GUIDE: Telling apart good Genesis 1s and Genesis 2s from bad ones.

[Eep386]

Well, the non-TMSS thing is really another matter of personal taste. Some enjoy the lack of 2-second delay before their game starts, others (myself) could care less, and actually almost depend on the TMSS message as an indication that their system on the troubleshooting bench is at least trying to read the cartridges. (Far more helpful than a static black screen...) The VA3, VA4 and VA5 Model 1s lack TMSS, but the VA6/6.5 are really just VA5s with the TMSS 'BIOS' embedded inside one of the ASIC chips. The VA6.8 is just a slightly modified VA6.5 with the same sound circuitry, but using the SOP CXA1145M or MB3514 encoders.

[JamesF]

Meanwhile I could test audio distortion with a DIY cable.

DIY genesis cable.jpg

Testing with my DAW (Sonar) and Studio headphones I could NOT hear or measure any distortion whatsoever from the Model2 VA4 with the loudest games and sound effects.
Mode 2 VA4 output is 0.3 db hotter signal than Model1 VA6 (headphones on 10) which is insignificant, and sounds just as clear, balanced and distortion free on high volume sounds as the Model1 VA6.

As for second-order filtering, it is noticeable only on high quality equipment like headphones or studio monitors with high frequency sounds and noises, but on TV speakers the audible difference is negligible.
Saying that one doesn't like the good sounding Model 2 because of second-order filtering is a mere personal preference and an exaggeration because it is not a big difference at all from a first-order filtering;
To me, Model2 VA4 sounds better than the Model1 VA6 on Hi-Fi equipment especially with the Genesis's low quality PCM sounds and sharp noises of the FM Synth.

YM2612 vs YM3438;
The YM2612 is the original chip with its badly designed DAC and loads of crossover distortion (ladder effect) which accompanies all sounds and especially audible with quiet sounds, it has its own dirty characteristic sound and I like it.
The YM3438 is a step up with better DAC but still has some minor crossover distortion but is much less audible than the previous chip, clean and quiet, an engineers dream.
My personal preference is YM3438 for its clearly audible superior qualities on a Hi-Fi system, but I do like the YM2612 for its gritty characteristic Genesis sound as already stated.

[Eep386]

Bear in mind I am testing them with music playing in the back (eg. in game). The sound effects by themselves won't clip from the sound test unless there's music behind them.

Of course it could be my cable too - I am using a cable that outputs Mono Out to Stereo Right instead of the proper output. I'd like to build a dongle to correct the Mono Out -> Stereo Right problem on my cable (should be relatively trivial, just run the right pin to the cable's Stereo Right pin) as apart from that it's a really nice, thick cable with gold-tipped contacts.

About the YM2612 vs. YM3438, some games deliberately try to incur that time-division multiplexer distortion to alter the timbres of some sounds. Streets of Rage 2's opening pianos fading out much sooner on an ASIC FM is a classic example, but Novotrade games also do some obviously deliberate things with TDM noise as well. Beyond Oasis explicitly depends on it for the song 'Evil Territory' - the exaggerated TDM racket adds an interesting 'whispering voice' like effect to the music, which the ASIC FM washes out.

On a marginally related note, one nasty little urban legend I'd like to dispel: the YM3438 does NOT increase the PCM 'channel' bit depth at all. The PCM data latch remains 8 bits wide on the YM3438 and ASIC FM. Changing it to a 9-bit depth would only heavily distort existing PCM samples unless they were resampled on the fly, which would break virtually every pre-VA7 Model 1 / VA0~1.8 Model 2 game that plays digitized samples. The difference in PCM response is likely due to the altered TDM DAC of the ASIC FM.

[JamesF]

Bear in mind I am testing them with music playing in the back (eg. in game). The sound effects by themselves won't clip from the sound test unless there's music behind them.

I understand.
Try Comix Zone, Options->Jukebox, most songs there are already louder than any other game (music+effects+PCM).
I hear no distortion on the Model2 VA4, sounds very HDG Model1 like but with 2nd-order filtering.

Of course it could be my cable too - I am using a cable that outputs Mono Out to Stereo Right instead of the proper output.

Mono out from the genesis to Right Stereo Channel only?
I don't quite understand what you did with the cable, can you please clarify?

About the YM2612 vs. YM3438, some games deliberately try to incur that time-division multiplexer distortion to alter the timbres of some sounds.

If you can, please give a list of games that use this "ladder effect" distortion so I can test them.
Currently, I found that on sounds that fade-out with quiet 'tail' the effect is the most prominent, but I would not say developers used it intentionally, instead they accepted it as an artifact of the technology.
I know Ace is a big advocate of the YM2612 and games that use its distortion as part of the sound effects.

[CrossBow]

I recently acquired a JPN model 1 VA4 and noticed something strange with its audio output from the headphone jack. First it isn't nearly as noisy as the US model 1's I messed with regarding the floor level noise and I was surprised by this given the lack of any EMF shielding on the JPN models at all! Anyway, I was playing some XGM compilations and noticed that the left and right channels appear to be backward from the headphone jack? Is this normal?

Now..to add more to that audio mystery, I then ran the 240p test suite on the system and did the audio test there. For reason's I can't explain, the audio test has the audio coming out of the correct channels during this test?! But on the XGMs the sounds are reversed to the channels being displayed as playing the given audio and it is noticeable.

So I then plugged in my US model 1 VA3 and tried the same tests. But this one has the audio channels matching XGM players playback correctly as does the 240p test suite tests?

So.. I'm wondering if there is some programming trick that the 240p test suite does and is able to know that the channels are in fact reversed on some model Genesis systems? I've read somewhere before that the channels were reversed on some models.

I did this several times to make sure it wasn't my Sam Adams altering my perception and indeed my wife confirmed what I was seeing and hearing between them.

[JamesF]

Noise floor measurement is meaningless without peak measurement.
The right ways is to measure Noise Floor in RMS and Strongest signal in Peak then calculate the Delta in db between them == Signal-to-Noise ratio; that's what we should compare between consoles/revisions.

Example:
Model2 VA4, Streets of Rage 2, BGM 25.
Noise floor(RMS) -79 dbFS.
Peak signal (Peak) -1.9 dbFS.
SNR = 77.1db.

As for the reversed channels... very peculiar, I assume it is the XGM player or compiler that is messed up.

[Eep386]

The reason my cable outputs Mono Out through Stereo Right, is because it's one of those 'eBay specials' and is wired incorrectly internally. It does that with every Genesis I plug it into. A dongle would fix this, and also afford me some space to fit a pair of ground loop isolation transformers as well. Of course I could just buy a proper cable, but I don't wanna waste things.

I am testing out Comix Zone on my VA4, and HOLY SHITE IT'S LOUD. (No joke, it's not this loud on any of my other consoles.) On this system and game, I hear the Mono Out channel clip slightly but noticeably my TV. It can get pretty crunchy toward the end of track 17. Hooked into my VA2.3 with corrected sound circuit, no clipping whatsoever.

Regarding the TDM-heavy songs, ask and ye shall receive, along with the songs that seem to modulate the distortion most prominently:
- Beyond Oasis (Evil Territory) - the TDM racket adds a really cool 'voice whisper' effect when the music fades out
- Ecco the Dolphin (almost every song) - mostly used to add subtle 'chorus' on instruments
- Ecco: The Tides of Time (sound number 5 (Sea of Darkness), 6 (Globe Holder, Lunar Bay) - often used to add a subtle 'chorus' to the instruments
- Puggsy (overworld map theme) - makes the synth pads sound 'deeper'
- Richard Scarry's Busytown (entire soundtrack) - this game rapes the TDM noise to make interesting ambient timbres, it sounds radically different on ASIC FM (and not for the better)
- Star Trek Deep Space Nine: Crossroads of Time (music 9, 18 and 24 most prominently) - on 9 it gives the strange 'tum-tum' instrument a buzzy roll-off, on 24 it's used to enhance the 'choir ahh' patches
- Streets of Rage 2 (BGM 10) - classic example, the pianos fade out much sooner on ASIC FM and the instruments sound flatter overall
- Toejam and Earl 2: Panic on Funktron (Boomin' Bump) - probably more an accidental effect here (GEMS music is usually pretty ignorant of TDM), but it sounds really cool on the synth pads

There are many others I'm sure, but I only own a relatively small number of games compared to some folks on this forum XD You'll notice these are some of the games I use to do the most extensive testing on my consoles while I'm working on them.

I realize the whole YM2612 vs. YM3438 debate is strictly an issue of personal taste, in the same vein as MOS6581 vs. MOS8580, but as for my own interests, I prefer my buzzy, distorted YM2612 because it achieves the obviously intended effects on the games I mentioned, and it's how I remember it sounding. To me the ASIC FM just sounds like a really good emulator. (And I also like the VA3 Model 2 quite a lot as well, as likewise some songs are optimized for the ASIC FM.)

[JamesF]

Thanks!

A few I want to add:
Streets of Rage 2, BGM-25 (my fav), there's an instrument completely missing right at the beginning when played with YM3438.
Aladdin, Turban Jazz & Camel Jazz, the high-hats sound is completely different, quick and sharp (closed Hi-Hat) with YM3438, but jazzy (brush sticks on HiHat) with YM2612.

I have no clipping on my M2 VA4 from Mono or Stereo out, the output voltage is practically identical to my the M1 VA6 headphones out on 10.
Both recorded into a Focusrite Saffire audio interface.
Maybe it's the inputs of your TV that are clipping? You should not use your TV for critical listening anyway.

[JamesF]

Streets of Rage 2 - BGM 25 comparison.

Here are quality recordings of Model1 VA6 and Model2 VA4 without volume adjustment between the two, just swap the cables and record.

http://www.mediafire.com/file/d82mo3..._-_M1_VA6.flac
http://www.mediafire.com/file/4usys6..._-_M2_VA4.flac

BGM 25 is considered a loud track and is a good sample to test if the console distorts.
It is also a good track to hear the difference between YM2612 and YM3438, the quiet intro reveals the difference.
M2 VA4 is 2db louder (peak) than M1 VA6, and has higher order filtering, but it is (once again) not distorting/clipping.

[Eep386]

Well, my TV is ultimately what I will be playing my games on. I use it as a reasonable benchmark for how others might hear the consoles I work on. (As a pessimist by nature, it's my testing methodology to use the worst, least ideal monitoring mechanisms, as it will give me a good idea as to how the 'worst case' will be like.) I also take direct recordings from the console from time to time, interfacing via a cheap Wal-Mart coiled cable to my PC's onboard Realtek.

The second-order filter thing isn't necessarily a bad thing - implemented correctly, it can do a much better job at getting rid of electronic noise than a first-order filter. Trick is, getting the second-order filter to not excessively impede the tonal range, which is difficult considering it has a much steeper roll-off.

Regarding the sound clips, I hear that instrument in the MD4 ASIC FM; it's just nearly inaudible. Plus the chorus itself sounds more 'precise' but dryer, flatter and less robust in general. However the VA4 Model 2 has considerably higher bass response, at the expense of a slightly poor high-end response. The reason for the higher bass response, I believe, is the fact that the VA4's 315-5684 has three low-pass filters it's subjecting FM sound to! There's a 1200pF low-pass pre-mix filter, a 5600pF FM low-pass filter (2.89 KHz corner), and an 1800pF output stage low-pass filter. Changing the 1200pF pre-mix filter caps to 270pF totally eliminates the second-order filter effect, making the rest of the sound much sharper. Afterward, replacing the 5600pFs FM filters with 4700pFs and the 1800pF output filter caps with 1600pFs gives a very VA2 Model 1-esque sound, minus clipping and distortion (though my TV begs to differ apparently).

[Ace]

That 1200pF capacitor is not a pre-mix filter cap, it's part of the second-order low-pass filter. It's still the same filter setup, but it's actually a little more aggressive than the second-order low-pass filter used on the junk discrete multi-stage design from the VA0 Genesis Model 2 to VA1.8 Genesis Model 2 (those maintain the 5600pF capacitors, but use 1000pF capacitors instead of 1200pF capacitors).

The thing with second-order low-pass filters, at least in my experience, is that they work best when kept as far out of the audible spectrum as possible, that is to say, keep them above 22KHz. This way, more high-frequency garbage gets filtered out, but the audio doesn't get massively muffled. I still need to experiment to see at what cutoff frequency a second-order low-pass filter starts to badly mess with the spectrum, and this is what I would consider to be the absolute lowest cutoff frequency I would accept for such a filter (except for some ADPCM audio hardware that just sounds awful without such a filter - this would be the case of the Sega CD's PCM audio, the Sharp X68000's ADPCM audio, NEC's uPD7759, etc., although in some instances, even a 3rd-order low-pass filter is required like on the Super NES due to a ton of high-frequency garbage that seems to make even the best op-amps crackle badly if omitted).

[Eep386]

I have pretty good results setting that second-order filter to 270pF, though I'd imagine you could get away with giving it somewhat more common 330pF caps for a 48.2 kHz cut-off without losing anything really noticeable in the audible range (aside from high frequency noise).

Oh yeah, speaking of the discrete VA0~1.8 systems, I think I figured out what's creating that awful tinniness that's present even with Sega's ridiculous 5600pF active filter caps on the IC10C/D outputs. It's caused by the mono output amplifier, or rather, the placement of the 10K resistor taps. They're placed on the outputs of IC10C and IC10D at R77 and R78. I tried relocating these resistors (using flying through-hole resistors) to a point between CE9/R44 and CE7/R42 respectively, and that completely eliminated the tinniness (and also gave a MUCH improved Mono Out sound quality), but it still won't sound very good until you address the rest of the circuit. Problem is, with this setup SL2/SR2 won't get mixed into Mono Out until you relocate those taps to an earlier stage as well.

Also, all '324 op-amps have really hideous dead-band crossover distortion.

[JamesF]

I have run a simulation and played with it to see how each component alters the frequency response:
Model 2 Filters.jpg

Active second order low pass filtering topology (like in the Model2): http://www.electronics-tutorials.ws/...r-filters.html
The Model 2 has 2nd-order LPF followed by 1st-order LPF.
2nd-order LPF is applied only to FM,, 1st-order LPF is applied to everything, including CD audio return (SL2,SR2).
Model 2 effectively has 3rd-order LPF for FM,, Model 1 only has 1st-order on all signals.

It is wrong to calculate the 10k+1200pF network inside the second-order topology like a independent first-order, all resistors and capacitors work together dependently to create the 2nd-order LPF.
Changing the 1200pF to 270pF changes the Q drastically and boosts by 8db at 12kHz, which is definitely not good.
270pF.jpg

The proper way to disable the 2nd-order LPF and turn it to a 1st-order is simply remove the 5600p in the loop and tune the 1200p (1.2nF) filter to match a Model1.
ie: 20k (10k+10k) and 2.7nF will give 2.9kHz cutoff frequency like Model 1.
Active first order LPF: http://www.electronics-tutorials.ws/.../filter_5.html


EDIT:
It seems that sega engineers forgot a decoupling cap between the MOL/MOR (FM) outputs and the following Opamp (IC9) on Model 2 V0-V1.8 thus creating terrible distortion which I confirmed in simulation.
Adding a Coupling capacitor (huge 100uF in my simulation) before the first 10k fixes the distortion.

All models after V1.8 have 0.1uF (100nF) coupling cap after the MOL/MOR outputs pins 55/56 on the main GOAC but they also have an additional opamp before the 2nd-order network.
So I assume adding 10uF coupling cap on Model2 V0-V1.8 (and Model1 VA7) should fix the distortion.

[JamesF]

Thought this might be relevant info for this thread.

Model1:
Model1.jpg

Model2:
Model2.jpg

Frequency response overlayed:
Overlayed.jpg

[CrossBow]

Noise floor measurement is meaningless without peak measurement.
The right ways is to measure Noise Floor in RMS and Strongest signal in Peak then calculate the Delta in db between them == Signal-to-Noise ratio; that's what we should compare between consoles/revisions.

Example:
Model2 VA4, Streets of Rage 2, BGM 25.
Noise floor(RMS) -79 dbFS.
Peak signal (Peak) -1.9 dbFS.
SNR = 77.1db.

As for the reversed channels... very peculiar, I assume it is the XGM player or compiler that is messed up.

I'm only witnessing the reversed channels using the XGM player on this JPN va4. On my va3 they sound correct on both XGM player and the 240p suite audio test as does my va2 is correct. So I'm wondering now..if there is something different this va4 JPN model is doing on the sound that is confusing my AV receiver? I need to hook the va2 and this JPN unit up to my computer and record them with known stereo panning tracks to be sure.