This JDM thing is outta control

[moogah]

Ok, first off I must mention that AFAICT all the links in the "links to JDM trouble" thread lead to a 404 error.

Secondly, no-one (with the exception of TK) who has credited their troubles to a specific motherboard or CPU/IO delay problem have fessed up to the specific details of which motherboards/CPU speeds worked/didn't work!

SO.  I'm working with:

XP SP1 system

ABIT NF7 motherboard

Athalon 1700 plugged in

I've also tried this on an old Compq Presario, 233Mhz running windows millenium.  Same results as below

I'm using a SmashTV board with an 8.2v zener/PIC18F452.  This board appears to be missing the 10k from pin 38 to pin 31.  My results have been the same even after connecting one to the bottom of the board

  Currently I have tried this with and without the 9v batteries, with little difference and no success.  I've used both COM ports, with the same results, and have tried every IO delay from 1-20 and several over 20, no noticable difference.  I've tried using API,  totalio.sys driver for Direct I/O and simply using API. with totalio running,  No noticable difference.

Without the PIC plugged in my readings are as follows

MCLR 12.6v / 13v with battery

VSS    4.6v / 4.8v  with battery. 

NOTE: with the battery connected VSS stays steady at 4.6 after it's been unchecked

clock  ~4v peak, rapidly drops.  same behavior with battery, but I suspect the peak is higher

Now, when I first check Data out, MCLR jumps to max and stays there so that the first check on MCLR makes no difference, unchecking it drops the voltage and re-cheking it them raises it again.  This seems normal to me.  Also, when icprog is not running MCLR is often at max, when it is started voltage drops to ~2v, when the hardware menu is opened the voltage drops to ~.5v and rises to ~2v after the hardware menu is closed.  With the battery connected MCLR is ~5v while not running icprog, no difference at startup, opening the hardware menu drops voltage to about 1v, closing it returns to ~5v.  Only small differences in voltage @MCLR with larger delay times using the totalio.sys.  If I remember correctly there is a difference in voltage with API, perhaps not tho. 

With the PIC plugged in and without a battery:

following the instructions to perform a Real All, MCLR varies from ~2v to ~10v MAX. 

switching on the battery:

MCLR jumps to ~13v, starting the read drops the voltage to ~10v where it remains much more consistant.  The voltage staying at 13v untill a read starts seems suspicious to me.  I've quadruple checked my diodes, several times.

NOW, If I start the program and do a Verify right away, it passes.  If I then load a .hex and do a Verify it fails at "0000h". 

Every attempt at doing a burn has failed at "000000h".  From my understanding this essentially means the computer was entirely unsuccessful in opening a connection with the JDM. 

I have never been faced with a ghost in the machine like this. 

QUESTIONS:

How do I test my Zeners?  Capacitors?  Transistors?

D5 shows 8v, D6 shows 4.6v

While icprog is running D4 reads .2v, D3 ~2.5v, D4 .3v, D1 ~2.5v

While a read is in progress D3 fluctuates with a peak of ~13v, D2 same with a peak of ~5v, D4 steady at .3v D1 peak ~3v

After the read D3 & D4 hold at 13v and 5v.  D1 & D2 hold at .7v

I've checked against the schematic again, pulled up the datasheet for the BC547's.  Everything is ok there.

While doing the hardware test everything checks out.  With or without the battery.  MCLR is still a bit on the low side tho, just below 13v.

I am 99% sure they are working after reading all the other threads here.  My problems are too similar and it seems the root is a combination of flakey drivers and inconsistant power to the JDM from COM.  Furthermore, there seems no correlation between success and failure from case to case.  The X factor is simply missing from peoples reports and seems as much attributable to VooDoo than any scientific effort.  It's late here, and my eyes are getting droopy.  More to come tommorow.

More questions:

Should MCLR remain steady while a read is happening, or should it fluctuate?

I've found that generaly it will peak at ~13v at the begining of the read and then fluctuate, often holding ~10v.  There seems to be a relation between what icprog is doing and the fluctuations, so I assume this is normal.

It seems to me that the "enable MCLR as VSS" option has no effect, and I believe it to be for a different chip/programmer anyway.

UPDATE:

I've got more info on the strange readings at CLOCK.  It seems that toggling the VCC checkbox causes Data In to toggle ON.  When this happens the voltage at CLOCK drops to -.7v.  Enabling CLOCK results in the situation described above, peaking at ~4v and rapidly droping to ~2, falling slowly from there.  Turning clock off returns the voltage to -.7

NOW what is strange is this:  while Data In is on I can get the clock readings to return to normal by toggling MCLR.  In fact, while MCLR is on, the clock voltage drops like described above.  When MCLR is turned off clock returns to max, around 5.5v.  Turning off data in or enabling data out, clock drops to 0v.  Enabling MCLR now drops clock to -.7v

[moogah]

*UGH* This sucks. 

Can someone with a working JDM try to replicate my measurements during a read all?

Can someone with a working JDM running windows XP post what type of motherboard they are using?

Perhaps I've wrecked my PIC?

[TK.]

Please forgive me, that I'm not able to give you an sufficient reply to all your questions/assumptions. I even don't find the time to complete the layout of MBHP_BURNER, which should replace MBHP_JDM due to all these strange problems (I'm neither the designer of the JDM circuit, nor the programmer of IC-Prog, therefore it's also VooDoo for me... the MBHP_BURNER design and the P18 programming software are much more "straight-forward") :-/

Most of the effects you described (even the strange ones like the slowly dropping voltage at the clock pin) are normal (I don't know why, I only know that this doesn't hurt, since the clock is normaly toggled very fast), but:

I've found that generaly it will peak at ~13v at the begining of the read and then fluctuate, often holding ~10v.  There seems to be a relation between what icprog is doing and the fluctuations, so I assume this is normal.

this seems to be the problem. The MCLR voltage must be stable during programming - maybe your two 9V blocks are not strong enough? Do you have a 15..20V PSU, or two PSUs which could be combined?

Best Regards, Thorsten.

[moogah]

Well.  I just tried it with a PSU set to 15v and tested all the way up to 25v.  I could get the peak voltage to imporve, but it still fluctuated, mostly between the space between read program and read data.