CurtPalme.com Home Theater Forum

[macgyver655]

Over the years and after many various tests I have a theory as to what is going on with ic421. It’s only theory because final testing and conclusion has not been done because I have lost interest in finding a solution and I also don’t think a solution is even needed at this point. But I thought I would post my findings and theory in case someone else may want to take it further.

First to look at the background testing. When you look at the image you see it jumping around along with changes in brightness and no apparent control. Now this is 1 clue but I’ll get into why later on when putting the clues together.

Now we know the problem is definitely in the chip itself since changing just the chip resolves the problem.

So to start testing I first removed power to the chip. This stops all shaking and brightness changes but as such also no control. This tells you the chip is still powering up and acknowledging but not properly.

Returning power back to chip and then removing data memory feeds HWR and LWR shaking again stops. This concludes that the chip is acknowledging data inputs and reacting to them but just not properly.

Removing chip enable also stops shaking which again indicates the chip is trying to function.

My next thought was the clock signal. Now we know the clock signal itself is good since changing the chip fixes it. But maybe the chip changes some how and doesn’t like the original clock frequency. So I experimented with supplying a different 20mhz signal source and then also changing the frequency going both up and down at various frequencies and this yielded “NO CHANGE”. So it’s not a clock issue.

Now testing the outputs would yield no information since we know the outputs ARE functioning since we are seeing changes on screen.

So what can we conclude at this point? We can conclude that the chip itself is not bad. A bad chip will, 1) pull supply power down, or 2) run extremely hot, or 3) not function at all. We know it’s turning on, we know its acknowledging incoming data and we know it is outputting data. It’s just not doing it properly.

So my next round of testing was to see if I could successfully read and/or write to the chip via its JTAG port. So I connected it to a PC and used an external power supply and tried various JTAGing software. After a few different software tries I hit one that gave me a read. However the data did not look correct. So I read it again and got a different read. My first thoughts were that I may be getting a phantom read so I remove power to the chip and did a read again, this time getting all F’s, which would mean a blank read. Powering the chip again gave me a different read again. This is another clue which I will also address later when putting the clues together. I then attempted to write to the chip but was a no go. So every read after that gave me a different code, every time.

So now we go back to the image itself. If you pull up a menu and watch the image you see that the image is not really shaking but rather the geometry, convergence and G2 is constantly changing. If you watch further you can see the changes are consistent and even more so, a cycle. You can pick any starting point and it will eventually return to that point and continue again and again to that point.

Keeping this in mind and the various jtag reads brings me to my conclusion.

The chip is LOOPED.

Some of you Sat guys from 10-15 years ago just went AH HA!

The chip is functioning but in a loop cycle. Which would also explain the different data reads.

Now a loop error is caused by an error in the chips firmware which can be triggered in different ways. And it would make since that it would happen early on in its life. Those that are still working of this serial code probably haven’t been triggered yet.

So why don’t the newer serial number chips have this problem. My guess is that Sony found out about this firmware glitch and rewrote the firmware for the newer chips. We know the chip is the same since it carries the same part number. If there were a change in the manufacture of the chip it would carry a different ending part number.

So can anything be done for the chips that don’t function properly? Well if you look back at those Sat days there was a device that could be connected to the looped chip, called an unlooper. It was also connected to a PC and special software was used to send various signals at the chip to try to get it to pause in its looping cycle just long enough to write a line of code that would stop the loop. The chip would then just be idle and a new code could be written to it and returning it to normal function.

So my next testing would have been to connect an unlooper and try some various software’s to see if the loop cycle could be broken and then just write all F’s to it. This would make it a blank chip. I would then try to make a minor change to the F code and write it to the chip and see if it would accept the write.

If this was successful then now it would just be a matter of getting a good read off a newer chip, which would be a simple 1,2,3 step. And then write this new code to the old chip.

Well that’s my theory but as I said I have no desire to continue. But if anyone else wants to delve into it I would gladly give my input along the way.

Well hope you enjoyed the read.

[barclay66]

[Curt Palme]

Fascinating! The talents of people on the forum here continue to blow my mind. Met a number of people at CLiff's place that had talents that I don't.



(no, wait, that was the midget stripper! )

[Curt Palme]

Fascinating! The talents of people on the forum here continue to blow my mind. Met a number of people at CLiff's place that had talents that I don't.



(no, wait, that was the midget stripper! )

[HaydnG90]

Truly fascinating. And with Halloween approaching quite timely. Maybe there's a way to bring these chips back from the dead.

[Nashou66]

I bet Craig would love to do the follow up as his Engineering background and extensive knowledge of the G90 would be a perfect fit.

This forum rules!!!

Nashou

[jeffslife]

Wow ! Great troubleshooting ! You people amaze me sometimes and thats not easy.

[CIR Engineering]

What would be interesting Mac would be to to a JTAG read on a chip out of one of the good lots, and then do a read off a working chip from the bad lot. If you are correct, the two firmwares should be different when compared.

craigr

[macgyver655]

[Curt Palme]

Oh would the two of you get a room... a lab specifically, and get on this?

I think both of you, and the rest of us would agree, that if this is do-able, there's a decent market for the boards.

I just sent Craig two bda IC421 YA boards, and have at least one more here and/or incoming as well. I've bought 20 G90s this year, and I'll bet I'll end up with 3-4 more before years end.

[Curt Palme]

Oh would the two of you get a room... a lab specifically, and get on this?

I think both of you, and the rest of us would agree, that if this is do-able, there's a decent market for the boards.

I just sent Craig two bda IC421 YA boards, and have at least one more here and/or incoming as well. I've bought 20 G90s this year, and I'll bet I'll end up with 3-4 more before years end.

[CIR Engineering]

Mac, can the FW on IC421 be read through JTAG while the chip is still on the YA board (I would think so). If so, I would consider trying to get a good read off one. Even if I have to remove an IC421 I might still do it.

Even if a bad chip is not looped, redoing the FW may repair the chip. It really depends on what actually is wrong with the IC.

What surprises me is that you were able to read the FW (you think) and that the IC is not locked.

craigr

[mp20748]

I've been doing something very similar the past year on a mini processor, that uses the MCB900 from Keil.

The MC900 uses "In Circuit Programming" where the chips are installed on the board and programmed using the MC900 PC interface to program them.


The MC900 will most likely not work with the G90 chip, but what I did find out in my research on this over the past year and in talking with a few chip manufacturers, is that all of the large scale SMD chips that require programming, it is being done external to the board itself once the board has completed manufacturing. The connection is usually serial, with a five wire connection (B-, Ground, Reset, RX, TX).

And these same chips all have a PC or Linux program software to get them started and for flashing.

On the few chips that I have been working with, I've been successful in re-flashing some of them and making the problem go away, to even have to replace the chip before re-flashing.

I agree with Mac in what I've been told so far and have experienced is that they chips are not really failing, they just seem to get locked up somehow, and what seems to have fixed most of them is a difference in programming. Much like how a firmware upgrade seems to solve a problem that was not noted before.

Anyway, it would be good to look into what software and programming interface the manufacturers used, because when they purchased the license key for the chips, they were provided access to the software, and what interface tool to use.

[macgyver655]

All my testing was done on a YA board with the chip still mounted. I was using an external 6v power supply and connecting it to ic421s 5v regulator. This was to insure a good clean 5v to the chip.

I connected a very small gauge insulated magnet wire to each jtag port and to an LPT port on the PC. Also before this I tested the PC and connection on a chip that I had jtagged in the past (not a Sony chip) and had no problem reading and writing to it so I know the PC and connection did function properly.

Now I was using a generic jtag program but should still be valid. It was reading something, I'm just not sure what. As I said if I attempted a read without powering the chip I got a blank read. Then just adding power I would get a full data read, just not a consistent one.

There is more advanced boundary scan software but a full working version is expensive. Plus most of these require a chips BSDL file which we can agree, getting stuff from Sony is no easy task. Now I thought about getting a similar chips BSDL file and altering that file for this chip but again, more work.

The process I wrote required to continue sounds simple, but it is a lot of ground work and a lot of testing to determine which direction to proceed. It's not a couple hour task.

[Tinman]

AH HA, Indeed. Makes perfect sense.

And I admit to NOTHING!

[macgyver655]