Patrick, RobS, Hugo:

This one has had me puzzled since we on the BBS first became aware of it. On many players, the cs4231a chip mysteriously "dies" at some point. Hijack automatically works around this, but a dead chip means no mic input, and no visuals for AUX/Tuner.

Apart from that, it's pretty harmless.

But what, exactly, has died? All software accesses to the chip return 0x80 under this condition. Which just happens to match what the datasheet says the chip should return when it is not initialized yet.

The chip has a PDOWN pin, which can be pulled low to "power down" the chip, causing it to return 0x80, until the PDOWN pin goes high again. The datasheet says that this pin can be tied to Vcc (as done in the empeg) if that functionality is not needed.

BUT.. I wonder if it *really* wants a pulse on that line to wake it up regardless? In the empeg, the chip's clock signals come from elsewhere, and don't become valid for quite some time after power-on, and may jitter for a short period there.

What if, this cs4231a is normally just working by accident, and after a while as capacitors and the like degrade slightly, things change just enough that the chip no longer comes out of reset properly after power-on + clock-on ?

What if, it merely needs a PDOWN pulse to awaken it under those conditions? If this were true, then it could save a lot of possibly unneeded repairs.

I have at least one player here with a "dead" cs4231a still onboard, and I'm thinking of modding that machine so that I can pulse that line under software control.

Or maybe I'll just try connecting it to pin 42 of the DSP, which is given a proper low reset pulse sometime after power-up. This could be all that it needs.

...

BINGO!!

The "dead" cs4231a in my Mk2 player now works again!



Note that the PDOWN line (pin 18 of the cs4231a) is normally wired to Vcc underneath the chip, so one must lift that pin in order to perform this modification.

At the other end, the RST (reset) line (on pin 42) of the DSP can be used as-is.

Some blobs of hot-melt glue help keep the whole thing together.

This surefire fix really beats the hit-and-miss approach of completely replacing the chip.

Cheers!

Smart!

Awesome. Well done, Mark.

Absolutely awesome. FAQ entry updated.

Heh.. two keyboards, passing in the ether..
Tony of course has already just updated the FAQ!

-ml

Hijack has now been updated.

v484: change "Could not find CS4231A" message to "CS4231A: not responding", and add hint about the PDOWN fix from this thread.

v485: add a leading '*' character in the Vital Signs screen (top left corner) to flag players with cs4231a problems.

-ml

Mmm! Most interesting! Good bit of detective work Mark (fancy coming to work here?)

Looking at the datasheet, that would appear to be the right thing to do (though freakishly, their example circuit shows the pin tied low which looks rather counterproductive)... definitely a better option than swapping the IC, though it's rather strange that units appear to degrade over time.

It doesn't look like this should affect the IDE bus though, which is one of the other symptoms that people see when the 4231 is playing up. Anyone got a unit that's unwell and both the 4231 and the IDE drives don't respond that they can try this on?

Hugo

That PDOWN pin ought to have been labeled as a RESET pin, since that's how it really behaves for all practical purposes.

Yeah, I wonder if it's the internal power-on-reset circuitry of the chip itself that goes bad over time. But why, yes.. why.


When that happens, I'm not so sure that the problem is really (or only) with the 4231 chip. Basically, if *anything* goes wrong on that bus, then the ethernet, USB, 4231, and IDE can all suffer, and it's not completely clear which ones are actually bad.

We've had quite a few instances of this where it turned out to be a particular data pin (D5) of the ethernet chip that gets stuck (low).

But, yes, it would be useful to hear if there are still any unfixed units out there with IDE problems that have been attributed to the 4231..

I suppose one possibility there, could be that the semi-initialized 4231 chip could have just gotten into a weird state where it's driving one or more data pins permanently until it gets a proper reset.

Cheers