The mainboard for the AD/DA-Box

Hello,

it is not the badest idea to open a new thread for discussing the mainboard for the AD/DA-Box, is it?
So here we go.
First of all the output connectors. The idea is to use D-SUB25 (Tascam-Format) for the analog i/o. IMHO this is the most economic solution for up to 64 i/o. But I know there will always be somebody how only needs 8 i/o but via 6.35mm balanced jacks or wants to connect the XLR of a microphone directly. Therefore, the DSUBs are placed on a break-away part of the pcb and connected by IDC cable/socket combination. If you want to use another connector then DSUB just remove that small pcb part and connect your favourite connector via the IDC socket (replacement boards with XLRs or Jacks can also be made).
Remark: I think if you look carefully on some pictures flying around on the internet from the Metric Halo LIO insight you can see a similar solution.

On the mainboard itself the modules are plugged into sockets that are connected to the front-end IDCs for I/O.

I attached a drawing of this. There is also some RFI filtering and a jumper for shield-to-ground connection on the break-away part.

Any thoughts about it?

R.

Rather than using the short ribbon cable jumper, perhaps consider using stacking board-to-board connectors (Samtec has a bunch of different types).

You have a female type on the main board. You still have the breakaway holding the DB-25s for the TASCAM format output. You make a little board with the XLR or 1/4" or whatever, and on the bottom of that board you have the mating male header which plugs into the main board.

Bonus would be to add mounting holes to the main board and matching mating holes on the XLR board, and put a threaded standoff between the two boards and screw them all together.

-a

Would you eventually offer the boards for this system built up or unpopulated? I imagine SMD components would be a turn-off for many a DIY builder...

ramshackles said:

Would you eventually offer the boards for this system built up or unpopulated? I imagine SMD components would be a turn-off for many a DIY builder...

Click to expand...

Which is sad, because with a magnifying glass, they're so much quicker than through hole.

Andy, which Samtec's do you mean?

Besides that, I'm now thinking: Does it make sense to have a computer interface to transfer digital audio? Something like USB, Firewire or Ethernet?

Ethernet: I don't like it because I could not find a nice solution for DIY. Either you need a lot of programming or unless you order 1000 pieces and you're a company they don't talk with you.

Firewire: Benefit is that MAC OS supports the DICE chips out of the box and there is a good support. Can be done DIY-friendly. But it seems that Apple decided not to continue with Firewire. Already on my MacBook Pro you can only have it via an adapter on the Thunderbolt-port and I've some stress with it because the adapter gives some trouble because contact is lost if you move the cable only a little bit.
The new Mac Pros will although have Firewire only via Thunderbolt.
Firewire can be used with Linux, too. About MS Windows I don't know. I'm not using it.

USB2.0 and above: Currently I prefer it most because it seems to be the future on Apple and Windows machines. Linux does support it, too. DIY-friendly solutions are available. If it supports UAC2 it runs via Plug'n Play on MACs. Again Windows I don't know.

What do you think?

R.

rkn80 said:

USB2.0 and above: Currently I prefer it most because it seems to be the future on Apple and Windows machines. Linux does support it, too. DIY-friendly solutions are available. If it supports UAC2 it runs via Plug'n Play on MACs. Again Windows I don't know.

Click to expand...

Probably USB2 / 3.  UAC2 will be plug and play, not only on Mac, but also Linux via ALSA/JACK. Also don't know how Windows handles UAC2.

But have you looked at Ravenna?  If I understand correctly, this is an open-source open-standard technology for audio over standard ethernet.  Merging Technologies have implemented this in their high-end Horus interface http://www.merging.com/horus

Yes I know Ravenna and I think it is a cool thing. So far I have only seen papers describing only the principle of operation => open standard. But I have never seen something like an evaluation board schematic or an IP-core for a FPGA. While I'm not scared with programming this stuff, the work load is too high. I can not do the software development and the hardware as a free-time project...  :-\
Therefore, I tend to solutions that are more ready-to-use.

R.

rkn80 said:

Andy, which Samtec's (sic) do you mean?

Click to expand...

Dunno, really -- depends on the density/pitch of the connection. And what you can buy easily.

-a

Hi,

I have some bad news. The firewire solutions leaves the party. Profusion now lists the DICE chips as obsolete.
I don't know any other easy-to-use solution that you can buy for diy. If somebody else knows something, let me know.

USB2: It turned out the XMOS has a easy solution. Only the chip is the problem. For a multi-channel solution you need a chip that is only available in a QFN package with pads you can not touch with a solder iron. You need a reflow oven. Somebody having one and willing to do the solder work?
But it seems that you can connect some xCores together and make them working together. The smaller ones have TQFP packages...
I'm trying to find out wether that is a doable solution.
Any other suggestions regarding a USB2 solution is welcome.

R.

I believe Rochey has a pick and place machine? Maybe he'd be willing to do a run of board with the QFN chip soldered on.

You can always reach out to a contract manufacturer to populate certain components if the volume is enough.  Some smaller shops, especially if you can find one interested in audio, might be willing to do something at a reasonable price for a short run of boards.

Or there are quite a lot of hobby builders using toaster oven reflow or homemade hot air, though those can still be hit or miss with high precision stuff.  One of the hardest parts is if it has a ground pad on the bottom (not just the edge lead pads) that needs to be connected.

rkn80 said:

USB2: It turned out the XMOS has a easy solution. Only the chip is the problem. For a multi-channel solution you need a chip that is only available in a QFN package with pads you can not touch with a solder iron.

Click to expand...

I thought they had devices in QFP which would do the trick.

Any other suggestions regarding a USB2 solution is welcome.

Click to expand...

This'll get ugly real quick, but:

a) Pick yr favorite ARM processor with an on-board High Speed USB interface and an external memory interface (NXP LPC18xx, Atmel SAM3U). Use a small FPGA (Xilinx XC3S50AN, or one of the small Lattice parts, perhaps) and bolt that to the ARM's EMIF. Write some code.

b) Use an Cypress EZ-USBFX2 part and connect a small FPGA to its Slave FIFO interface.

I think that Cypress should rot in hell for all eternity, so I would avoid option b.

-a

What are the channel count limitations on USB2? Between that and lack of remote deployment possibilities, I've personally never found it to be very compelling.

grantlack said:

What are the channel count limitations on USB2?

Click to expand...

Given the wire speed of 480 Mbps and 48 kHz sample frequency and 24 bits per sample, a couple hundred channels.

But basically the limitation is in the host computer and not the interface itself. It runs up against that sort of limit with FireWire, too.

Between that and lack of remote deployment possibilities, I've personally never found it to be very compelling.

Click to expand...

What do you mean by "lack of remote deployment possibilities?"

-a

grantlack said:

What are the channel count limitations on USB2?

Click to expand...

XMOS specifies up to 40channels and I think that is a limitation of their software.

XMOS has devices with QFP packages that you can diy. The problem is then not the processor's computational power. There are simply not enough i/o pins for ~32 i/o + clock lines via i2s.
Two options then:
1) go via TDM and let a small FPGA translate it to i2s. 
2) XMOS devices can be easily slaved to a master processor via a link port. By doing this you build a multi-core system that looks like one big processor to the software. But I could not find out yet how that works.

I would prefer 2) because 1) introduces another device that needs to be programmed. And not all devices I've used so far for the modules support TDM=>you need a translation.

Andy, I looked into an Atmel solution, too. There is the audio widget project for the AVR32. But I don't really want to go there. As you said you need then an extra FPGA and a lot of software needs to be developed. Besides that if USB2.0 is used I would like to go via USB Audio Class 2 because no driver is needed then, but how wants to program all the UAC2 stuff? That's why I prefer a ready-to-use software library.

R.

rkn80 said:

grantlack said:

What are the channel count limitations on USB2?

Click to expand...

XMOS specifies up to 40channels and I think that is a limitation of their software.

XMOS has devices with QFP packages that you can diy. The problem is then not the processor's computational power. There are simply not enough i/o pins for ~32 i/o + clock lines via i2s.
Two options then:
1) go via TDM and let a small FPGA translate it to i2s. 
2) XMOS devices can be easily slaved to a master processor via a link port. By doing this you build a multi-core system that looks like one big processor to the software. But I could not find out yet how that works.

I would prefer 2) because 1) introduces another device that needs to be programmed. And not all devices I've used so far for the modules support TDM=>you need a translation.

Click to expand...

I think once you start thinking about adding stuff to an XMOS design, then you start thinking about replacing the Xcore processor with something else. I think the XMOS parts are pretty cool, but I haven't -- at least in my limited scope -- seen a compelling use for one. That said, they ARE popular for High Speed USB 2.0 audio designs for companies where a BGA isn't a problem.

Andy, I looked into an Atmel solution, too. There is the audio widget project for the AVR32. But I don't really want to go there. As you said you need then an extra FPGA and a lot of software needs to be developed. Besides that if USB2.0 is used I would like to go via USB Audio Class 2 because no driver is needed then, but how wants to program all the UAC2 stuff? That's why I prefer a ready-to-use software library.

Click to expand...

Honestly, I know nothing about the AVRs, so I can't compare an AVR32 with an ARM Cortex-M. My guess is that the two families are fairly similar in features and horsepower, so it comes down to tools (compiler/debugger) and support and whether a specific processor has a specific feature you need.

It seems like in either case you still need the FPGA to translate the processor's external interface bus to multiple I2S and/or ADAT (or MADI) links.

I prefer the ready-made software (ARM firmware) solution. For me, the FPGA is the easy part.

One caveat about USB Audio Class 2.0 (which is separate from the larger USB 2.0 spec) -- Windows still (even in Windows 8!) doesn't have a native driver. (NB: OS X has supported it since 10.5!) There are a handful of companies which sell a driver but for DIY that's a complete non-starter.

-a

Andy Peters said:

One caveat about USB Audio Class 2.0 (which is separate from the larger USB 2.0 spec) -- Windows still (even in Windows 8!) doesn't have a native driver. (NB: OS X has supported it since 10.5!) There are a handful of companies which sell a driver but for DIY that's a complete non-starter.

Click to expand...

I looked into those XMOS parts a little and that's the part that had me wondering.  I've looked into options for audio interfaces from time to time and the driver side always seems to be the least available as a ready to go solution, and it's also the part I know the least about. 

Are there any good options for professional multichannel audio without full custom drivers?  Cheaper bought solutions for lower-volume production in a one-man operation scale of capital and room for ROI (not DIY friendly, though, but still of interest to me)?

I'd say let's go for BGA package and solder the chips in plant that will be doing PCBs. This will add to the cost, but benefits of using one chip solution AND ready made firmware are well worth few extra bucks.

mattamatta said:

Are there any good options for professional multichannel audio without full custom drivers?

Click to expand...

Use a Mac, and USB Audio Class 2.0 drivers are provided.

Petition Microsoft, and ask them when they'll provide this sort of support. Don't hold your breath.

Otherwise, I'm not sure if there are any other options.

-a

just want to say - SMC's don't scare me - im about to buy a hot air rework station - they're around 150 USD
If learning/buying a hot air station was all that stood between me and DIY (quality) DAC - sign me up!

question:
If we were to say use the same firewire chip as a commercial manufacture - could we use their software for the interface?