Given that the default configuration of dosbox is with a SB16, it might make sense to post the filter for that as well..
Could you post it so I can check how complex it is ?

James-F wrote on 2020-01-09, 08:13:

Yes I can. 😃

ME wrote:

Same for OPL3 lowpass, use Adlib Tracker 2 player called adt2play.exe to play the noise.a2m file and capture with spectrum analyzer.

Re: FM and PCM Volume Balance of SB Cards

I may be misinterpreting but the spectrum graph on that post is near identical to the SB Pro filter analysis for the PCM audio above?

@Qbix

SB16 is much harder to implement hence PCem uses this 2nd order LPF for SB16 too, it sounds good.
The SB16 will change its filter depending on the sampling rate of the PCM source it is playing (each game is different),, see first post.
For 11kHz sample rate PCM it will cut everything above 5.5kHz,,,, for 22kHz sample rate it will cut everything above 11kHz.

So you propose an incorrect filter instead of no filter (except for what weird up and down stuff) ?

Why incorrect?
A 2nd order filter like everything prior to SB16 is a ton better than no filter at all, and imo sounds better than the sb16 brickwall filtering.

You can implement SB16 filtering if you can read the game PCM sampling rate.

for the SB16 that filter would be incorrect right ? (despite maybe sounding better).

It's interesting that the filter of SB2pro sounds better according to you compared the SB16 one.

For starters, any filter is much better than no filter, and I'm sure no one will complain about it not being SB16 accurate. 😀

And yes, SBPro2 2nd order sounds better for dos games than a brickwall filter.

@appiah4

Yes, it is for PCM (SBPro2 and PCem) but the measurement process is exactly the same for OPL, I even attached a AT2 file for that.

Qbix wrote on 2020-01-09, 08:26:

for the SB16 that filter would be incorrect right ? (despite maybe sounding better).

It's interesting that the filter of SB2pro sounds better according to you compared the SB16 one.

I can also vouch for the SB Pro filtering being better and more nostalgic to me as an SB Pro 2.0 user in the early 90s. The two cards definitely sound different, with the Pro having a deeper sound with more bass response.

James-F wrote on 2020-01-09, 08:27:

For starters, any filter is much better than no filter, and I'm sure no one will complain about it not being SB16 accurate. 😀

Don't underestimate the purists 😀

Seriously filters would be great and to me these should be implemented like scalers. Optional setting in the config, even though I know the developers try to keep the config as small as possible. This could mean either just one filter (filter = true) or several (filter = 2nd pass or oplsomething or or...)

I am a purist myself, and have a big-ish collection of old SB cards.
In my opinion lowpass filtering is a huge part of a proper DOS experience, and painfully missing part of DOSBox, call me a nagger. 😀

hmm.. looks like DOSBox already does 1st order lowpass filtering similar to SB2.0.
How can that be? I recall Qbix saying there is no code in dosbox for filtering whatsoever.

How exactly are you sampling the audio from DOSBox?

That's Doom chainsaw gun captured with frequency analyzer.
Other smaples I use is White Noise which is flat when viewed in a frequency analyzer without filtering.
I'm not going to get into details, that would take hours, you can click my signature and do a little research.

But if you're not capturing the output directly (using the wav capture function), you're testing the frequency response of your output device rather than the emulator...

I'm using a DAW with VST plugins to capture DOBox response, it stays in the digital domain on the same computer.
Sorry, you realize I'm not new at this , right?

Yet you're somehow seeing a filtered output from software that does no filtering.
The wav capture is exactly what DOSBox outputs, and what it has control over; using any other data and saying it's wrong doesn't really help as who knows what happens once it's handed off to SDL and DOSBox can't do anything about that.

I'm talking to Qbix about it, dosbox has linear interpolation as Resampling and it results in lowpass filtering an some aliasing with lower sampling rates like most games.
But, the good news is it looks very similar to what the SB2.0 is already doing, and the SBPro just adds a filter on top of that.
So everything is clear now.

To @James-F:

I'm curious to know how the recently added SB16 emulation to VMware Fusion compares to the real thing. I don't know how really it is in Windows version of VMWare Player/Workstation (these versions had it since years), since I have never used them. Is it any better than DOSBox?

Even if it was which I doubt is the code it uses open source? The virtualization products have never had good sound quality, features or compatibily compared to emulators mostly due to that's not what those products are intended for.

The ideal brick wall filter can't be implemented without infinite terms, and a truncated brick wall filter gives ringing. A windowed sinc filter gives a smooth transition with an adjustable transition band. Performance requirements are higher than a second order HPF, although It can be implemented in integer with a LUT for the sinc function. The transition band could even be extended to resemble a simpler passive filter.
https://www.analog.com/media/en/technical-doc … p_book_Ch16.pdf
https://github.com/bensherlock/sherlock-code/ … FilterFactory.h