maxtherabbit wrote on 2023-10-16, 02:01:

Maelgrum wrote on 2023-10-16, 01:39:

From Duke Nukem 2:

Details

Copy code to clipboard

1D1          - Enable Speaker
240 38       - Set Time Constant 5000 Hz
375 15 09    - DMA DAC, 4-bit ADPCM Reference
4D0          - Halt DMA Operation, 8-bit
540 38       - Set Time Constant 5000 Hz
675 15 09    - DMA DAC, 4-bit ADPCM Reference
7D0          - Halt DMA Operation, 8-bit
840 38       - Set Time Constant 5000 Hz
975 15 09    - DMA DAC, 4-bit ADPCM Reference
10D0          - Halt DMA Operation, 8-bit
1140 72       - Set Time Constant 7042 Hz
1275 3E 05    - DMA DAC, 4-bit ADPCM Reference
13D0          - Halt DMA Operation, 8-bit
1440 39       - Set Time Constant 5025 Hz
1577 25 05    - DMA DAC, 2.6-bit ADPCM Reference
16D0          - Halt DMA Operation, 8-bit
1740 72       - Set Time Constant 7042 Hz
1875 3E 05    - DMA DAC, 4-bit ADPCM Reference
19D0          - Halt DMA Operation, 8-bit
2040 06       - Set Time Constant 4000 Hz
2175 3A 03    - DMA DAC, 4-bit ADPCM Reference
22D0          - Halt DMA Operation, 8-bit

Uses ADPCM 4 and 2.6 bit

So any "click build" testing we did with duke2 should be invalidated

This also means 4.16 should be retested with DOTT on a CT1740

Is here any difference in clicking with this ? delay lowered from 100 to 10

Log from Prince of Persia 1.4:

1E0 C6       - DSP Identification
2E0 AA       - DSP Identification
3E1          - DSP Version
4D1          - Enable Speaker
5D1          - Enable Speaker
640 A6       - Set Time Constant 11111 Hz
714 9B 15    - DMA DAC, 8-bit
8------------- IRQ8
940 87       - Set Time Constant 8264 Hz
1014 E3 03    - DMA DAC, 8-bit
11------------- IRQ8
1240 87       - Set Time Constant 8264 Hz
1314 E3 03    - DMA DAC, 8-bit
14------------- IRQ8
1540 A6       - Set Time Constant 11111 Hz
1614 77 10    - DMA DAC, 8-bit
17------------- IRQ8
1840 87       - Set Time Constant 8264 Hz
1914 E3 03    - DMA DAC, 8-bit
20------------- IRQ8
2140 87       - Set Time Constant 8264 Hz
2214 E3 03    - DMA DAC, 8-bit
23------------- IRQ8
2440 87       - Set Time Constant 8264 Hz
2514 E3 03    - DMA DAC, 8-bit
26------------- IRQ8
2740 A6       - Set Time Constant 11111 Hz
2814 51 06    - DMA DAC, 8-bit
29------------- IRQ8
3040 A6       - Set Time Constant 11111 Hz
3114 77 10    - DMA DAC, 8-bit
32------------- IRQ8
3340 A6       - Set Time Constant 11111 Hz
3414 51 06    - DMA DAC, 8-bit
35------------- IRQ8
3640 87       - Set Time Constant 8264 Hz
3714 E3 03    - DMA DAC, 8-bit
38------------- IRQ8
3940 87       - Set Time Constant 8264 Hz
4014 E3 03    - DMA DAC, 8-bit
41------------- IRQ8
4240 87       - Set Time Constant 8264 Hz
4314 E3 03    - DMA DAC, 8-bit
44------------- IRQ8
4540 87       - Set Time Constant 8264 Hz
4614 E3 03    - DMA DAC, 8-bit
47------------- IRQ8
4840 A6       - Set Time Constant 11111 Hz
4914 77 10    - DMA DAC, 8-bit
50------------- IRQ8
5140 87       - Set Time Constant 8264 Hz
5214 E3 03    - DMA DAC, 8-bit
53------------- IRQ8
5440 A6       - Set Time Constant 11111 Hz
5514 77 10    - DMA DAC, 8-bit
56------------- IRQ8
5740 87       - Set Time Constant 8264 Hz
5814 E3 03    - DMA DAC, 8-bit
59D0          - Halt DMA Operation, 8-bit
6040 A2       - Set Time Constant 10638 Hz

6114 8F 04    - DMA DAC, 8-bit
62D0          - Halt DMA Operation, 8-bit
6340 A7       - Set Time Constant 11235 Hz
6414 15 04    - DMA DAC, 8-bit
65------------- IRQ8
6640 A6       - Set Time Constant 11111 Hz
6714 8F 04    - DMA DAC, 8-bit
68------------- IRQ8
6940 A8       - Set Time Constant 11363 Hz
7014 15 04    - DMA DAC, 8-bit
71------------- IRQ8
7240 A6       - Set Time Constant 11111 Hz
7314 51 06    - DMA DAC, 8-bit
74------------- IRQ8
7540 A6       - Set Time Constant 11111 Hz
7614 37 18    - DMA DAC, 8-bit
77------------- IRQ8
7840 87       - Set Time Constant 8264 Hz
7914 E3 03    - DMA DAC, 8-bit
80------------- IRQ8
8140 A6       - Set Time Constant 11111 Hz
8214 7F 12    - DMA DAC, 8-bit
83------------- IRQ8
8440 87       - Set Time Constant 8264 Hz
8514 E3 03    - DMA DAC, 8-bit
86------------- IRQ8
8740 87       - Set Time Constant 8264 Hz
8814 E3 03    - DMA DAC, 8-bit
89------------- IRQ8
9040 87       - Set Time Constant 8264 Hz
9114 E3 03    - DMA DAC, 8-bit
92------------- IRQ8
9340 87       - Set Time Constant 8264 Hz
9414 E3 03    - DMA DAC, 8-bit
95------------- IRQ8
9640 87       - Set Time Constant 8264 Hz
9714 E3 03    - DMA DAC, 8-bit
98------------- IRQ8
9940 87       - Set Time Constant 8264 Hz
10014 E3 03    - DMA DAC, 8-bit
101------------- IRQ8
10240 87       - Set Time Constant 8264 Hz
10314 E3 03    - DMA DAC, 8-bit
104------------- IRQ8
10540 87       - Set Time Constant 8264 Hz
10614 E3 03    - DMA DAC, 8-bit
107------------- IRQ8
10840 87       - Set Time Constant 8264 Hz
10914 E3 03    - DMA DAC, 8-bit
110------------- IRQ8
11140 87       - Set Time Constant 8264 Hz
11214 E3 03    - DMA DAC, 8-bit
113------------- IRQ8
11440 87       - Set Time Constant 8264 Hz
11514 E3 03    - DMA DAC, 8-bit
116------------- IRQ8
11740 87       - Set Time Constant 8264 Hz
11814 E3 03    - DMA DAC, 8-bit
119------------- IRQ8
12040 A6       - Set Time Constant 11111 Hz
12114 51 06    - DMA DAC, 8-bit
122------------- IRQ8
12340 87       - Set Time Constant 8264 Hz
12414 E3 03    - DMA DAC, 8-bit
125------------- IRQ8
12640 87       - Set Time Constant 8264 Hz
12714 E3 03    - DMA DAC, 8-bit
128------------- IRQ8
12940 87       - Set Time Constant 8264 Hz
13014 E3 03    - DMA DAC, 8-bit
131------------- IRQ8
13240 87       - Set Time Constant 8264 Hz
13314 E3 03    - DMA DAC, 8-bit
134------------- IRQ8
13540 87       - Set Time Constant 8264 Hz
13614 E3 03    - DMA DAC, 8-bit
137------------- IRQ8

This is interesting:

1------------- IRQ8
240 87       - Set Time Constant 8264 Hz
314 E3 03    - DMA DAC, 8-bit
4D0          - Halt DMA Operation, 8-bit
540 A2       - Set Time Constant 10638 Hz
614 8F 04    - DMA DAC, 8-bit
7D0          - Halt DMA Operation, 8-bit
840 A7       - Set Time Constant 11235 Hz
914 15 04    - DMA DAC, 8-bit
10------------- IRQ8

1. Game changes samplerate, and starts multiple buffer playings, interrupted by game events
2. All operations - Set Time Constant, DMA DAC 0x14, Halt DMA operation - is long in term of time on SB16, compared to SB Pro

Out of curiosity, does the clicking happen for Sound Blaster 16 games or just those played in Sound Blaster Pro mode on a SB16?

Kahenraz wrote on 2023-10-16, 05:31:

Out of curiosity, does the clicking happen for Sound Blaster 16 games or just those played in Sound Blaster Pro mode on a SB16?

All games designed for Auto init DMA will have no problems on SB16. Auto init dma was supported in fw 2.xx and later, so even simple SB (not Pro) can support it, but may be for compatibility with first SB (1.0/1.5), or because of developers laziness, some games uses single cycle dma.

MAME runs the DSP code in their own 'I80C52'-emulator:

1I80C52(config, m_cpu, XTAL(24'000'000));

https://github.com/mamedev/mame/blob/master/s … us/isa/sb16.cpp

isn't in that regard MAME better to collect logs from their 'I80C52'-emulator directly, i.e. logs that are coming from the execution of the DSP code itself via that 'I80C52'-emulator and what exact DSP code was executed?

With DOSbox aren't all you get as logs is what the game is doing, i.e. what it's sending to the DSP as requests? then I guess, based on those requests and the DSP code itself, you assume what actual DSP code was executed. I don't know, maybe, that assumption is equally good to what MAME logs will show...

Afaik MAME isnt cycle accurate, I dont think MAME SB16 is clicking. Might be good for tracing 8052 firmware execution, but not for definitive verification of potential fixes.

rasz_pl wrote on 2023-10-16, 09:16:

Afaik MAME isnt cycle accurate, I dont think MAME SB16 is clicking. Might be good for tracing 8052 firmware execution, but not for definitive verification of potential fixes.

It is rather accurate. Which cycles do you mean? Those of x86 or those of other emulated CPUs?

mattw wrote on 2023-10-16, 08:40:

MAME runs the DSP code in their own 'I80C52'-emulator: […]
Show full quote

MAME runs the DSP code in their own 'I80C52'-emulator:

Copy code to clipboard

1I80C52(config, m_cpu, XTAL(24'000'000));

https://github.com/mamedev/mame/blob/master/s … us/isa/sb16.cpp

isn't in that regard MAME better to collect logs from their 'I80C52'-emulator directly, i.e. logs that are coming from the execution of the DSP code itself via that 'I80C52'-emulator and what exact DSP code was executed?

With DOSbox aren't all you get as logs is what the game is doing, i.e. what it's sending to the DSP as requests? then I guess, based on those requests and the DSP code itself, you assume what actual DSP code was executed. I don't know, maybe, that assumption is equally good to what MAME logs will show...

Re: The Soundblaster DSP project
Re: The Soundblaster DSP project

rasz_pl wrote on 2023-10-16, 09:16:

Afaik MAME isnt cycle accurate, I dont think MAME SB16 is clicking. Might be good for tracing 8052 firmware execution, but not for definitive verification of potential fixes.

This Mame SB emulator is based on dump research, not on real hardware.
Few hints from this source code was useful.
But it doesn't emulate Bus interface chip properly,
and no assumptions can be made on possible problems.
It's a toy. Very nice toy, but toy.
It's like using dosbox for solving hardware problems.

be interesting to see how it is with any demos using GoldPlay because it does single byte dma and does funky timing with the dsp.

To understand that happening on hardware side, you need:
1. Capture x-bus exchange
2. Capture raw DAC data
3. Capture DMA data
All of this - simultaneously, and on large data set.
32 bit logic analyzer with very large memory, or streaming real-time, with some additional dac-decoding circuit.

Maelgrum wrote on 2023-10-16, 12:13:

It's a toy. Very nice toy, but toy.

So are the sound blaster cards themselves 😉

Maelgrum wrote on 2023-10-16, 12:26:

To understand that happening on hardware side, you need: 1. Capture x-bus exchange 2. Capture raw DAC data 3. Capture DMA data […]
Show full quote

To understand that happening on hardware side, you need:
1. Capture x-bus exchange
2. Capture raw DAC data
3. Capture DMA data
All of this - simultaneously, and on large data set.
32 bit logic analyzer with very large memory, or streaming real-time, with some additional dac-decoding circuit.

Or use brain and/or be the original author of the device.

$5 salea clone snooping on /DACK, DRQ, IRQ together with I2S stream and something to decode port 220 access will be plenty to start with. Might not even need I2S, just seeing how long the pauses are between DRQ and port 220 reads/writes would tell a lot.

Maelgrum wrote on 2023-10-16, 02:48:

Is here any difference in clicking with this ? delay lowered from 100 to 10

Sounds the same as version 5.

S95Sedan wrote on 2023-10-16, 15:51:

Maelgrum wrote on 2023-10-16, 02:48:

Is here any difference in clicking with this ? delay lowered from 100 to 10

Sounds the same as version 5.

To my partially deaf from aviation ears and used from childhood to Apple2 sounds these examples are good enough and I hear no defective cracking, but you are the experts. Shall I attach quality headphones to my laptop to hear what is bothering you?

georgel wrote on 2023-10-16, 16:29:

S95Sedan wrote on 2023-10-16, 15:51:

Maelgrum wrote on 2023-10-16, 02:48:

Is here any difference in clicking with this ? delay lowered from 100 to 10

Sounds the same as version 5.

To my partially deaf from aviation ears and used from childhood to Apple2 sounds these examples are good enough and I hear no defective cracking, but you are the experts. Shall I attach quality headphones to my laptop to hear what is bothering you?

I can hear the clicks on my super tinny monitor speakers.
The best example is duke where it clicks after every sound. in the DOTT example i can just hear some major clicks (when he says "I dont think you should drink that.".and in a suction cup sound after he said "aggressive" The last example also clicks after every repeat of the soundeffect.
I am judging from my ears, but i am sure that you could see the spikes in the recording when opened in Audacity for example.

S95Sedan wrote on 2023-10-16, 15:51:

Maelgrum wrote on 2023-10-16, 02:48:

Is here any difference in clicking with this ? delay lowered from 100 to 10

Sounds the same as version 5.

Yeah I definitely hear clicking in the DOTT sample

Maelgrum wrote on 2023-10-16, 12:13:

It's a toy. Very nice toy, but toy.

it might be so, but if it can be even of minimal help then building MAME from source and thus be able to add debug messages, etc. turned out to be ultra simple, I documented it here:

Re: MAME Roland MPU-401 Emulator

(I guess the command line will need to change to something like:

mame ct486 -hard1 msdos622 -isa2 sblaster_16

for SB16 support and files "ct1741_v413[80c52].bin" and "ct1741_v413_xor" need to be added to the roms directory)

So, if someone need it - then even spending the time to install 'Ubuntu 22.04.2 LTS' (I saved that time in my case, because I already had such installation) is probably less than 1 hour time and most of the time is just waiting - in my case relaxing and drinking several cups of green tea.

Also, I guess that is helpful if someone is interested to improve/fix MAME support for SB16 based on all and any new findings - I guess they did best based on their knowledge and information at the time.

Today I went back and retested all of maelgrum's test builds from source with DOTT. Unfortunately they are all still clicking. I can't really notice any difference between the clicks and stock either