Digital audio sample rate converter: S/PDIF for N64, MV1C

The place for all discussion on gaming hardware
L-Train
Posts: 29
Joined: Wed Jun 11, 2014 6:09 am

Re: Digital audio sample rate converter: S/PDIF for N64, MV1

Post by L-Train »

NewSchoolBoxer wrote: Thanks for including the BOM. I see the CS8406 your, FirebrandX's mod and older version of borti4938's mod still for sale was obsoleted with an end of life notice 12 months ago. Not surprisingly of stock.
Last year's EOL was for the SOIC version of the CS8406 and I thought we'd be safe with the TSSOP version used in my mod, but unfortunately the day after your reply Cirrus Logic EOL'd the TSSOP version as well. If anyone's planning on building one of these things please place your orders for the CS8406 while they're still available. I plan on redesigning the board to use the DIT4192.
NewSchoolBoxer wrote: I assume the [more expensive] $10-12 USD CS8416-CZZ is the intended replacement.
The CS8416 is a digital audio receiver and isn't suitable for this application.
NewSchoolBoxer wrote: --Can you measure the power consumed and ambient operating temperature at the highest resample rate or is that not meaningful?
This is something I was meaning to measure but forgot to do it. Thanks for reminding me! The next time I have one on the bench I'll try to remember to do this.
NewSchoolBoxer wrote: --Is the mod in standby mode or something akin to that when Super Game Boy or MSU-1 audio is running?
No, there's no standby mode and it continuously resamples and outputs whatever is produced by the APU.
NewSchoolBoxer wrote: --Can you confirm if 1CHIP has inferior audio to 2CHIP? I'll test this myself but I'm biased in thinking 1CHIPs are cheap cost reduction machines with the SNES Jr/Mini being almost a clone console. Important distinction when 2CHIP digital mods are much more difficult.
plgDavid has done extensive measurements and his video about the research behind chipsynth SFC indicates that the digital audio they produce is practically identical. I did a quick check too, just a simple inversion test with a regular SNES with a DIT4192 and a SNES Jr. with a CS8406. Due to the differences in state at boot (explained starting at 10:42 in the video) the cancellation isn't perfect, but there are sounds that are cancelled completely which implies that they can be bit perfect.
NewSchoolBoxer wrote: --Can I ask what we refer to in American English as the 800-pound gorilla in the room question? At sub-CD audio with arguably high noise floor, which is to say SNES and N64, is the digital reduction in noise significant enough to make any audible difference? I was discussing that with Ikaruga11 and I predicted definitely not for 3.56:1 audio compression ratio, expected 32 kHz DAC SNES. Maybe yes for N64 from the rare game coded to use 44 kHz sampling. By extension, I think going up to 24-bit, 192 kHz sampling, 24 Mbps is silly but not going to hurt anything.
By digital reduction in noise do you mean the difference between the analog audio produced by the NEC DAC inside the console versus a pure digital signal? If so, the most audible differences are the elimination of 50/60 Hz hum and 15.7 kHz interference, and the lack of low pass filtering. A lower noise floor can be audible when using a high quality DAC, but is less noticeable. If you're asking about a potential reduction in noise from the 24-bit output of my mod, no, the noise floor is limited by the 16-bit PCM samples from these consoles.

My apologies for the lateness of my reply!
User avatar
NewSchoolBoxer
Posts: 369
Joined: Fri Jun 21, 2019 2:53 pm
Location: Atlanta, GA

Re: Digital audio sample rate converter: S/PDIF for N64, MV1

Post by NewSchoolBoxer »

L-Train wrote:
NewSchoolBoxer wrote: --Can you confirm if 1CHIP has inferior audio to 2CHIP? I'll test this myself but I'm biased in thinking 1CHIPs are cheap cost reduction machines with the SNES Jr/Mini being almost a clone console. Important distinction when 2CHIP digital mods are much more difficult.
plgDavid has done extensive measurements and his video about the research behind chipsynth SFC indicates that the digital audio they produce is practically identical. I did a quick check too, just a simple inversion test with a regular SNES with a DIT4192 and a SNES Jr. with a CS8406. Due to the differences in state at boot (explained starting at 10:42 in the video) the cancellation isn't perfect, but there are sounds that are cancelled completely which implies that they can be bit perfect.
Really interesting video, thanks for showing me. You've already tested DIT4192, that's great! I was thinking of audio going out of the analog ports being worse. I can imagine 1CHIP/Jr. being extremely similar or nearly identical from the digital sampling stage, barring some known audio glitches.

I 100% agree with video about prioritizing the first model's audio that was the most widely produced. Video was cringe on the science end. Selling the synthesizer with the claim the SNES model revisions all sound the same makes his work much easier. plgDavid states the wrong cutoff frequency for the preamp, which looks pretty bad. The sim.okawa-denshi.jp site I use too doesn't give the cutoff here. I believe the transfer function is correct from cascading the passive RC and active Sallen-Key 2nd order RC. Can find cutoff on WolframAlpha by:
Spoiler
enter: magnitude of 5.2174638951498E+14/(s3+141414.14141414s2+14921946740.129s+5.2174638951498E+14) where s=2*pi*11624*i
copy answer and set equal to sqrt(2)/2 for -3 dB point: sqrt(2)/2=(5.2174638951498×10^14)/abs(-248.05021344240 i w^3 - 5.582806529909×10^6 w^2 + 9.375735651209×10^10 i w + 5.2174638951498×10^14)
answer: w≈11623.921139745785 = 11.624 kHz
11.624 kHz cutoff to drop ~30% of voltage level is extreme. I like your reasoning of a filter wanting to suppress 15.7 kHz noise that implies a very low cutoff but this can't hit 50/60 Hz hum. I simulated the preamp filter and got the same 3 dB point:
Spoiler
(click for full size)
Image
LM324 preamp filter: Q factor of 1/(2*zeta) = 1.118 > 1 is pretty wonky when 3rd order Butterworth filter has Q of 1.000 on the opamp and Bessel of 0.691, with some tolerance for non-exact R and C values. Playing around in Micro-Cap 12 and Analog's Filter Wizard and comparing group delay, I'm fairly confident Nintendo went for a 2/3 Butterworth 1/3 Bessel hybrid filter. Is a legit design strat but forcing matching resistors versus letting R3 be 10x greater messes up the damping that reduces the passband. Lowering manufacturing costs by reusing same resistor value must have been more important.

The video's audio measurements are lacking when they only show MDFourier. I don't expect team with niche market to have a MATLAB or Adobe Audition license but affordable digital oscilloscopes have built-in THD+Noise and FFT and can easily measure channel separation.

He's giving himself an unfair advantage by recording digital capture from box from his own APU that sits outside the console to cut electromagnetic interference and avoid the unfiltered power supply. Compares this hyper clean audio with his own software tuned to match it, then against accurate emulators that would have used real console samples. In theory, they emulated the filter and other analog circuitry that plgDavid is skipping. He could reassure me with analog capture comparisons.
L-Train wrote:
NewSchoolBoxer wrote: --Can I ask what we refer to in American English as the 800-pound gorilla in the room question? At sub-CD audio with arguably high noise floor, which is to say SNES and N64, is the digital reduction in noise significant enough to make any audible difference? I was discussing that with Ikaruga11 and I predicted definitely not for 3.56:1 audio compression ratio, expected 32 kHz DAC SNES. Maybe yes for N64 from the rare game coded to use 44 kHz sampling. By extension, I think going up to 24-bit, 192 kHz sampling, 24 Mbps is silly but not going to hurt anything.
By digital reduction in noise do you mean the difference between the analog audio produced by the NEC DAC inside the console versus a pure digital signal? If so, the most audible differences are the elimination of 50/60 Hz hum and 15.7 kHz interference, and the lack of low pass filtering. A lower noise floor can be audible when using a high quality DAC, but is less noticeable. If you're asking about a potential reduction in noise from the 24-bit output of my mod, no, the noise floor is limited by the 16-bit PCM samples from these consoles.

My apologies for the lateness of my reply!
It's fine, I disappeared for a few months. Great and helpful responses! Was what I was asking, how much better, however 'better' is defined, is pulling digital audio directly from the chip going to sound versus going through the SNES DAC + filter? Noise floor limited by console sampling bit-compressed data makes sense. Can sound better in the sense of not using an ancient DAC and arguably poorly designed filter with ancient LM324 amp. The 15.7 kHz horizontal sync interference, I wouldn't have thought of that but seems obvious now.

I guess there's two ways to do this: extract digital audio that can sound better/cleaner or record the analog audio that can sound worse/accurate. I have the prosumer recording equipment for the latter to see if 1CHIP RCA audio is same or worse than earlier revisions. Limitation is having one 1CHIP.
Post Reply