Just got my ALS100 (non plus), I'm going to test it thoroughly.
AmoRetro says it has 1:1 clone of the OPL3 (LS262), we shall see.
Also, I'm going to test self-noise, SBPro, and SB16 behavior and all their aspects.

EDIT:
My card is dead on arrival, so there will be no testing. 😵
Sorry.

Oh well, that's a pity. Sorry to hear. 🙁
Anyway, thank you very much for your reply!
If I can test one or two games for you, just tell me. 😀

When i searching about GUS clones i found archived website of ASOUND / ASONIC. They started under name DnB and produced GUS clones but from 1995 switched to ALS chipset and sold six different cards with different ALS chips:
http://web.archive.org/web/19970412114603/htt … 0/catalogs.html
(German Tango 100 card with Crystal Dream wavetable is same as ASOUND)

Nice writeup Jo22. I got no ALS ISA cards myself, but I got an ALS4000 card which compares to the ALS120/100 in many ways: SB16 compatibility, nice FM synth and noise (it's the noisiest PCI sound card I got). It also works nice with later SiS and VIA AGP chipsets in DDMA mode.

Hi, all.
Sorry to bump a very old thread. I would like to share my experience regarding ALS120 based soundcards noise suppression.
There's a lot to tell about, but I will try to keep it as brief as possible. So here it goes:

I've got 3 different revisions of an AS9708 PCB and one of them had been with defective IC, until I replaced the chip.
I used the defective one as a reference, desoldered all components, took high-res pcb scans, measured values and redrew a replica card in Sprint Layout.
Then I ordered new old-stock IC's to repair the defective card and take this chance to examine it.
After some consideration, I also decided to design a 4-layer pcb pc/104 soundcard based on this chip to compare the noise and other things (see pics below)

To record the noise for spectrum analysis I used a battery powered hi-fi headphone amp connected to a soundcard line-out and a battery powered WAV sound recorder with a line-in input.
Only the minimum amount of components were soldered during tests. mic\line-in\gameport\cdrom\amp components were removed and the soundcard itself was powered directly via 5V battery, no PC connection.

Here are the brief results I got:
1) PSRR voltage regulation will have little effect. I hooked the (removed from PC) card directly to a 5v battery. The noise is still there even without initialization or any connection to PC. Card initialization in DOS and mixer will only add a small portion of noise to it.
2) This noise is a "1/f" noise in nature (pink noise or flicker noise) and is insanely similar to "by the book pink noise generator" (see the spectrum of the noise below), which makes me think it's a design flaw rather than silicon\grounding etc
3) Ground loop isolation transformer will have no effect (unsurprisingly), but the impedance mismatch causes the noise to be reduced. Sound volume also suffers greatly, but if amplified again by the speakers, you can get somewhat decent volume and quality (by ear). I suppose this also can be done with resistors, but not worth doing anyway.
4) My 4 layer design PCB did not perform any better noise-wise (I'll do some spectrum comparisons later). But all PC "thinking" noises are gone. I used (analogSIG\power<>ground<>digitalSIG\power<>digitalSIG) PCB layers layout. The traces layout is completely different. Only DAC filter values are the same. The idle noise is absolutely similar to #1
5) VREF1 and VREF2 IC outputs measure about 20mV peak noise (I'm sorry, but I could not make any pics, you'll have to take my word for it there). I cannot tell the spectrum either, all of my oscilloscopes are useless when it comes to audible band FFT
6) Disconnecting ALS120 VREF outputs and hooking the filter circuits directly to a LM4040 2V voltage reference did not improve things very much either (by ear). Interestingly, disconnecting VREF1\2 completely has no drastic effect either 0_o. (it's because vref external circuits are basically just filters)
7) Bipolar electrolytic 10uF vs 10uF Y5R MLCC as coupling line-out caps perform absolutely identical
8) Shunting floating inputs via 10K to ground vref - no effect

I may add some additional tests later (e.g. -0.3v digital to analog voltage rail diff, ground pin lifting, etc ), but I'm not sure there's any point.

UPD: additional tests will be added there:
8) 5V analog side ON \ 5V digital OFF = NOISE ; 5v analog side OFF \ 5v digital ON = silent (which is logical, since I disabled the mixer section completely)
9) Lower digital voltage below analog voltage level and vise versa = no effect. Lowering analog voltage decreases useful signal levels.
10) Lifting GND pins, so only one pin of each plane is connected and far apart= no effect + pc thinking noises return
11) Dry ice freezing (just for giggles and in the name of science). No effect, but I think I overhead some silicon cursing and threatening to be back.

My conclusions (please, take #2 with a grain af salt, It's only my theory):
As much as it pains me (the soundcards itself are not that bad at all.Good compatibility and OPL), I'm afraid there's not much that can be done in terms of supressing the noise.

1) Bad PCB grounding, EMI or amplifiers are not the culprit. I could not find anything out of the ordinary there. Amplifier capacitors, which are out of spec are simply either buffer ones or gain. But of course, they contribute their small part too.
2) Based on a sketchy datasheet info and my experience, I'm considering the culprit to be the internal IC's reference voltage source. It is the only thing I found drastically different between ALS100 and noisy ALS100+\ALS120 (apart from integrated OPL on ALS120, but ALS100+ also has similar noise with external OPL chip). At some point, when probing VREF outputs, I somehow managed to internally shunt it. It muted the sound effects completely, leaving the music playing with much less noise. Here's the brief spec:
ALS100 - 100uA 2.25v VREF (very much like some opamp based reference)
ALS100+\ALS120 - 1.5mA 2v VREF (1.5mA makes me think of some cheap-clASS zener reference)
Zeners are prone to noise and ARE actually used as a source for 1/f noise generator circuits.
Next thing I assume, that for a cheap IC, resistor ladder was probably used as a simple DAC. vref noise there may get attenuated to form a 3db point.
In noise generators, resistor ladder is used to filter and attenuate the curve of the noise spectrum.
Of course, you cannot exclude bad op-amps/DAC, but I have not yet found such absurdly noisy ones (on the other hand, maybe I have just found one).
UPD: Poor dynamic range and output gain contributes to the noise problem.

Any additional input, criticism, ideas are highly appreciated!
I've got a lot of files, which need sorting first, then I'm planning to add them there such as:
ALS120 replica gerbers or sources, high-res pcb scans, component values and maybe I'll draw a schematic also, because the datasheet is missing one.

P.S.
Some noise reduction points, not much, but nevertheless:
You can only reduce it to some reference level. Below is the picture of the similar soundcard, where I pointed out the most affecting modifications:
1) A place to solder 7805 regulator, TO-92 package.
1a) Ferrite choke, which connects motherboard's +5v to the soundcard analog power rail. Should be removed when soldering 7805 to avoid shorting their voltages.
2) Ferrite choke, which connects analog and digital grounds to a single connecting point.
3) vREF bypass\buffer capacitors
4) This capacitor will significantly reduce whistling\thinking noises. Many cards miss it. (Monkey Island 2 for instance, make astounding high-pitch whistling without this cap)
5) Digital +5v ferrite choke and bypass caps
6) When crystal is not used\soldered, 100 ohm resistor is soldered there instead to carry signal from motherboard. The crystal itself uses 1.8M resistor in parallel in addition to caps.

P.P.S.
Link to the scans and sprint layout file
https://drive.google.com/file/d/1xrRGQaRi8glw … iew?usp=sharing

Kudos, excellent work! 😎👍

Your contribution is wholeheartedly welcome, thank you very much! 🙂

Very nice effort ! Thänk you ~

One thing that may help is forcing your own VREF from a clean source into the VREF pin(s) and particular voltage can make a significant difference too.

Thank you for researching this so thoroughly.

Excessive noise is the main reason I stopped using my ALS100. Good to know the full story behind it.

Thanks everyone, I do really hope there may be a redemption for it yet )

One more interesting thing, new chip fails to be detected by Avance Logic Drivers 0_o.
Unisound works flawlessly. Maybe there's some ROM info missing. IDK

Tiido wrote on 2022-10-05, 06:53:

Very nice effort ! Thänk you ~

One thing that may help is forcing your own VREF from a clean source into the VREF pin(s) and particular voltage can make a significant difference too.

Thanks for the suggestion!
I've also thought of that, but at the time I was worried that reverse voltage would damage the circuitry. And if it's a shunt based ref - it would just be regulated back with no effect.
Now that I have some IC's to spare and a prototype pcb, I may try it. I have jumper pins there, so it should be easy to test.

UPD:
Here's the result. recorded WAV file, software amplified
First 10 sec - idle card with both vref intact.
10-20sec - just attached the wire to vref 1 pin (some wire jiggling noise) to the output of the powered off power supply. Noise reduced already. If I put the wire out of the powered-off power supply - the noise get louder again.
20-30sec - forcing 0.8 volt on the pin, futher noise reduction
30sec and further - switching between 1.5v and 1.6 volts. Above 1.6v the noise becomes louder.
At exactly 2.4v on VREF1 pin the great portion of the noise is reduced. IDK how healthy it is, though.
The datasheet is sketchy - it points out 2.0V in one place and 2.5v in the other
Forcing voltage to vref2 has little effect. Forcing the voltage to both vref1 and vref2 adds buzzing, maybe ground loop

All I have to do now is to replace LM4040 2V on my prototype PCB with another ref chip and hope I will not fry something
Also, I think I'll replace the stock values of the low-pass output filter from 820 ohm\10nf to 200 ohm\40nf to reduce impedance on the output

Interesting, I need to build some reference bench, because I've already lost track how exactly things get better or worse compared to stock card or others.
The noiuse is still there, but some of it gets reduced.

UPD2. I've added the link to the Sprint Layout files and scan of the PCB to my first post

I don't know if it might help? But I was able to extract the reference card schematics for the various ALS100 cards a couple of months ago: Re: Sound Blaster 16 Clones
Maybe they show some variation from the production cards that would give some indication of the cause.

What you've found so far is great though, I'll have to try out a couple of these modifications on my ALS100+ (or was it ALS200?) card that was extremely noisy.

Thermalwrong wrote on 2022-10-05, 16:32:

I don't know if it might help? But I was able to extract the reference card schematics for the various ALS100 cards a couple of months ago: Re: Sound Blaster 16 Clones
Maybe they show some variation from the production cards that would give some indication of the cause.

Yes, I've downloaded them already, great work by the way!

Thermalwrong wrote on 2022-10-05, 16:32:

What you've found so far is great though, I'll have to try out a couple of these modifications on my ALS100+ (or was it ALS200?) card that was extremely noisy.

I would rather say it's just adding the missing components which were meant to be there in the first place. Nothing of a breakthrough really, it keeps being noisy.

Anyway, I've recorded two 1-minute long samples of idle noise. Default mixer settings
My prototype pc/104 PCB and a reference card with most of the components in place, excluding 7805 regulator.
The reference card appears to be more silent than mine, but also has some harmonics, which the regulator should suppress a little.
Extra point goes for split A/D grounds vs my lousy solid plane prototyping 😀

Wow, almost 4 years like it's been yesterday...
Anyway, this is probably my final thoughts and developments regarding the topic:

UPD 22.04.26
I think I FINALLY got a pretty damn good result!
All that's left is to design a card around all my findings. Schematic related changes will be in a separate thread

UPD 25.04.26
Did a cleanup of the post.

TL:DR \ Current Status (22.04.26)
I've figured out both flaws of this chip and in the process of making a finished project.
I'm now 100% certain in the result.
See below for more details.

---The 2 main problems of these ICs ---

1) Noisy Weak high-z VREF.
VREF pins on this IC are weak and high-z, meaning they act lake vacuum cleaner for all the noise around it.
Avance Logic engineers actually did a good job separating A\D grounds, but cutting corners on components (plain absence) on most of them killed that effort.
I actually completed a full circle debugging that in my design, but it's a story for another time\thread.

2) Noisy FM Synth.
Second main source of the noise. It is a constant hiss noise coming from FM DAC\Core itself and does not relate to any volume settings.
I measured 10-14mV noise to 321mV test 1000Hz sinewave. That's makes an appaling 27-30dB SnR.
Without a way to boost the signal, the noise will always find it's way through, LPF capacitor can only mask some of it sacrificing the quality.
UPD 22.04.26 - SOLVED!
I've found the remedy. As it turns out, mixer register 52h bumps up dynamic range (Vref-P/Vref-N) and makes FM loud enough to lower the noise floor!
Because the noise level generated is constant by FM core and is unaffected, this basically means a bump in SnR! Gain staging is now also a possibility.
It has 4 settings:
Bit 1,0 Vref-P/Vref-N control
00 1.68V (Default)
01 2.11V (+25%)
10 2.53V (+50%)
11 2.95V (+75%)

---What can be done to stock ALS120 based cards (ALS100+ were not tested, don't have any)---

A lot of cards are missing basic filtering components and don't have any voltage regulators, so results vary.
A new design is needed to squeeze every bit of SnR.
But you can still lower the noise floor by A LOT.
Fully populated cards show best results (almost on par with my curent flawed design, actually.).

1) Initialize the card. Keep MASTER about 82, FM at 82-95 and WAVE at 65 (WAVE is way louder on this card).
You can fine tune this to get minimize whistling noise or bump up volume.

2)Copy debug.exe utility from DOS and als_set.bat from the archive below to the same folder.
Run the script. It will prepare a temp file for debug.exe to launch.
It assumes the base address of the card is 220h .
You can check the script yourself and also read more in registers.txt file.
It will set the following:
a) Set mixer L\R gain to 1 to minimize the overall noise level.
This setting is NOT the volume setting of the mixer, but it resets every time you use unisound or set volume.
So you'll need to run the script every time you setup the card, otherwise it will be reset to default (2).
You can also leave gain as is and use gain staging method to get even louder FM with absolutely tolerable noise.
b) Disable mixing to MONO Output.
This setting is mostly obsolete now. I used this to mute MONO Output and then use it as a noise reference for INA2134.
But it does not hurt to have less floating I\O.
c) Bump up the dynamic range (VREF-P\VREF-N).
The HOLY GRAIL setting of this card. The script will bump up FM range to max (+75%)
Making it loud enough to lower the overall noise floor.

3) Next, prepare the gain staging.
a) Disable TEA2025 and use line-out output.
b) Get yourself a good external high-z input preamp, which controls it's volume by using variable resistor at the input.
c) Connect you preamp to initialized card, setup input and gain, until the idle noise you hear is comfortable.
It can be even bet set low as inaudible, but stock cards lack power filtering for that.
Active speakers may also do the trick, but they usually lack good quality amps.

---UTILITIES---
I've attached an archive, which contains:
1) My tiny utility ISA-MIX.EXE to display mixer registers map.
Basically, it should work on any SB Compatible card set on 220h base with mixer setting and data at 224h+225h
Very handy for comparing changes.
2) Batch file ALS_SET.BAT for debug.exe to setup mixer registers, contains comments what does what.
Basically, it creates a temp file, append commands for debug to launch, then deletes the file.
It will set output gain of both channels to 1, disable mono input and set FM gain.
Just copy debug.exe to the same folder with this bat file
After initializing the card and setting volume using unisound.exe, you'll need to lauch this batch file.
I was too lazy to make a programm for something so trivial.

IMPORTANT - VREF setting is commented out as obsolete, you don't need it
---
REM echo o 224 4E >> temp.dbg
REM echo o 225 3C >> temp.dbg
---

3) Text file, which is a copy from ALS4000 about mixer datasheet registers info.
There is none in ALS120 datasheet, but ALS4000 seem to share common register map with ALS120.
This and a tiny hint in ALS120 datasheet about signal boost made this mod possible.

Well, that's it. And the memento of my prototype card (ALS120+INA2134), which should have died a thousand times, but didn't.
Due to personal circumstances, I'm still struggling to bring this project to completion, progress is painfully slow.
At least I can share what I can. Who knows, maybe someone will find these usefull. Cheers!
In time, I will clean up this post and leave only neccessary details.

This will probably be my final post in this thread regarding the project. I'll create a new one once I make a complete working prototype (I hope) and clean things up.
Here are three well known tracks recorded directly from my card, both of which have soft moments to check on noise levels and to compare how the chip sounds now.
Sorry for rather agressive LP filter (my card has 4.7nf LPF FM cap, which makes about 3400hz cutoff, intead of 1nf and 16khz. I plan to make LPF selectable).
But you can still hear the difference in noise level, and also poor crosstalk isolation in "Stalker" (PCM was set too loud) 😀

P.S. Funny and ironic thing is - I spend so much time tinkering and hearing same DOOM sound effects over and over, that I cannot go back to my ES1869 now.
Because ESS can only do 22khz in SB mode, while ALS120 is pseudo-16bit.
Anyway, I'm pretty satisfied with the result. Almost after 5 years, I've found my closure... At Last: Silence...