internal caps
> > charge. This requires special circuitry if you want to get rid of it.
> > 8) If you can try to put a SPDIF interface in. More and more
> stuff is going
> > this way. There are chips from Crystal and AKM that you should
> be able to
> > put in parallel with the DAC. If you want to go hard core you can put in
> > word or external sync also.
>
> S/PDIF is almost free....may as well have it if you can get it :=)
>

Here we get back to the comsumer vs high end/prosumer audio. I don't know of
any high end/prosumer converters that have SPDIF built in. They might be out
there though. Technology keeps changing and I don't keep up with this field
as much as I did. For high end audio you'll probably still have to use a
SPDIF chip or do it in a FPGA.

If you are just hooking two SPDIF device together then it is pretty simple.
But if you are designing a system around digital audio then it gets
complicated. All those crystals and clocking systems for each converter
system don't really match up. For example if we have a +/-5% tolerance on
crystals and 2 cd players of the same model with one crystal at -5% and the
other at +5%, 10% delta between the two clocking systems. The playing time
of same 60 minute CD will be 6 minutes different. So unless you have some
way of syncing them all together and running off the same clock you get some
occasional clicks and pops. In audio land that is typically done with what
they call word clock. That's a reference clock at the sample rate which is
converted to a master clock via a PLL and divided down for all the other
clocks at each 'converter system'. A converter system maybe a stereo
converter or something like a multi-channel ADAT or firewire box. Less
common is a master clock of some frequency that can be divided down. In TV
land they usually run a single sync signal throughout the facility called
black burst or black sync that they can derive a clock signal from. Some
might ask why do this. Why not just run it though DACs & ADCs all the time.
That depends on who many times you are going to convert things from digital
to audio and the quality of the system. After a few conversions you will
start to notice 'artifacts' from the conversion process and aliasing. You
typically hear something not right, tinnyness (sp?) or high freq whine. If
you are putting together a digital audio studio you really need to think
about how you work to get the most out of it. Do you use analog or digital
mixing, plug-ins vs effect boxes, converters, digital/analog routing.....


> BTW, I'm coming at this pretty new, relatively speaking - translated:
> whereas I know audio a thousand times better than the average PC gamer
> lamer, I bow to the more experienced minds on the list. I hope to learn
> much more than I already know.
>
> - Colin Hinz
>   Toronto, Canada
>
>
>
>