How would one get these graphs?

There graphs are recorded with a profession audio interface and a studio recording software along with VST plugins in the recording software, by no means an easy task for the non recording-hobbyist average Joe.

The most effective thing one can do to lower the noise is to zero the Mic, Line-In and CD in the Mixer, DOS or Windows.
The CD cable is a noise monster.

Tinkering with the ES1688 Mixer I could get a distortion free signal at -4dbFS Peak with -80dbFS RMS noise floor, that's 76db of SNR which is nothing short of amazing for a DOS sound card.
The SB-Pro emulation Left and Right are swapped, but when I set it to ESS (or in windows98) the channels are correct, so it looks like they emulated the SBP with all it's flaws.

Neat 😉

James-F, what do you use to produce white noise in DOS?

How would you test the frequency response of FM? Because that would expose if any filtering is limited to digital, or affects the entire output. And maybe show bass-heavy cards?

James-F, which has better SB-16 emulation. The Sound Blaster Live or CMI-8738?

Also, does YMF-718 have Voyetra Super SAPI FM drivers for Win 3.x?

PhilsComputerLab wrote:

James-F, what do you use to produce white noise in DOS?

Use Mpxplay (best dos wav, mp3, etc player) to play the attached audio files in DOS.
https://sourceforge.net/projects/mpxplay/file … pxplay%20v1.61/
Download the D version (DOS).
Note that Mpxplay leaves the SB-Pro in Stereo mode, so use the sbpmono.com fix from the SBP bug thread after you quit.
SB16 is fine.

To actually measure the spectrum you have to use a VST spectrum analyzer plugin, and a sound card which has quieter inputs than the subject under test (important!).
You can use SAVIHost to load a VST plugin without having a full blown DAW software, make sure you set the sound card input (Devices->Wave).

Voxengo SPAN is currently the best freeware frequency analyzer which I and countless others use.
http://www.voxengo.com/product/span/

Use the settings in SPAN I provided for accurate reading.
Block size makes the analyzer more accurate but slower.
Slope has to be zero for flat reading of the white noise.
RT Max, for real time reading.

Red arrow is where you take the noise floor reading when not playing anything.
Blue arrow is where you take the Peak reading when playing the 440_odb.wav file with Mpxplay.
Do a subtraction (delta) to get the SNR value.

To maximize signal to noise ratio you have to play with the SB mixer.
Typically, raising the WAV (SB) volume does not raise the noise floor, but the AUX (wavetable) does.
The MASTER volume raises the noise floor and might created distortion at high volumes.
The 440_0db reading should be pure as possible, if you see other frequencies there is distortion.

.

Here are readings from the ES1688 in SBP mode:

Noise floor:

Equal frequency White Noise:

Pure 440Hz sinewave:

The input noise floor of my audio interface is 20db lower (-102db RMS) than the noise floor of the ES1688 (-82db RMS) so it should be fine for testing.
In the first picture of the noise floor you can see more noise in the lower frequencies which it typical for old sound cards and PCs.
Second picture you can see the ES1688 has a steep lowpass filter at 10k.
In the last picture you can see some harmonic distortion along with the pure sine wave, you can also see the peak db reading.
Subtract Peak 440 from RMS noise floor to get signal to noise ratio (SNR).

Actually, I can set Mpxplay to ESS mode and the ES1688 becomes a 44.1kHz 16bit filter-less audio card just like SB16.
So if the game supports ESS, it actually very similar to SB16 or WSS in DOS.
But as you can see the WAV output is not the cleanest, it has some even (2nd, 4th) harmonic distortion (also in SBP mode), which is not that bad sounding as odd harmonic distortion.
We actually want to see nothing but the pure 440Hz tone with the noise floor, which is the case with modern sound cards.

Another excellent program to analyze the spectrum and show the THD+N is REW (Room EQ Wizard):
0.07% odd harmonic distortion is nothing I would worry (or hear) about, especially in SBP mode where the 8bit noise is high.
ES1688 in ESS mode in DOS.

colpoz wrote:

It would be interesting to know if the ES1868 is as good as ES1688... it seems to be easier to find. 😀

The ES1688 is natively supported in some later DOS games for what its worth, where the ES1868 isn't. And in my experience the ES1688 is less prone to nasty pops when the system is powered on and off, but that probably just comes down to implementation more than anything else. It's common to find ES1868 based cards with a real OPL3 on them, but I've never seen this on a ES1688 card. There's also the ES688 chipset to throw into the mix - this can be found with a real OPL3 but the wavetable header needs a TSR to work, where the ES1688 is implemented in hardware (you need to initialise it with software, but no TSR required).

I was also wondering if the ES1688 could be found with a real YMF262.

I was never impressed with the ES1868, always preferred my trusty YMF-718 cards. But now i'm curious about the ES1688, although i'm not too excited about its "enhanced FM". It's nice that the ES1688 includes the ADPCM modes that make it work correct on the games that need it. This was never a problem for me because i don't like Duke Nukem II, but i always wondered which gaes require these modes to work correctly.

Besides the grungy instead of buzzy noise of one specific sound of the ESFM due to slightly different feedback mechanism, the ES1688 card is basically the perfect SB-Pro2 with Wavetable header.
Grab one while they still available.

I want to add that the IRQ should be set to 7 for better compatibility with older dos games.
http://www.oldskool.org/guides/oldonnew/sound

Also, lower FM volume in the ESSVOL to 5 (from 😎, this makes the FM sounds more balanced with the SB sounds in volume.

MaxWar wrote:

CybersphereDemo - Here we have a genuine OPL3 game, it makes full use of the OPL3, not just for stereo like Descent. It cannot run on a OPL2.

According to suggestions on Vogons, I downloaded the Cyber Sphere Tunes Demo from the official website:
http://web.archive.org/web/19971008033943/htt … ellmanc/pc.html
Download cybtunes.zip for the OPL3 tunes demo only.

This is a great little demo for testing the OPL3 and its clones.
The tracks are one of the most complex OPL3 implementation I've heard and sound fantastic.
I ABX the tracks and could not hear any difference between YMF719 and ESFM, more precisely I could not hear the use of the single specific noise sound the ESFM differs to the YMF.

In the name of better understanding I used Adlib Tracker II to manipulate the ESFM and OPL and test the difference of the feedback behavior.
There are 7 levels of feedback in the OPL chip with each wavefrom, so I have recorded all 7 on Square and Sine with the ESFM and Nuked-OPL*.
The Operator configuration is set to parallel with #2 (Carrier) being completely off, so you hear only one Operator (Modulator) with various feedback levels.
* The Nuked-OPL is a bit perfect emulation of the YMF262 made by nukeykt.

Inside the ZIP there is an ESFM folder with 0-7 wav files, and a OPL folder with 0-7 wav files.

Square:
0 sounds identical between the two because there is no feedback.
1-3 gradually beginning to sound different on the ESFM to OPL.
4 is where the Stunts track from the first post would be, still musical on the ESFM but barely on the OPL.
The biggest difference is 5, it sounds completely broken on the OPL but quite good on the ESFM. 😐
6-7 also sound identical between the two because the feedback is so large there is practically no sound difference.

Sine:
0-3 sound identical.
4 beginning to sound different.
Again 5 is where the biggest difference is, on the OPL there is nothing but a thin glassy hiss, but the ESFM is much more musical.
6-7 have little difference in noise.

Conclusion:
The ESFM has a slightly different feedback distribution curve than the OPL.
In most cases the OPL sounds thin and buzzy/glassy with high feedback levels but the ESFM retains a more pleasant musical sound across all feedback levels.
I certainly don't mind the difference with DOS games that the ESFM produces (rare as they may be), in most cases it sounds even better (blasphemy I know). 😀
The reason the difference is so small in games is because the developers did not like/use the buzzy/glassy sounds produced with high feedback except for noises.
I guess this thread puts the nail in the coffin for the great ESFM vs YMF262 mystery, and allows the user to finally hear for themselves and decide if the ES1688 is good enough for dos gaming.

I also attached a YMF715x pdf which explains how the OPL chip works for informational and archival purposes.

Mate I'm loving your commitment to this comparison - I always suspected the AudioDrive was a good thing but it's great to see it scientifically proven 😎

So... I thought why the heck not and installed both cards in my Pentium.

ES1688 on: A220 I7 D1
YMF719 on: A240 I5 D0

ES1688 mixer: FM OFF, Line-In ON.
YMF719 mixer: Everything OFF, FM ON.
Loop from YMF719 Line-Out to ES1688 Line-In.
Adlib is on Address 388 on both by default.

Now I can freely choose between anything I want using the mixers and game setups.
ESFM, just mute the YMF OPL and point the game music setup to the standard ES1688 address.
Adlib automatically addresses to port 388 so both cards work at the same time with older games and newer applications like Adlib Tracker, I mute FM in the mixer of the card I don't want to hear.
For OPL, I select in game music setup SBPro2 and choose address 240 where the YMF lives or choose Adlib in port 240.

Older games go straight for Adlib on address 388 so they simply work with sound from the ES1688 and OPL music from the YMF.
Newer games I can freely choose what I want from the setups.
There are problematic games that insist using the same address for music and sound or alter the mixer, in these games I have to use ESFM or use the YMF addresses and its sound.
Some games require a change of SET BLASTER= variable to have sound from the audio card I choose as the primary, most old games want 220,7,1 in any case.

I can confirm the Adlib Tracker II sends signal to address 388 so I can freely choose what OPL I hear from what card using their mixers.
Anyway, it's way more hassle but I kinda like having the best of both worlds and full control, the perfect SBPro2 clone from the ES1688, and the real OPL3 from the YMF719.

EDIT:
One more thing, I had to change the MPU address on the YMF to something else than 330 because when both set to 330 it's a hanging note festival. 😎

Any other games that use ADPCM modes besides Duke Nukem II? I still prefer the authentic YMF sound to the enhanced ESFM, so if' its only DNII i can't be bothered about this to make the switch.

I'm still inclined to the pairing of CS4232 + YMF262 being the best option for DOS. The Turtle Beach Tropez cards fall into this category i think.

I agree about CS4232 + YMF262 being the crème de la crème but the ES1688 can stand on it's own without missing a thing for games, most people will not notice any difference without a direct comparison.
The YMF71x card is also a fantastic fully functional SBPro2 card with real OPL3, but I'm bothered by the sharp filter-less sound more than not having a real OPL3 in the ES1688.
Another thing is none of the above can set the wavetable return volume in the mixer besides the ES1688.
Currently I have settled on using both cards together so I can freely choose what OPL I'm hearing.

James-F wrote:

I agree about CS4232 + YMF262 being the crème de la crème but the ES1688 can stand on it's own without missing a thing for games […]
Show full quote

I agree about CS4232 + YMF262 being the crème de la crème but the ES1688 can stand on it's own without missing a thing for games, most people will not notice any difference without a direct comparison.
The YMF71x card is also a fantastic fully functional SBPro2 card with real OPL3, but I'm bothered by the sharp filter-less sound more than not having a real OPL3 in the ES1688.
Another thing is none of the above can set the wavetable return volume in the mixer besides the ES1688.
Currently I have settled on using both cards together so I can freely choose what OPL I'm hearing.

I haven't done it yet but thanks for the detailed comparisons. Very much appreciated. I will try to find a good deal on a ES1688 Audiodrive. I have a Guillemot Maxi Sound Lite with a ES1868 and was unimpressed by it. I wonder if the ES1688 and ES1868 are pin compatible and require the same logic. That way i could desolder the ES1868 on the card and solder a ES1688 on it.

You should do a comparison with the PCI YMF-724 cards, if you have a mobo that supports DSDMA.

I have a YMF-718 on my 486 machine and I've noticed that that card seem to not be able to manage the FM's volume adjustments into videogames.

For example, the first music in Doom2 menù should start at a low volume and then increase... it does that on all my Sound Blaster cards but with the Yamaha the music starts and remains flat at the maximum volume.

Another example is in FIFA International Soccer config menù, when testing the sounds it allow you to adjust the volume by pressing left-right arrows button, but on Yamaha the only result I got is to mute the card or to have full volume.

I think this is a considerable limitation of YMF-71x...

Games can't adjust the YMF71x mixer.
Wolfenstein3D will not have Stereo sound because of this.
Doom, Duke3D will not be able to adjust FM Music volume within the game.

EDIT: There is a fix for that in THIS thread.
The mixer problem is no more.

The ESS 1688 is not perfect either.
It has some noise from the Wavetable header amplifier.
I also can hear strong noise when lowering all Doom sounds to zero.

Very interesting. I am actually going to be building my 'final' time machine in the foreseeable future and I just grabbed one of these cards for $20 shipped. It looks exactly like the one you have. I really want to inspect its layout and remove the two LM386s and replace them with a low noise opamp for even less noise.

EDIT:

this is the card