GrüsseOK the plot thickens even further....
I was playing around with it, scope connected, measuring the delay. I set the sample rate to 44.1 to check that the delay was still one sample, which it was. Fiddling about some more and suddenly the DAC started outputting the left and right channels with no delay. Wow, that is strange I thought. So I powered off and on, and it went back to the delayed mode.
Some more playing about and I figured out what I'd done to make it work correctly.
Now let me back up a little and explain how this DAC is designed. It has a bunch of inputs, 4 x SPDIF (2 RCA, 1 x BNC, 1x toslink). Plus mine has 2 x I2S (some have USB input card). All these inputs convert to I2S, and they are switched. Then they go into this big DSP chip that does oversampling, etc. Then onto the DAC chips.
I found that the DAC has two 'modes' of operation - the 'faulty' mode where one channel is delayed one sample. And the 'correct' mode where channels are not delayed.
When it is powered on, with I2S input selected, running at 192kHz, the dac goes into FAULTY mode. If I select a different sample rate on the PC and play something, it switches to CORRECT mode.
But CORRECT mode is fragile. If I switch to another input it goes back to FAULTY mode. Then it needs the sample rate changing again before it goes into CORRECT mode. (The setup I have to drive this is HQplayer into Acourate Convolver, output via USB to Gustard U12 USB convertor. This outputs both SPDIF and I2S, so I can use either output. But I can change sample rates easily using HQplayer)
This applies to all inputs, so suppose I have got it working in CORRECT mode and switch to SPDIF input 1, it goes back to FAULTY mode. If I then change sample rate on the SPDIF input, it gets into CORRECT mode on the SPDIF input.
If I then switch back to I2S mode, it again goes to FAULTY mode.
The last thing I found before I went to bed - if you have SPDIF input selected with 44.1 sample rate, and power off/on, it does come up in CORRECT mode. But will revert to FAULTY mode if you change inputs.
To me it's clear that there is a bug in the DSP software. It's a pain in the ass having to change sample rates manually every time I change channels.
I think you'll agree that this is a pretty weird bug...no wonder it took some time (well, a couple of days) to figure out what was happening.
Thanks for your help and ideas! Not sure I have got to the bottom of this - there may be other combinations of things that affect this.
Other answers :
-- yes I measured directly at outputs - single ended outputs
-- Audio - GD master 11 (not NOS 11)
-- I had all the jumpers off, which is factory default - 8x oversampling, etc.
Uli