bear said:
There was an AES paper about Advanced Segment DAC architecture
when the PCM1738 came out. I'll try and dig it up.
Meanwhile there is some info here about the architecture and
in general a really interesting paper about DACs.
http://www.iet.ntnu.no/~ivarlo/files/School/PhD/Report_audiodac.pdf
cheers
Terry
Terry,
thanks for the link!
... am downloading it now... dial-up is slooooooowwwww...
Andrej,
I thought I downloaded the datasheet and didn't see much about the "theory of operation" on the one I got - they just said something about their "bla bla bla Advanced Segment...". I figured there must be another paper somewhere on their site that explained it.
The delta-sigma thing gives me pause.
Dunno if there is any reason for it, but all the delta sigma based DACs that I have heard have "something funny" about the sound when A/B'd against the "ladder dacs"... I dunno. Maybe by going back to a current output the issue is avoided - assuming the internal opamps in most of the delta-sigma dacs is the source of the problem I've noticed??
Ah well, guess I have to read more to try to understand it...
_-_-bear
thanks for the link!
... am downloading it now... dial-up is slooooooowwwww...
Andrej,
I thought I downloaded the datasheet and didn't see much about the "theory of operation" on the one I got - they just said something about their "bla bla bla Advanced Segment...". I figured there must be another paper somewhere on their site that explained it.
The delta-sigma thing gives me pause.
Dunno if there is any reason for it, but all the delta sigma based DACs that I have heard have "something funny" about the sound when A/B'd against the "ladder dacs"... I dunno. Maybe by going back to a current output the issue is avoided - assuming the internal opamps in most of the delta-sigma dacs is the source of the problem I've noticed??
Ah well, guess I have to read more to try to understand it...
_-_-bear
Two basic technical problems with delta-sigma modulators are noise and limited resolution. They work esentially on the same principle as pulse-width modulators: intermediate levels are achieved by operating the switch at higher frequency (64x for PCM1794), and keeping it open n cycles out of 64 to give n/64 output level when integrated. Since it switches the full reference level all the time, there is a significant amount of noise in the output. That is shifted away out of the audio band by the real delta-sigma algorithm (which is a bit trickier - it uses a digital feedback loop to shape modulation noise). If you look at the spectral noise density plots above 20kHz, you'll see the noise floor raising - by as much as 30dB for PCM1794. Also, to achieve high resolution delta-sigma modulators rely on long integration. If you are integrating output over just 1 sample - you have resolution of 1/64 - roughly 6 bits worth.
As a side effect of switching strategy, they are also more susceptible to jitter and transient distortions in I/V analog stage. They are generally only usefull for low frequency applications. In RF and telecomm, ladder DACs are the norm.
On the positive side, delta-sigma modulators offer better linearity and are less expensive to make. The later does not apply to pricing of PCM1794, somehow.
As a side effect of switching strategy, they are also more susceptible to jitter and transient distortions in I/V analog stage. They are generally only usefull for low frequency applications. In RF and telecomm, ladder DACs are the norm.
On the positive side, delta-sigma modulators offer better linearity and are less expensive to make. The later does not apply to pricing of PCM1794, somehow.
bear said:
It measures a lot better. 1/2 distortion at 0dBFS and 3 x dynamic range.
It appears 1704 is the limit of R2R linearity in a chip and BB must
have felt they needed to come up with an answer to competitors
dacs such as AD1853 etc.
My gut feeling is the 1704 is easier to get VG results with, lower
OP current, very low OOB RF noise into I-V, less susceptibility to
jitter.
Maybe the 1794 is a jewel waiting to be uncovered but only with
very careful implementation???
Whilst you are in dac research mode it's worth a look at the
Lavry DA 924, 24 bit / 96k fully discrete ladder dac.
The manual is a very interesting read.
http://www.lavryengineering.com/white_papers/DA924m.pdf
Dan's a very smart guy.
cheers
Terry
There are many reasons to use PCM1794. First, it can be mono or dual channel, whereas 1704 is only mono. From a practicality standpoint, not everyone can afford to have dual mono operation. Second, PCM1704 is obsolete, while PCM1794 is currently in production. Third, PCM1794 accepts I2S and is very easy to connect to modern receivers and ASRC's, whereas PCM1704 needs some external logic. Obviously, these are all technical issues, that may or may not affect sound, but these are important to take into consideration, especially if you are DIYing, and FIYing (F=Funding). Also, if cost is a consideration, you may have to compromise on your implementation of 1704, whereas with 1794 you may be able to have a better design at the same price point.
ezkcdude said:
What's your source for your claim the PCM1704 is obsolete? ‘Not recommended for new designs’ is not the same as obsolete. In fact, TI is in the process of converting the PCM1704 process to lead-free and new parts are scheduled for production next month. If the part was truly obsolete, why would TI bother making it lead-free?
Since when are DIYers concerned about avoiding ‘obsolete’ parts? The TDA154x DACs have been obsolete for years but they still enjoy cult status amongst the DIYers here.
. I didn't know an ROHS version was coming out.ezkcdude said:
That’s true only if one is afraid of digital logic, must have an I2S interface to the DAC chip, and accepts the damage done by ASRC, upsampling, and sigma-delta modulators. But this is DIY and the usual goal is performance beyond the datasheet implementation. The addition of a single ‘1G04 can’t be the end of the world because that’s all you need to interface most ‘modern’, as you call them, receivers and ASRCs to the PCM1704.
jbokelman said:
Actually, you need a shift register. If you just invert the wordclock on the other channel, it will be out of phase by half a sample.
BTW, speaking of pushing the envelope, I'm tempted to build an audio DAC using AD9726. That part is probably two orders of magnitude better in transient response and jitter characteristics than anything you'll find commonly talked about here, and will wipe the floor with TDA1541 and such. If one implements a LSB delta-sigma modulator to drive it, one can probably do significantly better than PCM1794...
.
This URL is death in the meantime. Here the currently URL:
http://www.iet.ntnu.no/courses/tfe08/slides06/pres_audiodac.pdf
This I want also to know exactly. My estimate is, that the requirement for the following I/U converter isn't too large by the use of PCM1794 (similar like AD1955).
BTW - what DAC IC types are currently clearly better than this types?
Last edited:
I listened to (and extensively modified) an AD1955 DAC for over a year. More recently I've started listening to a TDA1541A based DAC. In general I much prefer the sound of the TDA except in the area of soundstage depth. The AD1955 wins out on depth but loses on everything else. In particular the TDA is more dynamic for want of a better word, there's more (cliché coming) foot tapping energy in the music. I'm not going back to the AD1955, rather going to see if I can add depth to multibit.
Thank you for this comment. I have heard the AD1955 against a DIY NOS-DAC equipped with TDA1543 without reclocking unit 2 weeks ago. Here it is so, that in cases of music material, where by the cd manufacturing the high frequency aera is too loud recorded, the NOS DAC is clearly the better choice. If the quality of recording in the upper aera is good (like that one from OPUS records), the stronger resolution of the AD1955 isn't a disadvantage and I prefer now the sound of this DAC IC.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.