Welcome! Need support, you got it. Or share you ideas and experiences.
Thanks for the good advice, this looks like an promising DAC project,
http://users.verat.net/~pedjarogic/audio/index.html has anyone attempted this design.
Pedja has some pretty interesting stuff available. Note that I implemented the "LPF with Peak" solution to the Non-Os HF rolloff in my original Adagio and ported it to the Valve stage for the TDA1545.
I do however find the parallel RLC resonance circuit to sound better, for whatever reasons.
His TDA 1541 Design seems competent enough.
I plan to try this Non-Os DAC board/module... It uses an AD (forgot number) receiver and the AD1856N-K dac chip (digital filter bypassed?). The I/V outputstage is pushpull using discrete transistors. Maybe an "upgrade" would be a passive I/V converter using Sowter 9762's... There seems to be a problem with their server right now but if it does come back up I hope you may check out the schemas and see if the they are competent designs. Thanksabunch.
> I plan to try this Non-Os DAC board/module...
> I hope you may check out the schemas and see
> if the they are competent designs.
I am quite familiar with their design. In terms of basic technicalities it seems okay, but I note a few fundamental issues:
1) The DAC is not "Non-Oversampling" in the traditional sense, the Input receiver uses an Asyncronous Sample Rate converter to upsample/oversample the signal to 96KHz.
2) The specific implementaion of the Receiver/ASRC used seems highly suspect, as there is NO PLL on the receiver, any jitter attenuation relies on the ASRC digital PLL, thus Bit Accurate receiving is not possible and any jitter outside the digital PLL's rejection range will be encoded into the data, iremovable.
In sort, the Reciever is the kind of part an engineer obsessed with measurements would use but is less likely to give good sound than a competently implemented traditional receiver.
> could you please comment one the TDA1545 DAC from diyparadise.
In general fine, but Yeo made a hash with his "improvements" to the DAC Current scaling (BTW, I have always advocated to operate the TDA1545 at 2mA Full Scale Current, as far back as the late 90's).
As long as you operate the DAC with nominal Supply voltages you need to be very carefull how you arrange the passive I/V, in fact, the TDA1545 is non too well suited to passive I/V.
The DAC on +V supplies is strictly limited to a 3V Peak-Peak output Swing centered around a 3.5V Midpoint. In other words what you get out is 1V RMS.
If you bump up the current by connecting 11K from the 2/3 +V node (NEVER EVER use 1/2 +V) to 2mA Peak-Peak your I/V conversion resistor must be below 1K5, simply to not exceed the output voltage compliance and also be connected to 2/3 +V.
You can of course operate the DAC at overvoltage, in which case each Volt overvoltage increases the Output Range by 1V Peak-Peak, the midpoint is always at 2V + ((Vdd - 3V)/2).
But of course, if you go above 5V you live dangerously and on your own risk.
Other than this BIG caveat things should be peachy.
I probably would still recommend my Valve Output Stage to be used after the TDA1545, making sure to get 11K from 2/3 +V as reference and the adjust the nominal 375R cathode resistor untill the output voltage on the DAC is as close to 3.5V as possible.
many thanks for your comments!
Any "glimmer" if Yeo's DAC will sound better than Scott Nixon's TDA1543 based one?
> Any "glimmer" if Yeo's DAC will sound better than
> Scott Nixon's TDA1543 based one?
Reasonably well implemented the TDA1545 will sound a little smoother and relaxed, but I am not sure how well the kit does.
BTW, with diodes from my "junk box" I measure around 0.65V at 50mA current, this suggests that 9 diodes are best suited, giving 5.85V +Vdd.
If we tap off the nominal 2/3 Vdd at the 5th diode (NOT the 6th) we get 3.25V as common voltage. We can swing 1.25V towards the ground and 1.6V towards +Vdd, pretty much as good as it gets.
Now we can swing 1.25V max and the worst case current is 2.2mA. Thus we would need around 1K2 I/V conversion resistance. It seems advisable to make the Reference resistor adjustable (at least intially) and to adjust it with a 'scope and a 0dbfs sinewave to give just no clipping with a 1K2 I/V resistor, a 22k multiturn trimmer should do.
One might consider experimenting with the exact settings for the I/V resistor a little more and seek the best compromise between dynamic range and clipping, but I suspect that 1K5 will be the absolute limit.
One would really have to build a DAC and test it to get this down to a "T".
Thanks for pointing out the ASRC/Receiver issue. The cost is attractive though (not that high for what you get)... I will see if it would sound "decent enuff" compared to a friend's full diy 16 paralleled tda1543's - battery powered dac that sounded really superb IMHO. It has a small fan to cool those dac towers.
> Thanks for pointing out the ASRC/Receiver issue.
Note, I have not heard that specific DAC and it may well be superb, I merely question some of the design choices based on experience with ASRC's and such stuff....
Another PS, just checked, this DAC does feed the ASRC in the Receiver a clock at 44.1KHz, so there is no upsampling, however the digital filter remains in place and thus as said, the Data recovery is not bit accurate, if this is audible or not is another story, anecdotal experiences suggest it may be.
I double checked the rockna-line webpage again. The schema indicates an AD1802 receiver, but the the product description says it uses an AD1892... The same with the receiver, schema has PLL1705 while the description says PLL1707... Maybe the schema is old and hasn't been updated??? Gotta shoot 'em email to confirm. Anyways, I checked the AD1892 .pdf datasheet, and on -Page 18- there is an option of an ASRC Bypass Mode. Hmmm...
The diagram above also indicates that the Asynch Sample Rate Converter can be bypassed and go direct to the Output Serial Interface.
Thanks again for sharing valueble empirical experience on this.
> I double checked the rockna-line webpage again. The schema
> indicates an AD1802 receiver, but the the product description
> says it uses an AD1892...
> The same with the receiver, schema has PLL1705 while the
> description says PLL1707...
The PLL170X part is a clock divider whcih generates multiple clocks from a single 27MHz crystal, aimed at DVD-Players and usually not terribly low jitter, even though the 50 - 75pS from the PLL170X series is not too bad.
> Anyways, I checked the AD1892 .pdf datasheet, and on -Page 18-
> there is an option of an ASRC Bypass Mode. Hmmm...
Yes, there is, but with one caveat, namely this:
"In bypass without an external PLL, jitter may be as
high as one MCLK."
Now for a 44.1KHz sample rate and MCLK = 512 X 44.1KHz 1MCLK equals 44nS. This means the peak-peak jitter after the AD1892 in ASRC Bypass Mode may be as high as 88000pS, compared to the around 200pS after a CS8412.
FWIW, I got this reply from the designer:
> > > "The ASRC of the chip is not bypassed - it's just set to the same output frequency as the input (44,1 khz). So jitter will be as low as 50 pS (the PLL jitter)..." < < <
You were right the first time about 44.1khz feeding the ASRC part of the receiver...Data recovery NOT Bit Accurate. A "non-os" dac with digital filter set in place!? 8^(
I was thinking of replacing the Crystal and/or PLL with XO products by Tentlabs if applicable....
> I was thinking of replacing the Crystal and/or PLL with XO
> products by Tentlabs if applicable....
You could do that (You need the 22.5792MHz variant), but I may forgiven for my prejudices, I just really do not like the concept used in the DIR.
Considering how simple a CS8412/PCM56 DAC is to make I see no reason not to to do it "dead bug" wiring, such hardwired versions tend to be better sounding anyway, I tend to use a small piece of copper sheet bonded to a piece of solid hardwood as "platform" on which to place my "bugs".
If you spend some time you can use 3D approaches to wiring that allow current loops to be a fraction of what is needed for PCB's. If you don't want to do dead bug wiring you can use matrix boards (prototyping boards) for one-off designs.
While loads of people get worried about the 22.7uS time offset between channels when inverting the wordclock to drive the DAC's I tend not to position either my head (and thus my ears) or my speakers to better than 1cm accuracy, so I do not worry about the < 1cm timeshift of one speaker too much.
In the DAC's I tried I never noticed any difference I would attribute to this timeing difference between the channels for the PCM56/63 Types compared to the TDA's.
So, what you need is the CD8412 set to the correct output mode, a pair of PCM56 and a 74HC04 inverter. The circuit is so brutally simple (and there is no observable advantage to the more complex circuits to split the signals between R/L Channel) that designing a PCB is a total waste of effort in my view.
Hell, my first Adagio DAC was fully hardwired, except the raw PSU Circuit.
Here the most simple PCM56 DAC possible:
This is from here:
The Author first a version with a more complex arrangement to direct the correct signals to L/R but appears just like I did to have found the simplest arrangement the best.
The DAC uses the on-board Op-Amp's for I/V etc, I would probably change this to passive I/V. Here a "hacked up" schematic that shows how to use the passive I/V:
The PCM56 is fully charaterised to use 1K I/V resistors, which means my original TDA1543 I/V with "anti sinc" filter will drop right in, in place of the two resistors to ground shown in the schematic:
Output would then be 0.7V RMS, this can be just enough to drive a preamp/integrated. Otherwise an Analog stage with a little gain would be desirable, around 3 times would be good.
Sounds to me like a pair of 2A3's (or 45's or 71A's) would be an interesting twist, run with a suitable Anode load choke 50H+/20mA and with a 250V HT and a 2k2 Cathode resistor, capacitor coupling to output. Anyway, many ways of skinning that particular cat.
Also here an example of a matrix Board circuit to which a copper groundplane was added using self adheasive copper tape:
...or come close to a TDA1545 based design?
> Will this PCM56 DAC outperform Yeo's ?
> or come close to a TDA1545 based design?
The PCM56 is IMHO a much better dac than the TDA1545 which in turn is a lot better than the TDA1543....
Any idea where to get the PCM56?
Tried several sources like rs-componets etc. but wasn't able to find it.
Looks like that the CS8412 will get hard to find too (was easy last year over here in Germany, but now...).
> Any idea where to get the PCM56?
Try TI's website, they list plenty of sources with the PCM56, remember, it is a currently active and readily made product. You might even be able to get a pair of free samples (ever since TI took over BB they are a lot less tightfisted about samples I noticed recently).
> Looks like that the CS8412 will get hard to find too
Use a CS8414 on adapter. Actually, the CS8414 is chunky enough for hardwiring anyway and there are few enough connections to make....
Our local RS-Comp has the PCM56P (I've yet to find a source for the Crystal Sem DIR).... Some questions though, did you try the schematic from the Japanese site or did you make your own design?Between the 5V to 12VDC supply, which specific V have you found "best" sounding (ie, sweet spot)? Thanks.
> Some questions though, did you try the schematic from the Japanese
> site or did you make your own design?
I used my own, differences in powersuplies where applied, the core circuit happend to be identical, I even used a 74HC04 as it happened.
> Between the 5V to 12VDC supply, which specific V have you
> found "best" sounding (ie, sweet spot)? Thanks.
I only used 5V.
> > I used my own, differences in powersuplies where applied, the core > > circuit happend to be identical, I even used a 74HC04 as it > > happened.
Thanks for the copper tape tip.. The matrix boards available here already have a copper grounding plane on one side. I'll keep the PM56 Dac in mind.
A young buddy of mine is good at making pre-sensitized pcb's - but so far only single sided types for his discrete ss phonostage (his own design). We plan to launch it (locally) as soon as the bugs at the MC section get "solved" (pp or differential *?*). So far the paralleled chokes after the ic V regulator helped a lot, he is yet to try a power opamp as regulator with some trick V reference. BTW, the whole PSU board uses a matrix board with a copper groundplane.
Anyways, the designer of the rockna DAC prefers 1:1 upsampling for the incoming data - counter to the more purist approach of almost all "non-os dacs". I still have to ponder if I'm going for a module or complete diy from ground up using available schematics like Pedja's (I can only do this with a buddy's help of course)...
Do you have a good schematic for a tube outputstage for 1A out type dacs like the AD1856, preferably the Gomez type compund circuit? I have seen a dac kit being sold in the UK using four 12SN7's for Gomez output. I have some 6SN7's lying around I could try.
Thanks for the exhaustive thoughts on DACs.
Does Curcio's CD12V appear to be reasonably weld executed?
There is anecdotal reportage that it can be configured for 1X or 4X oversampling. The schematic is apparently the 4X configuration.
> Does Curcio's CD12V appear to be
> reasonably weld executed?
Probably better than most.
> From looking at a few schematics, I
> am wondering that if the digital signal is iput
> through one line, how is channel separation
The input is digital, its all numbers. As long as you know which, number belongs to which channel and at what time you can reconstruct what was encoded.
PS - Just found another diy td1543 non-os dac board/module with opamp outputstage:
Modified (other site is a bit more extensive):
Wow! lots of good info I think I will try the DIy paradise kit looks like a good one to get my feet wet with, I might try the 1541A kit after that and compare the two. Wish me luck
Thanks for all the help and useful URL's
If you want to get a non-os DAC up and running quickly while sourcing your 1541DAC parts you could try:
This is a basic non-os DAC that uses the TDA1545. The PCB looks well thought out and a full parts kit and PCB cost only ~ $85 USD! PCB only ~ $25 USD.
Good luck and please keep us posted on where you find the best TDA1541 DAC kit/PCB.
This post is made possible by the generous support of people like you and our sponsors: