Exactly: the whole argument in favour of a balanced DAC was shot down... Not just in post 27 but in this earlier post:
=======
Originally posted by hagtech
I don't get it. What's the benefit of two PCM1704 per channel? Why not just take one output and convert to balanced? I would think this is better that two converters that are not matched.
You don't get any more bits of resolution. Maybe 3dB of SNR. What am I missing?
===========
There was no reasonable answer given to that question -- other than more claims (from knowledgeable people like Guido Tent) supporting his point.
So why, after this thread's entire premise was destroyed, did everyone suddenly start begging for a balanced PCB?
=======
Originally posted by hagtech
I don't get it. What's the benefit of two PCM1704 per channel? Why not just take one output and convert to balanced? I would think this is better that two converters that are not matched.
You don't get any more bits of resolution. Maybe 3dB of SNR. What am I missing?
===========
There was no reasonable answer given to that question -- other than more claims (from knowledgeable people like Guido Tent) supporting his point.
So why, after this thread's entire premise was destroyed, did everyone suddenly start begging for a balanced PCB?
Balanced
- because it is less cable-dependant, more noise-immune, - and uses twice the amount of parts
- biggest problem imho is that most converters(inverter-circuits) adds its own signature, and most simple balanced-chip-converters sounds lousy.
Arne K
- because it is less cable-dependant, more noise-immune, - and uses twice the amount of parts
- biggest problem imho is that most converters(inverter-circuits) adds its own signature, and most simple balanced-chip-converters sounds lousy.
Arne K
Re: Balanced
That is somewhat dubious for a start. The dac itself does not drive the cable and a balanced output stage does not need a balanced dac.
And the basis for this?
Cobra2 said:- because it is less cable-dependant, more noise-immune, - and uses twice the amount of parts
That is somewhat dubious for a start. The dac itself does not drive the cable and a balanced output stage does not need a balanced dac.
And the basis for this?
I'm not arguing against balanced audio systems -- I used balanced cables, and a balanced preamp in one of my systems.
But this thread is about a balanced DAC. In other words, separate DACs for the phase and antiphase signals.
It was demonstrated earlier in this thread that this approach doesn't work -- the benefits of a balanced system are better obtained by using a single-ended DAC (i.e. one or several parallel DAC chips all in phase with each other) with a phase splitter in the analog domain -- either an op amp/tube or a transformer.
The board we're talking about ordering would use two sets of 1704s in parallel -- the very approach we've discredited.
But this thread is about a balanced DAC. In other words, separate DACs for the phase and antiphase signals.
It was demonstrated earlier in this thread that this approach doesn't work -- the benefits of a balanced system are better obtained by using a single-ended DAC (i.e. one or several parallel DAC chips all in phase with each other) with a phase splitter in the analog domain -- either an op amp/tube or a transformer.
The board we're talking about ordering would use two sets of 1704s in parallel -- the very approach we've discredited.
'We' haven't discredited anything. 'We' have simply questioned whether the gains are worth the expense. You, OTOH, are free to make what you will of it all but I doubt it will make much difference one way or the other to those the feel the gains are worth the expense.
No, the question was not whether the gains are worth the expense -- for people like me, even small gains are always worth the expense.
The question was whether there are any gains at all. And the response was: No. There are no gains, according to all the responses on this thread, but rather an actual balanced DAC would provide *worse* audio than a conventional DAC with a balanced analog output.
There was no counterargument offered -- it went from "No gains" to "Can I get a board."
If you *are* able to make the case that there are gains to be obtained by using a balanced DAC, then I will immediately support one and build one.
But that case has not been made on this thread -- not once.
The question was whether there are any gains at all. And the response was: No. There are no gains, according to all the responses on this thread, but rather an actual balanced DAC would provide *worse* audio than a conventional DAC with a balanced analog output.
There was no counterargument offered -- it went from "No gains" to "Can I get a board."
If you *are* able to make the case that there are gains to be obtained by using a balanced DAC, then I will immediately support one and build one.
But that case has not been made on this thread -- not once.
Hmmm.
PCM1702 ? Co-phased in one box.
Sounds great. Sorry if this has been pointed out in a previous post. I'd say it works, but then again both internal dacs use the same R string.
-Chris
PCM1702 ? Co-phased in one box.
Sounds great. Sorry if this has been pointed out in a previous post. I'd say it works, but then again both internal dacs use the same R string.
-Chris
dougigs said:
Perhaps I am reading a different thread.
hagtech said:
Jocko Homo said:
As I said it is either worth it or not but don't let me stop you convincing yourself there are no advantages, however small.
I think you misunderstood what Jocko was saying.
The PCM1704 has an SNR of 120db -- that is, any signal it puts out is a million times larger than any noise. A 3db improvement is inaudible -- by making that point, as I read it, he was somewhat sarcastically dismissing the notion of any genuine benefit. There are a lot of easier ways to get an extra 3db.
(EDIT: Just to put some perspective on that --
At 120db, a 1V signal would meet a noise of 0.000001V or 1uV
At 123db, a 1V signal would meet a noise of 0.0000007V)
And I should be clear: I am not prima facie and a priori against the idea of a balanced DAC; quite the contrary, I have spent the past decade assuming that this was the logical next step in advancement. But this thread has consisted of good arguments against my long-held beliefs -- and they weren't followed up by any rebuttals. Not good ones, not bad ones, just none at all.
So now is the time to rebut. Why should I still believe in this topology?
The PCM1704 has an SNR of 120db -- that is, any signal it puts out is a million times larger than any noise. A 3db improvement is inaudible -- by making that point, as I read it, he was somewhat sarcastically dismissing the notion of any genuine benefit. There are a lot of easier ways to get an extra 3db.
(EDIT: Just to put some perspective on that --
At 120db, a 1V signal would meet a noise of 0.000001V or 1uV
At 123db, a 1V signal would meet a noise of 0.0000007V)
And I should be clear: I am not prima facie and a priori against the idea of a balanced DAC; quite the contrary, I have spent the past decade assuming that this was the logical next step in advancement. But this thread has consisted of good arguments against my long-held beliefs -- and they weren't followed up by any rebuttals. Not good ones, not bad ones, just none at all.
So now is the time to rebut. Why should I still believe in this topology?
dougigs said:
There is some misunderstanding here WRT DAC dynamic range and
SN.
Standard measurement of of a DAC's DR is done by measuring the
OP noise with a -60dB signal present. This noise level is then
compared to 0dBFS signal.
Generally speaking a DAC's OP noise level will modulate with
signal level. So the background noise with a -1dBFS signal
will be much higher than at -60dBFS.
The PCM1704 actually has a DR of -112dB.
I believe the SN spec is done with a signal OP level of -90dB
or below.
Cheers,
Terry
Let's back up and look at his again.
First, I'm a big proponent of balanced. I have been pushing balanced topologies for most of the 20 or so years that I have been building high-end gear.
But in this case, it doesn't make sense to do it the hard way. You are going to have to shell out an extra $36 or so for the additional '1704 and more for all the extra circuitry. Just to pick up an extra 3 dB, so that information that probably does not exist on any of our source material is not buried in the noise.
OTOH......if you switch to say...........a '1794..............you get a DAC that already is balanced, and will cost >$50 less to implement. To say nothing of the fact that it has hot outputs, so that you may actually realise that extra SNR.
OK.......at this point, the more astute of you will point out that while it is a balanced DAC, it is more akin to a DSP engine that somehow takes an R-2R ladder DAC and a sigma-delta DAC, and mysteriously combines them to give us current outputs. And does so in a non-DIY friendly SMT package.
Sorry guys.........that is the future, so get used to it. Easier to build and test, uses less supplies, SMT..........any wonder why TI is pushing them?
I see it as a choice between one '1704, and making the anti-phase with a handful of components, or going fully balanced with a '1794. The odds of finding a replacement '1794 years down the road will be much better than that for a '170x series.
Take it from a guy who designed lots of gear with AD1862s................
Jocko
First, I'm a big proponent of balanced. I have been pushing balanced topologies for most of the 20 or so years that I have been building high-end gear.
But in this case, it doesn't make sense to do it the hard way. You are going to have to shell out an extra $36 or so for the additional '1704 and more for all the extra circuitry. Just to pick up an extra 3 dB, so that information that probably does not exist on any of our source material is not buried in the noise.
OTOH......if you switch to say...........a '1794..............you get a DAC that already is balanced, and will cost >$50 less to implement. To say nothing of the fact that it has hot outputs, so that you may actually realise that extra SNR.
OK.......at this point, the more astute of you will point out that while it is a balanced DAC, it is more akin to a DSP engine that somehow takes an R-2R ladder DAC and a sigma-delta DAC, and mysteriously combines them to give us current outputs. And does so in a non-DIY friendly SMT package.
Sorry guys.........that is the future, so get used to it. Easier to build and test, uses less supplies, SMT..........any wonder why TI is pushing them?
I see it as a choice between one '1704, and making the anti-phase with a handful of components, or going fully balanced with a '1794. The odds of finding a replacement '1794 years down the road will be much better than that for a '170x series.
Take it from a guy who designed lots of gear with AD1862s................
Jocko
That did the trick for me.
PCM1794 it is for my next project... and there are actually lots of decent DIY boards for it already out there (I think...)
PCM1794 it is for my next project... and there are actually lots of decent DIY boards for it already out there (I think...)
Jocko Homo said:
Everyone and his dog churns out a dac chip with balanced outputs these days but unfortunately they come bundled with a digital filter. This leaves the PCM1704 in a unique position, if you want to roll your own filter. And if you have gone that far, you might as well grab the potential improvement on offer by a balanced topology that includes the digital domain.
Now we're getting interesting.
First off: The '1794 really needs to be implemented in what its datasheet calls "Monaural Mode." (i.e. balanced in the digital domain) -- which is where the chip becomes interesting. Are there DIY PCB designs out there that implement it this way, with 2 chips?
And RFBRW: Note the item on page 22 of the '1794 datasheet labelled "application for external digital filter interface" -- it seems to have your concerns in mind.
(Though I would like to hear a succcinct outlining of the merits of a hand-sewn digital filter -- what advantages can be gained? I'm quite interested).
First off: The '1794 really needs to be implemented in what its datasheet calls "Monaural Mode." (i.e. balanced in the digital domain) -- which is where the chip becomes interesting. Are there DIY PCB designs out there that implement it this way, with 2 chips?
And RFBRW: Note the item on page 22 of the '1794 datasheet labelled "application for external digital filter interface" -- it seems to have your concerns in mind.
(Though I would like to hear a succcinct outlining of the merits of a hand-sewn digital filter -- what advantages can be gained? I'm quite interested).
I think they had DSD in mind, and maybe they remember the HDCD filter. But I do not know of any reason why you can't use their rebadged NPC filters.
I think one of the notorous clock mongers has PCB that replaces the '5843. Maybe it is the '5842......I don't recall. That should also work.
Jocko
I think one of the notorous clock mongers has PCB that replaces the '5843. Maybe it is the '5842......I don't recall. That should also work.
Jocko
Jocko -- By 'their rebadged NPC filters,' do you mean that the 8X oversampling filter *inside* the '1794 is equivalent to a '5843... or are you saying that an *external* '5843-or-equivalent is an improvement?
DF1704........the old DF1700..........
Remember, all the B-B digital audio stuff is designed in Japan. At least my buddies who used to work there said so.
Jocko
Remember, all the B-B digital audio stuff is designed in Japan. At least my buddies who used to work there said so.
Jocko
dougigs said:
Long story short: you can make your own filter profiles. Current filters tend to emphasize flatness to 20 kHz before dropping like a stone, which creates a long transient response, to wit pre- and post- ringing (whether this is audible is a subject for argument over numerous beers). An external filter could be designed flat to 18 kHz, say, and reduce the ringing by about half.
Old farts like me could even design the filter to be flat to 16 kHz and reduce the ringing further.
Cheers,
Francois.
dougigs said:
Far from it. The delta sigma modulator is still in play.
Jocko Homo said:
One could dispense with the great god 'computational efficiency' and optimise the filter for individual sample rates.
That would be the SM5847.
Jocko Homo said:
That would make sense. The patent for the Sign Magnitude architecture used in the PCM63 and PCM1702/4 is assigned to BB Japan.
- Status
- Not open for further replies.