VOGONS


Sound Blaster 16 Bugs and Deficiencies Summary

Topic actions

Reply 60 of 90, by Cloudschatze

User metadata
Rank Oldbie
Rank
Oldbie
GL1zdA wrote:

Soem SB16 cards have the AK4501-VS DAC - how does it compare to the CT1701 and the various CT1703 chips?

CT1701 = AKM AK4501-VS

The CT1703 is speculatively an AKM AK4503-VS. This would need to be confirmed with decapsulation.

GL1zdA wrote:

...is OPL audio routed through the CT1701 on the CT2760? On the early SB16s the OPL3 audio would go through the Yamaha DAC to the mixer. But I can't see such DAC one on the CT1747 cards.

On applicable cards, the OPL output from the CT1747 is digitally routed through the EMU8000 and its corresponding TDA1387 or TDA1543 DAC. Hence the need to initialize the EMU8000 (via AWEUTIL) to get FM output, and the ability to apply EQ/Reverb/Chorus to those sounds.

Reply 61 of 90, by GL1zdA

User metadata
Rank Oldbie
Rank
Oldbie
Cloudschatze wrote:
GL1zdA wrote:

...is OPL audio routed through the CT1701 on the CT2760? On the early SB16s the OPL3 audio would go through the Yamaha DAC to the mixer. But I can't see such DAC one on the CT1747 cards.

On applicable cards, the OPL output from the CT1747 is digitally routed through the EMU8000 and its corresponding TDA1387 or TDA1543 DAC. Hence the need to initialize the EMU8000 (via AWEUTIL) to get FM output, and the ability to apply EQ/Reverb/Chorus to those sounds.

Does this mean, that on the SB16 cards with CT1747 FM goes to the CT1701/CT1703?

getquake.gif | InfoWorld/PC Magazine Indices

Reply 62 of 90, by Cloudschatze

User metadata
Rank Oldbie
Rank
Oldbie
GL1zdA wrote:
Cloudschatze wrote:
GL1zdA wrote:

...is OPL audio routed through the CT1701 on the CT2760? On the early SB16s the OPL3 audio would go through the Yamaha DAC to the mixer. But I can't see such DAC one on the CT1747 cards.

On applicable cards, the OPL output from the CT1747 is digitally routed through the EMU8000 and its corresponding TDA1387 or TDA1543 DAC. Hence the need to initialize the EMU8000 (via AWEUTIL) to get FM output, and the ability to apply EQ/Reverb/Chorus to those sounds.

Does this mean, that on the SB16 cards with CT1747 FM goes to the CT1701/CT1703?

No, on such cards, the CT1747 is directly connected (I2S) to either a TDA1387 or TDA1543 DAC.

Reply 64 of 90, by Schyz

User metadata
Rank Newbie
Rank
Newbie

Hi,

My experience, if it helps.

- The clicking DMA noise is not as bad, compared with the overall quality of the digital audio of the time, I'd say it's several orders of magnitude less annoying than other issues like the hanging note bug or the fake FM sound.

- "Stuttering with high-rate samples", I could never reproduce this issue, tested with AWE32 3900 (CT1747 chip) and both XG50DB and CM-500, and the games Duke Nukem 3D and Ultimate Doom, which (I think) have above 11KHz samples. Let me know if there is any solid way to reproduce these issues. I remember reading another user with a 3900 reporting the same.

- There is an additional issue, not exclusive to Creative but maybe important to consider in this topic: OPL3 compatibility with fast computers
In fast machines the OPL3 music will generate horrible noise. Some cards are worse than others, my 2760 starts behaving at 133MHz on some games, my 3900 is rock solid on 550Mhz with 100Mhz bus. Some motherboards, Asus especially, have an option on BIOS to set "8-bit I/O recovery time", and this can mitigate the issue, but the option is usually missing from other manufacturers.

I suspect one of the reasons for many sound card manufacturers to go for fake FM sound is to mitigate this issue. The CQM present in late AWE32 and AWE64 cards allows those to work flawlessly on faster machines of the time.

One last thing, I think this topic should be pinned in the audio section, the work done by James-F documenting the Sound Blaster issues is top quality and I wish I knew this info many years (and cards) ago.

Reply 66 of 90, by The Serpent Rider

User metadata
Rank l33t
Rank
l33t

- "Stuttering with high-rate samples", I could never reproduce this issue

It's not exactly stuttering. The bug will cause brief but very noticeable slowdown during the initial loading of the music track, after that performance will return to normal. It's somewhat annoying, but manageable.

to go for fake FM

Different implementation =/= fake. That said, the difference between some cheap ass CQM Vibra and AWE64 Gold can be quite drastic.

In fast machines the OPL3 music will generate horrible noise

Never had any problems with generally quiet OPL3 cards on a 1ghz+ system.

Get up, come on get down with the sickness
Open up your hate, and let it flow into me

Reply 67 of 90, by Schyz

User metadata
Rank Newbie
Rank
Newbie

About the noise on fast machines, I'm not talking noise as in "there is some [ssshh] sound in the background, like a white noise", it's more like "why there is a 4 year-old kid messing with the electric guitar".

It depends on the games as well, some DOS games support fast machines better than others. Also it's more frequent to happen in real DOS mode than under Win9X. Give it a try with a real OPL3 and Monkey Island, running "monkey.exe a", bus 100 MHz and 1GHz CPU speed.

I've been able to reproduce this issue even with Adlib Tracker 2, but it has a setting that allows to increase the latency which fixes the problem("opl_latency=1"), of course games never offer this option.

Reply 68 of 90, by Scali

User metadata
Rank l33t
Rank
l33t
Schyz wrote:

I've been able to reproduce this issue even with Adlib Tracker 2, but it has a setting that allows to increase the latency which fixes the problem("opl_latency=1"), of course games never offer this option.

Sounds like the intrinsic problem in the design of the OPL2/OPL3 chips:
Writing to a register is fire-and-forget.
That is, the write on the bus completes quickly, but the synthesizer needs to complete a full cycle of processing before the written value is applied correctly.
Yamaha has documented the delays between register writes in absolute time.
Problem is, most software doesn't actually perform delays in absolute time, but rather performs a sequence of dummy register reads, under the assumption that a read takes a certain amount of time.
On very fast machines, these dummy reads complete too quickly, which means that the next register write is executed before the previous one was completed by the FM synthesizer. This leaves the chip in an undefined state, leading to all sorts of glitches and random noise.

http://scalibq.wordpress.com/just-keeping-it- … ro-programming/

Reply 69 of 90, by Schyz

User metadata
Rank Newbie
Rank
Newbie

I agree that this is probably more to blame on the game developers of the time than on the hardware, but at the same time it could be a factor to consider when purchasing a Sound Blaster, an AWE64 could be a better choice for a fast computer.

Also, I hope that awareness on the issue will prevent some OPL3 Sound Blasters to end up in the bin.

Reply 70 of 90, by Eep386

User metadata
Rank Member
Rank
Member
Cloudschatze wrote:
GL1zdA wrote:

Soem SB16 cards have the AK4501-VS DAC - how does it compare to the CT1701 and the various CT1703 chips?

CT1701 = AKM AK4501-VS

The CT1703 is speculatively an AKM AK4503-VS. This would need to be confirmed with decapsulation.

I tried swapping a CT1701 on a SB16 SCSI board with a CT1703-T, and got no PCM sound, so I don't think it's compatible at any rate.

Life isn't long enough to re-enable every hidden option in every BIOS on every board... 🙁

Reply 71 of 90, by Eep386

User metadata
Rank Member
Rank
Member

Apologies for reviving this ancient post, but I wanted to mention that the CT1703 series parts seem to be made by ROHM. Looking a little closer it seems Asahi Kasei might have also used ROHM as a fab/foundry, but I don't think the CT1703 is compatible with the CT1701.

Also, it is possible to de-CQM some cards with CQM synthesis. If there is a CT1978 present and there are empty pads for a YMF289 plus YAC516, the card in question can be de-CQM'ed. This will involve adding additional components as well, such as resistors, capacitors and in most cases a 33.868 MHz oscillator. (Some cards, such as the CT3600, instead need an 18 ohm resistor to connect the YMF289B chip's clock input to the onboard clock chip.) Unfortunately virtually every AWE64 in existence is forever stuck with this blight as the CQM function is built into the main chip.

Life isn't long enough to re-enable every hidden option in every BIOS on every board... 🙁

Reply 72 of 90, by GL1zdA

User metadata
Rank Oldbie
Rank
Oldbie
Eep386 wrote:

Apologies for reviving this ancient post

I actually hope this thread won't ever die, I always like it, when people report peculiarities of their SB16ss.

Eep386 wrote:

Also, it is possible to de-CQM some cards with CQM synthesis. If there is a CT1978 present and there are empty pads for a YMF289 plus YAC516, the card in question can be de-CQM'ed. This will involve adding additional components as well, such as resistors, capacitors and in most cases a 33.868 MHz oscillator. (Some cards, such as the CT3600, instead need an 18 ohm resistor to connect the YMF289B chip's clock input to the onboard clock chip.) Unfortunately virtually every AWE64 in existence is forever stuck with this blight as the CQM function is built into the main chip.

I've seen a CT2940 (SB16) in both versions and both the CT3600 (SB32) and CT3990 (AWE32) have the solder pads, but I've never seen a configuration with a YMF chip. It would be especially useful for the CT3990, since it would make it on par with the CT3980 and a close second to the CT3900 (like CT3980 but no PnP).

De-CQM thread

Last edited by GL1zdA on 2019-09-05, 07:52. Edited 1 time in total.

getquake.gif | InfoWorld/PC Magazine Indices

Reply 73 of 90, by GL1zdA

User metadata
Rank Oldbie
Rank
Oldbie

Reading another thread (Re: Duke Nukem 3D stuttering with Dreamblaster X1), I've begun to wonder what the other sources of bad audio are. The old CT1701, CT1703-T and AK4501 are known to be noisy, but there's still the problem with the MIDI sources. Is there a possibility, that the 3403 op-amps could be the source of the hiss? I've seen at least three different ones. One from STMicroelectronics, one from On Semiconductors (the MC3403) and one from Fairchild (KA3403). I know at least one is used for FM MIDI (the YAMAHA YAC requires one to bring the signal to line-level). But there are two to four of them on other cards. What uses the other one?

Does the CT1745A mixer operate at line-level? I know it's an analogue mixer, but I want to know where the analogue signal is modified. I assume the volume control is inside the CT1745A?

getquake.gif | InfoWorld/PC Magazine Indices

Reply 74 of 90, by auron

User metadata
Rank Member
Rank
Member

does shadow warrior fall under the category of the vibra distortion bug described in the OP? because in an attempt to diagnose the random crashes in the game i had described in this topic, i swapped in a ct2800 vibra card and discovered that the sound in the game is severely broken, see attached recording (8 voices, 16-bit 22khz recorded from DOS). the exact same thing happens on a ct2860 as well.

putting ingame volume at minimum does zero to help this, using 8-bit and lower sample rate obviously changes it but even then i can hear some distortion in the muffled audio. interestingly, in a quick test i was not able to spot this issue in duke nukem 3d.

incidentally i did reproduce the ringing bug in tyrian on the ct2860 and that can be annoying as well when it's not drowned out by sfx/music, but shadow warrior is downright unplayable on vibra cards.

Attachments

  • Filename
    sw vibra ct2800.mp3
    File size
    914.28 KiB
    Downloads
    23 downloads
    File license
    Fair use/fair dealing exception

Reply 75 of 90, by Joseph_Joestar

User metadata
Rank Oldbie
Rank
Oldbie

It appears that my Sound Blaster AWE64 Value (CT4520) has the Vibra distortion bug. I checked Skyroads and Shadow Warrior and experienced the same issues as presented in this thread. This distortion also shows up in Mortal Kombat 3 during certain moves like Sub-Zero's ice shower.

I'm curious if this affects all AWE64 cards or just the highly integrated CT4520 model. Can anyone with an AWE64 Gold test this? It's easiest to check in Skyroads, just crash the ship into the railing. The game is now freeware and can be downloaded from here.

Build #1: Celeron 466 / Abit ZM6 / Voodoo3 / AWE64 / YMF744 / SC-155
Build #2: AthlonXP 1700+ / Abit KT7A / GeForce4 / SBLive / ALS100
Build #3: Athlon64 3700+ / DFI LanParty / 9600GT / X-Fi Titanium

Reply 76 of 90, by Tiido

User metadata
Rank Oldbie
Rank
Oldbie
GL1zdA wrote on 2019-09-05, 06:49:

Does the CT1745A mixer operate at line-level? I know it's an analogue mixer, but I want to know where the analogue signal is modified. I assume the volume control is inside the CT1745A?

That chip certainly plays a role as all the signal go through it, it is what applies the volume controls etc. and will have effect on noise and distortion performance, though to what extent I don't know. I don't think any measurements have been done either. There are some revisions of it and the chip is also made by multitude of manufacturers.

T-04YBSC, a new YMF71x based sound card & Official VOGONS thread about it
Newly made 4MB 60ns 30pin SIMMs ~
mida sa loed ? nagunii aru ei saa 😜

Reply 77 of 90, by foil_fresh

User metadata
Rank Member
Rank
Member

i got a SB16 CT2830 recently (this is some ASP model or something, but the sound processor chip slot is empty so i dont know if that matters a lot or not) and yeah i got a question about the dma clicking bug.

is there any way to tweak mainboard memory settings to alleviate the amount of clicking and popping? or modding firmware (if it has anything to flash)?

does the DMA clicking bug occur on slow and fast PCs?

i really love the sb16 quality as well as the nice output. the FM sounds strong and the bass is punchy. the DMA bug is not really bad but its certainly noticable. if the high and low DMA channels are set the same, will this make the bug worse/better/nothing will change?

cheers

Reply 79 of 90, by Eep386

User metadata
Rank Member
Rank
Member

One more fault of many models of the SoundBlaster 16:
At least two models suffer from a rookie design mistake: unterminated (floating) portions of quad op-amps. These cards have at least one part of a quad amp floating with no defined input or feedback, so they generate insane amounts of noise. A link to a topic on why floating op-amps are so bad for noise: https://www.maximintegrated.com/en/design/tec … tes/1/1957.html

(The reason unterminated amps do this, is because without defined input levels and proper feedback, the amps are effectively running with extremely high gain on noise currents, so they flop in the electronic breeze like laundry in a windtunnel. You can imagine this will inject huge levels of noise into the power rails, and often disturb the operation of other amps in the package as the amps in most quad-package amps are not completely independent.)

Fixing this issue is simple but requires some flying wires: on your typical garden-variety SB16 with dual supplies (7805 and 7905 regulators), it's a matter of running the non-inverting input to ground, and bridging the output and inverting input pins together to turn them into unity-gain followers holding steady at around 0V. (Conversely, if you were to fix this problem on a card running the amps in single supply - test the voltage rails with a multimeter first to see if one side is connected to ground - you'd instead hold the non-inverting inputs to 1/2 Vcc to avoid railing the amp low.)

On the first-revision (TDA1543T DAC) CT2760 AWE32, the entire left half of U22, the right half of U27 and the bottom half of U30 are floating. This is significant as it's one of the cards with the 'noisy' CT1701-T mixer chip, and I've a hunch that this is why this card seems so noisy, rather than it being merely the mixer chip. Note the amp's proximity to the CT1701-T as well.

CT2760-Float1.jpg
Filename
CT2760-Float1.jpg
File size
272.22 KiB
Views
212 views
File license
Public domain
CT2760-Float2.jpg
Filename
CT2760-Float2.jpg
File size
356.73 KiB
Views
212 views
File license
Public domain
CT2760-Float3.jpg
Filename
CT2760-Float3.jpg
File size
455.43 KiB
Views
212 views
File license
Public domain

Ironically the cheapy CT3600 SB32 is free from floating amps. If only they all were like that... 🙁

On the CT2940 ViBRA16S, U5 and U7 have floating, unterminated amps at pins 12-14 each. Note that I've already fixed the amps in this picture, just note the pins circled in red.

CT2940-Float.jpg
Filename
CT2940-Float.jpg
File size
254.99 KiB
Views
197 views
File license
Public domain

Unfortunately there are other SB16/ViBRA cards with floating amps, but these two are the ones I have convenient that I know for sure have floating amps. (So please don't ask me where the floating amps are on a CT2770 or CT2230 etc., because I don't know at the moment - I am still finding out just how many of these cards have this problem.)
In any case, finding the 'floating' amps requires removing the amps to see which ones are not tied to anything at all. If there are no traces or vias connecting pins on an amp, odds are extremely good Creative just left that part of the amp to float.

Last edited by Eep386 on 2021-01-03, 20:27. Edited 2 times in total.

Life isn't long enough to re-enable every hidden option in every BIOS on every board... 🙁