A recent thread went over the maximum permitted power on a device.
So why don't we have a sensing diode (well, a transistor with collector tied to base works better) on the die somewhere? It's fairly easy to do, according to my VLSI acquaintances, and with faster and faster IO [implying as it does faster and faster logic switching as well], it would make sense to see these on any largescale (more than 100 pins perhaps) FPGA package.
Comments?
Cheers
PeteS
It is there, at least in Virtex 4.
/Mikhail
I'd like to see it in all the larger devices :)
Cheers
PeteS
PeteS,
There is a diode in every (Virtex) device.
If you have an old data sheet, you might want to get a newer one.
If you find a data sheet without a diode, let me know, and I will find which pins are the diode.
It seems that there are those who do not understand how to cool our devices, as well. For help on heatsinks, etc. we have a lot of information that would allow up to (and perhaps more) 25 watts per device of heat. That is basically 100% of everything switching, at or near the BUFG clock maximum frequency.
The heatsinking has to be able to remove the heat such that the 85C (commercial) or 100 C (industrial) junction temperature is not exceeded,
Aust> A recent thread went over the maximum permitted power on a device.
A little awkward but what about forward biasing one of the input protection diodes with a mA or so?
Peter Wallace
Peter,
It is the substate to s,d pn junction, but it does not resemble the
1N914 that is the "standard" for measuring temperature.I have no idea if it would work, or not.
Before we had temp diode, people would calibrate the frequency shift on a ring oscillator, with nothing else in the part, forcing the temperature.
Then, with the design running, all you had to do was measure the frequency of the ring oscillator to get a reading.
Since the temp diode, the ring oscillator method has been abandoned.
Aust> >
Junction or case?
regards, Gerhard
Junction,
Since we have no clue what the FPGA will do, we must specify junction temp.
This is different from most other devices where the dissipation is completely known, and case temperature is specified (because that is obviously easier to measure).
Austin
I'll have to do a little temperature chamber test then. For my (room temperature) measurements looks like a standard ~.7V @ 1 ma forward junction to me.
Peter Wallace
Peter,
Typically, people use a two current technique, which removes process variation from the equation (lierally).
The National, Linear, etc. chips measure voltage drop at I1, and then at I2, and then solve for temperature.
There are two numbers you need to know: the ideality factor, and the series resistance.
With those two numbers, you can then solve for temperature using the two current technique (by hand, if you like, as the ICs will not like being connected as you will be doing if you use the IOB intrinsci diode).
Just look up the IC app notes for these temperature sensors, like:
(Maxim, in this case)
In the V4 data sheet we state the two constants, as some ICs allow you to program these constants into them.
Austin
More useful link,
see page 17
Austin
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.