My customer is using the TVP5150AM1 TV decoder. Having a problem where the color carrier being digitized in addition to the regular TV signal. Normally this color carrier is filtered out before being digitized. It causes oscillations on the luminance data at the color carrier frequency when its not being filtered.
Any ideas for a possible reason!
Thanks-
Hello Saeb,
This question makes no sense as worded.
The entire composite video waveform including the color subcarrier is always supposed to be digitized by the TVP5150AM1.
Please describe the observed issue in detail.
There will always be some false color and dot crawl artifacts due to comb filter limitations.
Regards,
Jim
Hi,
Let me try and clarify the question. And this could all be just a Register mistake. I've attached as much info as I can.
Here you can see the inner block diagram. At the top "Composite" is where the Luma data is processed. The bottom is where the Cr and Cb enter. Chrominance is re-modulated and is subtracted from the Luma data. Now the luma data should not have color, and is then digitized. In S-video, the path goes through the mux without being subtracted. For NTSC, it is subtracted, filtered, and then digitized.
The problem almost looks like it's set for S-video, where the Color burst information is riding on the Y data. That shouldn't happen I don't believe, when setup properly. In the end, the digital data coming out should just be separate in the form YCrCb 4:2:2.
So I've created a video frame with a single yellow line on it. I've attached a picture. You can see how there are pixels are changing brightness.
I've captured the YCrCb data coming from the decoder where the line is. I find the Cr and Cb are pretty much constant as you would expect. Y has an oscillation on it. You can calculate that oscillation frequency to be about the color burst frequency. I've attached pictures. The video resolution is 720x242.
Using the Data I've graphed the amplitude of the Y data.
When I move the plug from the DVD player composite output and plug it into the "Y" portion of the component output I get this. And is in vary good shape.
Unfortunately, I can't seem to figure out exactly what the problem is. I've reviewed the registers, and they look ok to me. Maybe a second set of eyes would help.
I've also attached the schematic. The video input "INPUTA_GREEN" is driven from an onboard video amplifier.
Also on this board is a VGA video decoder which is fed data from the same source. It does not have any issues like this.
Anyways, Please help if you can. Thanks.
James McNeill
3073.TVP5150 RegDump.txt
3021.lastFrame.zip
6165.18_2101v2_2101v2.pdf
It looks like some of my pictures got removed.
TVP5150 Internal Diagram
Line With Color Oscillations
Line with Video connected to DVD Y of component output.
Hello James,
I haven't found anything yet that would explain what you're seeing, but I did find a few things that concern me.
According to registers 33h and 83h, you're not using the latest patch. Using the latest patch is strongly recommended.
The latest patch (v048CAA) can be downloaded from the following web site.
http://software-dl.ti.com/dsps/dsps_public_sw/dsps_swops_houston/ANALOG_VIDEO/Analog_Video_Decoder_Versions.htm
The source impedance of the two 300 ohm resistors (i.e. 150 ohms) is much higher than recommended (~37.5 ohms). The TVP5150AM1 clamp performance degrades with increasing source impedance.
Based on your schematic, I'm concerned that you're not following our PCB layout guidelines which recommends using a split ground plane (i.e. separate AGND and DGND). You basically have everything referenced to AGND1.
http://www.ti.com/general/docs/lit/getliterature.tsp?literatureNumber=slea081&fileType=pdf
Also, you don't appear to be using the recommended anti-aliasing filter.
http://www.ti.com/general/docs/lit/getliterature.tsp?literatureNumber=slea087a&fileType=pdf
Hi Jim,
Thank you very much for the help. I'm still in the process of implementing some of these changes. I had another question. If I buy new chips today, will they come with the newest firmware? Also, when I buy the chips, is there a way to identify what firmware version is in them from the packaging or part marking? The chips I have (TVP5150AM1IPBS) are marked with 5051l. The TI website says they should be marked with 5150Q.
Thanks for your help.
James
No, there are currently no plans to update the on-chip ROM code.
The latest version of the TVP5150AM1 ROM code is v04. The ROM code version can be read from I2C register 82h.
Recommend that you always use the latest patch from the patch download website.
I noticed you said the latest patch code was v048CAA. There are patch code versions with higher version numbers there. v04.90.02 and v04.A1.02. Is v04.8C.AA the latest? They all seem to have been released on the same day. Do all newer patch codes include the changes from the older ones?
James,
Please read the patch description.
The other patches are for different timing standards (e.g. square-pixel timing).
Improper video termination and signal level attenuation could lead to signal levels higher than that recommeded for the TVP5150AM1. This could lead to chrominance clipping at the ADC. We have seen this result in banding and artifacts when viewing certain colors.
Larry
About the Patch code.. The pdf document SLEA093 "TVP5150AM1 Patch Code Download Guidelines" says
Check the version number in registers 0x82 and 0x83. The values you will see are 0x04 and 0xXX,respectively. XX refers to the patch code revision number.
Since the patch code is v 04.8C.AA. What is the value that should be read back from 0x83 once the patch is done? Would it be 8C or AA or something else.
I've followed the download procedure from SLEA093 section 1.2. The device returns ACKs to everything, but I still read 0x00 from address 0x83.
Can I read back the code to be sure it was written correctly?
Never mind. I found my error. And Register 0x83 displays 8C for the current version v04.8C.AA.
The patch code did not fix the problem I am having. The Data coming from the decoder still appears to have color information in the Luma data.
This can be proven in data that I captured. The noise is there when a color picture is used and not when a black and white picture is used. Besides the termination and Anti-aliasing filter, do you have any other suggestions?
Is there an Eval board that I can test? There is the TVP5150AEVM but there is no stock anywhere.
Tomorrow I'll try adding the Anti aliasing filter and changing the termination.
James.
Thanks for your responses, but we have tried everything suggested and the problem still persists. You can see from the traces I provided that the chip is not separating color from luminence in certain cases. This is a very big bug that is unlikely to be fixed by a few tweaks to external components. How TI could sell this chip with this problem is a mystery. We've wasted weeks trying to get this fixed.
At this point, I can only see one solution to fix the problem, and that is to remove the chip and use a different one.
Make sure bit 5 of REG03h is set to 1 to make pin 27 an output. If pin 27 is floating and left configured as an input, color processing issues could result. The symtpoms you have decribed do not seem to be related to this, but this should be tried to make sure.
You tried attenuating the input as recommeded, right?
Have you tried multiple TVP5150AM1 devices?
Do you have a TVP5150EVM to try with your video source?
If you view a color bar test pattern, are colors mostly correct and stable, and just certain colors have vertical bandiing present?
Hi Larry,
It looks like register 0x03 bit 5 is already being set to 1.
I did try attenuating the input and lowering the value of the termination resistors. I also patched in the anti-aliasing filter. And updated the patch code. Overall, I see no difference. Not even a little difference. I previously attached a file "3021.lastFrame.zip " . It shows one frame of video. It's black with a single line in the middle.
This is the test because in some of our applications, there is just lines and text. This is how the lines show up.
Now, if you play a movie or moving video of some sort, you can't see a problem at all. Regular active graphics looks fine.
We have tried other TVP5150AM1 chips. This IC is used on 4 of our products. Each product uses 2 or 4 of these devices. All our products exibit this problem. Some are connected to an H.264 encoder, some to an FPGA. All have this problem. So we've tried multiple chips across multiple circuit cards.
We have not tried the Eval board because it's not in stock anywhere. It is actually for the TVP5150A and not AM1. I downloaded the software to checkout the register settings. There is definitely something strange there. Check out page 3-8 of the TVP5150AEVM User's Guide ( SLEU041 ). These Register settings don't make sense when compared to the Memory Map for the 5150A or 5150AM1. So the EVM board may not be useful, and the reason it's not stocked anywhere.
BEGIN_DATASETDATASET_NAME,”TVP5150A, Auto-switch, CVBS CH4A, 656 Out”// Initialize video encoder using an include fileINCLUDE, EncoderNTSC656_RTC.INC// Program TVP5150A registersWR_REG,VID_DEC,0x01,0x00,0x0C // Input: CVBS CH4WR_REG,VID_DEC,0x01,0x02,0x00 // Auto−switch modeWR_REG,VID_DEC,0x01,0x04,0x3F // Auto−switch MASKWR_REG,VID_DEC,0x01,0x06,0x40 // Pedestal offWR_REG,VID_DEC,0x01,0x31,0x05 // RTC modeWR_REG,VID_DEC,0x01,0x33,0x40 // ITU 601 (extended range)WR_REG,VID_DEC,0x01,0x34,0x11 // OUTPUTS ACTIVEWR_REG,VID_DEC,0x01,0x35,0xFA // GLCO and FID ACTIVEWR_REG,VID_DEC,0x01,0x36,0xAF // HS and VS ACTIVEEND_DATASET
We have not tried the color bar test. I could. The banding I see is Horizontal and not vertical as in the ...lastFrame.zip file.
What is the difference between the TVP5150AM1 and the TVP5151? They appear to be pin / register compatible. Would it be any use to try the TVP5151?
Thanks for your help. Talk to you later.
Best Regards,