Hi,
Ar2r4eg gave this EWS64 board to me for diagnostics.
So, what i tried to do:
- found out that the signal when playing wavetable MIDI originates from the SAM9407 chip in the form of a digital line I2S (Philips serial protocol for audio), specifically from pin 48 - DABD1 (data). Also, with SAM9407, a clock and a WS signal (word select) for switching the right / left channel are output. And all these three digital signals go straight to the DAC TDA1314T. The DAC output is connected through 10μF electrolytes to the eighth pins of the TEA6320, one channel to one, the second channel to the other. I didn't track it further, because the oscilloscope at the DAC output shows a signal with an overload, the same as heard in the speakers.
- connected a Saleae logic analyzer to the I2S bus and captured ~ 30 seconds of playing the DOOM E1M1 MIDI file, which is output from the SAM9407 chip to the DAC. Here is it: https://yadi.sk/d/qiCdU6369Ox56g (Saleae logic analyzer format *.logicdata). It seems that it was approximately possible to convert the data captured by the logic analyzer from the i2s bus into a regular .WAV file, here it is:
doome1m1.mp3
- checked the clocks - the MK1413S synthesizer-generator is correctly controlled from the mixer, when you switch between 32KHz, 44.1, 48KHz - it gives out the necessary frequencies according to the datasheet, depending on the states of the S0 and S1 inputs.
- Memory .. Soldered another DRAM KM416C1200AJ-7 chip removed from one of my SIMMs. The behavior has not changed - the midi hiss in the same way. Soldered back KM416C1200BJ-5 as it was. That is, it's not about memory...
- Sketched connection GAL or PAL, which with a sticker 172 000 832. It turned out something like this:
P1130911.jpg
From this it follows that it is mainly needed to service the SIMM slot, in much the same way as the GAL in the SIMMConn adapter for AWE64, a similar principle of forming the necessary / RAS2, / RAS3, / CAS0, / CAS1 for the SIMM slot. It does not participate in the life of onboard memory, except for the / CAS signal passing through it into the combined / UCAS + / LCAS signals on the DRAM chip itself. I looked with an oscilloscope, this signal corresponds to the input to the GAL and the output, that is, it simply buffers it, as it should be judging by the scheme for connecting a simple DRAM to the SAM9704. It turns out that suspicions are also removed from this PLD.
- Some addresses are buffered through the two 74ALS244, I checked them too - they work as expected.
- Found a short description of the 'Programming Manual' on SAM9407, tried to write a simple Pascal program that would read the contents of AWS64 memory into a file. I thought everything seemed to be fine, at least at the beginning the firmware is located, then the name of the bank file and the samples went ... I tried to write my data after the firmware, well, just the constant 0x5a5a at all addresses and then read it back to the file - everything is fine exactly until 1MB (offset 0x100000), and then there is some garbage in the read file. I guess I made some mistake in the program. If someone has a similar EWS64 XL soundcard, please try running my program and send me the dump file it creates, please.