SparkFun Electronics Forum

Here's what I found on the subject:

"MJPEG in AVIs having the MJPG fourcc, is restricted JPEG with a fixed -- and *omitted* -- Huffman table. The JPEG must be YCbCr colorspace, it must be 4:2:2, and it must use basic Huffman encoding, not arithmetic or progressive. "

This means it should be quite easy to get the cam to output right format as long as you can force the huffman table to standard and then omit it in the stream.

rgbphil wrote:a FIFO RAM might be of more use than a SRAM

You're absolutely right. Thanks for pointing me to this chip.

I edited my previous message about HCOUNT and VCOUNT calculations.

Here is what I added (although it still may not be correct):

buffercam wrote:It seems that the HCOUNT calculation is based on counting the clocks in a single row of image data at 1/2 the DSP clock frequency.
So for the Full Mega size, note that the data output is output at 1 x DSP frequency. This means you divide the number of clocks by two when calculating HCOUNT.
Smaller images are output at 1/4 the DSP clock rate, so you need to *multiply* the number of clocks by 2.
The result is that for every picture size, HCOUNT comes out to be approximately the default value of 713 (after you subtract the offset of 677). (I actually calculate 715.)

I also wish there were better docs. All we know are whats in the appnote together with the list of names and addresses in the datasheet.
what HCOUNT actually does appears to be a bit vague.

I have issues with my cams not delivering the entire frame and suspect this to be a small setting-error simply due to poor docs.
Best I can do is adjust so I get about half a image. Still, the image is then completely wrong colorwise. It's like it's YUV or something even though RGB is active.

Kreature,

Are you sure your camera is outputting at 15 fps (the maximum output rate specified in the datasheet)?

My camera tries to output at 30 fps when the frame rate is set to the max.

We have three settings for FRMSPD (specified in the Application Note):
Max frame rate
Half frame rate
Quarter frame rate

The datasheet only specified two frame rates: 15fps and 7.5fps.

Based on what I've measured, I think this is the relationship between the two:

Max frame rate = 30 fps, doesn't work properly?
Half frame rate = 15 fps
Quarter frame rate = 7.5 fps

Perhaps they initially designed the camera to do 30 fps, but couldn't get it to work in hardware, so they changed the datasheet to say 15 fps (instead of 30 fps)?

At any rate, try setting frame speed to "Half frame rate."

-David

I have been using lowest framerate as I am focusing on still capture to start with. What settings are you guys using for the rest of the parameters? Like RGB/YUV and audo-contrast/exposure...

After a lot of attempts:

Address Data Comments
0x0B 0x00 White Line OFF
0x58 0x20 Exposure Time
0x05 0x00 Frame Rate Quarter
0x1A 0xFF
0x1B 0xB3
0x11 0x4A
0x14 0x33
0x04 0x0D RGB, 352x288, OUT ON
0x1F 0x0B
0x1E 0xC3
0x0E 0x1E



Best regards,

ma4826

Thank you and congratulations.

Where did you find info on this secret register?
0x0B 0x00 White Line OFF

EDIT: I'm starting to get the feeling you captured this image with the TCM8230MD. Is that the case? EDIT: Nevermind.

Agreed, that is brilliant!
I will be testing this soon.

I've been writing structs for this cam and will incorporate all the details I find.

Anyone know how to define structs endian?

Thank you and congratulations.

Where did you find info on this secret register?
0x0B 0x00 White Line OFF

EDIT: I'm starting to get the feeling you captured this image with the TCM8230MD. Is that the case?

Testing the registers one for one until 0x19.
With the TCM8240MD.

With 0x0B default value:

OMG! That is the problem I have been having!
Basically, the pink stuff on the left side is over-exposed and with wrong colors. If you take into account me not setting the exposure that leaves the right side black on my shot and the left hidiously pink!

I think I'll have to play around a bit to get this right but wohooo!
Finally the missing piece!

Thanks for clarifying.

How did you calculate SPCOUNT?
The AppNote says SPCOUNT = 2 x (H_COUNT + 191) for that PICSIZ. That's different than what you used.

Also, what is Register 0x0E?

With the values of appnote I got the same results as Kreature.

I got the correct values with these automatic tests.
By changing these values the image moves within the "picture" and changes the color.

For example:

with 0x0E = 0x3C



Best regards,

ma4826

Very helpful, thanks.

Haha. To do that rigorous testing would kill a year for me.
I have been reading my images out with a ATmega88!
It takes time and then I have to view them manually.