Nice project!

Keep it going!

Very nice ~

Very cool!
AFAIK CS9233 is a relabeled or license-produced Dream SAM9233. "This device is available as the result of a technology partnership between Crystal Semi-conductor Corporation and DREAM S.A. of France."

I had a 4MB SAM9233 daughter board failing spectacularly a few months ago, when a tantalum capacitor burned out. I still have to repair it, maybe I can learn something here.

gerwin wrote on 2022-11-24, 00:02:

Very cool!
AFAIK CS9233 is a relabeled or license-produced Dream SAM9233. "This device is available as the result of a technology partnership between Crystal Semi-conductor Corporation and DREAM S.A. of France."

I had a 4MB SAM9233 daughter board failing spectacularly a few months ago, when a tantalum capacitor burned out. I still have to repair it, maybe I can learn something here.

Thats good! Not that it's broken, but because you will have to look it more closely. I have 4MB version coming and may need some help with it. Already fixed couple of minor layout issues today. It's kind of swedish "tårta på tårta" design with optional 1MB or 2x2MB ROMs and effects processor in two different packages. Would be a miracle if it rocks.

fool wrote on 2022-11-24, 18:24:

Thats good! Not that it's broken, but because you will have to look it more closely. I have 4MB version coming and may need some help with it. Already fixed couple of minor layout issues today. It's kind of swedish "tårta på tårta" design with optional 1MB or 2x2MB ROMs and effects processor in two different packages. Would be a miracle if it rocks.

Sure, hope I can help. I have two different Terratec 4MB models and one 1MB model. Another such 9233 synth is on a Terratec Maestro 32/96 sound card.

gerwin wrote on 2022-11-24, 00:02:

Very cool!
AFAIK CS9233 is a relabeled or license-produced Dream SAM9233. "This device is available as the result of a technology partnership between Crystal Semi-conductor Corporation and DREAM S.A. of France."

I had a 4MB SAM9233 daughter board failing spectacularly a few months ago, when a tantalum capacitor burned out. I still have to repair it, maybe I can learn something here.

Yes, they are one and the same. Crystal licensed the design, software code, and MIDI instrument sounds from DREAM (Atmel) in the early 90's, including the knock-off Roland GS bank that triggered a lawsuit from Roland at the same time Roland had just licensed their GS sounds to Microsoft and Apple and console gaming studios that wanted to use MIDI in their games, while also beginning to offer their Virtual Sound Canvas software.

RetroGamer4Ever wrote on 2022-11-24, 20:22:

Yes, they are one and the same. Crystal licensed the design, software code, and MIDI instrument sounds from DREAM (Atmel) in the early 90's, including the knock-off Roland GS bank that triggered a lawsuit from Roland at the same time Roland had just licensed their GS sounds to Microsoft and Apple and console gaming studios that wanted to use MIDI in their games, while also beginning to offer their Virtual Sound Canvas software.

Yes, I heard about that lawsuit.
I actually like the pre-lawsuit Dream illegal soundbanks a little more then those made after the Dream/Roland-agreement. The former are my personal favorite Hardware Midi boards. Cancun FX being an example of the latter, which has a little too much bass. But all of them are great.

@fool
Below are large images of the hardware I mentioned. I don't have the back-sides on photo now.

EDIT:
Here are more such images by vogons member CHiLL72: https://www.waveblaster.nl
Here are more such images by vogons member Salient: https://www.wavetable.nl/daughterboards
The Netherlands and Belgium are the place to be now, for Dream-based midi sound 😉...
Also this place, by vogons member kira, now only an incomplete backup: https://web.archive.org/web/20110710232559/ht … ble.web.fc2.com

Build couple of pieces with rev2 PWB. This has options for CS8905 effect processor in two packages and 2+2MB or 1MB ROM configurations.
This revision has also "better" opamp, newer DAC and "audio grade" DC-block capacitors. Uploaded couple if DOOM 2 tracks, but haven't analysed those yet. Probably difference is close to zero. Biggest difference is that CS8905 version volume level is only half of. CS9233 mixes main synth and effect channels together multiplied by 0.5. Because there's no effects in DOOM 2 tracks the output level will be half. I amplified recordings some level but still much quieter than previous recordings.
DOOM 2 ENTRYWAY: https://on.soundcloud.com/FxMgD
DOOM 2 ENDGAME: https://on.soundcloud.com/eYqbG

I heard that 4MB ROM needs different firmware, nevertheless I had to try how it sounds. (board in the second picture) Produced "music" was strange like modern art. Some DOOM 2 tracks were still recognizable. Still that wasn't fun enough to actually record it what was my original intention. Logically next step is to find proper firmware.

Build this "latest latest & final" rev C. I really hope it will be the final revision.
This time I used SAM8905-1 because it looks badass and I need to verify it's functionality.
According to Dream datasheet it should be the second revision issue, 64MHz. Same as Crystal CS8905A, but in different package. So it's just oversized chip doing much nothing, but looks cool.
There should be also third revision of that chip, but I've never seen one.

Everything is hand soldered in about 2 hours - headache follows.

fool wrote on 2023-01-14, 12:19:

Build this "latest latest & final" rev C. I really hope it will be the final revision. This time I used SAM8905-1 because it loo […]
Show full quote

Build this "latest latest & final" rev C. I really hope it will be the final revision.
This time I used SAM8905-1 because it looks badass and I need to verify it's functionality.
According to Dream datasheet it should be the second revision issue, 64MHz. Same as Crystal CS8905A, but if different package. So it's just oversized chip doing much nothing, but looks cool.
There should be also third revision of that chip, but I've never seen one.

Everything is hand soldered in about 2 hours - headache follows.

Wow... I'm completely impressed.

It'd take me a LOT longer than 2 hours to solder all of that... and it would not have turned out that well.

Nice job!

Looks good ~

Finally the 4MB ROM version is rocking!
Here's some very first samples that were easy to record.

(sorry about the mandatory Doom E1M1)

UFO Enemy Unknown Intro https://on.soundcloud.com/73p5A
Duke3D Lunar - Reactor E2M7 https://on.soundcloud.com/ca55E
Doom E1M1 https://on.soundcloud.com/2LhKm
Doom II Entryway https://on.soundcloud.com/sdDv4

First thoughts... It certanly more alive compared to 1MB. Some highs are a bit too loud and sounds weird...not in balance. Maybe it needs low-pass filter tuning.
But I'm glad it works after all.

oh this does sound nice , kudos!
I have the bad version with the 1mb rom 🤣 now I wish there were pads to upgrade

Well...I would not say it's bad. There are also some horrible ones 😁 It lacks some detail but still performs quite well in many titles. "Lower quality, well balanced" I would say.
Some intrumets are even better, or easier, to the ear. I have found many weird ones in 4MB that should be remapped.

Regarding this period correct game with mediocre multimedia speakers some of the drums are quite pointless. 4MB wave table is more colorful and has some really nice sounds, as long as listening from quality monitors or headphones.

Very nice ~

Awesome job!

I awesome work fool. I still havent looked at that board with the two SAM9707's on I PM'd you about a while back 😐

fool wrote on 2023-01-14, 12:19:

Build this "latest latest & final" rev C. I really hope it will be the final revision. This time I used SAM8905-1 because it loo […]
Show full quote

Build this "latest latest & final" rev C. I really hope it will be the final revision.
This time I used SAM8905-1 because it looks badass and I need to verify it's functionality.
According to Dream datasheet it should be the second revision issue, 64MHz. Same as Crystal CS8905A, but in different package. So it's just oversized chip doing much nothing, but looks cool.
There should be also third revision of that chip, but I've never seen one.

Everything is hand soldered in about 2 hours - headache follows.

Nice work on the smd parts! Two hours is a long time to focus when dealing with tiny components. I'm guessing you finished around C9 and were starting to fight against fallibility.