You need to populate C18, R21, R22, CB2 and T1 to get a proper Coax SPDIF output capability.
C18 and CB2 values are not critical, both can be 0.1µF. R21 and R22 depend on exact pulse transformer used, but likely in the vincinity of 75 ohm.
If you need Toslink you need to find some 5V transmitter, wire up power and get TTL SDPIF signal for its input from the lower pad of C18.

Tiido wrote on 2024-07-02, 06:07:

You need to populate C18, R21, R22, CB2 and T1 to get a proper Coax SPDIF output capability.
C18 and CB2 values are not critical, both can be 0.1µF. R21 and R22 depend on exact pulse transformer used, but likely in the vincinity of 75 ohm.
If you need Toslink you need to find some 5V transmitter, wire up power and get TTL SDPIF signal for its input from the lower pad of C18.

Coax SPDIF will do just fine. What kind of pulse transformer it might be? I guess it's indicated with a big square next to SPDIF unsoldered connector in the top right corner, correct?
If you mentioned optical by Toslink I wonder where do I get TTL signal? What kind of 5v transmitter? Unfortunately I'm not familiar with this sort of electronics. I know basics but not this particular topic, it is out of my knowledge. I'm not familiar with digital devices. Analog mostly.

I also wonder what U3 and U7 also not present on the PCB could be and what its purpose.

For all intents and purposes this is all analog domain stuff. You got 5Vpp and ~6MHz squarewave signal that needs to end up as galvanically isolated ~1Vpp 75ohm impedance signal with as rectangular shape as possible. There are dedicated SPDIF pulse transformers, and this card likely used one of them. I don't know any of the part numbers that would fit the existing footprint. I did mention the exact spot where the signal comes out of...

U3 is an EEPROM, for holding conf info. Perhaps to hold PCI subsystem ID for OEM solutions or something.
U7 is alternative to U4, you use one or other depending on parts at hand.

EDIT: Fixed SPDIF signal level, it is 1Vpp in the coax.

Tiido wrote on 2024-07-04, 19:59:

For all intents and purposes this is all analog domain stuff. You got 5Vpp and ~6MHz squarewave signal that needs to end up as galvanically isolated ~0.5Vpp 75ohm impedance signal with as rectangular shape as possible. There are dedicated SPDIF pulse transformers, and this card likely used one of them. I don't know any of the part numbers that would fit the existing footprint. I did mention the exact spot where the signal comes out of...

well my guess is that signal goes out of a rivet hole near J19 from the back side of the PCB and goes to C18 which must be filtering out DC component. R21 should attenuate somehow transformer input. CB2 probably cleans out EMI or something that may appear in transfrormer windings. R22 adopts what is left to SPDIF standart. Just guessing. need some research of dedicated pulse tranformers that may fit

Galova wrote on 2024-07-07, 20:26:

Tiido wrote on 2024-07-04, 19:59:

For all intents and purposes this is all analog domain stuff. You got 5Vpp and ~6MHz squarewave signal that needs to end up as galvanically isolated ~0.5Vpp 75ohm impedance signal with as rectangular shape as possible. There are dedicated SPDIF pulse transformers, and this card likely used one of them. I don't know any of the part numbers that would fit the existing footprint. I did mention the exact spot where the signal comes out of...

well my guess is that signal goes out of a rivet hole near J19 from the back side of the PCB and goes to C18 which must be filtering out DC component. R21 should attenuate somehow transformer input. CB2 probably cleans out EMI or something that may appear in transfrormer windings because it goes to a ground plane obviously. R22 adopts what is left to SPDIF standart. Just guessing. need some research of dedicated pulse tranformers that may fit

By the way. About pulse transformer. Is it 1:1 or step-down?

Probably 1:1, but you can use other stuff by adapting the resistors.
This looks like it might fit the footprint : https://no.mouser.com/ProductDetail/Pulse-Ele … yrwjtKK7g%3D%3D

Tiido wrote on 2024-07-07, 20:59:

Probably 1:1, but you can use other stuff by adapting the resistors.
This looks like it might fit the footprint : https://no.mouser.com/ProductDetail/Pulse-Ele … yrwjtKK7g%3D%3D

Oh thanks, should take a look at it.
I think I need to take oscilloscope and check if there is any signal at all. Who knows.

I'm not sure what is exactly going on in this picture but I found it while surfing on this subject.

That particular mod seen puts 5Vpp directly on the coax and there are devices that will die in such a condition, such as ESI Juli@ on which I actually fried the SPDIF input this way and had to replace a chip (and conjure up a protection circuit too). Coax has near 1Vpp level on it and devices don't necessarily handle beyond that. Other aspect is that the logic signal that outputs the signal is not meant to drive a 75ohm load and can get damaged from it also. The very least it will need a series resistor of about 330 ohms so that the 75ohm in the receiving end forms a divider that gives close to expected 1Vpp and also reduces load on the logic output on transmitting side. But since the impedance is not quite right, longer cables are probably not gonna produce an output, nor will there be galvanic isolation so there's ground loops potential and that can manifest in variety of ways too, including the SPDIF connection itself not functioning.

I have read that SPDIF carrier frequency is somewhat around 3 MHz and 1V peak to peak.
If the signal generated by yamaha chip is 5Vpp then 1:1 isolating transformer doesn't meet specification, as it should be stepdown 5:1 this case. Output impedance should be 75R as well as transmitting coax cable impedance. Somewhere inside of SPDIF recording device input terminal should be at 75R as well.
Forgive my ignorance, I may be wrong.

As for the picture above I don't think one can simple throw away the transformer and use such connection as it may be unsafe for receiving device. I don't think galvanic isolation is kept in mind here just for lulz

The frequency depends on sample rate. SPDIF has 64 bits per sample in each data frame which is then manchester encoded. I had misremembered that it doubles the frequency but it can work inside existing frequency since it only changes when the bit flips within the same clock period. YMF724 has locked 48000Hz sample rate so the output is 480000 * 64 = 3.072MHz. But yes, 1Vpp into 75ohm load with 75ohm line impedance.

Ideally there's a voltage divider before the transformer as I see in many application manuals of various digital audio chips but this particular card seems to just yolo it and rely on whatever winding resistance and those external resistors produce. Impedance will be wrong even if the output level is right, but unless you have long cables it won't really matter. This is not a Yamaha made card or even a reference design, if there's a reference design it is Yamaha Waveforce 192XG sound card. That one doesn't even have the SPDIF exposed on it 🤣.

Tiido wrote on 2024-07-08, 16:57:

This is not a Yamaha made card or even a reference design, if there's a reference design it is Yamaha Waveforce 192XG sound card. That one doesn't even have the SPDIF exposed on it 🤣.

The Yamaha Waveforce 192D has an optical out although its a different pcb layout

Funny thing is For that Labway card they list the spdif as an option but I have never seen a spdif actually populated on the a301-g50.

I didn't even know there was a 192D, but I found a VOGONS port about it with photos : Re: Question about the Yamaha Waveforce WF192D SPDIF.

Optical output is easy, you just wire up power and the TTL signal to the transmitter module and you're done, no transformers or anything else to worry about.