S/PDIF / AES3 switch and connector converter

[sg-o]

Hi,

I designed a circuit that switches 4 S/PDIF inputs to a single output and also provides 4 outputs from a single input. There are 2 Toslink I/Os, one RCA I/O and 1 BNC I/O.
I designed it to allow multiple S/PDIF devices to be connected to a single sink / source (in my case a digital mixing desk).

I'd be grateful for any suggestions on how to optimize the schematic. I'm especially concerned with the input and output sections of this design.

--SG-O

[sg-o]

I changed some resistor values of the coax input comparators. This should reduce the likelihood of the comparators resonating.

Here is the updated schematic and the PCB

[floobydust]

Why is PWR_FLAG going to U12, U13? Maybe an LED somewhere to tell me power is present.
I use optical or transformer isolation to a mixing console, otherwise noise can get in on the multiple grounds this box is tying together. I don't allow PC's (USB) GND to connect to a console ground, to keep SMPS noise out and avoid a ground loop.

[sg-o]

Why is PWR_FLAG going to U12, U13? Maybe an LED somewhere to tell me power is present.
I use optical or transformer isolation to a mixing console, otherwise noise can get in on the multiple grounds this box is tying together. I don't allow PC's (USB) GND to connect to a console ground, to keep SMPS noise out and avoid a ground loop.

Thanks for your  reply!
The power flag is only for ERC to indicate that a net is supplying power.
My plan was to add a LED to the enclosure. The power is taken from the Switch connector.
I'll add transformer isolation to my next version.

[Doctorandus_P]

I like to use RS485 transceiver for digital audio.
These have plenty of Uhmpf do drive a cable or an transformer, and on the other end they are good at signal recovery.

Also 100Mbit Ethernet only uses 2 of the 4 wire pairs of a Cat5 cable, and RS485 drivers are an excellent match for CAT5 cables.
I built an audio distribution network around this. One node puts digital audio on a CAT5 cable, and multiple other nodes can use it to make the audio audible again. They all have the same data, so no messin' with synchronisation etc.

[sg-o]

I like to use RS485 transceiver for digital audio.
These have plenty of Uhmpf do drive a cable or an transformer, and on the other end they are good at signal recovery.

Also 100Mbit Ethernet only uses 2 of the 4 wire pairs of a Cat5 cable, and RS485 drivers are an excellent match for CAT5 cables.
I built an audio distribution network around this. One node puts digital audio on a CAT5 cable, and multiple other nodes can use it to make the audio audible again. They all have the same data, so no messin' with synchronisation etc.

Thanks for your response!

My current design is loosely based on a application note by ST (https://www.st.com/content/ccc/resource/technical/document/application_note/group0/c5/67/84/35/7e/4c/4c/19/DM00431633/files/DM00431633.pdf/jcr:content/translations/en.DM00431633.pdf)

I used the comparator receiver design as I don't have much experience using a differential line transceiver in this use case.
The buffer based output stage is sufficient in my opinion as it provides 24mA source / sink capability and is simple and inexpensive to implement.

This SPDIF splitter won't connect to devices that are more than 2m away from my mixing console (which I use like a sound card with lots of IO). This makes using Cat5 cable a bit ridiculous.

[floobydust]

I'm not sure why the ST app note has such poor waveform quality, it looks like breadboard quality in some scope traces 

2.2.3 "In this case hysteresis (implemented, for example, by a positive feedback loop) is not recommended as it could influence the distance between the signal edges."
But you have it in your design. I would not bother with the comparators as the signal is strong at 0.2-0.6Vpp and I'm not sure if fidelity gets affected. It seems jitter issues fall into audiophoolery, cable voodoo and long winded arguments. Most SPDIF designs are using 'HC04 gates as the receivers.

For 192kHz (12.28MHz) I would be adding resistors at the IC outputs to lessen transmission line effects with your PCB layout. I start with 33R, you can put in 0R if you want but I leave a provision for termination resistors, usually they are needed to stop ringing and overshoot from being a problem.

Other people here know more and probably suggest how to keep the waveforms crisp.

[sg-o]

I'm not sure why the ST app note has such poor waveform quality, it looks like breadboard quality in some scope traces 

2.2.3 "In this case hysteresis (implemented, for example, by a positive feedback loop) is not recommended as it could influence the distance between the signal edges."
But you have it in your design. I would not bother with the comparators as the signal is strong at 0.2-0.6Vpp and I'm not sure if fidelity gets affected. It seems jitter issues fall into audiophoolery, cable voodoo and long winded arguments. Most SPDIF designs are using 'HC04 gates as the receivers.

For 192kHz (12.28MHz) I would be adding resistors at the IC outputs to lessen transmission line effects with your PCB layout. I start with 33R, you can put in 0R if you want but I leave a provision for termination resistors, usually they are needed to stop ringing and overshoot from being a problem.

Other people here know more and probably suggest how to keep the waveforms crisp.

I now have added pulse transformers (DA102JC)  To all Inputs and outputs. I'll look into differential transceivers as inputs over the next days.

Adding Resistors to the IC outputs is a good idea and I'll add them to the next version.