You need a pseudo differential input for CD :

LABS wrote:

How do their cd-ins behave by the way? Are they noiseless?
Anyone dealing with the same problem? Any suggestion will help. Thank you.

Good lord, no. Nothing in that era was noiseless. Are you kidding? PC audio is a case-study on how NOT to design audio circuitry.

Let's start with dubious-quality analog outputs from optical devices on the other end of a metal cage with an EMI-production plant in the middle, powered by a switching supply that is probably only JUST good enough not to explode, and tie grounds together at like a hundred different places. The audio cable can be a 5% shielded cable of 24AWG wire, draped over the power supply, RAM, CPU, video card, and IDE card, wrapped up in a coil, and set on top of the 74-series logic ICs on the address decoder of the sound card.

Honestly, we're lucky we heard ANYthing but noise.

SirNickity wrote:

Honestly, we're lucky we heard ANYthing but noise.

This is my experience with CD-ROM Audio cables 20 years later using at least twice as high quality components. No wonder I had a seperate CD-Audio music system all the way back in 1996..

Tiido wrote:

You need a pseudo differential input for CD :

Yea.. didn't want to add more opamps, but looks like it is the only way to go:

Works for me, -94dB RMS. Thanks a lot, man! 😀

That's a huge improvement !
Many single chip things and Codec chips have built in input like that just for the CD input.

Tiido wrote:

That's a huge improvement !
Many single chip things and Codec chips have built in input like that just for the CD input.

Interesting, didn't know that about single-chip cards. This clears up a lot 😀

LABS wrote:

Ok, just measured the output noise Used RME Multiface for input @ 44100hz/32-bit float & WaveSpectra for Win […]
Show full quote

Ok, just measured the output noise
Used RME Multiface for input @ 44100hz/32-bit float & WaveSpectra for Win

idle.jpg

-94db RMS - not bad, right? With DSP, OPL2 and PC-speaker knobs set to max

But after connecting CD-audio and set its knob to max - I get this

cd-in_idle.jpg

Can hear silent, but audible constant high-pitched noise and variable noises from cd-rom motors 😠
Looks like a gnd loop.
Already tried messing around with gnd and R/L signals, connecting/disconnecting/buffering of cd-out on dedicated cd-gnd plane/checked solder quality/lp filtering - nothing works, the noise is in audible spectrum with peak @~8khz. If I connect cd-out via buffer to the power amp directly with a 3m cable - no noise at all, very clean signal, so there is a hope.

Unfortunately I do not have any hi-q cards with cd-in to check them out for possible solutions. Tried to trace hi-res photos of Aztec and Gravis cards and as I could see - nothing fancy here, just some filtering of inaudible freqs and gnds are connected, but maybe I just didn't notice something. How do their cd-ins behave by the way? Are they noiseless?
Anyone dealing with the same problem? Any suggestion will help. Thank you.

Honestly, those former noise figures are impressive, especially as the analogue sections aren't canned.

As an aside on the previous suggestions, you could also make a properly screened cable which would massively cut down on noise instead of using the stock CD-Rom one which is essentially as others have pointed out, a big-ol antenna.

Vaudane wrote:

Honestly, those former noise figures are impressive, especially as the analogue sections aren't canned.
As an aside on the previous suggestions, you could also make a properly screened cable which would massively cut down on noise instead of using the stock CD-Rom one which is essentially as others have pointed out, a big-ol antenna.

Yes, it can be done, but actually the goal was to hit theoretical -98dB SNR of cd audio - the highest quality sound source of all 4 - without extra shielding etc. Considering crappy DACs of cd-roms - it's accomplished. I think there is no need for lower noise floor in a system of such fidelity, where even theoretical noise of any sound source is higher (and much higher) than the real noise floor of card's analog path.

LABS wrote:

Vaudane wrote:

Honestly, those former noise figures are impressive, especially as the analogue sections aren't canned.
As an aside on the previous suggestions, you could also make a properly screened cable which would massively cut down on noise instead of using the stock CD-Rom one which is essentially as others have pointed out, a big-ol antenna.

Yes, it can be done, but actually the goal was to hit theoretical -98dB SNR of cd audio - the highest quality sound source of all 4 - without extra shielding etc. Considering crappy DACs of cd-roms - it's accomplished. I think there is no need for lower noise floor in a system of such fidelity, where even theoretical noise of any sound source is higher (and much higher) than the real noise floor of card's analog path.

Fair enough, And the fact they send it analogue to the soundcard via a shitty cable after shitty DAC is a whole another discussion 🤣

I could immediately tell you weren't US based by the noise signal too xD good ol 50 Hz mains hum.

LABS wrote:

Works for me, -94dB RMS. Thanks a lot, man! 😀

Wow, that's some very impressive graphs! 😳

I was really happy to see digital audio connectors on the back of CD-ROMs in the latter half of the 90s. And sad that it was just then becoming a common thing, when nobody cared about CD audio anymore. (And those that did simply streamed it off the disc and into the sound card as PCM data.)

94dB is definitely a respectable achievement. Better than that era of sound cards, CD-ROM audio outputs, and the majority of consumer CD audio products we would've been using at the time. I fear your accomplishment is going to be wasted on most systems it's installed in, but hey -- at least it's not the bottleneck, and that's a good place to be. No need for further heroics, for sure.

SirNickity wrote:

I fear your accomplishment is going to be wasted on most systems it's installed in

I have 5 machines with ISA slots I'm going test the card on: 386SX-33, 486DX2-50, P166MMX, P200MMX, P3-1GHz.
Currently it is installed in a 22 y.o. P166MMX with original not recapped PSU which I use as a workbench machine. Interesting how others will behave. Will post the results.

Vaudane wrote:

As an aside on the previous suggestions, you could also make a properly screened cable which would massively cut down on noise instead of using the stock CD-Rom one which is essentially as others have pointed out, a big-ol antenna.

waaait, arent cdrom audio cables shielded? they certainly look the part

^ not all of them , cheaper ones are just sleeved in plastic with no shielding

Some CDROM cables are shielded, others are just unshielded.

Never saw one that was properly shielded, they all just have cheap plastic grey sleeves.

A bunch of mine have one of the ground pins on the connector connected to a heat-shrink-tube'd bare stranded wire that is wrapped around the L/R conductors. I'm sure it's about as effective as printing "LOW NOISE" on the cable jacket, but it's ... kind of ... technically shielded ... if you squint and don't ask too many questions.

Made some rough cd-in noise tests on 2 more machines - 386 and P3. Considering the layout (diff amp on a breadboard and all that wires hanging around), the results are quite good:

386DX-33 (29 y.o., no recapping):

P3-1GHz (20 y.o., no recapping):

Wow, very nice.