Anyone ever had a 0021 as a result of jumping j2d2.4 and .7?

[crossedflags]

I've done way over 25 jtags and never had this problem. (I say 25 because thats the amount of keys 360 flash tool lets me keep)

Falcon mobo, When I jump j2d2 -4 and 7 I get a 0021.  I pull one leg and it goes away and boots fine. Its has orginal NAND image on it, (I did have xell installed but I restored the original image for troubleshooting purposes. Just for giggles I put a 220 ohm resistor between the j2d2 4 and 7 points to see what it does. it ringed me with 0021.

Any Ideas?

[Blackaddr]

The jumper wire in question de-asserts the JTAG reset pin by tying it off to 1.8V, this allows the JTAG port to function on the GPU.  Without the jumper, MS has a pulldown resistor to hold the JTAG in reset all the time.

Either state will allow an original NAND to boot, so the wire itself cannot cause the problem you are having.

I suspect there may be an accidental solder bridge somewhere, or your console is simply having a RROD for more conventional reasons.  I'm not sure if 0021 is actually related to GPU or not.

[l_oliveira]

I've seen situations where simply pulling the reset pin to 1.8v on FALCON causes it to 0020/0021/0022. I usually put a diode (It causes a drop of voltage but the voltage is still enough ro cause a logic level 1 be detected.)

It's either excessive current being drawn by the pin or it running internally at a lower voltage and in both cases we're overdriving it.  And this only happens with the "GPU RHEA" (FALCON) chip.

[crossedflags]

The diode did the trick,  I inserted a 4148 with the cathode side facing pin 4.

Thank you so much!

[Blackaddr]

It's either excessive current being drawn by the pin

The reset pin is a MOS input so it's very high impedance.  The leakage current will only be a few tens of uA at the most.

or it running internally at a lower voltage and in both cases we're overdriving it.

If you look at the schematic, the JTAG is clearly running on V_MEM for I/O voltage which is 1.8V.

Clearly some people are having an issue tying it off though.  It's typical to bring up your core voltage rail before you bring up your I/O voltage rail.  I suppose it's possible when the I/O rail is coming up that there is a glitch on the reset line which causes problems in the TAP controller.  Adding a resistor or diode could be filtering out that glitch.  It would be interesting to remove the diode, replace the jumper wire, but place a capacitor on the line to try to filter out any glitching without inducing a voltage drop.  10N, 100N, and 1U would be a few good values to try.

I've never had access to a falcon that has this problem so I can't investigate it myself.  Perhaps someone else try the caps?

[yaywoop]

hey Blackaddr, is your theory that its acting like a 'transmission line', bouncing along the reset wire, thus capacitive terminations would match the line impedance and prevent reflections?
I know this can occur in high frequency digital circuits.. and the behavior can seem random, as it does in this case.
or where you thinking of something totally different?

anyway I have a falcon showing intermittent 0021 errors after a year being jtagged. its got a diode on it at the moment. might try replacing this with a resistor and capacitor, doing some experiments. although because the 0021 occurs very infrequently, might be difficult to test.

[Blackaddr]

In this case, I don't think it's a transmission line issue, but perhaps more of a power up issue.  In your case, the fact that it started acting up after a year makes me think it's a real hardware issue where something is going flaky (perhaps a RROD is on it's way).

(Regarding high speed digital signals, keep in mind that transmission line reflections are not based on data rate, but slew (or edge) speed.  It's the data edges that contain the highest frequencies so even slow signal rates can cause nasty reflections if the edges are very fast).

Anyway, it's very difficult to say what is causing this issue without scoping one while doing it.  The JTAG reset pin is optional (from a JTAG spec point of view), so the only reason we need to do anything with it is to make sure reset is NOT asserted (pulled high).  In other words, you should not need the reset to activate at all in order to get things going.

MS has a pull down resistor on the jtag reset.  By jumping the reset pin to 1.8V, we are pulling it high to prevent reset (it's active low remember).

My suspicion is that when the voltage rails are coming up, the GPU core voltage is likely up before the I/O voltage, and as the I/O voltage comes up the GPU may see glitches on the reset that cause a *partial* reset in the GPU TAP controller, this would create a fatal logic error in the JTAG circuit.

Here are two suggestions that may be worth trying.

1) Remove the MS pulldown resistor.  This reduces the chances that a reset low is detected causing partial reset.

2) Add a RC filter to the reset line to provide a short clean reset followed by clean de-assert.  So you'd have something like

1.8V -> Resistor -> nTRST
                       |
                     CAP
                       |
                     GND

[yaywoop]

I should have said it can wreak havoc on hight speed digital circuits.

With the RC circuit you propose, this will effectively hold the jtag in reset for the first few nanoseconds or however long.. Its been a while since I read into the hack so you will have to excuse my ignorance.. Is this desirable? At what time after power up is the jtag actually used to deliver the hack? So the reset state of the jtag will not effect the rest of the system?

I think I get it.

Anyway regarding my problem, what leads me to think it is related to the jtag modifications is because i have had strange ROL behaviour, 2 red lights diagonally from each other, while the xbox still boots normally, followed by a 3rrod with 0021 (which I believe is a southbridge error) after a reset. Then after leaving it unplugged for a few minutes it would boot fine..
This all has only happened over the last month, a couple of weeks apart.
I shot a video of the 2ROL once http://xboxdoc.com/yaywoop/2rol.mp4

last night I applied the transistor-jtag hack, using the AUD_CLAMP wiring (previously had the original diode wiring)
I soldered in a chip socket so it'l be easy to experiment without taking the board out. http://xboxdoc.com/yaywoop/jtagtrans.JPG

Although this isnt going to be very scientific because by reassembling my xbox (I had a hybrid-fix installed as a preventative measure) and changing the screw tightness I may have fixed any physical gpu problems.

Cheers.

[l_oliveira]

I have an working 2006 Xenon console which is still using the original xclamp and TORX screws. The only preventive fix I did on it was put the fans on 12V when it was new.