darry wrote on 2020-10-29, 01:58:

The DreamBlaster X2, when playing through the Orpheus digital out actually has more audible noise than when playing through the Orpheus analogue out . To be clear, I do need to apply a high degree of gain on my headphone output to perceive a difference. The noise, which is present even when nothing is playing and muting the synth mixer input eliminates it (all other analogue inputs are already muted).

This may be entirely avoidable. There are a few possible sources of this noise, and they should be mostly manageable in software.

The CS4237 offers quite a range of gain values on most audio sources, but beyond a certain point, it is analogue amplification, which tends to introduce a certain amount of noise. I don't think there is any amount of careful layout/routing work that will eliminate that, but managing the range of available values in software should help. Have you been using UNISOUND exclusively for DOS?

The other problem is, in part, due to the fact that the CS4248 and CS423x lines have quite a long history, with new features regularly "stapled" on (sometimes somewhat awkwardly), in an effort to preserve "perfect" compatibility. In general, this works relatively well, but creates challenges when trying to manage the later chips' features to extract more quality and capability. When using software that mostly treats the chip as one of its predecessors (like Crystal's own tools) or takes certain shortcuts (because that's "good enough"), the result can be added noise. Managing the register values carefully can go a long way toward minimising that noise, but can be a rather tedious process to find just the right balance of quality, features and simplicity (for the people who will use the card, not its creators).

So, if you've been using CWDMIX for DOS, stop that! 😀 That nasty kludge can be a source of all sorts of noise. The next thing to try for now is to reduce the input volume on the synth/wavetable header to a maximum of a few notches (or percentage points) below maximum. See if that helps a little. If you have time to test some more, please continue sharing your experiences. All constructive feedback is helpful.

640K!enough wrote on 2020-10-29, 22:02:

This may be entirely avoidable. There are a few possible sources of this noise, and they should be mostly manageable in softwar […]
Show full quote

darry wrote on 2020-10-29, 01:58:

The DreamBlaster X2, when playing through the Orpheus digital out actually has more audible noise than when playing through the Orpheus analogue out . To be clear, I do need to apply a high degree of gain on my headphone output to perceive a difference. The noise, which is present even when nothing is playing and muting the synth mixer input eliminates it (all other analogue inputs are already muted).

This may be entirely avoidable. There are a few possible sources of this noise, and they should be mostly manageable in software.

The CS4237 offers quite a range of gain values on most audio sources, but beyond a certain point, it is analogue amplification, which tends to introduce a certain amount of noise. I don't think there is any amount of careful layout/routing work that will eliminate that, but managing the range of available values in software should help. Have you been using UNISOUND exclusively for DOS?

The other problem is, in part, due to the fact that the CS4248 and CS423x lines have quite a long history, with new features regularly "stapled" on (sometimes somewhat awkwardly), in an effort to preserve "perfect" compatibility. In general, this works relatively well, but creates challenges when trying to manage the later chips' features to extract more quality and capability. When using software that mostly treats the chip as one of its predecessors (like Crystal's own tools) or takes certain shortcuts (because that's "good enough"), the result can be added noise. Managing the registers values carefully can go a long way toward minimising that noise, but can be a rather tedious process to find just the right balance of quality, features and simplicity (for the people who will use the card, not its creators).

So, if you've been using CWDMIX for DOS, stop that! 😀 That nasty kludge can be a source of all sorts of noise. The next thing to try for now is to reduce the input volume on the synth/wavetable header to a maximum of a few notches (or percentage points) below maximum. See if that helps a little. If you have time to test some more, please continue sharing your experiences. All constructive feedback is helpful.

I have been using UniSound exclusively (never cwdmix) .

My initial impressions about there being extra noise with the X2 playing back through Orpheus' ADC and digital out seem to have been wrong as I can no longer reproduce that issue .
I have decided to test digital out directly connected to an X-FI Platinum in bit-matched mode and everything sound perfect . I think my initial erroneous impression was due to the confusion regarding the SRS slider and its mute button .

To clarify and correct my previous statements regarding the X2 (waveblaster) and analogue vs digital out on the Orpheus:

When using Orpheus analogue out under Windows :
The synth slider and mute button control the volume of the X2 . SRS slider MUST be set to 0 AND muted (or extra noise and distortion will be generated).

When using Orpheus digital out under Windows :
The synth slider controls the volume of the X2 . SRS slider MUST set to 0 (or extra noise and distortion will be generated) BUT both SRS and Synth must be unmuted for the X2 to be heard .

The reason why I was getting extra noise on the X2 was probably because the SRS slider was not set at 0 .

Under DOS, I have yet to run more detailed tests, but will probably do so soon.
EDIT : It sounds fine in DOS over digital out too .

EDIT2: Since digital out is cleaner than analogue out, but I am all out of digital inputs on my VS-880EX (the only one I have is now used by an SB Live! ), I will try using the Orpheus with an outboard S/PDIF DAC . I have one of these https://www.amazon.ca/D03K-Digital-Analog-Aud … r/dp/B009346RSS on order, which I should receive by Saturday . I realize that this may be no better than the Orpheus' already excellent analogue out, but for 35 CAN$, it's worth a try .

In the meantime, tomorrow I will give the official Orpheus drivers under Windows 98 SE another go and record exact error messages .

darry wrote on 2020-10-30, 01:28:

I have one of these https://www.amazon.ca/D03K-Digital-Analog-Aud … r/dp/B009346RSS on order, which I should receive by Saturday . I realize that this may be no better than the Orpheus' already excellent analogue out, but for 35 CAN$, it's worth a try .

If you can still cancel and re-order, I would recommend taking the model with volume control instead. I have been using something similar as a guinea-pig device for months, and it's actually decent. I had to modify it to get decent audio quality out of it, but overall, it was worth its price, at under $40.

640K!enough wrote on 2020-10-30, 02:53:

darry wrote on 2020-10-30, 01:28:

I have one of these https://www.amazon.ca/D03K-Digital-Analog-Aud … r/dp/B009346RSS on order, which I should receive by Saturday . I realize that this may be no better than the Orpheus' already excellent analogue out, but for 35 CAN$, it's worth a try .

If you can still cancel and re-order, I would recommend taking the model with volume control instead. I have been using something similar as a guinea-pig device for months, and it's actually decent. I had to modify it to get decent audio quality out of it, but overall, it was worth its price, at under $40.

Could you please share the brand and model number of the device you have ? And a quick summary of the mods done to it ?

I chose a Fiio device because of my good experience with the E10K and I was not expecting volume control to be useful as the intended use is to connect to a Roland VS-880EX , which does have analogue input gain control .

What is the use case for volume control on a device with line level output (presumably -10dB?) ?

darry wrote on 2020-10-30, 03:12:

Could you please share the brand and model number of the device you have ? And a quick summary of the mods done to it ?

I chose this one. It seemed generally more capable, included a cheap optical cable and has built-in volume control.

The unit I received had small ceramic capacitors soldered over the through-hole footprints for the main output capacitors. The sound was generally flat and lacking. I won't detail all of the testing I did, but I settled on two 470 µF Nichicon UES as replacements. Once soldered, it was rather... interesting getting the board back into the small metal case; let's just say the board isn't quite as flat as it used to be. Nonetheless, outputs and volume knob still line up, and the sound is dramatically improved.

darry wrote on 2020-10-30, 03:12:

What is the use case for volume control on a device with line level output (presumably -10dB?) ?

For me, it turned out to be essential, particularly with my testing of the Orpheus design. I wanted the output to be as pure and unprocessed as possible, which meant going straight to headphones, with an easy way to manage volume levels in an emergency. Doing the wrong things and/or in the wrong order can produce some fairly ear-shattering noises; without volume control, I can assure you that I'd be deaf by now.

640K!enough wrote on 2020-10-30, 03:39:

The one I chose is this: B0746H2LJ1 (enter that ASIN into your local Amazon search field). It seemed generally more capable, i […]
Show full quote

darry wrote on 2020-10-30, 03:12:

Could you please share the brand and model number of the device you have ? And a quick summary of the mods done to it ?

The one I chose is this: B0746H2LJ1 (enter that ASIN into your local Amazon search field). It seemed generally more capable, included a cheap optical cable and has built-in volume control.

The unit I received had small ceramic capacitors soldered over the through-hole footprints for the main output capacitors. The sound was generally flat and lacking. I won't detail all of the testing I did, but I settled on two 470 µF Nichicon UES as replacements. Once soldered, it was rather... interesting getting the board back into the small metal case; let's just say the board isn't quite as flat as it used to be. Nonetheless, outputs and volume knob still line up, and the sound is dramatically improved.

darry wrote on 2020-10-30, 03:12:

What is the use case for volume control on a device with line level output (presumably -10dB?) ?

For me, it turned out to be essential, particularly with my testing of the Orpheus design. I wanted the output to be as pure and unprocessed as possible, which meant going straight to headphones, with an easy way to manage volume levels in an emergency. Doing the wrong things and/or in the wrong order can produce some fairly ear-shattering noises; without volume control, I can assure you that I'd be deaf by now.

Thank you for your suggestion, but as I will be using the Orpheus with a Roland VS-880EX (an old pro-sumer multi-track recorder, so supposedly very good ADC performance) , which has analogue gain control, the ability to isolate an input and a dedicated headphone out with a volume control, I think I will try the Fiio first, for now . The Fiio's noise performance at 44.1KHz is supposed to be quite good .
https://www.audiosciencereview.com/forum/inde … t-20-dacs.1316/

I had the chance to play a bit with the Orpheus official Windows 9x drivers and captured the error messages that I get upon installation and the near identical ones I get after a reboot post installation .
I have also observed something which I believe is strange with the official driver set ( ORPHEUS_dos_95.rar ) : the presence of iusv2860.cat and most of the other files provided with it suggest that the Orpheus driver is based on the pv2860.zip VXD reference driver , but CWDAUDIX.VXD is a much older version than that which is provided with pv2860.zip . I tried replacing CWDAUDIX.VXD with the one from pv2860.zip , but that results in a non-functional MIDI out and an unstable system . Using the pv2860.zip reference driver set as is works fine for me under Windows 98 SE .

cwdaudix.vxd versions :

errors upon official Orpheus driver installation :

errors after reboot with official Orpheus driver installed :

Here is the driver I am using in Windows 98 SE with the Orpheus sound card .

Digital (S/PDIF) output is not enabled by default with this driver . To enable it, the final "0" on line 265 of cwdaudio.inf must be replaced with a "4" . Alternatively, if the driver is already installed, the corresponding registry key must be modified and the system rebooted to enable Digital (S/PDIF) output .

thanks darry for the info , I still have not tested this tbh but marmes also got the same error
I guess at some point I packed the wrong driver files while testing stuff ...

keropi wrote on 2020-10-31, 08:12:

thanks darry for the info , I still have not tested this tbh but marmes also got the same error
I guess at some point I packed the wrong driver files while testing stuff ...

No worries. The reference driver can be used in the meantime, as a workaround. I did not test any DOS software under Windows with that driver installed, as I tend to avoid DOS under Windows as much as possible, but Windows apps seem to work fine in my very limited testing. The only exceptions so far are Foobar 2k and Audacity in Directsound mode at 44.1KHz and my AWE64 has the same issues with Directsound. There is a WDM reference driver for the CX4237 that works better with 44.1KHz Directsound playback, at least in Foobar, but it is unstable on my system and has a strange issue with audio recording being too fast, even in MME mode . The SB Live! in the same system is a better choice for Directsound anyway and seems to work fine with Creative VXD drivers .

Other than branding and enabling S/PDIF output by default, what tweaks/mods does the official Orpheus driver have ?

darry wrote on 2020-10-31, 08:53:

[...]
Other than branding and enabling S/PDIF output by default, what tweaks/mods does the official Orpheus driver have ?

none atm as we can't really mess with the windows drivers unless someone that knows this stuff can hack them or something....

btw the instabilities with the WDM driver are expected if there are vxd and wdm drivers installed in the system at the same time - it is something os-related so nothing can be done in this case unless someone only uses WDM drivers in his system but I am not sure this can be achieved in 98SE

keropi wrote on 2020-10-31, 09:48:

darry wrote on 2020-10-31, 08:53:

[...]
Other than branding and enabling S/PDIF output by default, what tweaks/mods does the official Orpheus driver have ?

none atm as we can't really mess with the windows drivers unless someone that knows this stuff can hack them or something....

btw the instabilities with the WDM driver are expected if there are vxd and wdm drivers installed in the system at the same time - it is something os-related so nothing can be done in this case unless someone only uses WDM drivers in his system but I am not sure this can be achieved in 98SE

Thank you for the info .

I remember having read that mixing VxD and WDM drivers, at least as far as sound card drivers go, would cause stability issues. That was actually my working theory as to why the built-in Windows driver and the Crystal WDM reference drivers were problematic on my system (with AWE64 VXD and now SB Live! VXD drivers) . Thank you again for confirming that .

I am quite content using the pv2860.zip VXD driver, as it works fine on my system, now that I have figured out how to operate the mixer when using S/PDIF output .

The only remaining (minor) issue that I have is non working mixer level output control using UniSound under DOS only when using S/PDIF output . Though it has been recommended to avoid it, I may end up experimenting with CWDMIX.EXE .

I will experiment more once my outboard DAC arrives .

Cheers!

darry wrote on 2020-10-31, 13:39:

The only remaining (minor) issue that I have is non working mixer level output control using UniSound under DOS only when using S/PDIF output . Though it has been recommended to avoid it, I may end up experimenting with CWDMIX.EXE .

Hi!

Thank you for reporting!! I'm working on a new version with some fine-tuning changes to mixer code for Orpheus.
In the meantime, you can adjust mixer (and Wave/PCM) volume properly if you enable internal CrystalFM ( /XOFi option )

JazeFox wrote on 2020-10-31, 17:42:

Hi! […]
Show full quote

darry wrote on 2020-10-31, 13:39:

The only remaining (minor) issue that I have is non working mixer level output control using UniSound under DOS only when using S/PDIF output . Though it has been recommended to avoid it, I may end up experimenting with CWDMIX.EXE .

Hi!

Thank you for reporting!! I'm working on a new version with some fine-tuning changes to mixer code for Orpheus.
In the meantime, you can adjust mixer (and Wave/PCM) volume properly if you enable internal CrystalFM ( /XOFi option )

Thank you very much .

Two quick (hopefully) questions, if I may.

I noticed that CWDINIT.EXE seems to load firmware of some kind when run .

What is the purpose of this ?

Does UniSound do something similar (silently) ?

I think I may have found a bug in UniSound's initialization of the Orpheus .

When digital output (S/PDIF) mode is enabled and used, whether using internal or external OPL3, choosing Sound Blaster as a music device in Duke Nukem 3D setup produces no FM music . Choosing Adlib works as expected.

As a comparison test, I tried initializing the Orpheus using the CWBINIT.EXE (and corresponding CWBAUDIO.BIN) for the Turtle Beach Malibu Surround 64, as that also inits the Orpheus in digital output (S/PDIF) mode, but obviously with internal OPL3 and that allows both the Sound Blaster and Adlib options in Duke Nukem 3D to work .

As a side note, when using analogue output, having digital output (S/PDIF) disabled seems to be slightly less noisy than having it enabled . I could only test under DOS (I can't dynamically switch under Windows as it needs a reboot). I don't know whether this is normal behavior for a CX4237B or if is due to a mixer or initialization issue . I could do 2 test recordings (S/PDIF on and S/PDIF off) under Windows and compare, if that could be helpful .

Also, an option to mute the Waveblaster header under DOS without muting OPL3 when using digital (S/PDIF) output would be a nice feature, as that would cut down on noise when not using a waveblaster daughterboard . The Windows driver allows that indirectly (SRS mute), obviously when using internal OPL3, so it should be possible to do the same in DOS and hopefully, also when using external OPL3 .

Cheers!

EDIT : Corrected typos

@darry,
This is mainly a card with DOS in mind, so Windows drivers are just like Malibu/Crystal drivers with SPDIF enabled. At least with me I can run most games thru SPDIF, but some games with their own sound mixers, may cause some issues with Orpheus sound card with SPDIF output. That we cannot do anything. From my experience with some games, those with no mixer work just fine, in both analog and Digital. Those that support WSS, also have no trouble. Games like Doom and DN3d
that have their own mixers must be patched or played in analog. Of course we are also having a great amount of support from Jazefox and 640k!enough in order to have the best sound available from this card. So we hope these small issues will be surpassed . Cheers!

@darry
Thanks again for reporting.
Some of the issues are currently being addressed (noise issues, wavetable, etc..). But as Marmes said, some things don't work the same way as in Windows (different codec modes and WSS/SB differences). Also, Crystal chips do have an internal auto-switching mechanism from/to WSS<->SB modes, that is causing some issues in DOS. We try to minimize them.

darry wrote on 2020-11-01, 13:59:

When digital output (S/PDIF) mode is enabled and used, whether using internal or external OPL3, choosing Sound Blaster as a music device in Duke Nukem 3D setup produces no FM music . Choosing Adlib works as expected.

Well, I tried to reproduce that behavior with Duke3D but I couldn't. It works well here, FM sound is working. What are the exact steps (from the cold boot to the exact command line commands written) you take?

As a comparison test, I tried initializing the Orpheus using the CWBINIT.EXE (and corresponding CWBAUDIO.BIN) for the Turtle Beach Malibu Surround 64, as that also inits the Orpheus in digital output (S/PDIF) mode, but obviously with internal OPL3 and that allows both the Sound Blaster and Adlib options in Duke Nukem 3D to work .

AFAIK, SPDIF activation is not present in any Malibu DOS driver (I tried every version I could before). Are you booting to pure DOS from a cold boot? (not windows -> restart in DOS mode), and no other DOS driver o initializer (including Unisound) is running before CWBINIT?

And about the .BIN used by official drivers, yes, it's a firmware patch file. It's used to enhance or fix possible issues present in the CS chip revision (internal ROM), that is uploaded to chip's RAM. In the first internal versions of Unisound, firmware uploading was added, with the last revision patch available, but after many tests, it made no difference in the card's performance or capabilities, so I removed it to reduce init time and size (it's still present for older cards like CS4232 that require firmware to be uploaded at initialization). Anyway if I detect any (positive) difference in the future, it can be re-added.

I also tried dn3d now. No trouble with external opl3 either.

Marmes wrote on 2020-11-01, 15:20:

@darry,
This is mainly a card with DOS in mind, so Windows drivers are just like Malibu/Crystal drivers with SPDIF enabled. At least with me I can run most games thru SPDIF, but some games with their own sound mixers, may cause some issues with Orpheus sound card with SPDIF output. That we cannot do anything. From my experience with some games, those with no mixer work just fine, in both analog and Digital. Those that support WSS, also have no trouble. Games like Doom and DN3d
that have their own mixers must be patched or played in analog. Of course we are also having a great amount of support from Jazefox and 640k!enough in order to have the best sound available from this card. So we hope these small issues will be surpassed . Cheers!

@marmes @Jazefox @640k!enough

I may have been unclear, sorry, and you may have partially misunderstood me . The Windows driver works very well and exposes all the features and functions required, once I deduced how to use them . Additionally, as you mentioned (and I agree), the Orpheus being mainly DOS oriented, I doubt anybody would really want more on the Windows driver front .

As for the DOS side of things, I am quite hopeful that the issue I described regarding Duke Nukem 3D in DOS is fixable as using the Malibu's init program does manage to work around the said issue . As for the "feature request" regarding muting the waveblaster header while leaving OPL3 on in DOS, that would really help with the noise that this input seems to pick up in my PC even when no waveblaster type card is installed . The only reason I mentioned Windows in this case is because the Windows mixer allows that kind of selective muting, which leaves me hopeful that this could be implemented in DOS as well .

I apologize if I sound like I am complaining or unhappy or a serial nitpicker (ok that last one may be true 😉 ). I am EXTREMELY happy with the Orpheus and my only regret is not having the skill sets that all you amazing folks have, which prevents me from being an active contributor to this project . I too am confident that any remaining possible software tweaks will become a reality with time and your hard work, for which plebs like me are eternally grateful.

If there is any way that I can be of assistance, please let me know .

Cheers!

JazeFox wrote on 2020-11-01, 15:39:

AFAIK, SPDIF activation is not present in any Malibu DOS driver (I tried every version I could before). Are you booting to pure DOS from a cold boot? (not windows -> restart in DOS mode), and no other DOS driver o initializer (including Unisound) is running before CWBINIT?

I will agree that this is a key point. The Crystal tools (at least the DOS versions) are somewhat sloppily implemented. If the card had been previously initialised, not all previous state is purged without a full system reset, and some register values may have been unintentionally retained. This can be problematic in some situations, going so far as to produce very unpleasant results (mixer feedback loops, loud noises, etc.), particularly when also combined with CWxMIX. They are passable for occasional testing, but I wouldn't recommend Crystal's DOS tools for long-term usage, especially with a card that offers as many features as Orpheus.

On the subject of retained state, using the Crystal tools after the card was initialised by the Win 9x drivers may offer some clues. The Windows drivers would have set up the mixer registers, while CWxINIT doesn't; you need CWxMIX for that (sort of; it still only does part of the job). If some of the previous setup was retained (S/PDIF being enabled, etc.), that may explain the apparent differences. DOS is a completely different world, and software may treat the card quite differently. Some software may perform a reset, obliterating any previous state, in which case, there's absolutely nothing we can do, short of trying some protected-mode TSR kludge that is likely to cause more trouble than it would solve. Other software will change the CODEC mode, which will affect the behaviour of the mixer. This is what I was saying earlier about the complications of additional features having been successively "stapled" on. They're pretty good for Windows, where a single driver generally handles the chip exclusively, but can make life under DOS rather interesting.

Another thorny issue is the "context" switch (Crystal terminology; no relation to the OS concept) between SB and WSS operation. This leads to all sorts of ugliness, and combined with firmware/design bugs, makes getting the right mixer setup an interesting and occasionally frustrating challenge. In my testing, I was unable to find a firmware version that improved this behaviour. Different chip revisions don't seem to have made any substantive difference in this regard, either.

All of this is actively being investigated, so all feedback and comments are appreciated, but being a small niche project involving products that were designated EOL decades ago, we can't make any promises. We can only try to make the best software lemonade we can with the Crystal lemons that we have. It should end up being relatively smooth, with some exceptions for software that tries to "help" too much (like issuing unwelcome mixer commands on start-up).

darry wrote on 2020-11-01, 16:21:

I apologize if I sound like I am complaining or unhappy or a serial nitpicker (ok that last one may be true 😉 ).

Do I know you from somewhere, fellow "serial nitpicker"?

darry wrote on 2020-11-01, 16:21:

If there is any way that I can be of assistance, please let me know .

There is: you can be of assistance by continuing to pick those nits and telling us about every single one of them. There is nothing worse than someone who sits around seething about the "junk" they bought, without ever constructively discussing the issues. Feedback is appreciated.