Raspberry Pi -A New DIY'ers Digital Hub?

[scopeboy]

So, have you tried fiddling with TXTHR? If your current results are with it set to 1, then 16 could give a big improvement. Assuming your driver code knows how to put in the right number of samples which will vary between 48 and almost 64.

I2S is just a synchronous serial interface with data, clock and word clock. SPI has data, clock and chip select. They are similar enough that many uCs have a serial peripheral flexible enough to do either. The lucky BCM2835 has one of each :-)

[parisnight]

Originally Posted by philpoole

It transpires that it interrupts on an empty FIFO, 16 left, 48 left, or if not full - depending how you set TXTHR.

Don't know if the sample refers to bytes, sample (one channel), or frame (both channels) but in any case there isn't a lot of permissible latency.

At 44100 samples/sec, 2 channels, 16 bits
48 bytes is 362 microsec
48 frames is 1.45 millisec.

Not much leeway for the OS.

[philpoole]

I'm referring to samples. So I actually write 32 bits (2 samples) at a time.
I agree that is a lot of interrupts.
For a first drop of code, with flash memory storage (not USB) it's not bad.

[scopeboy]

Yes, I agree that 16 samples doesn't allow for much interrupt latency. I would want 512. I'm just trying to figure out how much interrupt latency philpoole has actually been using up until now. If the FIFO was set to interrupt on empty, then the OS would only have one sample's worth of time to respond to the interrupt. It would drop a lot of samples which would explain the "warbling" sound that was reported.

[parisnight]

Originally Posted by scopeboy

It would drop a lot of samples which would explain the "warbling" sound that was reported.

I think the only bugger was left and right channels were occasionally swapped. Maybe hang an LED or scope off a GPIO and light it if underflow bit set after interrupt. That would at least confirm underflow. But the cool thing is it basically works at 44.1/16 with a miniscule buffer and it is a legitimate driver. It's so tempting .. I've got some old TDA1543s somewhere.. humm but I also have so many unfinished projects..

[philpoole]

Yes, I did have warbling (probably mentioned in the RPi forum), and it was resolved by changing txthr to 1 (i.e. interrupt when only 16 samples, or quarter of a FIFO are left).
So, it's not there yet, but at least proves the concept.

[scopeboy]

Another report on my experience with the QNKTC USB DAC:

I got rid of the USB keyboard from my system. Instead I'm using some buttons connected to the GPIO port, with the "pikeyd" GPIO keyboard daemon by mmoller, and also the mpod remote app.

In this configuration, I get about one stutter per 5 minutes at 24/192. At 24/96 and below it seems fine. The project continues

The USB DAC contains a quite powerful microcontroller that is basically just converting USB to I2S, so it's not nearly as elegant (or cheap ) as using the Pi's own I2S port. However it is a working solution, and allows the Pi to be easily changed out for another single board computer later.

[parisnight]

Thanks for the update and GPIO tests through pikeyd. I admit I'm a Luddite and have resisted the latest gadgets that could be used to interface to the RPi mpd. The QNKTC DAC must sound good driven through the RPi in lieu of direct I2S. At 44.1/16 the ALSA driver could almost sound good if the jitter is low.

[merlin2069er]

I've got a rasperry pi with raspbian installed on it.

Thinking of setting up just a quick digital radio setup and dump my am/ fm radio.

I've only listened to a few AM stations anyways- what audio software would you guys recommend?

Hopefully it will recognize my tenor dac.

[merlin2069er]

Raspberry Pi Wifi Internet Radio Player | Fuzzy Logic Robots

hey, that's pretty cool.

I've got a lcd display kicking around, I'm going to hook that up to it.