My first DAC on PCM1794A - help needed.

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi fanatics! Although I have quite decent experience in electronics in general, my expertise in audiophile products is limited - I'd like to ask you to shed some light on areas that are a bit dark for me. I've red a lot and tried to find the answers myself - and most of them I've found, but some need clarification. Or maybe just the confirmation that I got them right, before I get my hands dirty :]

I'm going to use Raspberry Pi with Volumio as direct I2S source, this whole project is based on RPi-DAC project from here: RPi-DAC - Dual Mono

As I expect this DAC to be superior to my Hegel's H70 integrated one - I'm building the dual mono version, including power supply. I've marked areas with questions below:

QwHTc.jpg


#1 - do I have to use separate LDO for generating 3.3V line, or I can take one from RPi. Without influencing the quality of course (even theoretically).

#1B - Can I use the same power supply (TOROID1) to power DACs on both channels? Or is it a big crime and a rude deviation from "dual mono" thing (to keep both channels as much separated as possible) and I should use separate transformer or its coil to the other channel?

#2 - Have I got the "merging" thing in PSU correctly (to create +15/0/-15 line from two +15 lines)?

#3 - the most important one. This last stage (is it correct to call it "output buffer"?) in original T-DAC/RPi-DAC design is mean to be used with headphones (TPA6120A used). I need it to be used with RCA outputs as a regular line input to my amp. How do I select this stage properly then? What are available designs that you would recommend for me? Will I be able to try several of them without changing I/V part too? Any info on this stage (both general and specific circuits/designs) would be a huge help.

Thanks in advance!
 
#1 I wouldn't recommend it. Everything I've heard is that the RPi PSU is being run near its maximum output, and drawing too much current from the RPi board causes instability.

#2 Without seeing the actual schematic, it looks right. Those are positive regulators, so the "ground" of the negative half will become -15V (because you'll hook the output to ground). You must use dual secondary transformers for this type of circuit, no center-taps allowed.

#3 You can use the headphone output as a line level output in most cases. As long as there isn't a lot of extra gain for the headphone output.
 
I consider #1 and #2 answered - nothing to add here, thanks.

#3 - Seems I'm missing a lot of info about this stage :] If it's fine to skip it altogether - why anybody is even thinking about putting anything in there?

I get "why" in headphones case - there's too little gain in I/V stage I guess. But is it really completely not required when using with a regular amp setup? I think my project just become a lot more easier then :]

On the other hand - let's say we still add it. What benefits would we expect from doing that (using different designs, by increasing difficulty)?
 
I should give more detail on #2. The drawing is incorrect. That's a dual secondary transformer providing power to 4 regulators. If you want to use 4 regulators, you need 4 secondaries. As drawn, it'll short the transformer.

As for whether to have a buffer stage (headphone amp), the buffer stage will reduce unwanted effects of having capacitors on the output of the I/V stage and capacitors or volume potentiometers on the input of the power amp. Headphone amps don't always have gain, they sometimes just provide more current.
 
#3 - Seems I'm missing a lot of info about this stage :] If it's fine to skip it altogether - why anybody is even thinking about putting anything in there?

I get "why" in headphones case - there's too little gain in I/V stage I guess. But is it really completely not required when using with a regular amp setup? I think my project just become a lot more easier then :]

In the headphones cases, it's mostly needed because the opamps used in the I/V and filter stage don't have enough current output for low impedance headphones and not enough voltage output for some high impedance ones. Adding a dedicated stage for the headphones also allows to easily place a volume control in between stage 2 and 3 if required.

With a line-out, any half decent opamp used in the stage 2 filter should be able to drive the interconnect cables and the amp input impedance.
 
Would it be a big crime to build all modules on separate PCBs and to run signals and power lines with short wires between them (as opposed by cramming everything on one PCB) ? As I think I'll go through the process of trying out various different modules for I/V and output (if any) stages, that would be much more convenient.

Also for PSU it seems I'll have to buy 2x32V secondaries toroid and then re-make it to produce 4x16V as I've never seen toroid with more than 2 secondaries. Maybe it's not without a reason and I should look into EI type (which, from my experience would be less ideal for PSU)?
 
Last edited:
I like to be through, that's the pleasure of doing things yourself, afterall :]

My DAC starts with a proper power. After doing a research on toroidal VS ei transformers, I decided to use EI in this particular project. Carefully placing and screening them of course will be done, but the question is:

Since I need so much secondaries for powering everything up, should I prefer less, but bigger trannies with more secondaries on them, or smaller ones, but more of them? My guess it's the second option, since smaller transformer (actually, a group of them) should radiate it's stray inductance in a smaller field, thus making the distance between it and the PCBs a better "isolator".
 
Member
Joined 2007
Paid Member
My DAC starts with a proper power. After doing a research on toroidal VS ei transformers, I decided to use EI in this particular project. Carefully placing and screening them of course will be done, but the question is:

Hi ...

Interesting project which I would say have many possible solutions depending on what you want to achieve sound-wise and/or technically.

Regarding transformers my personal consideration would be EMI since you are using regulators for each rail to the DAC. It likely is of less importance in this case if the transformer is capable of delivering high impulses (which is what I remember EIs are known for being able to do - I may be mistaken here, though). Personally, I would just get transformers with the required voltages and windings, possibly separating the digital side from the analog side. In my experience the PCM1794 is especially sensitive to VCC and VDD voltage fluctuations so a clean supply would be my choice (possibly prefiltering the TPS7A ..)

Regarding an output (your #3) directly to an RCA I would personally use the LME49990s directly. It's capable of driving 600 ohms and I suppose the input your are feeding is not lower than this ... (?)

In terms of splitting up the boards it of course can be done - but you may wish to consider any additional noise between the connecting wires, a split ground plane etc. I would say that it depends on how you do it.

Finally, having designed a couple of PCM1794 based DACs I'd like to mention this implementation as an alternative to an opamp based I/V conversion:

http://www.diyaudio.com/forums/digi...-nos-192-24-dac-pcm1794-waveio-usb-input.html

http://www.diyaudio.com/forums/digi...-upgraded-single-board-pcm1794-nos-dddac.html

I'm not saying it's better than an LME49990 based design but IMHO it's quite musical not least when given a very good power supply. Also, it is by default designed to work with a USB-to-I2S converter board which uses low jitter oscillators - a factor which in my experience is quite key to a DAC's sound.

In either way: Good luck with your project ;)

Jesper
 
I'm back to this project - and I have some questions at my hands :] PCM1794a has a de-emphasis filter (pin3). By default in original project of the author it's enabled (pull-up resistor), but defeatable with jumper to ground.

Do I need to keep it enabled at all? I wasn't able to find enough info what this filter does, just that it's somehow related to a very rare use cases. Do I need to worry about it at all. Maybe simple short it to gnd and disable?
 
Hm, ok, one more. The original designer of this project put 2.7n caps in I/V stage opamps. What he also did - is added a note about using 6.8n. This led me to some confusion which ones am I supposed to use.
1) What's the influence of this choice? I see some notes with resulting frequencies, but no idea how to interpret them.
2) If I go with 6.8n, does it apply to all 8 of them, or only the 4 ones with note "6.8n/16V" next to them (in the right-side part of the I/V schematics)?

If this is influenced somehow by 1K...1.47K resistor choice, I already bought some metal film 1.2K ones and would like to use them. As for 360R/715R ones, they're in the basket but will be ordered together with caps.

oyMSct.jpg
 
Did you actually make it work as the drawing?

From pi/volumio to dac via i2s without USB is not going to work since 1794@ doesn’t has internal pll

Sorry, I2S and USB? Two different topics. The PCM1794A is connected via I2S directly to the RPi digital audio interface. The PCM1794A is clocked by the I2S SCK signal, with an automatic clock detection. A PLL is not really needed as long as the I2S SCK clock is stable and not with so much jitter.
Even with the "jittery" signal out from Raspberry Pi - the DAC follows well the I2S clock. The jitter might just create a bit of distortions but the DAC works for sure as shown in the schematics. I have not realized an audible effect based on the (acceptable) jitter generated by the RPi.
 
Hm, ok, one more. The original designer of this project put 2.7n caps in I/V stage opamps. What he also did - is added a note about using 6.8n. This led me to some confusion which ones am I supposed to use.
1) What's the influence of this choice? I see some notes with resulting frequencies, but no idea how to interpret them.
2) If I go with 6.8n, does it apply to all 8 of them, or only the 4 ones with note "6.8n/16V" next to them (in the right-side part of the I/V schematics)?

If this is influenced somehow by 1K...1.47K resistor choice, I already bought some metal film 1.2K ones and would like to use them. As for 360R/715R ones, they're in the basket but will be ordered together with caps.

oyMSct.jpg

All these RC circuits on the OpAmps create low-pass filters. They are "needed" (actually suggested) in order to get rid of the high frequency noise generated by this Delta-Sigma DAC (due to the "noise shaping"). Otherwise, noise which is not audible (way above 24 KHz) could confuse a following power amp.
For the filter calculation - you can check here: Op Amp Low Pass Filter | Active Filter Circuit | Radio-Electronics.Com

So, the filter corner frequency is given by: fc = 1 / (2*Pi*R*C) .
You can vary R and C and set a corner frequency as you like, as long it is above approx. 24..30 KHz.The resulting corner frequencies for the R and C used are given in the schematics.
If you double C (or R) the corner frequency is lowered by one-half. So, with 1.2K and 6.8n you would get 19.5 KHz which might be a bit too low!: corner frequency has already a -3dB attenuation, so you would cut a bit from the audible audio spectrum (if your ears are able to realize 18 KHz or higher, quite unlikely if you are not so young anymore).
BTW: do not make the R too small: it sets the input impedance of the OpAmp circuit and acts as a load on the DAC output.
If you make R way too large you have to consider that it will act as a thermal noise generator. (a too large R needs a very small capacitance but think about tolerances which will result in a "wrong" corner frequency).

If you change - change all in in the same 'stage' in the same way: see that we have two low-pass filters in series (gives you a second-order filter). Both filters have different corner frequencies (the second one filters very high frequencies left by the first one again).

How do you set the corner frequencies is up to you. Just bear in mind they should just remove the not-audible part of noise which might still come out from the DAC (above 30..50 KHz). It is just to 'protect' the following power-amp: often such one has a low-lass filter also on its input, but in case not and the power-amp is very wide-band (e.g. able to amplify also frequencies above 30 KHz) the amp could run into saturation and will burn power by amplifying 'just' high-frequency noise (which will not come out to the speaker as audible signal but the amp could 'run away' by such noise).

Metal-film resistors are a good choice (lower thermal noise). Also to use good caps, e.g. as MKT (but not available as SMD), e.g. as NP0 (great temperature coefficient). I prefer often X7R, X6R, X5R. Sometimes also a need to optimize for low ESR but on these tiny caps it does not matter so much.
More details to find here: Basics of Ceramic Chip Capacitors

Nothing wrong with the schematics, room for changing R and C, just to cross-check the resulting corner frequencies - just cut outside of your audio spectrum (e.g. assuming 22..28 KHz).
BTW: actually, it depends a bit which sample frequency do you want to run: 44.1/48 KHz gives you a corner frequency as 1/2, so 22..24 KHz. With 192 KHz sample frequency on play via DAC - you could think to set 96 KHz, but you cannot hear anyway. So, the given values here in schematics are a good fit for not so drastic filtering somewhere in between.

My suggestion: go with nominal values, e.g. 1.2K and 2.7n first: if you need more drastic noise filtering (which depends on the following amp) - you can just solder a parallel cap on top of the SMD cap (make C a bit larger by adding a parallel C, with parallel R you could just make it smaller which increases frequency).
 
Last edited:
Sorry, I2S and USB? Two different topics. The PCM1794A is connected via I2S directly to the RPi digital audio interface. The PCM1794A is clocked by the I2S SCK signal, with an automatic clock detection. A PLL is not really needed as long as the I2S SCK clock is stable and not with so much jitter.

I guess the problem raised by 24century is that the pcm1794 needs 4 I2S lines, including a master clock, and that the PI only provide 3 I2S lines, missing the master clock. DAC ICs with an internal pll, such as the pcm5102, will run fine on 3 lines, not the pcm1794.

It seems that you deal with the problem by tying bck and sck. Since the pcm1794 works at least with a 128fs system clock, you need a 5.6448MHz clock at least. A 44.1KHz/16bit signal will have a BCK of 1.4112MHz so I'm not sure how the digital filter can properly work...

BTW, X7R (or even worse X5R) caps should never be used in filters as their capacitance is voltage dependant. Among ceramic caps, NP0 only should be used in the signal path. X7R is good for PS decoupling.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.