Today I was looking through the post and I realized that there are components that get very hot and need big heatsinks, therefore I assume the components once they change in temperature they also change their characteristics (not good). I remembered an old article I read about computer overclocking (making a cpu run faster than is supposed) and the way the authors dealt with heat (one of the main limiter factor while overclocking)… by submerging the whole board in a below 0C non conductive fluid !! 
Since there are people who will even argue about the importance of the brand of the tennis balls they are using as a dampening material for their speaker cords (so the cord doesn’t touch the ground..better rf isolation ? ) I though that maybe some people here may find it interesting for their next super-watt power amp project…
if not, at least is interesting to see the pictures..
Maybe freezing the power amp is the next big thing after peeling the caps..hehe..
The article: http://www.octools.com/index.cgi?caller=articles/submersion2/day1.html
Since there are people who will even argue about the importance of the brand of the tennis balls they are using as a dampening material for their speaker cords (so the cord doesn’t touch the ground..better rf isolation ? ) I though that maybe some people here may find it interesting for their next super-watt power amp project…
if not, at least is interesting to see the pictures..Maybe freezing the power amp is the next big thing after peeling the caps..hehe..
The article: http://www.octools.com/index.cgi?caller=articles/submersion2/day1.html
Not so crazy.
Cryogenic amplifiers are used. Usually for very high gain high frequency RF amplifiers (To get very low noise levels). The main reason for using them is that Brownian/Johnson noise can be reduced. Heat after all, is just atomic kinetic energy, and the main source for "Noise" in an amplifier. (The kinetic motion of atoms will bump electrons loose, causing current to flow)
I did a little experiment a while back (somewhere in the thread "Thermal Noise Study"), with just such the idea in mind. I did some noise measurements of an OP-AMP heated to about 130F, and cooled well below the freezing point of water. There was basically no detectable change. I did see some very small changes in RMS noise voltage, but not enough (IMHO) to be called evidence.
Now, From the point of damage to a component, this would have siginificant merrit. All semiconductors are limited by thermal ratings (Not voltage or current). Take the maximum voltage, times the maximum current for any transistor and it will be more than the peak power rating (remember P = V*I). This is because the die itself can carry lots of current, or withstand high voltage, but is limited by power disipated in the silicon. Again, we need to remember that the power rating is really set by the maximum temperature the die can safely withstand. With a given thermal resistance, cooling the device will 'suck' more heat out, and thus reduce the die temperature (And therefore increase the power rating of the device). This is true for all semiconductors, when cooled with a refrigerator or heatsink.
Now in the case of computers, Switching losses (When the bits go from 1 to 0 or viseversa) are one of the main sources of heat. So, if you increase the switching frequency, you increase swicthing losses, and therefore heat. Cooling is the obvious solution.
Transistor gain characteristics change with themperature. So, cooling the should affect the linearity. The quesiton is if that change in linearity would affect sonics. Your guess is as good as mine!
Maybe someone needs to look at active temperatue control to fix the temperature of the die at a constant temperature, and therefore make distortion/nonlinearities constant. Anyone ever play with this? (I know that it is done with solid state laser diodes, and othe temp critical applications).
Warrents investigation, IMHO.
-Dan
Cryogenic amplifiers are used. Usually for very high gain high frequency RF amplifiers (To get very low noise levels). The main reason for using them is that Brownian/Johnson noise can be reduced. Heat after all, is just atomic kinetic energy, and the main source for "Noise" in an amplifier. (The kinetic motion of atoms will bump electrons loose, causing current to flow)
I did a little experiment a while back (somewhere in the thread "Thermal Noise Study"), with just such the idea in mind. I did some noise measurements of an OP-AMP heated to about 130F, and cooled well below the freezing point of water. There was basically no detectable change. I did see some very small changes in RMS noise voltage, but not enough (IMHO) to be called evidence.
Now, From the point of damage to a component, this would have siginificant merrit. All semiconductors are limited by thermal ratings (Not voltage or current). Take the maximum voltage, times the maximum current for any transistor and it will be more than the peak power rating (remember P = V*I). This is because the die itself can carry lots of current, or withstand high voltage, but is limited by power disipated in the silicon. Again, we need to remember that the power rating is really set by the maximum temperature the die can safely withstand. With a given thermal resistance, cooling the device will 'suck' more heat out, and thus reduce the die temperature (And therefore increase the power rating of the device). This is true for all semiconductors, when cooled with a refrigerator or heatsink.
Now in the case of computers, Switching losses (When the bits go from 1 to 0 or viseversa) are one of the main sources of heat. So, if you increase the switching frequency, you increase swicthing losses, and therefore heat. Cooling is the obvious solution.
Transistor gain characteristics change with themperature. So, cooling the should affect the linearity. The quesiton is if that change in linearity would affect sonics. Your guess is as good as mine!
Maybe someone needs to look at active temperatue control to fix the temperature of the die at a constant temperature, and therefore make distortion/nonlinearities constant. Anyone ever play with this? (I know that it is done with solid state laser diodes, and othe temp critical applications).
Warrents investigation, IMHO.
-Dan
Do not forget that all components have an upper limit AND an lower limit for wich they are specified. To notice the effect of cooling on noise, you have to calculate in Kelvin. And you have to halve the temperature in K before the noise halves. So you will have to go to about 150K to halve the noise present at room temp. That is about -123 deg centigrade.
Also most stuff works in a feedback loop (general or local), and this will keep everything as designed. Also are most references temperature independent just to counter this effect.
Also most stuff works in a feedback loop (general or local), and this will keep everything as designed. Also are most references temperature independent just to counter this effect.
Havoc said:
Yes, I guess that the Kelvin temp scale is a good point to make. I tried the temp ranges I did because I was able to actually achieve thoes temperatures easily (With Theremoelectric coolers).
I'm not sure what the lower temperature implications are. I'm guessing that mechanical stresses due to dissimilar materials (Silicon and copper) being connected on a planar surface and being cooled must be a major factor. That is, copper and silicon don't expand/contract by the same ammount with variation in temperature. (Do they?)
Your thoughts?
-Dan
mekanoplastik, can you please explain the influence of that third ear on your back, what does your system sounds like now ???
We have some very good deals on cloning, tuning here in florida. If it's really an improvement I will set up a group deal with the Real sect here in florida.
Ps. the dutch saying : having an ear sown on(een oor aangenaaid), has a different meaning than extending your hearing.
(jean-paul please...)
We have some very good deals on cloning, tuning here in florida. If it's really an improvement I will set up a group deal with the Real sect here in florida.
Ps. the dutch saying : having an ear sown on(een oor aangenaaid), has a different meaning than extending your hearing.
(jean-paul please...)An application I know about is NMR cryoprobes. Basically it's a RF circuit tuned for the frequency of the chemical element under analysis. The S/N increases by a factor of ~3. Basically all the RF components, wiring, coils and caps (no semiconductors), are cooled to liquid temperature (~4 Kelvin) by a Joule-Thompson apparatus.
The drawback is that once the probe is in place and cooled it cannot be warmed up without braking it.
To reduce Brownian motion the temperature really needs to be LOW was already made, I am not sure that liquid nitrogen is sufficient.
Dan, at what temp did you do your experiments? Is a semiconductor still such at liquid N2 temperature? How large is the S/N reduction of Brownian noise with respect to the intrinsic noise of the semiconductor, their "cost of doing business" so to speak?
The drawback is that once the probe is in place and cooled it cannot be warmed up without braking it.
To reduce Brownian motion the temperature really needs to be LOW was already made, I am not sure that liquid nitrogen is sufficient.
Dan, at what temp did you do your experiments? Is a semiconductor still such at liquid N2 temperature? How large is the S/N reduction of Brownian noise with respect to the intrinsic noise of the semiconductor, their "cost of doing business" so to speak?
grataku said:
I did my experiments on the other end of the scale. I got down to around 6F, and up to 130F. In reality I was looking for an increase in noise due to increasing temperature. (Same effect, just moving in the other direction). Didn't really find anything concrete, but there was a very small change (maybe few percent or so).
Not sure about when semiconductors stop being so at lower temperatures. I could be way off on this, but I was under the impression that heat was not required. (Unless, you hit 0 Kelvin, and stop all molecular motion). Seems to me the Cryogenic RF amplifiers run at around 15 Kelvin. I think these are also special HFET devices. Cryogenic cooling may get you an order of magnitude reduction in noise!?! (something like that, anyway)
http://www.sofia.usra.edu/det_workshop/papers/session4/4-05gaier_cr_edjw021022.pdf
The thing to realize is that RF amplifiers usually have gains that are much much higher than audio amplifiers (Many orders of magnitude larger). For noise, cooling an audio amplifier probably isn't worth the trouble (Because of the gain).
Now for power disipation, that may be a different story. However, more practical methodologies such as liquid (Water) or forced air may prove to be the most beneficial (And cost effective). Possibly even a thermoelectric cooler.
-Dan
Dan,
At 4 K you get the ~3 times more sensitivity. At 6 F I wouldn't expect any change. I believe I remeber stories about people dunking op-amps in liquid N2 (~160 K) and hearing reports that things were working.
Anywho,
there is no way to stop all molecular motion, that would violate the uncertainty principle. Zero point energy is always present.
At 4 K you get the ~3 times more sensitivity. At 6 F I wouldn't expect any change. I believe I remeber stories about people dunking op-amps in liquid N2 (~160 K) and hearing reports that things were working.
Anywho,
there is no way to stop all molecular motion, that would violate the uncertainty principle. Zero point energy is always present.
grataku said:
I agree, since we cannot get to absolute 0, we cannot stop molecular motion.
For a conductor, Given we can calculate noise power (N) in a given bandwidth (B), equil to KTB. (This would not hold true for infinite bandwidth systems, because noise power would be infinite. In application however, it works for limited bandwidth systems. )
B = Bandwidth in Hz
K = 1.38e-23 (Boltzmans constant)
T = absolute temperature (K)
Assuming a 100Khz bandwidth: We can calculate the noise power.
At -6F, 3.4781x10^-16 Watts
At -130F, 4.5208x10^-16 Watts
Moving the thermal noise sources from 130F to 6F would gives a reduction of about 23% (or 1.1dB) in thermal nosie power. Going from room temperature (68F, 4.0454e-16 Watts) to 15K (2.07e-17), gives a 12.9 dB reduction in noise power.
In theory there is a change, even from 130F to 6F. However, in my experiment I was unable to reduce the temperature of every component in the system, so there were still noise sources at room temperature while cooling, and heating the system. (Meaning much of the noise would remain constant). Maybe I should repeat the experiment in my oven, and freezer.
In reality, looking at the noise power of in the range of
10^-16Watts, there is very little noise. However, if this noise power (4.5208e-16 Watts) is across a resistance of 100K Ohms at the input of an amplifier, we would see a RMS voltage of 6.72uV RMS. (Or 21uV across 1meg Ohm). Multiply this voltage by the gain of the amplifier (Say 30 to 50dB) and you could start to see noise in the millivolt range. (Which is typical for many amplifiers).
The thing to remember, is that if you increase the bandwidth of the amplifier, you would see more noise. Decreasing the bandwidth would reduce noise as well.
-Dan
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.