VOGONS


DOSBox ECE (for Windows & Linux)

Topic actions

Reply 100 of 1550, by KainXVIII

User metadata
Rank Member
Rank
Member
James-F wrote:
Nope, it's not the OPL patch, adlib sounds fine, only PCM affected. Happens in Vanilla 4008 (Feb 2017). […]
Show full quote

Nope, it's not the OPL patch, adlib sounds fine, only PCM affected.
Happens in Vanilla 4008 (Feb 2017).

https://drive.google.com/drive/folders/0BxEUS … S2tCbEJRUXpnNTg
Please rollback from 4007 to 4000 one by one to find where the problem began.

PCM works fine with DOSBox SVN r4000 enhanced (without FLAC support).7z and garbled with rest of versions that i found by this link

Reply 101 of 1550, by James-F

User metadata
Rank Oldbie
Rank
Oldbie

I confirm, something went southwards between 4000 and 4003.
I would assume in 4002 something in the mixer.cpp code went wrong (there are a lot of changes).
I opened a ticket on sourceforge.

4001:
Rewrite gus emulation a bit to make things a bit clearer

4002:
Rewrite some mixing code to fix issue with equal rates having aliasing issues

4003:
Updated header for mixer changes


my important / useful posts are here

Reply 102 of 1550, by YaketyYakumo

User metadata
Rank Newbie
Rank
Newbie

Sorry if this is a dumb question, but is there any way to increase the output resolution of the emulated Voodoo? With builds based on the old gulikoza patch this was handled by the wrapper, but I don't see any way to do this here.

Reply 103 of 1550, by Yesterplay80

User metadata
Rank Oldbie
Rank
Oldbie
James-F wrote:

I confirm, something went southwards between 4000 and 4003.
I would assume in 4002 something in the mixer.cpp code went wrong (there are a lot of changes).

So, this is a general problem? If so, does it occur in the regular SVN as well?

YaketyYakumo wrote:

Sorry if this is a dumb question, but is there any way to increase the output resolution of the emulated Voodoo? With builds based on the old gulikoza patch this was handled by the wrapper, but I don't see any way to do this here.

I'm afraid you can't change the resolution with this build / patch, it performs as it would have done so with a real MS DOS pc.

My full-featured DOSBox SVN builds for Windows & Linux: Vanilla DOSBox and DOSBox ECE (Google Drive Mirror)

Reply 105 of 1550, by Qbix

User metadata
Rank DOSBox Author
Rank
DOSBox Author
James-F wrote:
My calculations were correct, 150 equals to 75, and 130 equals to 65. […]
Show full quote

My calculations were correct, 150 equals to 75, and 130 equals to 65.

150 is the correct value if I match to 440Hz as with the hardware sound blaster at that frequency.
But in DOSBox 150 still sounds just a little louder than a SB1.0 13020B because it has no filtering on top of the FM sounds.

Using games that mix FM and PCM as reference: SB1.5, SB2.0, SBPro2, SB16 - Audio Comparison
For DOSBox to audibly match a Sound Blaster 1.0 or 2.0 (CT1320C or CT1350B respectively) I need to set to 130 (mixer FM 65).
I chose SB1/2 as reference because DOSBox filter resembles the filter on these cards, the SBPro is much too filtered compared to DOSBox and the SB16 has completely different kind of filter.
Moreover, Adlib games that don't have in-game volume controls for music/audio usually belong to the SB1.0 or 2.0 era.

I use quality studio monitors and studio headphones for critical listening, they usually have more treble than typical desktop speakers, or should I say more "correct" treble.
Behringer Truth B3030A
Audio Technica ATH-M40X

I suggest keeping it at 150 for typical desktop speakers (Creative, Logitech, etc...), ones that have tone controls on the speakers.
But lower to 130 if you have speakers with flat frequency response that go up to 20kHz, at ear level, or headphones.
Once DOSBox incorporate filtering on the FM sounds it would be much more straight forward comparison and 150 should be the default setting.
On my setup I use 130.

I should mention once again than the difference between 130 and 150 is only 1.25db, so most people will not notice the difference at all.
But vanilla DOSBox default 200 and 130 is -3.75db difference which is undoubtedly audible.
A 3.75db quieter music might make all the difference between intelligible speech vs drowned-under-music speech.

I have set it to 1.5 (150) as default now, mostly because that is the closest to our old value, "best" for speakers and most future proof with regards to filtering.

Thanks!

Water flows down the stream
How to ask questions the smart way!

Reply 107 of 1550, by Yesterplay80

User metadata
Rank Oldbie
Rank
Oldbie

So, this makes your little patch more or less obsolete, doesn't it?

My full-featured DOSBox SVN builds for Windows & Linux: Vanilla DOSBox and DOSBox ECE (Google Drive Mirror)

Reply 108 of 1550, by James-F

User metadata
Rank Oldbie
Rank
Oldbie

@YP80, please keep it for me for a little longer.
I am sorry I clog the ECE thread with non related testing posts, it obviously belongs to the PCM vs FM thread.
I will continue the discussion over here: Re: FM and PCM Volume Balance of SB Cards
* I re-posted this there with a little example how I measure the "d" value.
EDIT: Or in the SVN thread.

@Qbix, I have found an even more accurate test (see below), and 140 will be closer to a SB1/2, although 150 will be leaning more towards the SBPro and exactly the SB16 value.
The resulting 140 is from a measured delta between PCM and FM of a SB1/2 and is below the filtered frequencies, so it is accurate.
Please consider to revise from 150 to 140, it's only 0.6db difference but for accuracy sake, it matters.
EDIT: Forgot to mention that the measurements done with "compat" as default, but "nuked" gives the same results.
EDIT2: Added SB16 CT2230; 150 in DOSBox matches the hardware SB16 but the SB16 has completely different filter (thus sound) than DOSBox or SB1/2, so it should not be taken as the reference card.

* MPXPlayer had some problems with the SB1/2 card on my system to test PCM and also there is difference between SB1 and SBPro stereo support.
So a constant rifle shot (mono) in Wolf3D reflects PCM behavior much better between cards, which in turn gave 1/100 db accuracy using a FreeG Channel Strip plugin in my DAW.

SB1 CT1320C PCM: -7.70 wolf3d rifle 3 OPL: -13.31 opl440 d: 5.61 […]
Show full quote

SB1 CT1320C
PCM: -7.70 wolf3d rifle 3
OPL: -13.31 opl440
d: 5.61

SB2 CT1350B
PCM: -10.07 wolf3d rifle 3
OPL: -15.80 opl440.
d: 5.73

SBPro2 CT1600
PCM: -16.74
OPL: -21.4
d: 4.66

SB16 CT2230
PCM: -6.57
OPL: -11.66
d: 5.09

DOSBox 200
PCM: -9.6
OPL: -12.04
d: 2.44

DOSBox 150
PCM: -9.6
OPL: -14.55
d: 4.95

DOSBox 140
PCM: -9.6
OPL: -15.14
d: 5.54

DOSBox 130
PCM: -9.6
OPL: -15.8
d: 6.2


my important / useful posts are here

Reply 109 of 1550, by Qbix

User metadata
Rank DOSBox Author
Rank
DOSBox Author
Qbix wrote:

Harekiet is checking the mixer problem out.

I have located the problem, however we are discussing if we can't do it better.
Commited in 4013, let me know if it helps.

Water flows down the stream
How to ask questions the smart way!

Reply 110 of 1550, by Qbix

User metadata
Rank DOSBox Author
Rank
DOSBox Author
James-F wrote:
@YP80, please keep it for me for a little longer. I am sorry I clog the ECE thread with non related testing posts, it obviously […]
Show full quote

@YP80, please keep it for me for a little longer.
I am sorry I clog the ECE thread with non related testing posts, it obviously belongs to the PCM vs FM thread.
I will continue the discussion over here: Re: FM and PCM Volume Balance of SB Cards
* I re-posted this there with a little example how I measure the "d" value.
EDIT: Or in the SVN thread.

@Qbix, I have found an even more accurate test (see below), and 140 will be closer to a SB1/2, although 150 will be leaning more towards the SBPro and exactly the SB16 value.
The resulting 140 is from a measured delta between PCM and FM of a SB1/2 and is below the filtered frequencies, so it is accurate.
Please consider to revise from 150 to 140, it's only 0.6db difference but for accuracy sake, it matters.
EDIT: Forgot to mention that the measurements done with "compat" as default, but "nuked" gives the same results.
EDIT2: Added SB16 CT2230; 150 in DOSBox matches the hardware SB16 but the SB16 has completely different filter (thus sound) than DOSBox or SB1/2, so it should not be taken as the reference card.

* MPXPlayer had some problems with the SB1/2 card on my system to test PCM and also there is difference between SB1 and SBPro stereo support.
So a constant rifle shot (mono) in Wolf3D reflects PCM behavior much better between cards, which in turn gave 1/100 db accuracy using a FreeG Channel Strip plugin in my DAW.

SB1 CT1320C PCM: -7.70 wolf3d rifle 3 OPL: -13.31 opl440 d: 5.61 […]
Show full quote

SB1 CT1320C
PCM: -7.70 wolf3d rifle 3
OPL: -13.31 opl440
d: 5.61

SB2 CT1350B
PCM: -10.07 wolf3d rifle 3
OPL: -15.80 opl440.
d: 5.73

SBPro2 CT1600
PCM: -16.74
OPL: -21.4
d: 4.66

SB16 CT2230
PCM: -6.57
OPL: -11.66
d: 5.09

DOSBox 200
PCM: -9.6
OPL: -12.04
d: 2.44

DOSBox 150
PCM: -9.6
OPL: -14.55
d: 4.95

DOSBox 140
PCM: -9.6
OPL: -15.14
d: 5.54

DOSBox 130
PCM: -9.6
OPL: -15.8
d: 6.2

Well the SB16 is the default in DOSBox, in that light the 150 isn't that bad right ?

Water flows down the stream
How to ask questions the smart way!

Reply 111 of 1550, by James-F

User metadata
Rank Oldbie
Rank
Oldbie
Qbix wrote:

Well the SB16 is the default in DOSBox, in that light the 150 isn't that bad right ?

DOSBox sounds nothing like a SB16, whether SB1/2 much closer to DOSBox with its passive filters.
Also, considering that most Adlib/FM games are from the SB1/2/Pro era makes the SB16 less favorable reference.
There is only 0.6db difference between 150 and 140, so it's not bad per-se, but 0.6db off target from the ideal balance in DOSBox in comparison to a SB1/2.

The exact middle ground would be (5.73+4.66)/2 = 5.2
Delta of 5.2 is 146 after peak testing, but that's down to splitting hair level.
If you decide to keep it at 150 I still be a happy camper, but not the happiest. 😁

I have located the problem, however we are discussing if we can't do it better.
Commited in 4013, let me know if it helps.

Thank you.
Also, where can I ask questions about SVN changes on Vogons without creating a new thread each time?

Last edited by James-F on 2017-05-10, 10:34. Edited 1 time in total.


my important / useful posts are here

Reply 112 of 1550, by Qbix

User metadata
Rank DOSBox Author
Rank
DOSBox Author
James-F wrote:
DOSBox sounds nothing like a SB16, whether SB1/2 much closer to DOSBox with its passive filters. Also, considering that most Adl […]
Show full quote
Qbix wrote:

Well the SB16 is the default in DOSBox, in that light the 150 isn't that bad right ?

DOSBox sounds nothing like a SB16, whether SB1/2 much closer to DOSBox with its passive filters.
Also, considering that most Adlib/FM games are from the SB1/2/Pro era makes the SB16 less favorable reference.
There is only 0.6db difference between 150 and 140, so it's not bad per-se, but 0.6db off target from the ideal balance in DOSBox in comparison to a SB1/2.

The exact middle ground would be (5.73+4.66)/2 = 5.2
Delta of 5.2 is 146 after peak testing, but that's down to splitting hair level.
If you decide to keep it at 150 I still be a happy camper.

What about the "fast" core ? (and optionally the emulated sbtype.)
With the information available now, I prefer to keep it at 150, but ideally I'd like to know the fast and if the emulated type has an influence according to your test.

Water flows down the stream
How to ask questions the smart way!

Reply 113 of 1550, by James-F

User metadata
Rank Oldbie
Rank
Oldbie

PCM sound doesn't change between SB types in DOSBox, I tried sb1, sb2, sbpro1, sbpro2 and sb16.
I use the Rifle (weapon 3) shot in Wolfenstein3D as a reference PCM, because it's a mono sample and this game works with all hardware SB types.

"fast" is exactly 0.6db quieter than compat or nuked, using the opl440.com software as reference.
OPL recording problem

DOSBox 200 Shot: -9.6 OPL3: -12.06, -12.04, -12.66 (nuked, compat, fast) d: 2.44 […]
Show full quote

DOSBox 200
Shot: -9.6
OPL3: -12.06, -12.04, -12.66 (nuked, compat, fast)
d: 2.44

DOSBox 150
Shot: -9.6
OPL3: -15.55, -14.54, -15.15
d: 4.95

DOSBox 140
Shot: -9.6
OPL3: -15.15, -15.14, -15.75
d: 5.55


my important / useful posts are here

Reply 114 of 1550, by Qbix

User metadata
Rank DOSBox Author
Rank
DOSBox Author

so 150 with fast (which is the current default setting), is the same as 140 with compat.

Water flows down the stream
How to ask questions the smart way!

Reply 115 of 1550, by James-F

User metadata
Rank Oldbie
Rank
Oldbie

Yes, but I tested only at 440Hz sinewave from the emulated OPL core.
I will test the new SVN as soon as YP80 compiles it, because I don't have the means.


my important / useful posts are here

Reply 116 of 1550, by KainXVIII

User metadata
Rank Member
Rank
Member
KainXVIII wrote:
James-F wrote:
Nope, it's not the OPL patch, adlib sounds fine, only PCM affected. Happens in Vanilla 4008 (Feb 2017). […]
Show full quote

Nope, it's not the OPL patch, adlib sounds fine, only PCM affected.
Happens in Vanilla 4008 (Feb 2017).

https://drive.google.com/drive/folders/0BxEUS … S2tCbEJRUXpnNTg
Please rollback from 4007 to 4000 one by one to find where the problem began.

PCM works fine with DOSBox SVN r4000 enhanced (without FLAC support).7z and garbled with rest of versions that i found by this link

Fixed in DOSBox ECE SVN r4014.7z, nice! 😎

Reply 117 of 1550, by valnar

User metadata
Rank Oldbie
Rank
Oldbie

I read that I needed to set this in DOSBox to make the levels correct for the Roland SC VSTi. I'm not sure if this is needed anymore, or if it should be tweaked.

mixer master 50 fm 75 spkr 25

Reply 119 of 1550, by Sammy

User metadata
Rank Oldbie
Rank
Oldbie

Sorry for asking, but is it planed to add ne2000 support ?

I need it when i play games like screamer Rally, and connect a dosbox with a real old PC over network.

And would i make things so unstable when its possible again to switch between glide=emu (build in patch) and glide=true for a wrapper (nglide) ?

Is it planed to add this things later ?

Dosbox ECE seems to be a great Project, but is missing some "must haves".

Ah, btw: where can i find a plain dosbox.conf for ECE to see what commands and variables are supported ?