FS-A1ST and FS-A1GT MSX Turbo-Rs sound fix

like+1000000 thanks!!

I published four lists of soundtracks comparing the after/before of the FS-A1ST audio fix, so anyone can check themselves the difference in terms of sound quality and how much detail was being missed. I hope you enjoy it.

• SCC after/before the audiofix soundtrack comparison
• PSG after/before the audiofix soundtrack comparison
• OPLL after/before the audiofix soundtrack comparison
• PSG+OPLL+SCC after/before the audiofix soundtrack comparison

EDIT: I used VGMplay for the non-Micro Cabin soundtracks, but I noticed that some tracks have some weird glitches, regardless of the fix. It's very noticeable in the beginning of Metal Gear-trk2, in the drumkit of Mr Ghost and in the opening song of Greatest Driver.

PS: Be sure to use a decent sound equipment. Notebook speakers won't do. :)

YES!! The difference is huge!!

Great work FRS!!

Absolutely great!!!! I can't wait until the moment to fix my computer!

Wow sd_snatcher, not only is the comparison very interesting, your selection of songs is also really enjoyable in itself.

My guess about the Greatest Driver glitches are warped parameters in pitchbending and "note linking" (new tone without retriggering the preset) commands.
Though it's puzzling why it only affects certain instances and doesn't seem to happen in the "original" version.

Maybe it just isn't...
(•_•)
( •_•)>⌐■-■
...using the greatest driver.
(⌐■_■)
YYYYYEEEEEAAAAAHHHHHH

The Greatest Driver glitches are from too fast OPLL access, although as you say I only hear them in the fixed ST recordings. Which is a bit puzzling, especially if that’s also the case for other songs. It suggests a machine-specific difference.

@sd_snatcher Did you perchance upgrade something related to CPU speed on the fixed ST? Or maybe the other ST was running in Z80 mode? Otherwise, if there are no differences to speak of, I might’ve timed things a bit too tightly according to my own GT’s tolerances. I’ve had an issue with this in the past for an MSX-Audio modded with a clock crystal (Meits’s).

That doesn’t explain the PSG issues though. Maybe there’s something I forgot to initialise.

I think the idea is that you record in openMSX including initialization (for PSG that may be the BIOS starting up) and cut out some silence parts out later.

How could the script add an initialization sequence if it doesn't even know what it should be for the music that will be recorded?

Wow, differences seem insane, like the originals almost sound muted? Is there actually such a difference in volume/clarity/loudness, or does the recording itself also play part in this? I can hardly believe the original MSX sounds that bad.

Thanks for the (huge! amount of) recordings (and indeed nice selection ) Gonna compare some stuff with openMSX's rendering when I get to it

Something I'm wondering about though: one may assume the various composers heard their work as well on an original (thus prolly flawed) MSX.. So now with the fix it may actually *not* sound the way it's intended right? (But then, there were already differences between models..)

Is it safe to assume then that the glitches in the other games are caused by that same thing?

So, in terms that laymen (like me) can understand, the data is sent to the chip faster than it can process, so that it uses previous frequency information, that lags behind the 'action'?

Okay I did a quick listen. I don't own a Turbo R (yet) so I don't have much to gain by this. Nevertheless I would give you my personal opinion, to help you maybe further improve this mod.

btw I disregard the difference in overal volume level, since that is most definitely an improvement.

PSG+SCC: Balance is improved, overal sound is good. Improves a lot of the tunes.
PSG only: Sounds like an improvement, things are more clear. It would seem some of the noise register sounds (hihats etc) were a lot softer before the mod. Sounds has more "texture" i guess (if that makes any sense).

However the OPLL seems to have way too much treble. I've talked on #msxdev about this briefly (like a few minutes), it would seem your unmodded turbo R (st) indeed sounds a more "muffled" compared to a GT (that lowpass filter thing i guess). So fixing could be needed.

But in my opinion the fixed versions of some of the songs just don't sound right. For example your comparison of PSG+OPLL+SCC (sorcerian tune), I prefer the original unmodded version. The fixed version has so much treble around 0:48 that it starts sound unpleasant for me ears if I dont turn down volume. Same for some of the OPLL Fray tunes. I guess it depends on how much "high" was present in the original, other songs sound alright.

And no, this is not me based on years of being used to certain settings/machines, since I barely know half of the tunes you posted.

Btw does your mod for the sanyo make sound about the same as this? (not asking/expecting you to do a recording)

This is what I am wondering about too. I dont know about the exact release dates of the games or development machines of Microcabin, but it seems that the "treble problem" that I talk about earlier seem to more affect "newer" microcabin games. I suspect they were made on a muffled turbo r, and the composers just upped the higher instruments' volume to compensate for the lowpass filter.

I like the fixed sounds too. Matter of taste here of course, could be compared to preferring CRT blurry display vs LCD sharp square pixels (in this case I prefer the CRT!).

Thanks for asking, I'll add this info to the article. If you can find polypropylenes that fit in there without the risk of ripping the pads out of the board, you can used them. This means: *only* if they are the small 50V type. Do not use the 100V rating because they're way too bulky to install there.

Plz also keep in mind that the recommended type of ceramics, if used, are not the crappy Y5V or the not-so-great X7R. Use the more expensive C0G (aka NP0) type as they are of excellent quality and much more appropriate for this application.

Cool, then I can order all-Panasonic components .

Here’s my DigiKey shopping list (with assortment of caps so I can compare different values).

Index,Quantity,Part Number,Description,Customer Reference,Available Quantity,Backorder Quantity,Unit Price USD,Extended Price USD,
1,10,PCF1299CT-ND,CAP FILM 1800PF 2% 50VDC 0805,,10,0,0.36100,3.61,
2,10,PCF1177CT-ND,CAP FILM 10000PF 2% 16VDC 0805,,10,0,0.34600,3.46,
3,10,PCF1174CT-ND,CAP FILM 5600PF 2% 16VDC 0805,,10,0,0.37400,3.74,
4,10,PCF1301CT-ND,CAP FILM 2700PF 2% 50VDC 0805,,10,0,0.32100,3.21,
5,10,PCF1287CT-ND,CAP FILM 180PF 2% 50VDC 0805,,10,0,0.35700,3.57,
6,10,PCF1286CT-ND,CAP FILM 150PF 2% 50VDC 0805,,10,0,0.35700,3.57,
7,10,P22KDACT-ND,RES SMD 22K OHM 0.1% 1/8W 0805,,10,0,0.30200,3.02,
8,10,P15KDACT-ND,RES SMD 15K OHM 0.1% 1/8W 0805,,10,0,0.30200,3.02,
9,10,P47KDACT-ND,RES SMD 47K OHM 0.1% 1/8W 0805,,10,0,0.30200,3.02,

Tomorrow I’ll see if I can find some other nice stuff that I’d like to have to get above the $60 free shipping :).

For me the “Stable Frequency Characteristics” graph in this overview was pretty illustrative.

For the resistors I picked thin film ones which are supposedly better for audio purposes (though a bit lower wattage, and I’m not sure the high-end audio characteristics will survive the rest of the MSX’s circuitry :D, so maybe just a bit of a waste of money). And the higher value caps are only rated for 16V, I assume that’s no problem either since line level is just 1.7V and before the opamps I guess it’s even less.

As for the Polyphenylene, I just saw “film” and “polyflufluflulene” so that was a bit accidental :D.

@Grauw, you could order some more components, so you could make diy sets for the turbo-r's.
Maybe include all capacitors so we could do a complete recap of the computer's.
Are the differences between the St send Gt "big"? Component wise? Then one set could be used for St or Gt.

And I noticed something strange, the microphone of the turbo-r's are not completely muted.
When you put on your computer and blow or whistle in to the mic. You can hear it in the speakers (audio out).
I can imagine that this causes some background noise no one whants. Is there a fix for this?

@Grauw,
Nijmegen is a good idea, but I want to fix my turbo-r's before that
And good to know that the caps are of good quality, I will leave them be then.

Today I modified my turboR (finally), and as promised, a review with comparisons!

I made recordings of the following six songs: 1) Greatest Driver - Opening (FM), 2) Xak - Opening 2 (FM+PSG), 3) Xak - Fort (FM+PSG), 4) Xak - Land of Flames, The Sky (FM+PSG), 5) Xak - Millie’s Theme (PSG) and 6) Space Manbow - Stage 1 (SCC+PSG).

The SCC is played on the MegaFlashROM’s SCC, a real SCC is comparable. Tip: load the audio files side-by-side in your favourite audio editor (e.g. Audacity) and mute / solo the tracks one by one.

TurboR FS-A1GT - Original

Filter chain:
OPLL-1: 4.7 kΩ / 33000 pF = 1026 Hz
OPLL-2: 56k / 470p = 6047 Hz
Global: 39 kΩ / 560 pF = 7287 Hz

TurboR FS-A1GT - WX filter

Filter chain:
OPLL-1: 4.7 kΩ / 10000 pF = 3386 Hz
OPLL-2: 56k / 150p = 18947 Hz
Global: 47 kΩ / 270 pF = 12542 Hz

Note: I modified FRS’s audio fix to use filter values similar to the Panasonic FS-A1WX. Although there OPLL-1 is 2258 Hz and OPLL-2 is absent, but close enough.

TurboR FS-A1GT - SFG filter

Filter chain:
OPLL-1: 4.7 kΩ / 1800 pF = 18813 Hz
OPLL-2: 56k / 150p = 18947 Hz
Global: 47 kΩ / 180 pF = 18813 Hz

Yamaha CX5MII

Filter chain:
OPLL-1: 2.2 kΩ / 2200 pF = 32883 Hz
OPLL-2: absent
Global: 4.7 kΩ / 1500 pF = 22575 Hz

Note: I used a Monster Sound FM PAQ for OPLL on the CX5MII.

=== Review ===

FRS’s audio fix consists of three main parts: 1. correction of the balance, 2. increased output volume, and 3. reducing the low-pass filter. I’m going to look at these one by one, comparing with Yamaha’s great CX5MII as a reference machine when needed.

Balance

The fix changes two mixing resistors; the PSG resistor from 56 kΩ to 22 kΩ, and the OPLL resistor from 47 kΩ to 15 kΩ. This means that relative to external sound chips the PSG and OPLL are both mixed in louder now, the OPLL also a bit more than the PSG.

The difference is quite noticeable in the SCC test (song 6). Before, the PSG drums would be barely audible, drowned out by the SCC music. After, the drums are mixed in correctly, matching the CX5MII. (Note, due to recording volume difference the original’s SCC song is louder, don’t get fooled thinking louder = better, normalise the volume manually if needed.)

The FM-PSG mixing change can be heard well in the second Xak song (song 3). Before, the PSG was a bit too prominent, whereas after it is much more balanced, square waves blending well with the FM tones which is so important in Micro Cabin music.

Output volume

Since the PSG and OPLL mixing levels were increased their total output has as well. Additionally, on the GT the resistor on the final op-amp is also changed from 39 kΩ to 47 kΩ, increasing the amplification, matching the ST and the WX. Indeed, on the ST this is already at the correct value so it gets less of a boost.

Because the volume used to be so low, I had to amplify the output on my mixer quite a lot, and the hum from the video interference became very noticeable. I always had to carefully pick mixing volumes, ducking the volume in silent parts of recordings, and I would use “color 12,0,0” to reduce the video interference noise as well.

After the fix, the output volume increased quite significantly (about doubled), and matches the output volume of my CX5MII. The noise floor is barely noticeable anymore. Great!

Filtering

Last, the low-pass filter. This has been the biggest source of debate. The turboR is well known for its excessive low pass filtering, as you can see in the numbers I mentioned for the original filter chain above. As a result, the sound output sounds very muffled, the OPLL in particular. It was such a known issue that it is noted on the Japanese Wikipedia page, stating that people would often remove the filtering capacitors as a quick fix (even though of course the filter is there for a reason).

A low-pass filter lowers the volume for higher sound frequencies above a so-called cut-off frequency. From there on the higher the frequency, the more they are attenuated. Given that a middle A is 440 Hz, as you can see the turboR’s 1026 Hz cut-off frequency is just barely an octave above that. No wonder it sounds muffled!

The low pass filter cut-off frequencies are determined by resistor-capacitor combinations (“RC” filter), and by increasing the resistor value or reducing the capacitor value you increase the cut-off frequency. The formula is pretty simple, and there are handy online tools to calculate it. Because there are three filters in series, they are applied cumulatively, increasing the slope of the attenuation.

The filter improvements by FRS crank these filter values up quite a bit, making all the muffling disappear. The settings are based on the filter in the Yamaha SFG-05 sound module, with all three filters kicking in around 18 kHz, just below the human audible hearing frequency.

Because these filter values get pretty close to the OPLL’s sampling frequency though (at 49716 Hz), and the OPLL also uses the filter for mixing its channels, there was some debate whether this would not reduce the mixing quality. For that reason I’ve also tried out an alternative filter setting, based on the values of the Panasonic FS-A1WX MSX2+ computer. This should definitely be a “reference” value that everybody will like :).

Either way, so after trying them out, I have to say it’s a huge improvement, the both of them. Already the WX-based setting sounds so much better, and if I compare them side by side the SFG-based setting sounds even more crisp and removes the last audible remnants of muffling. Which is better is subject to opinion (which I’m looking forward to hearing about!), but I think there’s no debate that either way it is an improvement over the original. Personally I think I prefer the SFG-type version.

I also hope the PCM audio quality will be improved significantly, it sounded quite filtered when I played Sega MegaDrive music with PCM before. I haven’t been able to try it yet though, since the keyboard isn’t attached yet.

Applying the fix

Couple of notes on applying the fix:

1. To open the turboR body, after removing all the screws, pry apart the side near the power button, then pull your hands forward to loosen the keyboard area, and once everything feels nice and loose, lift and push the top half of the body backwards to get it to release the clamps in the back. Just use your hands, no need to damage the plastic with a screwdriver.
2. To unsolder the resistors and caps, I first removed the existing solder with solder braid, and then alternatingly heated both sides of the component while gently pressing against the side of the component with a pointy tool (it works best to press at the end that you’re soldering). At some point they start to move a bit, then repeat a few more times and it should be loose. Be careful not to force it, or you can damage the solder pad. (I did this once before I got the hang of it, luckily it just came loose and was not detached, I added a tiny botch wire anyway just in case.)
3. I ordered all components from Digi-Key, using Polyphenylene Sulfide capacitors, see the shopping list I posted earlier in this thread. As a nice touch, I got all-Panasonic components :). To keep the price down I recommend ordering 10/20 of each, and you can share some sets with MSX users in your neighbourhood.

Conclusion

This is a very worthwhile fix to apply, it really has improved the sound quality very significantly. This is not just one of those things only audiophiles can appreciate. Really overall a big improvement on all accounts.

Done!

Really nice!! Thanks Grauw.

Thank you for this detailed review, Grauw! It will help to clear the remaining doubts, so everyone can enjoy the beautiful soundtracks of the MSX games with the best possible quality on their real machines.

Some friends insistently asked me for a similar fix for the Sony MSX2+ machines, so wait for some news about this soon.

syn: Thanks for your opinion! I was hoping you would reply. Glad that the WX-type filter values I tried are to your liking. If you ever want to fix the audio on your turboR (if you have one, I forgot), then now you know what capacitor values to use to match your preference .

Indeed a difference between the SFG and WX filters can be heard in that drum, I noticed that too; since the high end of the snare is filtered away more on the WX-type, it ends up sounding punchier. But at the same time on the WX-type everything sounds a bit muffled compared to the SFG-type. I’m a bit on the fence on which I like more as well, so I can understand if opinions go either way . It’s the details.

Today I tried one more thing; I stacked two pairs of 2700pF capacitors to get a 6271 Hz OPLL-1 filter. Somewhere inbetween. Since I can’t really decide, am tired of keeping my turboR opened up, and would have to open up one of my sets prepared for others if I change it again, I’ll stick to this compromise :D.

p.s. I note there is no audible difference between the SFG and WX type filter settings for the other sound chips (PSG, SCC, etc.). Makes sense since there’s just half an octave difference on the global filter, and looking at the CX5MII neither SFG or WX-type values are out of the ordinary.

Grauw I forgot to post a follow up: When I heard that last sound clip (in december) I thought it would be the best settings/values

I dont know if that would reflect my current opinion though (since its been so long ago and I dont feel like comparing them again right now )

The sets are gone now.

we need more Audio Fix Kit available please!!!!!!!!!! someone!!!!!!!!!!!

Thanks Grauw for fixing my ST and GT. I am very happy with it. A very big improvement! Enjoying Space Manbow and XAK tunes at the moment on the GT through VGMPLAY. Also the removal of the jailbars on my screen makes my Turbo R GT so much better. Thanks again and also sd_snatcher for inventing this mod.

1chipMSX w/ OCM-PLD v3.6 - Default Volume (Level 10 out of 13)

@sd_snatcher: great fix! Is it also possible for you to do the audiofix for emulators as an IPS patch? Probably yes for audio balance and volume? Don't know for filtering. Then we can switch easily back to original and audiofix
If possible please to a fix for both FS-A1ST and FS-A1GT for fs-a1(s/g)t_firmware.rom and separate BIOS images like MSXTR.ROM, MSXTREXT.ROM and MSXTRMUS.ROM.

@Retrofan
That will be a no go. The AV characteristics are board design features/flaws. The horrible 8250 (and up) audio mix and jailbars didn't translate to the emulator for this reason either. It's not written in ROM what components the developer uses.

Just a small errata on the A1GT soundfix: There was a typo, and where C10 was written, C18 should be the correct ID. It's fixed on my page now.

Could someone help me with this fix on my turboR machines?

Find OmegaMSX. He will fix your stuff.

Hi, if i´m right this is in A1ST, if helps anyone:

Hi, i just tried it, sounds very powerful and louder and more balanced it seems to me...to be honest, reviewing the mod i can say that original A1ST sounds be sinked into the water, like if the A1ST was play sounds sinked into the water, with the FRS fix...is like you ear the msx out of the water.
I get the kit prepared from Console5 web, i had to get some parts for another stuff and i get the kit from that web.

WARNING: If you want to make by yourself the mod take a lot of care, i used hotgun to remove and tradicional solder iron to replace the caps and resistors using liquid Flux, the problem here is calibrate ok your hotgun, the TurboR mobos are pretty delicate if you use a hotgun, also i installed 4 sockets for 512K memory and prepared the machine for 1mb expansion, take a lot of care with yours heat gun, practice in another boards first.