I've got a pair of GI AY-3-8910 chips, and in both cases, I'm having trouble getting the noise generator to work (acceptably, anyway). The tone generators sound great, and I can specify amplitude and envelopes successfully, etc. But the noise generator just produces a very faint hiss and maybe a few pops, and there is no discernible difference in the sound of the noise through the entire range of values (0-31).

For audio output, I'm using the circuit from the user's manual (http://f.rdw.se/AY-3-8910-datasheet.pdf, Fig. 1). I've tried other circuits as well, based on projects others have posted online, but the result is always the same. I'm using a 2 MHz TTL oscillator for the clock, and I've verified it's outputting a clean 2 MHz signal.

I've definitely enabled the noise output in the mixer and have the volume up. I've tried using headphones, an amplified computer speaker, and also an amplifier board with attached speaker for output.

I bought the chips from (I think) a reputable retailer: https://www.bgmicro.com/1B1576.aspx.

Anyone have any insight?

The second page suggests using a 1MHz signal rather than a 2MHz one, specifically for improving the noise output. But it seems like it should just be shifted up an octave from the normal range.

That multiplexed bus looks very inconvenient to interface to a 6502…

Thanks for the suggestion. I did try a 1 MHz oscillator as well, but no luck.

For anyone that may come across this in the future...

I was able to get the noise generator to work correctly by simply resetting the chip (which sets all registers to 0) before running my code. I'm still not sure *why* this worked, because I set all relevant registers to the values I wanted, and the remaining registers were irrelevant/unused in this scenario...but in any case, it solved my problem.

There are probably internal registers which the reset signal reaches but don't have an address on the CPU bus. One of these is probably the RNG for the noise generator. It's common for them to be designed in a way that requires them to be non-zero to function - but on power-up they might be zero.

I think you're right. I wrote a routine that would zero out all addressable registers and that doesn't do it. Only bringing the reset pin low gets it working.