Extreme Temperature Electronic... stuff

[gamozo]

I was thinking of picking up a dewer of liquid nitrogen and having a play around with some electronics. One of the things I was thinking of, is trying to overclock some 7400 series logic... I've never seen it done. Since they're so cheap, it would be fun to just drive them until they are destroyed. Perhaps I'm missing out on something and there would be no way to overclock 7400 chips with liquid nitrogen? I know it's overkill... but it would be a lot of fun to play with. Anything else that would be fun to work with if I had some LiNo, maybe this idea is just stupid and it really wouldn't do anything fun/noticeable?

Proper safety procedures will be taken

-Brandon Falk

[don.r]

Not sure about 7400 series but temperature is not always the limitation. Sometimes capacitance can soften the clock edges until they approach a limit. Obviously, the capacitance is also a function of the internal temperature.

[ejeffrey]

TTL logic (74F) will not work at all in liquid nitrogen, nor will anything with bipolar transistors.  The temperature is too low for dopant states to be thermally excited, so there are no conduction electrons.  This happens at about 100 K.  BJTs also have diminishing current gain with decreasing temperature, so they can start to misbehave at higher temperature as well.  MOSFETs work OK at low temperature because they rely on the field effect to create carriers.  CMOS logic (74AC/74HC) should continue to work, although the operating voltages may change.  I don't know if they will go particularly faster or not.  Generally I would expect their performance to be optimized for operation around room temperature.

Voltage regulators tend not to work at low temperature due to their use of BJTs as pass transistors.  This means you will have to feed in your regulated DC from room temperature.

Some circuits (especially amplifiers like CMOS opamps) may also have startup problems at low temperature due to changes in bias currents and turn-on thresholds.  Sometimes you can solve this by increasing the voltage until it starts, then lowering it back down to the operating point.  Other times you need to keep the circuit powered during the entire cooling process.

Most resistors will work fine in liquid nitrogen, with a relatively small change in resistance.  At lower temperatures (liquid helium) carbon and metal oxide resistors have dramatic increases in resistance -- so much that they are commonly used as thermometers.  Capacitors, especially high value capacitors will have *dramatic* changes in value.  Electrolytic and high-k ceramic capacitors may lose more than 90% of their value.  NP0/C0G caps fare much better, but still lose a reasonable fraction of their capacitance.

[gamozo]

All of my chips are 74HC series, so I'm assuming they would be fine. If anything, I could just overvolt them a bit to generate a bit more of a core temp to a temperature that they could operate at. Unless they wouldn't be able to run to the point that they could start generating heat in the first place, but I doubt that. I've never really gotten this straight... is the only reason that chips cannot be overvolted is for heat reasons? At least that's what I thought (maybe only to an extent). So I was thinking of running a 5V chip at 12V-30V or something. Who knows... I have about 300 74HC chips, so torching 20-50 of them for the sake of entertainment and science wouldn't be an issue... if anything, I could always buy some more. According to the data sheets for the 74HC00, the higher Vcc, the quicker the fall time. I'll probably plot this over a voltage range then. It might be hard to actually test this all out though, since I have a breadboard which will play into the rise and fall time quite a bit, and a 100MHz scope that will probably only go down to 3-4ns... new scope time?! I think so!

Perhaps the solderless breadboard wouldn't be much of an issue as long as I use short enough wires?

I guess I've never worked with high frequency stuff. Perhaps anyone has some links for me? I'd love to learn how to work with high freq stuff, for measuring it and for building with it.

[alm]

I wouldn't build a fast circuit on a solderless breadboard, even the capacitance between the tracks will present a fairly low-impedance load to the high frequencies from the fast edges. Plus all the inductance of the wires and the large loop areas making them work well as antennas.

Just use perfboard, or for higher frequencies, construct the circuits on a piece of copperclad, using the copper as ground plane.

[gamozo]

Would copperclad or perfboard be faster/more responsive for high frequencies? I've actually never done permanent builds to save money (you can argue that you can just reuse perfboards though), but now that I'm out of school, I can actually put some money into projects.

Sidenote (alm):
"Posts: 1999"
Hehe.

[ejeffrey]

Yeah, solderless breadboard is pretty hopeless for high speed.  It has bad parasitics of all types: parasitic capacitance, parasitic inductance, leakage currents, contact resistance, and crosstalk.  You can prototype a lot of stuff on it, but if you are trying to go for high speed you will be completely dominated by the breadboard.  An etched copper clad board or a well done perfboard will be much better. 

Another big issue is the off-board wiring.  The maximum toggle rate is basically limited by the capacitance you have to charge.  The device has a certain internal capacitance that gives you a baseline, but anything you add on top of that will slow things down.  30 cm of twisted pair wiring will add about 20 picofarads of load.  You can get around this by using transmission lines terminated in the characteristic impedance, but many logic gates don't have enough output current to drive a 50 ohm coax.  One trick is to use a hex inverter with 3 gates wired in parallel.  They will have no problem driving 50 ohms, although you need really good supply bypassing here.

You will have the same problem with the input signal, depending on your signal source.  Function generators will be designed to drive a 50 ohm terminated transmission line, but do you have one fast enough?

So what I would recommend is to make a small PCB with a ring oscillator (an odd number of inverters connected in a ring) with the output then connected to a ripple counter used as a divide by 64 counter (i.e., tap the 6th bit).   You can now measure with a cheap scope or frequency counter how the supply voltage affects the oscillation frequency.  That will basically be related to the propagation delay through the inverters.  You can then dunk it in liquid nitrogen and see how the output frequency vs. voltage curve changes.  This won't necessarily tell you the fastest toggle frequency possible, but it should tell you if the device gets 10% faster.  Since the only signal leaving the board is the DC voltage and the divided signal that is < 10 MHz, you should have no problems with cable loads.

It may take some experimentation.  You might need to use a separate fixed supply for the frequency divider and a variable supply for the counter.  You might need to capacitively couple the signal (you will then need a voltage divider to bias the voltage to the counter at mid-supply). You might try mixing logic families: for instance put a 74AC series ring oscillator with a 74HC series counter.

Record what you do and let us know the results.

[gamozo]

So what I would recommend is to make a small PCB with a ring oscillator (an odd number of inverters connected in a ring) with the output then connected to a ripple counter used as a divide by 64 counter (i.e., tap the 6th bit).   You can now measure with a cheap scope or frequency counter how the supply voltage affects the oscillation frequency.  That will basically be related to the propagation delay through the inverters.

That's a bloody brilliant idea. Thank you so much. I found out I can get 5 liters of liquid nitrogen for $50, or 5 pounds of dry ice for $2 a pound (those are the minimums of the supplier). Dry ice will last half a day to a day given good storage, where the nitrogen would last for about 5-7 days. I think I'll do a few runs of using dry ice and alcohol for cooling, then later scale it up to some nitrogen.

Will record the data. I'll probably get some dry ice towards the end of the week. I've got to put in an order for some breadboards and some other bits and bobs I've found I've needed for my parts bin.

-Brandon Falk

[ejeffrey]

Make sure to do the measurements at room temperature first so you know everything is working and have something to compare it to.

One more thing: see if you can put a pt1000 or pt100 platinum resistance thermometer on the board.  It costs a few dollars, but then you can then measure the temperature of the board.  This allows you to do things like hold the PCB slightly over the liquid level to measure at intermediate temperatures.  You can also use a silicon diode (or diode connected transistor) which is basically free but not as accurate without calibration.

One more thing for my piece of mind: be careful with that stuff, they can be dangerous.  Don't store them in sealed containers, wear eye protection, and don't use or store them in areas without excellent ventilation.

[Psi]

So what I would recommend is to make a small PCB with a ring oscillator (an odd number of inverters connected in a ring) with the output then connected to a ripple counter used as a divide by 64 counter (i.e., tap the 6th bit).   You can now measure with a cheap scope or frequency counter how the supply voltage affects the oscillation frequency.  That will basically be related to the propagation delay through the inverters.

That's a bloody brilliant idea.

I second that. Awesome idea, it gives me chills just reading it.
Make the ring oscillator in a physical ring on the pcb so it resembles a tiny haltron collider. 

[gamozo]

One more thing for my piece of mind: be careful with that stuff, they can be dangerous.  Don't store them in sealed containers, wear eye protection, and don't use or store them in areas without excellent ventilation.

Thanks . I plan on using dry ice for a few weeks, before I'm comfortable enough to try some liquid nitrogen.

I'm gonna try a few set ups for the ring loop. Perhaps I'll be able to get by using just one chip. 6 inverters on one chip, I'll chain those together using a very thin layer of solder, it would probably have the least resistance and capacitance of a setup I could do.

I'm going to pick up some mineral oil, and then drop the entire IC in there as well as a few chunks of dry ice. From what I can think, this will probably be the most optimal way of cooling. The only downside is that mineral oil has a dielectric constant of ~2.3 vs 1.0 of air. This could lead to a bit of capacitance, but I think the level of cooling with this idea would provide outweighs that (and it's not like 2.3 is really that high anyways).  I can also try using a heat sink which I can put on top of the IC with some thermal compound for connection and then pour some water and dry ice chunks in the heat sink (given it's hollow to provide a place to let the liquid sit). Any other ideas of how to cool it?

For temperature measurement I'll probably just end up using my thermocouple on my multi.

[ejeffrey]

Any other ideas of how to cool it?

Hmm... If you have a small die-cast aluminum box you could put thermal grease on the ICs and mount the board upside down with the ICs pushed up against the bottom of the box.  If you clamp the wires to the edge of the box as strain relief the thermal grease should be strong enough to hold it in place well enough.  Then just float the box as a little aluminum boat in your coolant.  That way you can use alcohol or acetone as the cooling liquid instead of mineral oil, which I believe will freeze well above the temperature of dry ice. 

For the liquid nitrogen you can just dunk it in the liquid.  Attach the board to a meter stick with some masking tape or dental floss.  Spiral the wires around the stick and anchor them periodically with floss and just dunk the whole thing in the liquid.

[gamozo]

I like the boat idea. As for alcohol... it's nice since it's got a lower freezing temperature, but it also got a much higher dielectric constant than mineral oil. Also, from what I've heard... the rapid release of CO2 from dry ice causes enough turbulence in liquids (even water), that it is not allowed to freeze, but it still can drop to lower temperatures than 0 deg C (can't find another source for this though... I might just have to try it)

Here's an interesting link:
https://en.wikipedia.org/wiki/Cooling_bath

As well as:
http://www.smartmeasurement.com/en/wizards/flowmeter/flmtr_mag_conductivity.asp

Hopefully I can find something with no conductivity, low dielectric constant (1-5) and freezing point below 78 deg C.

[ejeffrey]

The purpose of the boat is that the circuit is inside the boat and dry, while the alcohol cools the aluminum which acts a heatsink.  That way you get the low freezing point of alcohol without the dielectric losses.

[gamozo]

That's very true... I was thinking just about submersion stupidly. I'll try a heat sink solution (probably just alcohol), and a submersion solution (still looking for a 'perfect' fluid).

Looks like Acetone might work well... it's easy to obtain, cheap to get, and given a 2cm IC, would have 25Mohm of resistance (0.02uS/cm). And the dielectric constant approaches 1 (that of air) at 32 deg F (0 deg C, and seems to have a downward trend as temperature decreases) [source: http://www.asiinstr.com/technical/Dielectric%20Constants.htm]. Acetone is also used for cooling baths, as it's freezing point is below the temperature of dry ice, thus it's possible to keep acetone at -78 deg C, which is our lowest temperature with sublimating dry ice. The only worry would be a spark igniting it, but its flash point is -17 deg C, which we would be below, so it would technically be safe (but obviously some could be above that temperature). Making a circuit and then properly protecting it would get rid of this risk (also the amount of acetone would be minimal for a remotely dangerous explosion/fire).

Any thoughts on acetone? It looks to be perfect.

[ejeffrey]

Acetone is the typical thing to use in dry ice cooling baths, but I think there is a typo or mistake in that chart.  I think the dielectric constant of acetone at 0 Celcius is 21.nnn, not 1.nnn.  1.01 would be substantially less than the dielectric constant of liquid helium (1.04).

Edit: another possibility is that that is the dielectric constant of the vapor phase... which would be about right.

[gamozo]

Edit: another possibility is that that is the dielectric constant of the vapor phase... which would be about right.

That's what I was thinking, as generally the constant goes up while temperature goes down (not sure what the exceptions are). I don't think there will be too many issues with the dielectric being fairly high and hopefully not, because acetone looks like the only easy-to-get, moderately safe, and dry ice coolable to -78 deg C, with very little ability to conduct electricity. Anyone think that a dielectric constant of ~24 (est. for -78 deg C) would really cause some issues in this experiment? Given it's just a single IC in terms of size and shape.

[tom66]

I got an ancient GoldStar (80's IIRC) 74LS00 (quad NAND gate) in a ring oscillator on a 130-in-one kit up to 35 MHz. Using long unshielded wires picked from my lab-bench floor, no decoupling capacitors and 4xAA batteries, and a 1X probe. I was later, with a few design alterations, able to go up to 75 MHz, although I started hitting my oscilloscopes' limitations, even with a 10X. This was at room temperature. Then I tried a modern 74HC00 and boo, it only went at a few MHz, but that was on a breadboard.

I wouldn't want to know how fast you could go with cooling below room temp.

[gamozo]

Did you just use the 4 nands, or did you use more? I really need to find like 10 seconds to try some stuff out on my own... I've been caught up in my serial-to-32-bit-parallel circuit (have to stagger the clocks for each of the daisy chained SIPO... otherwise the shift isn't done when another picks it up... etc. And I'm doing it all in 7400 series logic for a bit of a challenge).

[tom66]

Three NANDs wired as a ring oscillator. Shortening the wire got it up to ~75 MHz but the amplitude was only 40mVp-p.