Originally Posted by edorr
I think we are overcomplicating this. When I did my proof of concept, I used a short $5 digital coax to XLR cable (made by Lynx). It worked beatifully and sounded fine. I also tried an adapted (probably the canare). It is restricted to 48/24 and it sounded like crap. I currently use the z-system switcher/converter (I switch multiple digital inputs into my Trinnov) which works fine as well. However, for a single source system (i.e. Oppo -> Trinnov) the short runs coax to XLR (S/PDIF to AES/EBU) will work OK. Curt @ Trinnov also used them.
Great to see the discussion on modded Oppo w MC2. I use it here.
I concur with Edorr. I've used the Canare transformers, spec’d at 6mHz bandwidth, are intended for passing 48k digital audio in broadcast applications, not more. Broadcast facilities have lots of 75 ohm running around, so the Canare adapter made moving digital audio very easy. I have four of them here. They are useless above 48/24 audio and indeed sound terrible at 96/24. Give it a try if you like. Someone can buy mine.
SPDIF to MC2 AES in
Trinnov has high bandwidth input isolation transformers (good for 192/24) with the standard AES/EBU 110 input impedance, and Trinnov has designed in significant input sensitivity specifically to accept SPDIF signals. The key is to use short cables. The higher the sample rate, the shorter the cable, and importantly, the less reduced jitter.
What's going on?
The easiest way to describe the impedance mismatch issue is that when impedance is ideal (seldom accomplished, usually some error of low magnitude), the signal flows from end to end with no reflections. If there is an impedance mismatch- from any source, be it a connector, the cable, or the terminations (the end source, in this case 75 ohm SPDIF, or the end = AES 110 ohm), there is a reflection caused. The reflection of the signal then gets mixed with the original signal, causing distortion. Long cables are best served by using an active impedance converter that has the bandwidth required, and short cables doesn't matter. Whats short or long? The typical rule, applied to radio frequencies, is that impedance mismatches don't matter at 1/10 the wavelength or less. For 6 Mhz, this would be 150m. As digital audio is square waves, the rule of thumb is 10x, so this would become 15m, still a considerable length. Of course, we're all looking for excellent performance, so the shorter the better. I believe the best route is to go well inside these guidelines. One can get away with 6' (2m) interfaces, but I prefer 1.5-3’, as it just means less jitter.
MC2 AES > SPDIF
For applications such as an outboard converter. Caution! The Impedance mismatch is easily handled, but the voltage difference is an issue. Where one drops voltages (SPDIF> AES), here you have AES (higher voltage) feeding an input that is easily over-saturated. The input circuit could be damaged by the mis-match. The easiest way to do this is with a resistor bridge that corrects for both impedance and voltage placed at the SPDIF side. Please PM me if you need an example of such a design.
Anyone done any AP measurements? I'd be interested it that...