Seiki SE50UY04 3840x2160 50" TV ($1300)

With each new firmware, the test pattern becomes clearer. The test pattern is in honor of the IBM T221.

closer_testpattern.JPG


almost_correct_testpattern.jpg
 
Looks like I may have damaged the panel during testing. One segment (1/12th of the screen) has half of the pixels always showing blue :/.

The test image is much closer to what it should be, but still many glitches.

closer_to_corrrect_pattern.jpg


Now to start working on making the gradients appear properly...
almost_correct_testpattern2.jpg
 
Last edited:
Two bugs remain. What is probably a hardware/assembly error on one of the blue lines on the left(upper) quadrant and an indexing error which should be an easy fix.

Currently testing at 4k60, will test higher only after the bugs are worked out.


testpattern_two_bugs_remain.JPG
 
Fixed the blue lines. Was a soldering issue on the output connector.

4k120 good test pattern (aside from the offset, but that's a minor fix)

4k120_almost_perfect_testpattern.jpg
 
integer scaling now works. Bits are now moving around nicely in the fpga. Next step is to bring up and debug the input block.
 
This is freaking insane. I didn't even know such a thing was possible, but here it is. I may need to pick one of these displays up on eBay...

Edit: We should see if we can put together a list of TV's with the appropriate panels for this custom tcon. Innolux makes a lot of variations on this panel with the same interface, there's got to be something out there that uses it...
 
Last edited:
So we've got a bunch of Seiki's newer SE42UM TVs at work that we use as monitors. Some poking around has revealed that they have a service menu very similar to the one in the SE50UY04 and SE39UY04 (though I can't find the engineering menu, they probably used a different code). Poking around that menu reveals that they use a 120hz Innolux panel with the same V-by-one interface, namely the V420DK1-QS1 (It seems Seiki really likes these 120hz Super MVA panels).

Said TV is currently $329 on Amazon, Prime shipped, and often touches the mid $200s. I think some tweakery is in order...
 
Yup! We've got a couple dozen at work because they're so cheap and they make great monitors if you can put up with 4:2:2 chroma (4K60 out of the box wheeeeeee!). Add one of Cirthix's boards and you've got a 42" 4K120 monitor for a fraction of the price of that $5000 30" OLED Dell was supposed to be launching. :D Only real disadvantage is the display technology, but S-MVA is still pretty good (though I am drooling over that OLED contrast and response time).
 
Last edited:
Either a pair of DP 1.2s or a heavily overclocked dual-link DVI. DP 1.3/1.4 hardware still hasn't reached maturity, sadly.
 
Either a pair of DP 1.2s or a heavily overclocked dual-link DVI. DP 1.3/1.4 hardware still hasn't reached maturity, sadly.

You lost me there. Wikipedia says that both NVIDIA and AMD have cards on sale with DP 1.4.

Using CVT-R2 one can squeeze 4k 80HZ inside a single DP1.2 signal. A combination of custom timings and overclock can theoretically fit even more. But you would need a short and fat custom DP cable in any case. This is the bleeding edge of display signal delivery. it feels safer to use 2 cables, so we are not limited to 3ft cables.
 
Wikipedia says that both NVIDIA and AMD have cards on sale with DP 1.4.
Like many times on wikipedia, this just because NVIDIA and AMD claimed they support DP 1.4

However, this largely unsubstantiated claim - because there are no working DP 1.4 endpoint you can actually test it with. I think there is big probability that with current drivers even if you get DP 1.4 endpoint/monitor right now, you won't get DP 1.4. They probably just know its easy for them to update drivers/output code for tweaks that DP 1.4 require when such monitors/endpoints actually arrive.

But from what I know DP 1.4 (and even 1.3) turned to be actually very hard to implement with any kind of reasonable stability and mass produce-ready price point for the whole producer-cable-consumer path, outside of very expensive experimental setups.
 
If i remember correctly, DP 1.2 was available on VGA long before DP monitors reached the market. cable length is a DP limitation, and DP 1.4 did not helped it at all. but i am not brave enough to claim that "you won't get DP 1.4 with current VGAs".

EDit: The First DP 1.2 graphics were the Radeon HD 6000 series, back on 2011. I have not found any mention of MST over DP 1.2 older than 2013, please correct me if i am wrong about the order to market.

DP 1.1 was the other way around, teh Dell 3008 come out almost 3 months before the first DP 1.1 VGAs reached the market.
 
Last edited:
If i remember correctly, DP 1.2 was available on VGA long before DP monitors reached the market. cable length is a DP limitation, and DP 1.4 did not helped it at all. but i am not brave enough to claim that "you won't get DP 1.4 with current VGAs".
Its not that hard to increase clock speed if you have capable output stage on your GPU.
But getting this signal across cable and have your monitor get it without corruption and/or loss of sync is another matter.
They pulled it off with doubling of bandwidth from 1.1 to 1.2. Then they (DP specifiers) proposed it will be almost doubled again. But then it suddenly appeared you can't just hope to double your bandwidth without eventually hitting very hard limits imposed by physical transmission layer (which was largely unchanged).
Unlike that, Thunderbolt hard sufficient reserves to up bandwidth (because much more complicated & expensive physical layer), and, unlike DP 1.3, actually reached maturity with TB3 hardware. The way things go, it may actually be the next standard gaining traction for high-bandwidth displays.
 
Here is a little on-screen display showing a realtime scrolling graph of the input framerate. Will be neat to see when driven with a freesync source. There are a few bugs that still stand in the way of the full input->output path, but it is almost entirely working now.

The tcon also supports 1:1, 1:2, 1:3, and 1:4 scaling.

fps_osd.jpg
 
Progress marches on.

Not going to say what "full speed" is, but it is operating at "full speed" here. There are still bugs and glitches to fix.


real_image_glitchy.jpg
 
HDMI 2.1 is out. requires cables capable of 48Gbps . time to finish this board and start working on the next one:

http://www.cepro.com/article/new_hdmi_2_1_specification_supports_8k60hz_with_hdr_4k120hz

Edit : the most relevant spec for gamers is *sync becomes an integral part of the standard:

Game Mode VRR features variable refresh rate, which enables a 3D graphics processor to display the image at the moment it is rendered for more fluid and better detailed gameplay, and for reducing or eliminating lag, stutter, and frame tearing.

http://www.fox34.com/story/34180243/hdmi-forum-announces-version-21-of-the-hdmi-specification
 
Last edited:
Oooooh, nice. Can't wait for that to start showing up on GPUs. Also, should be compatible with this board since HDMI is handled by a separate board.
 
Rough idea of availability and price on these alternate driver boards? I have a 42" Seiki I was planning to return, but if it is suddenly 120Hz capable with next to no input lag...
 
Some updates, good and bad. There has been an enormous amount of work put towards this, even in the times that I fell silent on this thread. 12+ hour days working on this project... for six months.

First, the bad:
I've found that the 39" panel is not 'split' and as such, cannot have left and right halves driven independently. This is a bummer because I had planned to use a pair of dp1.2 as the interface to the outside world. 2* 1920*2160 halves now will not work. Sync within one line is necessary for proper operation with two halves. I am working on sourcing a DP1.3 input board, but this could take some time.
4k144 isn't going to work on this panel, not with this FPGA.

Now the good:
It is pretty much usable now. The only hardware change that I made to the board was to add an input pin for the OSD enable/disable. The board design worked 100% on the first try. Yay for engineering.

The image looks great, no stability issues at full speed with the dp1.2 board. No loss of image quality when running at 120Hz... or in other more 'special' modes ;). With the dp1.2 board, 4k60 works fine, 2160*2160 is about the highest that it can take 120Hz at due to using only four of the eight input channels.

Input timings are now "DE-only", meaning that the hsync and vsync signals are completely ignored. Only the horizontal active, horizontal total, vertical active, and vertical total matter.

The TCON now can accept a pretty wide range of inputs in terms of both frequency and resolution. It automatically reconfigures itself to support video input at ~10MHz~165MHz per input channel (there are 8 channels). Getting a wide range to work is important for compatibility reasons.

There is a cool little scrolling graph that displays the current configuration and graphs the incoming framerate. This can be turned on or off through an input pin.

Error and warning messages now exist. If there is a problem, a nice big error message will come up on top of the test pattern telling you what the problem is.

All of this is done without adding latency ;).

And the interesting:
Found a bug in the gtx980, sometimes lines are skipped within the video frame (eg: 1920*1080 becomes 1920*1078 for one frame, then resumes normal operation) when operating DVI beyond the expected spec . This lineskipping behavior happens well before the link starts showing other visual glitches, so the bug is in the display logic and not the transmission or reception. To be fair, you have to use the pixelclock patcher to be able to go fast enough to trigger this behavior. The same 980 does not have this problem when using a displayport output. An R7 260x did not exhibit this problem.

A neat find, I spent two days trying to fix signal problems within the fpga before figuring out that it was actually the video source at fault.


Once I make the board boot from itself instead of being manually booted from JTAG, I will use it as a primary display for a while to see how things go. Nearly at this stage :)


PS: The "extra modes" are probably just as interesting as 4k120.

Do I release now with dp1.2 and offer dp1.3 in the future? It might take some months for that.
 
2160x2160 120hz sound interesting.

How much wider can we go if we drop the refresh down to 100hz?
 
No idea what 1.3 would add when you just single handedly invented 4k 120hz on a seiki!!!

I'm kind of confused, this is a thread for the 50" Seki. Is this new board for the 50"?
 
No idea what 1.3 would add when you just single handedly invented 4k 120hz on a seiki!!!

I'm kind of confused, this is a thread for the 50" Seki. Is this new board for the 50"?

39". 50" won't be supported since it has a different driving configuration.

I've pretty much taken over this thread as a build-log.

1.3 is necessary because using two 1.2s is not an option unless they are synchronized, and I don't want to add a frame buffer for this purpose because it would add latency. I originally thought that the panel was split and two mesochronous streams would be sufficient.
 
....

Do I release now with dp1.2 and offer dp1.3 in the future? It might take some months for that.

It is hard to answer, without any price indication.
If I understand correctly current board is:
- single DP 1.2
- 4k @ 60Hz
My questions are:
- is 1080 @ 120Hz still possible ( I think so, but to be sure)?
- What about freesync?
In my view, even in current state, it is nice upgrade to Seiki, and for right price, I'll buy it.
 
Answering your question with questions:

Q-Can you release with DP 1.2?

A-Sure, most of the compatible panels are currently running at 4k30Hz. The board ups the ante to somewhere around 4k 60-80Hz


What do we buyers wanna know before purchasing?

A-What is your stable refresh rate at 1080p from an HDMI 1.4/2.0 source( insert a HDMI>DP adapter)? Same question for DP 1.2 source. 4k speeds?

B- What TV functionalities will be preserved? will i lose sound? what about the ability to connect a HDMI set top box, android TV, etc?

C- Will i have the ability to fine tune color by gain and offset? I was in love with the panel acuraccy after calibration, mainly because unlike all the competition, this one had an engineer menu with plenty of color calibration options.


From my buyer perspective:

A- I really need to know how fast it goes with these sources. 1080p at 120Hz is enough- the VA panel can not handle more anyway. 4k above 60hz anyone? how far have you taken it with DP 1.2? 80Hz?75Hz? THIS IS THE MAIN VALUE BEING OFFERED!!!

B- having some cheap way to preserve sound is good (but not essential). accepting HDMI inputs from non-PC sources is a great bonus to avoid wife complains of "losing the TV".

C-Probably the least important because most buyers do not have a hardware color calibration device, but the ability to adjust color gain and offset opens the panel as a viable option for professional uses, as this one is among the best sRGB panels ever built at this price point.
 
Last edited:
As an Amazon Associate, HardForum may earn from qualifying purchases.
39". 50" won't be supported since it has a different driving configuration.

You really should've made your own thread.. This has nothing to do with the 50 then? Or are you planning on doing one for the 50", relevant topic model, too?
 
He can make his own thread when everything is done and ready. The topic was made three years ago and has since expanded to include a number of Seiki models. (Still, I am holding out hope for the 50" as well.)
 
actually, if you've had followed the thread since it started you would have seen that the 39" soon became the most discussed Seiki TV [H]ere, because it was smaller and more suited to be used as a PC monitor.
 
Back
Top