Amplifier temperature compensation example

Click here to go to our main page on thick-film resistors

Click here to go to our main page on attenuators

Click here to go to our page on thermistors

New for February 2022. Themopads® were patented in 1994, by inventors Joseph B. Mazzochette and John R. Steponick ,employees of EMC in Cherry Hill NJ (now part of Smiths).The US patent is number 5,332,981.  Click here to download the patent. Joseph B. Mazzochette is on LinkedIn and lists 37 patents. John R. Steponick, was born in 1937 and apparently resides in Cherry Hill NJ as of this writing. Tell them we said thank you!

The word Thermopad is a registered trademark, so you are supposed to put this symbol with it ®. But like Xerox, Legos and other popular products, the Thermopad® is a victim of its own success and people the word even when referring to competitors' products. The title "thermo-variable attenuator" is more generic but takes longer to say or type... another descrption is "temperature variable attenuator."

The invention's idea is to make an attenuator that has a controlled temperature coefficient, but stays close to its impedance. The patent explains how two different thick-film thermistor pastes are used to pull this off.

In most applications, Thermopads® are used to compensate amplifier gain-versus-temperature effects.  Amplifier gain mostly decreases with temperature, so you will want an attenuator that decreases attenuation with increasing temperature. In the parlance of Thermpads®, this means an attenuator with a negative temperature coefficient (NTC), not a positive temperature coefficient (PTC). In microwave design, you will note that the transmission coefficient S21 of a NTC Thermopad® decreases with temperature, a source of confusion that you will have to live with. In practice, NTC up to -0.007 is about the limit for a temperature coefficient.

A figure from the Thermopad® patent showing a tee attenuator network and thick-film resistors

Below is a collage of Thermopads® supplied by Smiths.  They come in flip-chip, wire-bondable and coaxial formats.


There are other ways to perform temperature compensation, you could use digital or analog variable attenuators and develop control circuitry or even firmware to make the chain behave.   You can use dual-gate amplifiers and use VG2 to tweak the gain. But as "Lord of the Files" author William Golding said,  once said, the greatest ideas are the simplest.


What are the down-sides to using Thermopads®? You need to make sure you stay below the recommended average power, to prevent self-heating.  And you will find that virtually no one has measured S-parameters of these devices that you can paste into your design. You'd better be sure that your Thermopad® is mounted in close proximity to the active chain it is compensating, so that it senses the correct temperature.  Thick-film technology seems to crap out at 18 GHz, so don't plan on using Thermopads® in your millimeterwave front-end. Put them in the IF path!

Example: compensating an amplifier 

Self heating can cause the dB value to drift, so be sure to check the thermal resistance and maximum input power in your design. Ideally you don't want the Thermopad® temperature to increase more than a few degrees during operation.

Here is the simple explanation that Smiths provides in their Thermopad® brochure. 

Now, let's dive into how a Thermopad® can fix a design issue. Let's start with a simple, ideal model of the Thermopad® itself.  It has two basic parameters, temperature coefficient, given in dB/dB/C, and attenuator nominal value, given in dB. Below, we have assigned "TCdB" as -0.006, and "dBnom" at 6 dB.

Below is a plot of loss of our ideal Thermopad® at -40, 25 and 85 degrees C. This is not rocket science, but it will get more interesting when we try to compensate and amplifier, we promise.

Below, we have inserted measured S-parameters at three temperatures, for amplifier ADL8107 from Analog Devices. We used resistors and some simple logic to switch the S-parameter files in a three-temperature sweep, think of them as a pair of ideal SP3T switches.


Before we continue, there are more than one way to connect up the three S-parameter blocks over temperature. Hadrien suggests that instead of the resistors he would have used an S-port with the Sn1 equal to 1 at the right temperature, equal to 0 at the others, and interpolated between temperatures. Sound like a nice idea, we'll try to follow up on this. A third way is to use the ideal SPDT switch element.

Here is the amplifier's gain at three temperatures. At 15 GHz, it changes almost three dB.

Next, we inserted our Thermopad® model into the design, and allowed the dBnom and TCdB parameters to be manually tuned (indicated by blue font).  We found that 3dB with -0.006 dB/dB/C was just about right to fix the problem.  Note that a higher value attenuator with a lower TC would also work, but we wanted to minimize the drop in gain.

Below is the compensated response as we left it. For wideband designs, you will never achieve perfect compensation across the entire band. You can quit when you get the temperature curves to cross over each other at band center. Note that at 15 GHz, the variation is less than 0.7 dB, as opposed to 3dB without compensation. The cost of fixing the gain/temperature response is 3dB at room temperature, you don't get something for nothing.

Other choices

Below is a competitor's product (Susumu). in this case, the design uses a bridged-tee attenuator, and a discrete thermistor is strapped across the bridge. There are more than one ways to skin a cat!





Author : Unknown Editor