Propeller RCFast oscillator as a function of temperature

Paul Baker

Here's a graph I threw together that shows how the RCFast's oscillator frequency changes over temperature. Since there is variability in the actual·values from chip to chip, I have normalized the frequency to the value produced at 25 C (room temperture).·I did this because I do not want people misinterpreting the graph, it is only meant to show how the frequency on any one Propeller will change over temperature. Those that want a fitting function, the following inverse quadradic was used to generate the line: y = 1 / (.9·+ .00385x +·5.39E-07x²)

The current sample set is 1, however this will be expanded before the graph is included into the datasheet.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Paul Baker
Propeller Applications Engineer

Parallax, Inc.

potatohead

I've a question about this.

Could one determine this in software somehow, thus know about temp changes with no additional hardware? I know there would be no absolutes, without some reference, but that still leaves the delta.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller Wiki: Share the coolness!

Paul Baker

An easy way would be if you had a RTC that output a seconds heartbeat, you would only have to count the cycles between each edge. You would have to do some calibration, but it's possible. Another possible method would be to use a RC circuit on an I/O pin where the cpacitor is a NP0/CoG type which is fairly temperature stable and compare the number of cycles a discharge occurs.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Paul Baker
Propeller Applications Engineer

Parallax, Inc.

potatohead

The RC circuit is almost no hardware and is totally not significant hardware. Thanks!

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Propeller Wiki: Share the coolness!

Paul Baker

No problem, let me know if it works out.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Paul Baker
Propeller Applications Engineer

Parallax, Inc.

QuattroRS4

Paul - was it your temperature tests that triggered your decision to compile this graph ?

Quattro

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Necessity is the mother of invention'

Paul Baker

The reason we released the datasheet as preliminary is because we had no means to test the temperature extremes in an extremenly controlled environment, now that we have an environmental chamber I am proceeeding with the series of tests so that the data can be included into a final datasheet. The video we posted was a first round "let's see how far we can push it", interesting but not very scientific. Now that we know the chip can perform at any temperture we are capable of throwing at it, I am now proceeding to do a series of measurements to fully characterize the chip over the entire military range of temperatures.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Paul Baker
Propeller Applications Engineer

Parallax, Inc.

Skogsgurra

I assume that we will also see some graphs describing Vth=f(temperature,Vcc) too.

That would be very nice to have for someone that is using the threshold voltage as a crude measuring tool. I have checked that for two batches and the typical voltage seems to be around 1.43 V at 3.3 V supply and not too much variation over the +10 to +30 centigrade range where I will use this technique.

(For those that think that I have gone mad: I do this to detect very fast spikes in the +/- 1000 V range. The detection circuit is a voltage divider and a bias network where I can adjust bias by outputting a variable voltage using PWM. I do not need better than around +/-50 V accuracy so the threshold method works quite well. And capturing the spikes using the counter makes it very fast. Faster than most high-priced comparators, actually.)

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

Gadgetman

What would have been nice was to access the RC-oscilator somehow(with a counter, probably) while driving the Propeller from an Xtal.

Just my brain running on idle the last few minutes before I close the Helldesk and begin my weekend...

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Don't visit my new website...

Paul Baker

The problem with Vth is it is not a well controlled statistic, for each chip it is well defined but that value can change according to process variations much the same way as the on board oscillators and brown out voltage are. So it wouldn't be very meaningful to provide a graph on it because the expected variation would swamp the data. I could do a similar delta graph of it as I did with the RCFAST as change according to temperature with respect to itself should provide more meaningful data.

Yeah my original thought of using the internal oscillator as a temperature gauge involved using a crystal until I realized that both would need to be accessible simultaneously.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Paul Baker
Propeller Applications Engineer

Parallax, Inc.

Skogsgurra

Thanks Paul,

Even if there are individual Vth variations between chips, it would be useful to have an "official" paper describing variation with temperature and also with supply voltage. In my application, there will always be one point where the thing is tested (manually) and then it can also be calibrated.

It has been mentioned that Vth is .5*Vcc, but that is not what I see. Most chips I have tested are 1.43 - 1.44 V at 3.30 V Vcc. It is difficult for me to test across many batches, I have only seen two so far. If there is an even bell shape around 1.65 V and I happened to find 1.43 - 1.44, will there also be 1.65 + (1.65-1.43) = 1.87 V individuals?

These data are important in some applications, but I do not expect any guaranteed values. Typical values and a possible range would be very nice, though.

Using the counters as glitch cathers is great. A lot faster than most comparators. Detecting pulses down to 25 ns width without any problems. Might even work for 12.5 ns. But haven't been able to test that.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

Paul Baker

You are correct that Vth is not 0.5Vdd, the reason is that there are a couple parasitic values pertaining to the N-wells for the PMOS transistors that were not included in the models, this means the drive between the PMOS and NMOS of the I/O is not even and therefore Vth is not 0.5 Vdd, this is also the reason the output drive curves for sink and source are not symmetric. These traits are universal for the Propeller and this is reflected by the fact that Vih min (0.6Vdd) is not a reflection of Vil max (0.3Vdd) as stated on page 25 of the datasheet. The next Propeller includes those parasitics and should have more evenly matched I/O stages.

Some·parameters people would like to have are simply not possible to provide, it is not just temperature but process which determines the boundries. There are many variables to a process each which has a particular effect on the performance of the chip. Unless the structure is small enough to where we can model it in SPICE using the corner models provided by the foundry, the only way we could fully characterize the boundries of a chip are to have wafers produced at each permutation of variable boundries of the process. This would be a monumental waste of money because all the wafers would have to be thrown away after a few dies of each wafer were tested because they represent the outer limits of acceptabilty.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Paul Baker
Propeller Applications Engineer

Parallax, Inc.

Post Edited (Paul Baker (Parallax)) : 7/27/2007 10:38:23 PM GMT

Skogsgurra

I hate monumental waste of anything. I am fully satisfied with the explanation. And the Prop. Spin is also getting a very comfortable language after some exposure to it.

The values I have measured seems to be fairly constant over the limited temperature range I will use the device, so the calibration can probably handle individual variations very well.

Thanks again.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

Paul Baker

I will do a spot check but I suspect that since both the NMOS and PMOS transistors will experience the same difference in temperature the Vth should remain very close to constant. The output drive capability however could very well change with temperature.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Paul Baker
Propeller Applications Engineer

Parallax, Inc.

Skogsgurra

Appreciate that!
Output drive is of much less concern. As long as you can drive an R2R ladder with R = 10 k without too much variation in Vhigh and Vlow, I am fine.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔