search

ika-musume / IKAOPM

A BSD-licensed YM2151 cycle-accurate Verilog core based on the die shot from siliconpr0n
BSD 2-Clause "Simplified" License
65 stars 5 forks source link

Modulation issue reported #18

Closed ika-musume closed 1 year ago

ika-musume commented 1 year ago

I can't confirm the details yet, but a user on Twitter reported a bug that seems to be having a problem with the modulation process. Will upload parameter settings.

ika-musume commented 1 year ago

The parameter settings of the reported tone.

OP0(M1), AR=31, D1R(DR)=5, D2R(SR)=0, RR=0, D1L(SL)=0, TL=21, KS=0, MUL=2, DT1=0, DT2=0, AMEN=0 OP1(C1), AR=30, D1R(DR)=18, D2R(SR)=7, RR=4, D1L(SL)=1, TL=1, KS=0, MUL=1, DT1=0, DT2=0, AMEN=0 OP2(M2), AR=0, D1R(DR)=0, D2R(SR)=0, RR=0, D1L(SL)=0, TL=127, KS=0, MUL=0, DT1=0, DT2=0, AMEN=0 OP3(C2), AR=0, D1R(DR)=0, D2R(SR)=0, RR=0, D1L(SL)=0, TL=127, KS=0, MUL=0, DT1=0, DT2=0, AMEN=0 ALG=4, FL=7, KON=0011

ika-musume commented 1 year ago

An OPAMP that supports the YM3012 in the X68000 is the MC3403, with a slew rate of 0.6V/us at ±10V supply. However, in the YM3012 datasheet, Yamaha recommends using the NJM4560 or an OPAMP with a slew rate ≥ 4 V/us. If the slew rate is too low, noise caused by excessive feedback levels of the M1 can be canceled.

The Twitter user who reported this issue played the same tune on an OPM standalone player equipped with the LT1804CS, and as I expected, the noise level increased more than the X68000 line output recording, and the number of local extrema of the noise matched a simulation result.

This suggests that the IKAOPM does not differ in behavior from the YM2151's digital output, but that the YM3012 DAC and passive components are attenuating some of the noise.

TODO: Still have to check the actual behavior of the YM2151, analyze the DAC circuitary, implement the x68000 sound filter in the long run

ika-musume commented 1 year ago

Another Twitter user sent me some tune samples using the actual YM3012, and I'm closing this issue because the characteristics of the modulation noise is exactly the same, even in tunes with high self-feedback levels. This difference is very likely due to the analog characteristics of the YM3012 and OPAMP.