Cooling PHEMT amplifiers

The following discussion about cooling PHEMT amplifiers took place in the Moon-Net reflector. I found the subject so interesting that I decided to put together all the messages in this page.

On 9-Mar-1999 G8WRB wrote:
Has anyone ever tried cooling a PHEMT (Pseudomorphic high electron mobility) FET amplifier, with a view to decreasing the noise figure (NF).  I have this bird-brained idea that there may be reasonably significant gains for modest drops in temperature (ie 30 deg C or so, that can be achieved with a Peltier cooler). (For those that done know, a Peltier coolier is like a mini fridge. You apply DC power and one side gets hot, the other cold. They cost only a few pounds). My reasoning is based on the physics of PHEMPTs, although I have never looked at it very deeply, or done any calculations. My basic idea is that for these PHEMPT devices, the NF drops with increasing electron mobility but electron mobility increases rapidly with decreasing temperature.

If anyone has tried this, or has the necessary equipment to try it, I would be interested. So I expect would others, as it would cost no more #15 (say $25) to do.

PS: PHEMPT's are sometimes called something else, depending on manufacturer, so the acronym you know may be a bit different, but the idea is the same.

(N.R.: You can visit G8WRB's web site where there is also a quite a large number of complete Eimac data sheets (not the 2-page summaries on their web site) for a large number of the common tubes used in amateur amplifiers.)

On 10-Mar-1999 K0XP wrote:
Yes; several cooled 1296 preamp experiments was described by Tom Henderson, WD5AGO, in the 1990 CSVHF Proceedings. I'l attempt to summarize the article's findings below.

Tom used a Peltier-effect cooler, describing several experiments that took place over a year or so and whose results were discussed at the 1989 CSVHF conference. Briefly, he built several preamps and concluded there was insufficient decrease in noise figure to "...justify a working system,".

But he was also able to have several experimental preamps cryogenically cooled. There was some thought that perhaps some device types were more responsive to such cooling, so he also built and tested devices in a test fixture. It should be understood that the preamps under test were already performing far better than those of the average 1296 EMEer; i.e., 0.3 dB noise figure. Tom chose 1296 for his tests as the sky temperature is lower than at 432 and so further reduction in noise figure due to the cooling should be more easily noticeable when pointing at sky noise sources.

The article in the 1990 CSVHF Proceedings contains several charts of noise figure vs. temperature for several devices, including the MGF1303 (cavity) and the common ATF10135 (series L). The charts for those devices is shown below (note that Tom cautions that each device was adjusted for a "reasonable" noise figure at Ta and might not have been the best attainable):

device    n/f       dbg      deg. Kelvin
MGF1303   0.62     15.8      298
          0.47     17.0      50

ATF10135  0.52     13.5      298
          0.33     14.4      100
          0.28     14.7      18

Other tests were performed using the test fixture with these devices:

device    n/f       dbg      deg. Kelvin
ATF13284  0.60     15.0      295
          0.43     16.1      200

ATF10135  0.51     12.2      295
          0.38     12.8      200

NE13783   0.64     12.0      295
          0.45     13.0      200

MGF1801   0.65     12.5      295
          0.56     17.3      200

MGF1412   0.54     14.3      295
          0.34     15.6      200
          0.27     16.3      100

MGF4303   0.50     16.0      295
HEMT      0.29     17.1      200
w/gnd S.  0.20     17.9      100

For your information, Tom listed the temperatures of some coolants:

Liquid Helium    4K (-455F)
   "   Nitrogen 77K (-320F)
Dry Ice         195K (-108F)
TED @ Ta        250K (-10F)
4#TEDs          205K (-90F)
Freeze Mist     220K (-55F)
Room Temp       295K (72F)

TED stands for Thermo-Electric Device, what's called a Peltier Effect cooler. These are widely available at many of the discount electronics part places such as Hosfelt, for around US$20 or so.

Tom's findings were that the minimum required cooling was on the order of 100 degrees Kelvin. He concluded that the most practical method of cooling was probably to use dry ice on the preamp's device at the feed, although a set of stacked TEDs showed promise; but the problem with the stack was dissipating the resulting heat. The power consumption of five stacked coolers was over 200 watts, far too high for a simple heatsink. He says that one major problem was that the thermal mass of the preamp itself is such that insufficient cooling of the device results. To that end, he described several methods of cooling the device while thermally isolating it from the other uncooled preamp components.

Hope this was of help, Dave; and of some interest to the other readers of the reflector. I have not found any other such articles in my small collection.

On 10-Mar-1999 W3IWI wrote:
Dave -- the basic answer is yes. In addition to the posting on WD5AGO's work, let me add some comments.

Cooled HEMTs are used all the time in the radio astronomy world. As an example of the current state-of-the-art, take a look at the plot on cooled FET/HEMT performance on the NRAO web site at

In my professional life, I have been responsible for the global network of radio telescopes used in Very Long Baseline Interferometry (VLBI) for high accuracy (millimeters on a global scale) geodetic science (see our web site at for some info). The geodetic VLBI network operates at S-band (2.2-2.4 GHz) and X-band (8.1-8.9 GHz) and all the stations use HEMTs operating at cryogenic temperatures (~20K). At S-band, the HEMT LNAs have amplifier noise temperatures < 5K, resulting in Tsys ~30-50K, and at Xband they contribute ~10K to the ~50K Tsys.

To get the cryogenic temperatures, we use commercial 2- or 3-stage closed- cycle Joule-Thompson refrigerators. These refrigerators are rather similar to conventional air conditioners, except that Helium is used as the "working fluid" -- Freon would freeze hard as a rock! The 1st "warm" stage cools the ~300K ambient temperature down by a factor ~4-5 to ~60-70K. The 2nd "cold" stage cools the ~60-70K by another factor of 4-5 to ~15-20K. Some receivers use a 3rd stage to get down to the ~4K level.

Note that I used absolute temperatures in this description. A large portion of the cooling improvement comes from the reduction of the kTB noise contribution, where T is the ABSOLUTE temperature. Thus cooling from ~300K ambient by ~30C (which is also 30K) results in only a 10% drop in the thermal noise contribution -- hardly worth the effort! Basically, the solid-state thermionic refrigerators just can't "pump" enough heat to make a significant improvement.

A much better approach has been used by optical astronomers for years to cool photomultiplier tubes (see, for example )

either use Dry Ice purchased at your local Ice Cream store -- it's even advertised at local "Seven-Eleven" neighborhood stores (see for an example in the Tampa, FL area!).

I'll tell an anecdote from ~15 years ago to illustrate how well Dry Ice works. It was at a Central States meeting with Barry (VE4MA) and I competing to win the 1296 NF contest. Barry brought his newest FET amplifier built with copper water pipe. I brought a 1420 MHz LNA we were using for radio astronomy. I put my 21cm LNA into a foam plastic box with only the coax & bias cables visible and filled the box with dry ice, and let it cool for a few minutes.

Barry was so proud of his LNA and was CERTAIN he would win. He was showing ~0.5 dB NF and ~20 dB of gain. My "black box" had more that a tenth dB better NF and about 40 dB of gain. It was also broad as a barn, with little difference anywhere in the 1200-1500 MHz range. My only "tweaker" was a gate-voltage bias pot.

Then Barry realized what I had done and decided to cool his LNA. Unfortunately, copper water pipe presents a huge thermal mass. And since his FET biases were optimezed for ambient temperatures, his amplifier was a BITCH to tune when it finally got cold. After a couple of hours of tweaking, Barry matched my LNA and we declared a tie.

So my advice, if you want to get a significant performance improvement, try putting your LNA into a foam plastic box; the kind that holds a 6-pack of beer is about the right size. To minimize moisture condensation problems, first dry the amplifier well first, then put it in your kitchen freezer. While still cold, seal the amplifier from moist air by putting it into a condom. Depending on how you bias the amplifier, you might want to bring the bias out separately so you can tweak it cold.

At the bottom of my premature posting was this URL:

This is the personal homepage for Darrel Emerson, AA7FV/G3SYS at NRAO in Tucson, AZ.    At NRAO, Darrel has done a lot of work on cryogenic amplifiers at millimeter wavelengths. His personal homepage has some fascinating background info you might find interesting.

To give you some other pointers to cooled amplifiers, got to the SNAP search engine at   and enter the search subject field "cooled hemt"  including the quotation marks. You will get back 3 pages of "hits", some of which may be of interest.

On 11-Mar-1999 W3IWI wrote:
I got several replies to my posting. Here are some answers:

(1) Yes, LN2 (Liquid Nitrogen) should work fine. Depending on how the components in your preamp are held together, there may be some thermal stresses that could break leads on the HEMT or discrete components. Most of the professional LNAs are built on Teflon PCB material which seems to be softw enough to absorb most of the stresses. Chip caps have been known to break loose on cooling. I'd make sure that the amplifier is well sealed so that the LN2 doesn't flow into the amplifier. Condoms have been used (even though they get brittle when cold, the do act as a seal). If you are going to use LN2, don't try to use 7805 regulators built into the box -- do all the power conditioning outside the dewar.

(2) Regarding biases: There are two common ways to bias FET/HEMTs.

One way is to self-bias the transistor with a resistor in the source lead. If you do this, then locate the resistor outside the dewar
because you will have to tweak it!

The second way involves a separate power supply (providing zero current) fed into the gate. Again locate any power conditioning widgets outside the dewar and be prepared to tweak the bias voltage.

To answer one question -- these bias/power supplies are just the DC suppllies needed to make the device operate.They don't add any noise.

ANY semiconductor you use is temperature sensitive and its parameters will change when cooled. You can demonstrate this with an ordinary LED. Bias it so that it glows dimly at room temperature and plunge it into LN2. You will be amazed to see how bright it gets, even at the same current flowing thru the LED.

(3) In general the FET/HEMT devices will perform very well when cold. The internal noise will drop, and the thermal noise added by the surrounding cold components will also drop.

Early HEMTs had one very strange property. Sometimes when cooled below ~50K, the devices exhibited a binary on/off mode (you might call it ON/MORON!) -- sometimes a device worked fine, and other times it died.

This behavior was traced to the fact that a few minority carriers were needed in the device for it to work. It was found that a little bit of light shined onto the ceramic case while being cooled would solve the problem. All our HEMT amplifiers have a LED in the case lid, with the LED in series with the device's drain lead. As noted above, ordinary LEDs work fine when cold.

On 11-Mar-1999 W8TN wrote:
There is an excellent article in the IV International EME Conference Proceedings (Trenton, New Jersey) August 10, 11 and 12, 1990, on this subject.  The article is called "Does Cooling Pay?" by Bill, AA4TJ, (now W4TJ.)

Bill goes into great detail in the 35-page article describing work done by NRAO as well as his own amateur experiments.  The article covers suggested devices along with tables and graphs of various amplifiers and devices at different temperatures.  Construction details of two different types of dewar are also included.

Should anyone want a copy of his article I am sure Bill would be glad to supply one.   His email address is:  <>

On 11-Mar-1999 K3PGP wrote:
I did some experiments along these lines many years ago when C-Band (3.7 to 4.2 Ghz) Satellite TV was still in it's experimental days.  Back then you BUILT your LNA and downconverter as there was nothing available commercially.

My first LNA used biploar transistors.  This became the second stage when I was able to afford a GasFet transistor to use in the front end.  I noticed that when the hot summer sun would beat down on the LNA the noise floor would come up and result in 'sparkles' in the received video as this system was somewhat marginal in SNR and slight decreases in performance were readily observable.  I also noticed that I always got my best performance on cold days, the colder the better!

I experimented with various sun shields out of aluminum foil.  This definitely helped and I recommend that anyone with a mast mounted preamp try to keep it out of direct sunlight to prevent heating.

I then began experimenting with some surplus Peltier coolers.  The results were pretty spectacular although I don't have a clue as to what my noise figure was with and without cooling.  All I knew was I could tell just by looking at the picture if the Peltier was working or not.

One problem I did run into though was moisture condensation inside the LNA. If you are considering using a Peltier cooler on your preamp you will definitely need to look out for this.  I ended up dumping some of those dehydration crystals (the ones that are blue and turn pink when damp) inside my pream then sealing it as best I could.  Many years later when I opened the preamp the crystals were still blue so it must have worked!   I suppose a vacuum could also be used to eliminate moisture inside the preamp just as long as you can maintain a good seal.

I have also experimented with cooling PIN diodes for laser reception.  You can find a graph of these results at the following url along with some other neat data on the home page.

(Note: I'm not showing any E-Mail address here in order to avoid them from being collected by SpamBots. You can possibly find the E-Mail addresses of the above OM at QRZ.COM.)

This discussion is still not closed. If you have any opinions or additional related information you would like to be published here, just send me an E-Mail

Read other Moon-Net discussions

Share: Twitter   Facebook   Home page Site map Radio Sherlock search engine Terms of service  Privacy  Cookies