Large Distance Optical isolation for DAC.

I want to optically isolate my digital section from the analog section by 2 inches.

I'll need 12 megabit for the stereo 192KHz, 24 bits. 2 Channels up(L/R & data), 1 channel down(clk). (using the CS43122 in I2S mode)

What's the cheapest way to optically TX 12 megabits? I've found Photo Pin diodes @ 25 cents which can do 40MHz easily, but most IR leds die at 1MHz. HP has a HEMT IR led which runs up to 15 MHz, but I do not want to rely only on HP for my IR transmitters.


(When I say cheap, I mean under 2$...)
 
If you want to isolate the two sections from each other's electrical noise, you may want to try differential transmitters and receivers (maybe some ECL logic parts?)... you can send the data streams down twisted pairs. That way you can achieve good timing without direct ground coupling. The differences between the ground potential of your two boards (caused by their relative separation and, presumably, single ground connection through a star-ground point) will not matter to the logic timing if there is sufficient CMRR at the receiving end. You may also want to consider the ISO-150.

Personally, I don't think it's really worth all the effort if the digital and analogue circuits are both resident in the same enclosure. In this case, it is my preference to put both circuits on a single PCB, with one continuous ground plane. This seems to go against much of the popular views on the subject, but it works... If you are careful in how you do your layout, you can make the quietest and most digital-noise-immune analogue circuit this way. The key is all in controlling ground return paths and EM noise radiating elements (eg ground plane splits). You must exercise careful layout, and decoupling of the digital ICs becomes extremely important. It's all about ground return paths at LF and HF, which are dealt with separately.

There was a good paper written about all this on the Audio Crafter's Guild website... anyone know wher ethis paper can still be found online?
 
--------------------
Don't do it...
unless the oscillator sits right next to the DAC without any optical insulation.
-------------------
capslock

Yes, the clock is on the DAC board. This is why I said there were 2 signals comming in (L/R) & data bits, and 1 going out (clock feeding back to digital side).


-------------------
If you want to isolate the two sections from each other's electrical noise, you may want to try differential transmitters and receivers (maybe some ECL logic parts?)... you can send the data streams down twisted pairs.
-------------------
Chad.

The ECL idea is good, except I'm trying to isolate everything, including RF & electrostatic. These guys will make it through the ECL, & any optocoupler & through data isolation transformers. If you want to know what kind of stuff this does to analog circuits, it's a different & new topic.

------------------
Personally, I don't think it's really worth all the effort if the digital and analogue circuits are both resident in the same enclosure.
------------------

Both sections are on different PCBs. They will both be coated with tuner stablant, the digital side will have a secondary coating of RF shielding epoxy resin, & to top it all off, both sections will be in their own double sided shielded compartment. (Like the ones on tuners.)

-------------------
You are fixing to go to a lot of effort for very little improvement. In fact, do it wrong, and it will be worse. All of the suggestions above are good, but I'm not fond of the ISO-150.
-------------------
Jocko

True. I'm making the extra effort to completely isolate everything. The idea here is that this USB2.0 6 channel sound card with built in analog pre-amp attenuator would not be affected by the fact that the PC equiptment is on 1 set of outlets from the fuse box, &, the analod side is conected directly to 3 PassLabs X350 amps, which are on a different line from the fuse box. Even though my USB sound card feeds will be balanced, the CMR of the Pass Labs amps is only -80db.

The idea here is to even better isolate the 2 sides than what would be possible with any AC 120v isolation transformers. Just that lousy ground pin alone from 1 outlet in my house to another reads between 2-8 vac, with ridiculous RF in it. This is due to my proximity to those mega huge power transmition lines & 2 AM radio station broadcaster antennas.
 
The AM interference is being created by the huge ground loop created by the wiring in the house. If I run my PC equipment from the same fuse line, the problem disappears. But, I am still left with lousy sound card performance.

One of my choices was to get a sound card with optical out only, &, on the AV side, get dedicated optical D/A converters with built in pre-amps, or D/A converter with a separate pre-amp. The prices here would be ridiculously out of my budget.


Here is how I plan to solve my problem with my home-made USB sound card:

-----------------------------------------------------------------------
The digital side is powered by the USB from the computer. It has a 16 Megabyte cache and only works at 192Khz x 24 bit by 8 channels.

So far, I am over the USB current limit by 190ma, I’m thinking of just cheating here.
-----------------------------------------------------------------------
The analog side has it's own 25+25v toroid for the analog & another 9v toroid for the digital parts of the DACs & reference crystal oscillator.

I'm using 4 Crystal CS43122 chips.

I'm still investigating the best I/V converter to use.

For the attenuation, I'm in the middle of figuring out the bets way to do it. Dual vactrols have caught my interest in that they supposedly behave exactly like a normal resistor without the mechanical noise & decay of a pot.
-----------------------------------------------------------------------

My friend, Mike, has agreed to make the drivers & SACD/CD/MP3 player for the project in trade for 1 unit. Sample rate conversion will be done by CPU. Not counting the analog I/V converters & attenuation, the project is currently at just around 180$, including professional PCB.

I’m also thinking of making the plans public here at diyAudio. But, unless, anyone here has an Altera Quartus license, they may not be able to program the boot prom. All the rest of the components should be easy enough to get.
 

Electro

Member
2002-04-12 4:41 am
Use laser diodes. They can transfer faster than 22 gigabits per second (about 2 terrabytes per second). Then use a very fast phototransistor.

USB version 1.0 will not give you 12 megabits. It is about 6 megabits. Use either USB version 2.0 or Firewire (IEE-1394).
 
I already found a good USB2.0 chip, except, Firewire offers a better supply current through the cable & TI's 400Mb/s 1394.a interface chip is only 2.04$

http://focus.ti.com/docs/prod/productfolder.jhtml?genericPartNumber=TSB41AB1

As for optical transmition, I found an 880nm IR led from sharp which has a 40MHz bandwisth. Price = 0.49$. As for photo transistors, they are too slow & the fast ones are too expensive. I'm going with photo-diodes. Typical 100 Mhz bandwidth photodiode costs 0.21$. +, with the transmition distance of 1 inch, just a resistor off of the photodiode is good enough to cleanly drive CMOS I/Os. Same goes for the IR emitting LED.

Try to find a laser-diode, or fiber optic emitter, that cheap.

40Mhz cheap LED!
http://smaecom1.sharpsec.com/sma/products/opto/data/pdf/en/oed/infrared_ed/gl496_e.pdf


My last headache is the analog side. I want a real pre-amp, non of this built in DAC pre-amp BS. I'm currently leaning toward a photo vactrol design.
 
Forget USB2.0, I found a 2$ 1394.a (fire wire) controler chip. Full duplex 400 Megabit. Made by TI. As for drivers, for now, I plan to use the port as a dummy com conection with the dummy com developement drivers from TI, &, Ill use my own CDPlayer software which will send straight to the com.


As for the vactrol design, I'm creating my own from scratch. I'm goind to work it out on a bread board this weekend. I believe I found the perfect way to absolutely compensate for all variations in the vactrols to always give you ferfect balance & perfect volume of 0 - 100%.

I think I'll start a new thread for this 1, under Source/Digital since this pre-amp will be driven from the balanced outputs of the Crystal DAC which I plan to use.
 
The best thing about 1394, is that you can draw 7 watts from the buss powersupply. Try doing that with USB.

For the all - digital side of my sound board, 7 watts is plentiful.

Items which need 2.5v to run:

2 x GL496 IR leds, 40 MHz BW - L/R clock & Data bits for I2S DAC
2 x cheap IR leds - sends controlls to the pre-amp DAC & I2C buss of audio dac.
Altera EP1K30 - PLD
Altera EPC1 - boot prom
TSB41AB1 - TI 1394/fir wire/i.link One port transciever.
MT46V32M4 - 128 megabit ram buffer.
--------------------------------------------------------------------
After the 1 inch opto-iso gap, everything here
is run off a separate supply, isolated from the PC.
--------------------------------------------------------------------
CS43122 - Crystal 122 dB DAC.
GPS grade 12.288MHz oscilator.
1 x GL496 - send clock back to the all digital side.
4 channel 16 bit I2C dac for attenuation/pre-amp control.
+ all of the analog stuff.
---------------------------------------------------------------------
(The digital side will be modified to fit another project, a BQ video grabber, but, that's for a diy video forum.)

My software player will be doing all the up-sampling to 192KHz. for cheaper, slower PCs, it can be run in triangular oversampling, for those of us with 1.7GHz & faster, there will be a true Fourier sampling transform with programable high frequency tapper off.
 
Here is how I did mine

I used a transformer to send the clock from the DAC to the CDP. I think 3E25 toroid core, 1.2 cm diameter, 10 turns oneither side.

And I used another transformer to convey the SPDIF back from the CDP to the DAC. SPDIF has no DC so it goes through transformers.

Of course the SPDIF receiver is connected directly to the DAC chip but I don't think it matters really. Most of the noise comes from the player and the transformer takes care of it.

It sounds so good.
 
If you insist on applying isolation, which for a lot of good reasons presented here, I would try to avoid, you may want to consider Analog Devices ADum1100. It is based on a mems transformer with signal conditioning. It will actually do 100mpbs, though it is more the jitter that is the issue. The data sheet does not have jitter specs that I can remember.

Alvaius