First post, by rpocc
Hi. Yes, I know that it's not the first thread of that kind, but I already have the data, so why not share it?
WIth the first intention just to test all of my ISA cards on the subject of their value for sale and malfunction detection, I've got deep into a rabbit hole and made a comparison of currently 22 sound cards with 18 of them different. The method is quite universal and repeatable, so probably I will add more later when I have more ISA cards at my workshop and enough time.
The testing procedure:
I have picked one particular song from the modules collection coming with the latest version of ADLIB Tracker 2, WHATSLOV by ENCORE. (adlib-adaptation of What Is Love? by Howard Jones) This song was played right after PC's boot using the bundled command-line player.
I used a Pentium MMX-166 test rig, feeding the soundcard's line output directly to Zoom H6N line inputs with -20 dB attenuator on and sensivity set to 4. Record format is WAV 24/48. Then all files's volume was precisely matched, using DAW and mastering tools at my studio workstation. All measurements were made at this point. Then all files were also normalized to the standard volume level for Internet -14 LUFS long-term using FabFIlter Pro L-2 and exported with 16-bit dithering, then converter to mp3 using BladeEnc @ 320 kbps. I could upload the WAV files directly but audio hosting I've found doesn't support uncomressed audio.
Measured parameters:
1. Absolute attenuation (reference is CT4520, as the loudest in the set, 3.3 dB lower than emulator's output recorded via digital feedback in my RME Fireface-800)
2. Relative noise level (after level matching, which I found more practically useful)
3. Average frequency response of a 30-sec (or so) material of the song.
4. Frequency response of noise. (uniform noise is less noticeable or annoying than complex noise with certain bands or frequencies emphasized)
5. Phase (which has practical mean for music production and speaker wiring)
6. On/Off clicks (not included, still consolidating, but mainly all of them have no mute circuits, so unsafe for good speakers)
Most of cards are in working condition but never recapped, altough with output capacitors within reasonable range for their original value. When the card is suspicious or any repair job is made, it's noted specifically.
TLDR:
The comparison sheet contains links to audio sharing service containing the full track, the same as below. Other archives contain spectrum analyzer pictures per each unit.
A little summary in advance: From my set the best CQM cards for adlib music are CT4500, CT4180 and CT3600. The best OPL3 cards seem to be CT2230, CT2290 and HP MM Pro based on Aztech OPL clone.
Now, let's proceed to the list!
A typical AWE64. The whole frequency response is a bit skewed towards the high-end with the center around 1.5 kHz. Quite noisy. Annoyingly loud ON/Off pop/click
All CT4520 have boosted bass and high-end. Probaby a marketing thing but actually it sounds great without any need in equalization. A true working horse.
Noise spectrum: 86 Hz with harmonic series + spike around 7K. Mediocre.
Phase: 180°
Attenuation: 0
Relative noise RMS: -69.2
A typical AWE64. The whole frequency response is a bit skewed towards the high-end with the center around 1.5 kHz. Quite noisy. Annoyingly loud ON/Off pop/click
Noise spectrum: 86 Hz and lots of harmonic spikes. Mediocre.
Phase: 180°
Attenuation: 0
Relative noise RMS: -66.9
A typical AWE64. The whole frequency response is a bit skewed towards the high-end with the center around 1.5 kHz. Quite noisy. Annoyingly loud ON/Off pop/click
I have measured this one the last because I had to replace 74LS245 buffer causing glitchy FM.
Noise spectrum: 86 Hz and lots of harmonic spikes. Mediocre.
Phase: 180°
Attenuation: -0.100000000000001
Relative noise RMS: -67.9
This one has a separate DAC: CT1703A unlike later AWE64.
Reference AWE64 CQM sound. Wide, rich and even spectrum, low noise. Has a slight peak at 12.9 kHz, common for all CQM cards.
You see an AWE64, black sockets? Just take and put in place of one you have with coloured sockets.
Noise spectrum: Uniform + peak at 400 Hz.
Phase: 0°
Attenuation: -4.1
Relative noise RMS: -76.7
CT1749-DAQ, CT1978-BAP, CT1703-A, CT1741 ver 413, CT1972 Rev A, EMU8000
LF below 95 are attenuated for 3.3 dB/oct. 100..1500 Hz band is boosted up to 1 dB, and 1.3..20 kHz is weakened up to 5 dB.
I owned this card having built-in 512K of memory back in early 2000-s and even written some songs using soundfont engine as a drum sampler. Not the best in the list but likely good than bad. Had to replace the damn cracky SIMM slot.
Noise spectrum: Uniform.
Phase: 0°
Attenuation: -6.5
Relative noise RMS: -78.2
CT5202, EMU8000, EMU8011-01, CT1978
The best SNR. FR is similar to CT3620 but with a shelf around 1.4 K, lowering HF even more for 1 dB. Quiet on/off click
Noise spectrum: Uniform + spikes at 8.45, 15.2 kHz.
Phase: 0°
Attenuation: -6.7
Relative noise RMS: -80.5
CT2502, EMU8000, EMU8011-01, CT1978, MK1467S
Close to CT3620 and 3600, but all above 330 Hz is boosted for 0.5 dB/oct, and below 330 Hz is attenuated for about 3.4 dB/oct. Quieter on/off click
Noise spectrum: Uniform, brighter + spikes at 8.46 kHz.
Phase: 0°
Attenuation: -7.1
Relative noise RMS: -78.8
It's quite confusing that AWE32 considered rare and profesiional-grade soundblaster actually sounds weaker than AWE64. On the other hand, I bet that replacing the output capacitors with a twice value should boost the low end almost the same way as on AWE64. Also, even original capacitors might loose their initial capacity. Nevertheless one of the main quality of these cards is great signal to noise ratio.
Creative CT2505-TDQ2
Frequency response is the closest both to emulator and CT2230. Noise level is moderately low.
This card performs surprisingly good for a very common Vibra16 card with no way as luxury look as AWE32 or old and long SB16 cards.
Noise spectrum: Couples of spikes around 160, 320, 3K, 6K, 15K, 18.5K.
Phase: 0°
Attenuation: -2.2
Relative noise RMS: -74.2
CT2502-SDQ, CT1978-TAP, marketed as Vibra16
Frequency response is identical to CT3630. Low noise. Quieter on/off click
Noise spectrum: Spikes at 616 + 15.2.
Phase: 0°
Attenuation: -6.8
Relative noise RMS: -79.7
CT2502-SDQ, CT1978-TAP, marketed as SB16
Identical to the other CT2940 with a bit less low-end. Low noise.
Noise spectrum: Spikes at 2.58 + 15.2.
Phase: 0°
Attenuation: -7.6
Relative noise RMS: -78.7
10. CT2230 b680757
CT1745A, CT1747, CT1741, CT1703A
Reference OPL sound, balanced, rich mix and even, professionally-looking spectrum. A peak at 16.3 kHz common for all OPL3 cards. Noise measured with some residue of undermuted voices.
This card at first had unbalanced output, so replacing the output 1uF electrolytic caps with 1uF Epcos film capacitors did a great work.
Noise spectrum: Mostly uniform.
Phase: 0°
Attenuation: -7.8
Relative noise RMS: -76.2
11. CT2230 b535876
SB16 CSP (CT1741, CT1745, CT1748, CT1747, CT1703-A)
The same as the reference CT2230.
Noise spectrum: Little spikes at 286, 1.45 and 12.2.
Phase: 0°
Attenuation: -8.7
Relative noise RMS: -76.1
12. CT2290 050234
CT1745A, CT1747, CT1741, CT1703A
The same as the reference CT2230, even the chips are the same.
Noise spectrum: Perfectly uniform.
Phase: 0°
Attenuation: -7.6
Relative noise RMS: -77.5
CT2501-TDQ, OPL YMF262-M, 0148229000
A Waste of Yamaha chip. Produces awful squeal at 10 kHz and shows lack of low-end. One of the worst soundcards, maybe it's even defective with no obvious signs of that.
Noise spectrum: Lots of very distinctive spikes. Horrible.
Phase: 0°
Attenuation: -6.3
Relative noise RMS: -60.4
Now, few cards from other manufacturers
Essentially, an Aztech card: AZT2316A, CS4231-KL, AZT816, YMF262-M
Kinda "tape-ish" sound, no extra-LF and HF. Close to emulator in the band 100-5000 Hz.
Noise spectrum: Harmonic series of 2.78 kHz. Quite annoying.
Phase: 180°
Attenuation: -12.5
Relative noise RMS: -73.1
15. SoundPro
HT1869V+
Inverted right channel, Incorrect mid-side balance (-9 side, +2 Mid), hense the lack of stereo-field, sharper transients in the mid.
This card has just two chips: HT1869 and mysterious CMA8300 appearing to be a dual OpAmp but with weird pinout. It also has S/PDIF In/Out pins. Very contradictional card bacause I like its sound as it comes from speakers but messed-up phase is the last thing you expect from something having "Pro" in its name. However, sometimes inverting one channel in a stereo output is used in electro-musical instruments, which helps to eliminate noise in the mid-channel, caused by interference in a long cable from stage to the mixer and inverting one channel back, but it's barely applicable for typical PC setup with 1.8 meter cable going to consumer speakers without any phase inversion button. So in this context to me it appears an error in schematic.
Noise spectrum: Even, perfect uniform noise with a slope of -2.4 dB/oct.
Phase: R: 180°
Attenuation: -6.5
Relative noise RMS: -74.4
Single Yamaha YMF719-S
A "tape-ish" sound: a bit saturated, but with wider spectrum. Closest to emulator. Very noisy.
Noise spectrum: Uniform, raising towards low end + spike at 8 kHz and some hum.
Phase: 0°
Attenuation: -10.6
Relative noise RMS: -58.3
17. HP MM PRO 16V-A
A cheap triangle-shaped Aztech: Aztech Systems AZT2320, "OPL"
A bit less ultra low-end and high-end is less detailed. But overall OK-ish. Less body. Close to emulator.
Noise spectrum: Very even, uniform noise + spikes at 11K, 14.9K.
Phase: 180°
Attenuation: -11.1
Relative noise RMS: -73.1
18. HP MM PRO 16Y-A
A cheap triangle-shaped Aztech: Aztech Systems AZT2320
Almost like 2230 but less rich, more squeezed. Close to emulator. By the way, Creative CSP resident hangs on the system if this card is installed, so it's a little bit different from the other one.
Noise spectrum: Very even, uniform noise + spikes at 11K, 14.9K.
Phase: 180°
Attenuation: -11.2
Relative noise RMS: -71.9
ES1868F, 8-bit-compatible ISA connector!
Awful, telephonish sound missing even a little sign of low and high end. Channels sound different and out-of phase. Most likely the card is defective or very, very bad. No 16.3 kHz peak though.
Noise spectrum: Uniform.
Phase: Uneven
Attenuation: -8.7
Relative noise RMS: -79.4
20. ESS AudioDrive SE258V544677
ES1868F
A bit less of low-end than with CT2810, without that awful hiss. Low bit depth. No 16.3 kHz peak though.
Noise spectrum: Uniform, raising towards low-end.
Phase: 180°
Attenuation: -9.4
Relative noise RMS: -77.2
21. SST-1868 4928610
ES1868F
Sounds a bit brighter than CT2810. Low bit depth. No 16.3 kHz peak. Low noise.
Noise spectrum: Uniform.
Phase: 180°
Attenuation: -11.9
Relative noise RMS: -78
