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.
Hi, thanks for the welcome  ;)

I acknowledge the points raised by all.  I'm no engineer, but I have done a bit of modding and maintenance work and am looking to expand my DIYing activities.  I learn best by doing, and was planning to order one or two extra kits for learning purposes in case I were to somehow completely botch up the first one.  I actually do have a friend who is an excellent tech that does a great deal of work for studios in the area that I could call upon should I find myself overwhelmed.

I'm pretty handy at precision work though and feel I can slog my way through the manual assembly - I just wanted clarification on the chain of components required to go from assembled individual kits to a completed functioning unit. 

Assembling existing designs with some instructions I can do - designing my own solutions is definetly beyond my range of experience right now.  :)
 
I do not know why people are always scared by SMDs. The is state of the art, Throug-Hole devices are almost dead.  :(  ::)
By the way my first DIY kit that I did many years ago was already a kit with SMDs. It was an USB soundcard with the main IC in TSSOP case. Suceeded the challange. It is not so difficult it only seems to be. :)

Regarding the interface: Yes there are intefaces to the outside in the pipeline. First of all there are Mikkel's ADAT boards and many S/PDIF boards can be found on the internet. I have an AES interface board ready to make and currently I'm designing a wordclock board which will come before the AES board because I got several messages, that people have an ADAT interface but no good wordclock in/out and no good main clock generation. So the wordclock board will hopefully fill this gap providing wordclock, masterclock (= Superclock in ProTools) and sync to externally applied wordclock.

And yes, all the boards I've successfully built and tested will be offered for a group buy. The only reason for restricting to a certain board is to minimize the risk of confusion. You can be sure that after ADC board a DAC group buy will be started, then wordclock, then AES interface and whatever comes then. There are so much possibilities. Even other people can jump into it and design boards for this box.

Raphael
 
Ok. I would like to discuss the wordclock module now.
I'm currently busy with designing the circuit. Here is what I specified:
- generate master clock for the box from a crystal on wordclock board
- generate I2S clocks from the master clock
- fan out the clock signals on 6 pins for each clock, each pin buffered so that you can connect many ADCs and DACs to the clocks
- send internal WCLK (wordclock) to the outside via BNC
- send internal MCLK (superclock) to the outside via BNC
The following features are switchable:
- replace/sync internal MCLK by externally applied superclock
- sync to externally applied wordclock
(in both cases the internal I2S clocks are generated from the sync'ed master clock)
- sync to externally applied AES signal

- the design shall use parts which can be easily ordered by a private person. I know that there are some high-precison clock generation and distribution chips but very often it is not so easy to get them in small quantities.
- the design should not use devices which need to be programmed (like MCs or FPGAs)

- the board shall be usable stand-alone: Makes it possible to build a master clock unit without the ADC/DAC stuff.

Anything missing or additional ideas?

Of course a more advanced board can be done later utilising programmable devices but for now I want to keep the board doable also for those who do not have the tools for programming devices.

Raphael
 
I am very interested in the upcoming clock circuit and appreciate the idea of keeping it free from programmable devices first.

I like to link to a small basic clock module that I stumbled upon when looking up suitable crystals (see some posts back). It could be an example how some of the specs could be fulfilled, e.g. easily available parts esp. the crystal. This basic clock core uses a low noise LT regulator to bring down the 5V rail to 3.3V: http://www.awdiy.com/index.php?page=miniclock-v2

Kind regards,
Martin
 
rkn80 said:
- generate master clock for the box from a crystal on wordclock board
- generate I2S clocks from the master clock
- fan out the clock signals on 6 pins for each clock, each pin buffered so that you can connect many ADCs and DACs to the clocks
- send internal WCLK (wordclock) to the outside via BNC
- send internal MCLK (superclock) to the outside via BNC

That's the easy part. Do consider source termination on the BNC drivers, possibly with a jumper(/switch) to short them out.

rkn80 said:
- sync to externally applied wordclock
(in both cases the internal I2S clocks are generated from the sync'ed master clock)
- sync to externally applied AES signal

That's the hard part, especially if you want to allow syncing to 44.1k, 48k and everything in between while still having good jitter performance. Bonus points for supporting Varispeed (yuck).

rkn80 said:
- the design shall use parts which can be easily ordered by a private person. I know that there are some high-precison clock generation and distribution chips but very often it is not so easy to get them in small quantities.

That's an admirable goal. Should be doable for parts like line drivers and distribution amplifiers, but I suspect the clock oscillator itself is safer to implement in a pre-canned version than as a discrete design due to the sensitivity of component variations and construction.

JDB.
 
jdbakker said:
That's an admirable goal. Should be doable for parts like line drivers and distribution amplifiers, but I suspect the clock oscillator itself is safer to implement in a pre-canned version than as a discrete design due to the sensitivity of component variations and construction.

I do not exactly understand what you mean eith a pre-canned version. For the design I'm doing (like most of my diy projects) the decision for components is made by the question. "Can I get the part from one of the bigger distributors liek digi-key or Farnell or do I have to order it in quantities of 1000pcs." If the last is the case I do not use the part... ;)
That is the rule for the wordclock module too. Therefore I decided to use an good oscillator and to derrive the needed clocks from it via a clock divider circuit. I don't want to use a PLL170x from TI or the clock generators from Cypress. I think a good crystal oscillator will outperform a PLL chip.

Raphael
 
rkn80 said:
I do not exactly understand what you mean eith a pre-canned version.

This is a pre-canned crystal oscillator, this is a discrete one. I suggest you go with the former.

rkn80 said:
That is the rule for the wordclock module too. Therefore I decided to use an good oscillator and to derrive the needed clocks from it via a clock divider circuit. I don't want to use a PLL170x from TI or the clock generators from Cypress. I think a good crystal oscillator will outperform a PLL chip.

No question about that.

But you said that you also wanted to allow locking to external word clock. I would urge you to specify the intended capture range, as that will not only have a major impact on the rest of your design but it'll also help people determine if your clock module (and with it the entire converter) will be usable in their intended applications. That's especially good to know given that we're at the board-ordering stage of one of the subsystems.

JDB.
[have a bit of a fever, so I might not be expressing myself too clearly]
 
Ah, ok. That what you call a pre-canned crystal oscillator I wanted to use, too, because that are those devices which I call easy to get. :)

Regarding locking on wordclock. I was thinking of capturing the range between 32kHz and 96kHz. About 192kHz I'm a bit unsure wether it is useful or not. I know that ADAT supports that vie SMUX IV but is it used by any soundcard? Same thoughts about AES. Is 192kHz multichannel output really in use? Or shsould we think of a more sophsticated board with a fpga or microcontroller on it if you really wnat to have this high-quality 192kHz signals.
But for this board it is important to find a good balance between features, price and effort to build it.

Get well soon, JDB!

Raphael
 
All of this sounds very good Raphael. I have added my responses IN CAPITAL LETTERS for what I need in my studio. This is just what I need. Other people have other needs. Just trying to contribute. DW.

Here is what I specified:
GOOD - generate master clock for the box from a crystal on wordclock board
GOOD - generate I2S clocks from the master clock
VERY GOOD. ABSOLUTELY NECESSARY - fan out the clock signals on 6 pins for each clock, each pin buffered so that you can connect many ADCs and DACs to the clocks
DON'T NEED - send internal WCLK (wordclock) to the outside via BNC
DON'T NEED - send internal MCLK (superclock) to the outside via BNC
The following features are switchable:
GOOD BUT DON'T NEED - replace/sync internal MCLK by externally applied superclock
DON'T NEED - sync to externally applied wordclock
GOOD (in both cases the internal I2S clocks are generated from the sync'ed master clock)
ABSOLUTELY POSITIVELY NECESSARY - sync to externally applied AES signal

Sync/lock to externally applied signal up to 192kHz is good but not 100% necessary. We run at 96kHz these days but our RME and Lynx cards and some of our external converters will operate at 192kHz. And like JDB said, pre-canned oscillator is the way to go.
 
If you're going to sync to external clocks, you should include a WC In option.

In studios with more than one converter or a WC generator it's a necessity to get
them running in sync.

Beyond 96kHz is not a concern here.

In my list, WC IN is in capitols.  8)

Mark
 
up till now i have only used $200-$300, 2-4 I/O sound cards. home studio to record myself and friends. from researching the internet i understand (correctly or incorrectly) that their is not much difference in audio fidelity in cards in this price range, but the A/D converters on higher end $$$  (apogee.lynx) are a vast improvement over the cheaper cards. if i build this project how much of an improvement can i expect to hear? is this on par with the stuff used in major studios,somewhere in between or i wont hear much of a difference from my m-audio card unless i listen really hard?  sorry for the divergence from all the tech talk (that i dont understand.....Yet!) but i forsee a great effort i my part to understand this circuit and im weighing whether it will be worth the effort in audible terms. i know learning is always a good thing but i have a staggering amount to learn about analog circuits allready. that said: rkn80 thanks for sharing a great project with others.
 
on the actual converters, yes, there is a signficant performance different. PCM4204 and PCM4104 are not used in toys.  ;)

What will make the biggest difference is the power supply and the clocking structure.

Insisting that the ADC itself is driven from it's own clock, and not PLLing to another master will make a huge difference.

Raphael - don't think I'm not watching this thread!  ;) Good Luck!
 
Tubemooley said:
ABSOLUTELY POSITIVELY NECESSARY - sync to externally applied AES signal

Biasrocks said:
In my list, WC IN is in capitols.  8)

nielsk said:
EXT WC in is an absolute must in all but the most analog dominated dominated situations. Any piece of digital gear is incomplete without it.

OK, so it would appear that people want to be able to sync with external clocks, be it AES or WC.

If I were designing this clock board, I would find it very useful to know how much lock range you folks require. Do you need full-range any-rate locking to any odd sample rate between 32k-192k, including oddballs like say 46.37ksps, or would it be sufficient to be able to lock to fixed incoming rates at 44.1/48/88.2/96/176.4/192ksps +/- 100ppm?

(Sure, in kid-in-a-candystore mode all of us would love full any-rate support, but it would make clocking much more complex and/or jittery)

JDB.
[I for one do mostly live location recording, running off my own clock, and the few times I have to deal with external digital streams and/or timecode I use Mr. ASRC, so for me personally external sync is Not An Issue]
 
JDB,

is it really that important to know which rates will occur? Let's assume we have a circuit that always multiplies the incoming wordclock by 256. Then you always have a masterclock dependent on the incoming clocking. From this master clock you then derrive the bit and the frame sync clock of the i2s and with these three signals you drive the DAC or the ADC or what ever. Due to this master, bit and frame sync clock have now a fixed relationship: MCK = 256fs, BCLK = 128fs, LRCLK = fs. If fs is 48kHz then the clocks are exactly that what is needed for a PCM4204 in single rate mode. For fs = 44.1kHz it is exactly the same scheme. So it seems to me that the ratio between the clocks is important (and this scheme is changed for double and quad rate mode).  The data sheet states that the PCM4204 supports master clocks between 6.144 and 38.4MHz for single rate configuration. So now let's assume we do not have exactly 48kHz but 47kHz. Then the master clock will be 12.032MHz instead of 12.288MHz but the PCM4204 will still lock on it. So it doesn't matter which frequency is coming in for the generation of the i2s clocks.
From a project some years ago I know that you can vary your master clock in a certain range and as long as the relations between the i2s clocks are ok your system will work because everything is "in tune" (of course there are limits and surely at a certain point the clock detection in the chips will be confused...).
So in fact everything is derrived from your incoming clock signal any change in the wordclock rate also occurs on all clocks keeping them still in the correct relation. Then you don't have to care about the rates so much. But you have to find a way to deal with the jitter on the incoming signal and get it free from jitter as much as possible because the jitter will then also affect the master clock and from there all other clocks and finally the whole system.

Raphael
 
rkn80 said:
is it really that important to know which rates will occur?

Yes.

The area where it matters a lot is jitter. Very simply put, the higher the Q of the frequency-dependent element of an oscillator the lower the phase noise and thus the lower the jitter. However, the wider the required tuning range the lower the Q, and thus the higher the jitter.

If you limit yourself to 44.1/48k and multiples +/- 100ppm, you can simply have two pre-canned VCXOs (one at 22.5792MHz and one at 24.576MHz) to cover all needed frequencies. This plus a (reasonably) simple PLL leads to a DIYable design where the clock jitter does not limit the converter's performance.

If you need to cover the whole range in one go you have to sacrifice either performance or simplicity. On one end of the scale is a simple 4046-based PLL using its internal VCO, with jitter performance as bad as a few ns. On the other end of the scale you can get down below 10ps jitter with a careful digital synthesizer (DDS) with clever spur cancellation, but that's hard to build and calibrate even for those of us who do have an electronics lab at our disposal. And even then the VCXO has better performance.

JDB.
[we've gone over it in other threads before -- for best performance with this class of converter chip you do want to keep jitter below 10ps]
 
jdbakker said:
[I for one do mostly live location recording, running off my own clock, and the few times I have to deal with external digital streams and/or timecode I use Mr. ASRC, so for me personally external sync is Not An Issue]

i second this.. i'd prefer to ASRC and leave an ADC run from it's own oscillator, instead of slaving it to a jitter WC+PLL.
 
jdbakker said:
Tubemooley said:
ABSOLUTELY POSITIVELY NECESSARY - sync to externally applied AES signal

Biasrocks said:
In my list, WC IN is in capitols.  8)

nielsk said:
EXT WC in is an absolute must in all but the most analog dominated dominated situations. Any piece of digital gear is incomplete without it.

OK, so it would appear that people want to be able to sync with external clocks, be it AES or WC.

If I were designing this clock board, I would find it very useful to know how much lock range you folks require. Do you need full-range any-rate locking to any odd sample rate between 32k-192k, including oddballs like say 46.37ksps,

Without getting into religious issues regarding external clocking I would say yes.

In addition to the standard rates (I go as high as 96kHz in my productions). I make use of the option in my WC generator to do vari-speed. Anyone coming from a world of analog knows how useful and valuable VSO can be in a production. Rarely, if ever would I go below say +/-1.5kHz.

Other folks may have different requirements.

And while I'm aware that we're not designing a commercial product, the bottom line is:

External clocking is used everyday in thousands of studios.  8)

Mark
 
Biasrocks said:
Without getting into religious issues such regarding external clocking I would say yes.

What do religious issues have to do with anything? As far as I'm concerned the arguments here are purely technical.

Biasrocks said:
In addition to the standard rates (I go as high as 96kHz in my productions). I make use of the option in my WC generator to do vari-speed. Anyone coming from a world of analog knows how useful and valuable VSO can be in a production. Rarely, if ever would I go below say +/-1.5kHz.

Then for your application there's a problem. That's not meant as a judgement on value or validity, just a technical observation.

As far as I can see, there are three DIYable clock sync paths:

1) VCXOs with a simple PLL. Cheap, best performance, but limited to +/-100ppm tunability.
2) Wideband PLL (like in the Behringer ADA8000). Cheap, crap performance.
3) Wideband PLL coupled with ASRCs. Good performance (slightly less than option 1), more complex implementation, more expensive (Rochey's masters have nice ASRCs, but they cost more than the converters, and can be hard to get). A microcontroller is required to program the ASRCs.

Anyone see a fourth option?

JDB.
 
Back
Top