Accessing an SX semiconductor junction?

Paul Baker

Im pretty sure this isn't possible, but do any of you know the actual transistor layout of·SX pins·to know if a diode junction can be isolated (access to an n-channel in a p-well or p-channel in n-well) by an external chip? Im just wondering if the LM86 can be used with an SX, heres an except of the specification:

Miner_with_a_PIC

Access to internal components of most ICs is pretty much impossible. If you were an engineer with connections within the SX chip packaging facility you may be able to pull this off. For the rest of the 99.999999% of us this is not possible. It would be helpful if you could include why you need to know the temperature of a specific junction. This is understandable if you intend on overloading the pin for a very short duration. Otherwise an external MOSFET could prove useful for high current applications without sacrificing too much speed. If you need to know the temperature as a function of load this could be empirically determined for the entire chip using thermocouple and various loads. Remember that exceeding specifications is not recommended and can result in chip failure.

Good Luck,

Mr. T

Miner_with_a_PIC

Another thought. Most I/O pins have clamping diodes going between the pin and the rails. No guarantees that the SX does not use a more excentric layout to protect the pins but possibly these could be used. Of course to use them you would have to take the pins into a state that is outside of their normal operating range. The I/O pin would have to be an input, I recommend going above VDD rather than below VSS. Doing this for a short period only would prove helpful to the SX chip. Now be prepared for latch-up conditions to occur and possible premature failure of the chip. This would be an interesting experiment to perform on your SX guinea pig but not something to impliement into design. Let me know how the project goes, I would be interested to hear about just how robust the SX chip is.

Paul Baker

I thought about the ESD diodes, but the LM86 places one terminal at .7 V and the other at 1.35 V, the ESD diodes are connected to the rails. I intend to see how far a SX can be pushed (speed wise) when TC cooled. If I could get a die temp the testing process would be much faster than waiting for thermal equalibrium at each testing point to take an external measurement, plus eliminating the need of another hole to measure the underside temp of the SX that needs to be condensation proofed. Ribbon themocouples are not made small enough to sandwich underneath a 52-PQFP.

Post Edited (Paul Baker) : 6/17/2005 11:39:06 PM GMT

Bean

Just a wild idea...
You might be able to use the internal 4MHz clock as a temperature monitor.
The internal clock changes with temperature and you could probably calibrate it at several temperature points then read the frequency and tell what the temperature is.
Or maybe not...
Bean.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95

http://www.parallax.com/detail.asp?product_id=30012

Product web site: www.sxvm.com

"What's the difference between ignorance and apathy ?"
"I don't know, and I don't care."
·

Paul Baker

Hmmm very interesting idea Bean, but how would I access it if Im not using it as the clock source? I will be using a programmable oscillator capable of 250MHz (I plan on having a 1 MHz clock driver driving an I/O pin as a frequency reference). I strongly suspect Im barking up the wrong tree with the die temp thing, but I thought Id ask.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

Bean

I guess you would have to use the internal 4MHz as the clock and be running a program that toggle an I/O to be able to measure the changes in the clock.

Hmmm,
How about somehow timing from the watchdog timeout ? That runs from a 14KHz internal oscillator that I assume will vary with temperature also (I think even more so than the 4MHz).
In the program you could Reset the watchdog and send out a pulse, then just wait for the watchdog to expire which will reset the chip and have the startup code send out another pulse.
The time between the pulses is inversely proportional to the watchdog clock speed.

Bean.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95

http://www.parallax.com/detail.asp?product_id=30012

Product web site: www.sxvm.com

"What's the difference between ignorance and apathy ?"
"I don't know, and I don't care."
·

Beau Schwabe

Paul,

The ESD diode method would only work if you completely disconnected everything else from the SX chip. Since the ESD diode's are reverse biased you need to flip your D+/D- terminals
around on the LM86 to do it this way. Not a practical approach since it defeats the purpose, sorry.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe

IC Layout Engineer
Parallax, Inc.

Beau Schwabe

Paul,

Mulling this idea over....what if you completely electrically isolated the LM86 except for what's allowed at the D+/D- terminals?
In other words it's own power supply with nothing shared. You still would need to reverse the polarity of your D+/D- pins to your
SX. If you setup a pin as an input, and connect D- to the input and D+ to the SX VDD or D+ to the input and D- at SX VSS, you
should "see" the ESD diode. As far as reading/driving the signal from the LM86, you will need to optically isolate them. The
SMBData signal might be tough, but a dedicated SX that just talks to the LM86 and then through an opto isolator to a PC might be
a solution.

One thing to note.... The ESD diodes are part of the PAD structure and every pin should have them (except rare special cases).
That said, there may be a considerable distance in terms of silicon from the PAD and the actual area of silicon generating the
most heat. I would map several pins to determine if there are indeed "hot spots" that a single pin may not be able to register.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe

IC Layout Engineer
Parallax, Inc.