v1.2.4 - Three SPWM drivers added - what's next?

[board707]

Hi all!
Finally I managed to get my panel with the FM6373 working, and I also figured out two more — those based on DP3264 and ICND2055 chips. All of these drivers operate on a similar principle, though they aren’t exact analogs. I added support for them to the library, so together with chips were already there, the project now supports six PWM‑type models: FM6353 / FM6363 / FM6373, ICN2153, ICN2055, and DP3264.

As we decode more drivers, a pattern is beginning to emerge. Comparing the operation principles of the drivers I’ve analyzed, I’d group them into three categories:

• First group — PWM timing and row switching are performed by the same signal, OE. (examples: FM6353, ICN2153)
• Second group — PWM timing is provided separately by a clock line, but row switching still requires a specific number of OE pulses, which links the two control signals. (example: FM6363)
• Third group — PWM timing is handled by the clock signal, while row switching is driven by OE pulses; the two signals are completely independent. (chips: ICN2055, FM6373, DP3264)

I plan to move on. This is only the beginning of a systematic classification. Decoding protocols for additional chips will expand the map of PWM drivers into new classes and types.
Unfortunately, adding new panels can’t be done from documentation alone — datasheets are often unclear or incomplete. Decoding each new driver requires hands‑on tests with the actual panel hardware. So far I’ve decoded every panel I had available; when I get new matrices I’ll try to add them to the library. My top priorities to learn next are chips SM16380 and SM16169.

It it seems to me also it would be useful to build a database or wiki for SPWM drivers, categorizing them by type — similar to the existing LED‑driver table but with richer chip descriptions, datasheet links, and logic‑analyzer recordings. The wiki could include illustrations explaining the general operating principles of driver families and the specific protocols of individual chips, like the resources has posted in the hzeller repo discussion.

We all need such a resource — PWM drivers have practically taken over the market, and we’ll inevitably have to work with them. I’m willing to gather this information, but I can’t do it alone, I need your help; my English is weak and I have limited technical‑writing skills.

[kingdo9]

Logic capture for

SM16380SH + 74HC245KAJ
128x64 1/32

https://mega.nz/file/qYEizb6C#5-YNLPRVJ_62qahELRQi4rzTeLkzORFCSRd480m-LAQ

@board707
new link
also included some comparison between SM16380 and SM16380SH , would need verifying.

[kingdo9]

@board707

Nice one,
If you need the full list, below are the register values from the rcfgx I was given from the provider for my SM16380SH display.
not sure whether the first config is 0000 like FM6373, I would need to check the start of the sequence again.

Sure ask in here about the visializer. It was only a test I was doing messing around. It would probably be easy enough to adjust parameters on how many LAT pulses there are etc to make it work with similar display captures other than fm6373.
Mainly though I can get the register configs from the rcfgx files now unless I don't have them.
The search how it works if I remember was hardcoded on where the FM6373 LAT pulses were.

1,021F,021F,021F
2,0300,0300,0300
3,0400,0400,0400
4,0500,0500,0500
5,0600,0600,0600
6,0750,0751,0753
7,0800,0800,0800
8,0900,0900,0900
9,0A02,0A02,0A02
10,0B0C,0B0C,0B0C
11,0C08,0C18,0C30
12,0D00,0D00,0D00
13,0E05,0E05,0E05
14,0F00,0F00,0F00
15,1000,1000,1000
16,1100,1100,1100
17,1200,1200,1200
18,1300,1300,1300
19,1414,1422,1432
20,1500,1500,1500
21,1630,1630,1630
22,1700,1700,1700
23,1801,1801,1801
24,1904,1903,1903
25,1A03,1A01,1A01
26,1B14,1B14,1B14
27,1C12,1C8F,1C8F
28,1D00,1D00,1D00
29,1E00,1E00,1E00
30,1F0C,1F0C,1F0C
31,2000,2000,2000
32,2200,2200,2200

[board707]

Hi @kingdo9

About the visualizer. I apologize for the slow exchange of messages...

On the image above I selected a four areas where I have a questions.

First - RED mark.
The bottom frames displays channel waveforms, located in the left and right halves. Sometimes, for easier viewing of certain signals, it would be convenient to move the border between the halves to the right or left, but I couldn't find a way to do this. Is this possible? All other frame borders move, but this one doesn't.

Second - a WHITE RECTANGLE
The label : "Left:Control | Right: Data..." - as I understand it, these refer to mouse modes. Holding down the left mouse button allows to move traces left and right, and holding down the right mouse button works the same way for me. Could you please explain the difference between "Control" and "Data" mode?

Third - Yellow mark
This is the data import field, right? After loading, it says something like "Loaded XXXXX events." And what are events?
I'd like to clarify where the question comes from. It seems to me that your visualizer might be most valuable for reading config register values. In a FM6373 driver, to read each config register you need to analyze a separate data load for the entire matrix, so to read several different registers, you have to load data files with millions of points. I encountered a problem with the visualizer crashing when loading large files (probably because I don't have enough memory; I only have a 32GB), so I tried reducing the number of points by excluding some traces—for example, leaving only one color channel. However, your visualizer can't read such files and reports that there are "No events found". What is this error and how can I get around it?

And last - a blue mark - a data field
I remember that you wrote that reading values ​​is strictly tied to the position of the LAT signals in the 6373 driver. How difficult would it be to change this? I think it would be very useful to be able to specify a "zero point" directly on the graph, starting from which signals would be decoded into bytes or words. Then your tool would be able to read values ​​in any format, regardless of protocol or driver type.

Please don't take the above as a complaint. I understand that you developed this tool for yourself and customized it to your needs. If you could answer some of my questions, I would be grateful.

[kingdo9]

@board707

It would be good to make it more adaptable for different driver configs, not just FM6373
The way it works I think would be hard to optimize for captures with multiple frames. It can only handle one frame unfortunately.
I have an alternative that might interest you at the end of this comment to extract registers from multiple frames !

Red mark - Should be able to add a frame adjustment for the channels

White mark - Left Control / Right Data - To be honest I cannot remember what that was for. I can remove that !
On the Visualizer if you have a mouse with a scroll wheel it allows you to Zoom in/out. Did you know about that?

Yellow mark - An Event is classed as one line of the KingST capture csv file - No events found would be it not detecting the FM6373 structure most likely. I would like to make the search more flexible to detect other displays.
Blue - Same as above.

I have a better tool for register extracting, again only FM6373 designed for, I'll try to adapt this also.
if you use the -h switch, you will see how to specify which RGB channel or which Block of the registers section to extract, Example: how the FM6373 has 5 "blocks" at the start and only the 3rd block changes right.

python reg_extract.py SM16380SH_4_frames.csv --channels R0,G0,B0 -o output.csv --frame-start-lat-pulses=3,14
python reg_extract.py FM6373_4_frames.csv --channels R0,G0,B0 -o output.csv --frame-start-lat-pulses=3,11,14

Register Extractor.zip

 

I don't use this much as Its easier just to get them from the rcfgx config files, if you need any specific display registers to test with let me know.

[board707]

Thanks for your quick response. I will glad to try your script.

[kingdo9]

Thanks for your quick response. I will glad to try your script.

@board707
this register extractor version now works with FM6363C, FM6373, SM16380SH and others I hope.
reg_extract_v2.zip

  FM6373:     --frame-start-lat-pulses=3,11,14
  FM6363C:    --frame-start-lat-pulses=14,12,3,14 --register-clock-sequence=*,14
  SM16380SH:  --frame-start-lat-pulses=3,14

[board707]

Thanks
I can't see it from the work, so I study them later

[kingdo9]

Logic capture for

FM6363C + DP32020A
128x64 1/32

https://mega.nz/file/PZk10ILK#AFqiaj5UVLBmrRx0v4etabdtUDRmRnxIIXHY1wkXFxo

Also do you have one of these displays SM16169 ? I think I can get one.

[board707]

do you have one of these displays SM16169 ?

I don't
I trying to obtain one, but no success so far

[kingdo9]

No more noise with this @board707

How do you do similar, splitting the wires to logic analyzer

From @board707 - sorry, I accidentally edit your message instead of replying to it and deleted the image. Please insert the image again.

[board707]

How do you do similar, splitting the wires to logic analyzer?

I most often use a direct connection to LA with standard flat cable (frame (1) on the picture). A 16-pin connectors with an angled pins (3) are ideal for this because they have thick pins and hold the wires very well. To ensure a secure connection to the STM32/RP2040 boards, I combine individual wires into multiple sockets (2).

[board707]

@andywwright , it's interesting.
I don't see a LA on your photo.

[andywwright]

That’s because it’s behind the masking tape, which prevents pins from falling out :)

[board707]

Do you have a link to the LA? I never seen working LA for 1$ :)

[andywwright]

Well, there’s no such thing anymore, as I bought mine more than 10 years ago, but the order of price is still the same. There are no links to “bundle deals” on Ali, as far as I know.

It reliably works at 8 MHz, I haven't tried more

sigrok-cli --scan -d fx2lafw                         01/26/2026 07:13:37 pm
The following devices were found:
fx2lafw:conn=8.31 - Saleae Logic [S/N: Saleae Logic] with 8 channels: D0 D1 D2 D3 D4 D5 D6 D7

[kingdo9]

I think when I was using a lower sample rate, I wasn't catching all the CLKs accurately. I use 50mhz now.

[board707]

@kingdo9
sorry, I accidentally edit your message instead of replying to it and deleted the image. Please insert the image again.

[kingdo9]

no problem.
Interesting, yes your setup looks very ideal. It seems a lot better than what I was doing on my first attempt !

You have a KingST also. I think they are very good for the price.