Modular multi channel DIY AD/DA Box

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Every order of a PCB or kit will come with a schematic and a bom of the current revision.
For the ADC module we hav enow revision B with minor changes like an additional socket for master clock in.
The schematic of revision B will be published, too. I think it is good to make a website for it so that all important information are concentrated into one place. Then you don't have to read all 17 pages of this thread to get all the infos you need just to build an ADC. Unfortunately, the next few days I'll be very busy and do not have the time for doing that. So please be a little bit patient. In two weeks I'll have the time to support the DIY-building of ADCs more helpful. But maybe someone else can manage a wiki?

Raphael
 
There was some brief talk about wikis last summer. I think Kato put one up here
http://wiki.nimbleswitch.com/Main_Page

there seems to be couple entries already, not hugely popular yet :)
 
Good news for some of you:
The 4channel AESout PCB and the parts for the prototype have just arrived. I'm going to stuff and test this board this week so you can expect that it will be available for the DIY community with in two weeks if there are no big mistakes in the design.

Raphael
 
jdbakker said:
rkn80 said:
Mikkel states that his boards do work with SMUX.

They do, sort of.

As far as I can tell, Mikkel's OptoRec is a direct implementation of the Wavefront ADAT receiver chip. This means that while it will accept S/MUX streams (and set a flag to show the stream is S/MUX), you still need an external converter to multiplex the two 48k channels into one 96k channel. This could be implemented in a CPLD/FPGA, a fast enough microcontroller/DSP or a lot of discrete logic.

JDB.

Regarding SMUX implementation I think the easiest way would be to use Wavefront DSP-1k because they also have the software and application note for the implementation.

For the OptoRec (ADAT receiver for the DAC's) look for the AN3101-10:
http://www.wavefrontsemi.com/UserFiles/File/AL_Info/AL31/AppNotes/WavefrontAN3101-10%20S-Mux%20Receiver%20for%20ADAT%20Optical%20Protocol.pdf

For the OptoGen (ADAT transmitter for the ADC's) look for the appnote AN3101-09:
http://www.wavefrontsemi.com/UserFiles/File/AL_Info/AL31/AppNotes/WavefrontAN3101-09%20s-Mux%20Transmitter%20for%20ADAT%20Optical%20Protocol.pdf

Link to other DSP-1K appnotes:
http://www.wavefrontsemi.com/index.php?id=11,13,0,0,1,0

Mikkel also happens to have a design for the OptoGen pcb (I know because I've got one such pcb from him).

Btw. Using similar approach you could combine output from 4 stereo ADC's @48k (8 mono channels) or from 2 stereo ADC's (4 channels) into single stereo channel so that you could output the data using regular S/PDIF or AES/EBU interface @192 kHz. To be able to frame the channels in ASIO (or WDM or WaveRT or ALSA or whatever) driver you had to sacrifice the audio data LSB as there is no other way to mark the beginning of the frame (in HW implementation you could use the User Data bit of course). The loss of one bit would limit the dynamics to about 6 dB*23 bits which is about 138 dB or so.

 
Update:

The AESout prototype has arrived and is stuffed now. Tommorrow I'm going to test this PCB.
This design uses the CS8406 because it can be fully hardware controlled and does not need a microcontroller like the DIX4192 used in the bigger design above.

Raphael

P.S. I mounted a the pin header because in my box I want to place the PCB a little bit away from the rear panel (I need to place several connectors on the rear panel...). But you can also mount two BNCs directly on that PCB or XLRs via the pin header.
 

Attachments

  • DSC_0033.JPG
    DSC_0033.JPG
    169.1 KB · Views: 289
It has not been discussed in detail, but I would like to make a start:

PSU requirements for all voltages needed in this AD/DA with AES-EBU project. Even for 16 channel I/O there won't be a ridiculous amount of current draw so this certainly eases up the choices. Let's assume it'll stay well below about 1.5A for all voltages combined, which is already overkill.

If we want to make this really easy a standard selection of 78XX regulators with "application note" implementations should suffice. Moving further in complexity, a selection of LM317/LM337 with all the standard bells and whistles is a slight improvement. Still cheap and easy.

But as far as I have understood from discussions elsewhere on the web, it is exactly these stock standard PSU implementations that are holding back the performance of more serious AD/DA solutions.

Let's say we want to move into over engineering territory, and use Jung regulators or similar opamp based linear regulators. For the audio 15V rails that would certainly make an improvement. What exactly are the 5V and 3.3V rails driving? What kind of performance is expected of these supplies? Jung regulators a good idea here as well?

Oh and, what is the VBIAS voltage input for each ADC input buffer opamp?
 
In my understanding the dac chip has separate digital and audio Vcc, and I think using Jung regulator for audio Vcc is in order.
As it happens with many of my projects, I even have Jung regulator pcbs but never got around to building one to see what happens...
::)
 
Kingston said:
Let's say we want to move into over engineering territory, and use Jung regulators or similar opamp based linear regulators. For the audio 15V rails that would certainly make an improvement. What exactly are the 5V and 3.3V rails driving? What kind of performance is expected of these supplies? Jung regulators a good idea here as well?

Oh and, what is the VBIAS voltage input for each ADC input buffer opamp?

All ADC's and DAC's are mixed-signal devices. Digital data either goes in or comes out on one side. And analog signal either go in or come out on the other side. Most all digital IC's running up in the MHz region make noise. Lots of noise. The noise can come shooting out the power supply pins on these IC's (and elsewhere too). If you are contemplating building this AD/DA modular project, I suggest you go download a couple of datasheets. Go download PCM4204 and PCM4104 at TI.com. The digital side of these devices needs regulated +3.3vdc. The analog side needs +5vdc. I am sure that various people will debate the requirements for the digital supplies. Low-impedance, low-noise supplies with proper decoupling caps right at the converters. And the analog side is either creating your analog output signal. Or feeding it into the ADC's for conversion. Jung regulators here are not going to hurt at all. In fact, I'm sure they will help........... And then to address the last question........... VBIAS is a reference voltage usually created by the ADC IC itself. VBIAS is set at one-half of the total input voltage range. Even though analog audio can sometimes go to pretty high voltage levels, the ADC's can only accept an input signal between 0 and 5 volts. So VBIAS is +2.5v and the signal fed into the ADC inputs are centered at +2.5 and can swing from 0 to 5v without causing any digital over's. DW.
 
Tubemooley said:
Even though analog audio can sometimes go to pretty high voltage levels, the ADC's can only accept an input signal between 0 and 5 volts. So VBIAS is +2.5v and the signal fed into the ADC inputs are centered at +2.5 and can swing from 0 to 5v without causing any digital over's. DW.

Full scale input voltage for PCM4202 is 6.0 VPP.
 
mhelin said:
Tubemooley said:
Even though analog audio can sometimes go to pretty high voltage levels, the ADC's can only accept an input signal between 0 and 5 volts. So VBIAS is +2.5v and the signal fed into the ADC inputs are centered at +2.5 and can swing from 0 to 5v without causing any digital over's. DW.

Full scale input voltage for PCM4202 is 6.0 VPP.

6Vpp differential.

For a differential signal with a common-mode voltage of 2.5V, each of the input pins must stay between 1V and 4V to avoid clipping, and between -0.3V and 5.3V to avoid damage to the chip.

JDB.
 
Kingston said:
But as far as I have understood from discussions elsewhere on the web, it is exactly these stock standard PSU implementations that are holding back the performance of more serious AD/DA solutions.


This is exactly what I have read many a time as well. It's poor power supply designs in prosumer commercial units that hampers the converter box and affects things like its headroom.  
 
deuce42 said:
This is exactly what I have read many a time as well. It's poor power supply designs in prosumer commercial units that hampers the converter box and affects things like its headroom.  

Not headroom. That is probably never the case. AD/DA units are always "low current" and regulators simply won't sag. If they did, your digital chips would simply refuse to work.

Rail noise is the real issue here. From noisy regulator outputs, or noise coupled between IC's that should never "see" each other through rails.
 
jdbakker said:
Full scale input voltage for PCM4202 is 6.0 VPP.

6Vpp differential.

For a differential signal with a common-mode voltage of 2.5V, each of the input pins must stay between 1V and 4V to avoid clipping, and between -0.3V and 5.3V to avoid damage to the chip.

JDB.
[/quote]

And that's why you have a input divider option on the ADC board. :)
 
Hi,

I'm happy to tell you that the AESout board passed the test in the prototype. So we can now add AES out to the feature list!

Raphael
 
Here I started the group buy of the AES out boards:

http://www.groupdiy.com/index.php?topic=38556.0

Raphael
 
This a great build idea , but is there a way to use DOa s on IN's and outs?
It seems that is where all the conveters fall short,I know that is what Burl audio is doing..
Keep up the good work.
 
Ok I am going to ask a very dumb question(s) now so apologies in advance.

These AES boards:

1) They have four channels just like the A/D boards do. So if you have ordered two A/D boards, you accordingly need two AES boards as well to make 8 channels?

2) These boards represent the input side of this project? So XLRS etc connected to these boards are your audio signal inputs?

3) I want to eventually connect this AD and DA project via lightpipe so I can use an M Audio lProfire lightbridge as a pro tools dongle. My idea is to use a good quality converter like this project, but the inputs and outputs of this project will be lightpipe connected to the ins and outs of the M Audio lightbridge.   Do these AES boards connect via lighpipe rather than XLR etc?

Once again apologies if this stuff is pretty obvious
 
deuce42 said:
Ok I am going to ask a very dumb question(s) now so apologies in advance.

These AES boards:

1) They have four channels just like the A/D boards do. So if you have ordered two A/D boards, you accordingly need two AES boards as well to make 8 channels?

2) These boards represent the input side of this project? So XLRS etc connected to these boards are your audio signal inputs?

3) I want to eventually connect this AD and DA project via lightpipe so I can use an M Audio Profire lightbridge as a pro tools dongle. My idea is to use a good quality converter like this project, but the inputs and outputs of this project will be lightpipe connected to the ins and outs of the M Audio lightbridge.   Do these AES boards connect via lighpipe rather than XLR etc?

Once again apologies if this stuff is pretty obvious
1. Not quite. "It can be directly connected to the AD board and accepts two I2S input streams and puts out two AES output streams. That gives in total 4 mono channels." Each AES digital signal is two encoded analog audio channels. Raphael's ADC board has two PCM4204 devices on it. Each PCM4204 is 4 ADC input channels. So two of those is 8 analog audio channels. Eight analog channels going into 2 PCM4204's will produce 4 I2S digital output signals from those 2 PCM4204's. So if you have ordered two A/D boards, you need 4 AES boards.

2. Incorrect. The signals coming off the AES boards are your digital outputs. You can use XLR's there to transmit the digital signal if you decide to do balanced AES instead of unbalanced SPDIF. I do AES. But those XLR's are transmitting a digital signal and not receiving an analog signal. Your analog signal inputs will be elsewhere.

3. Yes and no. Mostly no. You can take any one of the AES output signals and transmit it via TOSLINK optical. But that will be one AES signal transmitting 2 encoded analog audio channels. That's optical SPDIF and not ADAT. ADAT is basically multiplexed optical SPDIF. Eight analog signals are digitized and all jammed into one ADAT stream (with sample rate up to 48kHz). I do not know the Lightbridge that well but I'm guessing it wants to receive ADAT signals and not multiple optical SPDIF signals. Most boxes don't support more than one optical SPDIF input. They could do that but the manufacturers are lazy and cheap and rarely implement that. In order to receive the I2S output signals from the ADC board and convert it to ADAT, you need another board which Raphael is not going to produce (not yet anyways). Someone else makes one of those. I think his name is Mikkel and he's mentioned much earlier in this thread.
 
ad 1)
Tubemooley, you are not right, deuce42 is right. Have a look on the schematic:
On every AD board is only 1 ADC taking four analog inputs and putting out 2 I2S channels.
The number of ADC boards and AES boards must be identical.
Some example configurations are:
1 ADC boards => 1 AES board
2 ADC boards => 2 AES boards
3 ADC boards => 3 AES boards
...

Raphael
 
Back
Top