VOGONS


First post, by Physikant

User metadata
Rank Newbie
Rank
Newbie

Hi guys,

I'm currently trying to upgrade the sound setup of my DOS-PC. It currently runs a Gravis Ultrasound Extreme, but even with its ESS chip, it has some trouble compatibility-wise.
Well, so i thought: why not get just the most standard sound card, a good old sound blaster.
I picked the SB16 because of it's true OPL3 in combination with a WTH, and my CTM1740 even has an old enough DSP so the hangig note bug is not present. But they call this thing the noise blaster! Thats painful.
Well, I have quite some experience in electronics development and reverse engineering. Perhaps something can be done here?

Enough of the stories, here is the reason for this post:
I'm thinking of options to upgrade my SB16 to reduce the noise blaster-ness of it. I've read that the DAC is one of the bigger problems here, especially the old versions like the one on my card. And the wavetable header is another source of noise.
So here is my question:
Do you guys have any ideas what could possibly be done to this card to improve its sound quality? I would start with replacing all its electrolytic capcitors with tantalum/ceramic types, just for its longevity. Another idea would be transplanting a newer version of the DAC from another card to this one (if the interface is the same? any infos here?). Perhaps adding some filtering to the output? Replacing the amplifier circuit?
I know that this is some kind of sacrilege doing such ugly mods to a good old card, but the SB16 is not that rare and for me, compatibility and authenticity of the sound are my main goals, but perhaps without the unwanted (and unintended!) noise.
I hope you guys can help me 😀

Nikolas

Last edited by Physikant on 2022-02-16, 22:17. Edited 1 time in total.

Reply 1 of 15, by TheMobRules

User metadata
Rank Oldbie
Rank
Oldbie

Before doing any hardware mods on that card I would start by disabling the internal amplifier, which introduces LOTS of noise. After all, you're going to use amplified speakers, right? So no need for it.

Just set the OPSR and OPSL jumpers to the 1-2 position and the amplifier will be disabled. It makes a big difference in output quality.

Reply 2 of 15, by S95Sedan

User metadata
Rank Member
Rank
Member

Why would you want to do so, is it even noisy in your setup? Its not a given, even with the internal amp enabled.

From what i know you dont want to go from electrolytic to tantalum, they both serve different purposes.

Swapping out the CT1701 to a CT1703 wont work it's not compatible;
Re: Sound Blaster 16 Bugs and Deficiencies Summary

It might have floating though opamps section which can contribute;
Creative ISA Sound Card Floating Op-Amps Running Log/List

Reply 3 of 15, by Physikant

User metadata
Rank Newbie
Rank
Newbie

Thanks already!
Sad that the DAC is not compatible.
I still think a solid recap is a good idea. Of course some care has to be taken in part selection, but I think for buffer caps in the audio frequency range, tantalum and electrolytic behave quite similar, besides differences in ESR, and there are pretty good tantalums with low ESR out there. Even if I can't match the ESR of these old electrolytics, some additional small ceramics in parallel will solve this.
For coupling capacitors, hybrid polymer electrolytic caps are a modern alternative. I will try to avoid the old wet electrolytic capacitors as much as possible.

The floating operational amplifier thing sounds so terrible. Thats something you learn so early on in analog electronics design. I will check my card for these errors. Perhaps the amplifiers and linear regulators can be swapped for more modern, low noise types, not the cheap crap creative used in the 90s.
Thanks for the hint regarding disabeling the amplifier. I will try that first and make comparsion recordings, and later try to improve the amplifier and activate it again.
Someone interested in these experiments? Suggestions for a good example of game audio for comparsion?

Reply 4 of 15, by Physikant

User metadata
Rank Newbie
Rank
Newbie

I read a little bit in the DAC-swapping thing. I'm still not conviced that it isn't a possibility to try. I know that EEP386 tried swapping them himself, but soldering is not always successful, so it can't be told for sure.
Looking at the CT2910, it is already using the CT1703 while having a pretty similar layout to the CT1740. Even the soundblaster overview chart by Burrito78 (https://docs.google.com/spreadsheets/d/1o0_u- … dit?usp=sharing) is mentioning the possibility that CT1780/CT1790 are using either CT1701 or CT1703. So the difference between them can't be that big interface-wise.

I found a datasheet for AK4503 (mentioned by EEP386 in the thread linked by S95Sedan) here:
https://datasheetspdf.com/pdf/541245/ETC/AK4503-VF/1
I did a few minutes reversing on a CT4380 I had lying around and identified a few connections on the CT1703-A used there:
2,28: +5V
3,9,16,18,27: GND
8,10,11,12,13,17: connection to CT18903 (the bus interface chip, so probably some digital input)
20,21,23,25: Not connected
14,15: each have a 10 Ohm resistor to 5V, each buffered with 2 caps

This doesen't fit the AK4503 pinout at all, so the assumption CT1703 = AK4503 is most definitely wrong.
I will try some reversing on the CT1740 and compare the pinout.

Reply 5 of 15, by Physikant

User metadata
Rank Newbie
Rank
Newbie

Okay, I checked the Pinout of the CT1701 on my CT1740.
The connections that I have measured in the previous post are almost identical on the CT1740/CT1701:
2,28: +5V
3,9,16,27: GND (18 is hard to trace, but not connected to GND)
8,10,11,12,13,17: connection to CT1746B (bus interface chip)
14,15: 10 Ohm to 5V

So it looks like at least nothing will be damaged if I swap the chips...

Reply 6 of 15, by S95Sedan

User metadata
Rank Member
Rank
Member

Still didn't really answer why you would need to do this swap.
Not only is a stock CT1740 not noisy at all (i own the same card) it will only complicate troubleshooting issues/faults with your specific card.

Reply 7 of 15, by Physikant

User metadata
Rank Newbie
Rank
Newbie

I made some progress.
First step was to remove all 54 old caps (first image).
I ordered them new. Most of them stayed standard electrolytic types, but I was able to replace the big ones with new hybrid polymer types. (second image).
Then I replaced all 4 old ST 3403 operational amplifiers with new MAX44252. They should be far superior in noise and linearity. After desoldering the old OPs, I found that one of them has one unused section without any connections, what could introduce some additional noise (third picture).

I made audio recordings prior and after all modifications.

The last step was trying to replace the original CT1701 (which is famous for its noise) with an CT1703-A from an AWE64. Contrary to what was stated before in this forum, it IS replacable and PCM sound is working with the new chip. While the samples seem to sound less noisy, there is a background chirp added that can be heard when the audio is silent. Perhaps the reason behind this is some wrong passives around the DAC, I could try and make a schematic of this area of the AWE64 and try to find the difference on the CT1740 I am using...

Another thing I'm planning to do is to try to fix the unconnected OP pins and see if this makes any difference.

Attachments

Reply 8 of 15, by Physikant

User metadata
Rank Newbie
Rank
Newbie

I found the error. I touched a 0805-Capacitor while changing the chips and it got disconnected from one side. After reattaching them, CT1703-A performs like CT1701, but with less silent noise. So the theory is right, CT1703-A IS less noisy than CT1701. But for my ears, the difference is not that big to justify the hassle. Stating that you just have to check your card if a CT1703 is present or not to determine if your card will be noisy or not is just wrong in my opinion from my experiences now.

Reply 9 of 15, by PC-Engineer

User metadata
Rank Member
Rank
Member

It would be very interesting to hear the audio recordings before and after the recap. What is your personal, subjective perception. I am planning a similar procedure with my CT1740 and a CT2760 and would be interested to the effect.

Last edited by Stiletto on 2022-02-18, 02:04. Edited 1 time in total.

Epox 7KXA Slot A / Athlon 950MHz / Voodoo 5 5500 / PowerVR / 512 MB / AWE32 / SCSI - Windows 98SE

Reply 10 of 15, by Physikant

User metadata
Rank Newbie
Rank
Newbie

I made some amount of audio recordings. I always recorded samples of Doom, Dune, Monkey1 and DOTT with disabled and enabled amplifier, all according to the rules given by MaxWar in this thread:
The Grand OPL3 Comparison Run!
The question is: how can I make them downloadable? Currently they take more than 200MB ...

Reply 11 of 15, by PC-Engineer

User metadata
Rank Member
Rank
Member

Ohh thanks a lot, but i have no solution for upload. Do you have a personal perception? Which characteristics changed?

Epox 7KXA Slot A / Athlon 950MHz / Voodoo 5 5500 / PowerVR / 512 MB / AWE32 / SCSI - Windows 98SE

Reply 12 of 15, by Physikant

User metadata
Rank Newbie
Rank
Newbie

Okay, this is the final update.

I fixed the missing OP connections, so now the card is as good as it gets. The only thing that is missing is the ASP/CSP chip I couldn't find without wasting another 100€+ for a SB16 with it ...
Summary:
- replaced all electrolytic capcitors. The 470µF-types were replaced with hybrid polymer electrolytic caps.
- replaced all ST 3403 quad operational amplifiers with new MAX44242 (4 pieces)
- changed CT1701 DAC to CT1703-A
- fixed the missing connections to the unused OP

In the detail shot, most of the modifications can be seen: the new CT1703-A, 3x the new MAX44252, the 2 solder bridges to the op in the center, and on the right side the 2 big hybrid caps.

As a conclusion I would say: especially the silent noise is reduced. But without knowing, the modifications are not that much audible to be something everyone should do ...
For me, this is now the ultimate DOS gaming ISA card. It is the most compatible card available, it is now very low noise, no hanging note bug and the sweet wavetable header!

For everyone to compare, I recorded 1:30 min of Doom, one sample prior to any modification, and one sample in the end. I think the biggest difference can be heard at the more silent part after 1:02.
Here the untouched card:
http://www.nikolas-becker.de/untouched.flac
Here the completely modded card:
http://www.nikolas-becker.de/modded.flac

Hope that satisfies everyone 😀
If anyone has another idea what still could be improved, I am happy to try it!

Niko

Attachments

Reply 13 of 15, by Physikant

User metadata
Rank Newbie
Rank
Newbie

Oh, a small warning for those who want to try this themselves:
The PCB quality of these old cards is quite bad. Removing the old caps and especially clearing the holes without damaging the contact barrels is difficult, especially since many of the buffer caps have contacts in the inner layers...