C64 SID test, differences between old and new version?
From: Asger Alstrup
Fredrik Claesson <e94 fcl@elixir.e.kth.se> writes:
>>>And All 6581s don't sound alike, this is especially apparent with the filters.
>> >>Oh, how do you proofed that it is the 6581 that is responsible for different sound and not the board? There are two ceramic capacitors involved for filtering and these capacitors' capacity varies about +-{$de}20%, which is likely the reason by any board sounds different. For testing I used 3 different SIDs in the same C64 board and the sound stayed the same.
>Actually, I did some testing a while ago, using several different SIDs in a C64 and a C128 respectively. Filter-specific sounds differed in no way when using a SID in the C64 and then switching it to the C128. But compared to one another, the SIDs sounded very different.
I have experienced differences in filter sounds of the type Andreas mentions (same SID in different computers), which are probably cause by the capacitors, but I have also experienced differences in the sound using different SIDs in the same computer. This was at a copyparty here in Denmark, where Link/Vibrants swapped his SID with another one which sounded much cooler - it had more bass and generally it had a more "full" sound.
Now, this was a long time ago, and thus before I started investigating the SID, so I don't know which revisions the SIDs were (sigh).
I only have one SID now (from 1982(!), I think), but I have noticed several features in the SID, that might be things to change in newer revisions:
When you trigger the test bit in $d404 bit 3, the waveform is supposed to reset as soon as the test bit is cleared. This reseting happens immediately when you use the triangle, sawtooth and pulsewaveforms, but with the noise- waveform, the waveform is not reset until several thousands of cycles later (this delay is different from time to time). This does not imply that the noise waveform doesn't output anything when you toggle the test bit, but rather that the noise waveform pattern does not start over until the above mentioned cycles. This means that you have to wait a few seconds after reseting the test bit before you start the noise waveform if you want to read the noise waveform from the beginning as I have done when I investigated it and discovered the underlying formula. I consider this to be a bug in the SID and I would imagine this is one of the things to fix in future revisions (and it shouldn't be hard to fix as it is only a matter of loading a 23 internal register with a fixed value).
Furthermore, interesting things happen when you mix waveforms such as triangle and pulse. The programmers reference guide states that the output should be the logical AND of the two waveforms, but this was be proven to be false (other statements of the appendix regarding the SID have also be proven to be false). The resulting waveform is rather complex and I have not been able to establish what happens. But I know that some newer SIDs can produce audiable output when using other waveforms than the $51 which is heard on all c64s I have encountered.
Known to most, is also the fact that a lot of the newer c64 can't play samples as audiable as the older c64s can.
Finally, the envelope is subject to irratic behaviour on my SID too. If I toggle the gate bit a lot in a few seconds, the resulting envelope value measure at a given time differs from time to time.
