How to test passively if a GaN transistor is damaged

Hi,
Have any of you used GaN transistors?
We are using a GaN pHEMT. This transistor was operated at a power well above the maximum for about an hour by mistake. This was realized later and subsequently the bias was adjusted to bring down the drain current. Initially for a while the transistor operated as per the design. Now it is giving a loss of 22 dB instead of the required gain of 8 dB. How do I test passively if the transistor is damaged or not (by measuring the input/output impedance etc).

Using an ohm-meter, if the transistor is good you should see a low ohmic value between drain-source and an high ohmic value between gate-source and gate-drain. Keep the negative voltage generated by the multimeter on source, when you conduct the test.

Hi,
Thanks for the answer.
A followup question though.
Is measuring the resistance enough? Isn't it necessary to measure the impedance?

I think a 22 dB loss is symptom of something breaking the continuity of the circuits. If you are sure all the other components and all the power supplies are OK, then the transistor shoud be broken, not just less performing. So I don't think you need to measure the impedance.

Thank you, albbg
Let me test further, first the discrete components if there is any problem with them. If this doesn't work, let me replace the transistor.

Usually, if the dc characteristics are still working, the RF is probably not too damaged. So watch the drain current on a current meter, and vary the gate DC bias voltage. Can you vary the draing current fully, like in a normal working FET, or is it always on or off?

Also, there is the venerable smoke test: if you touch the device, and it burns your finger tip, it might be fried! If smoke should come out, note the color.

Hi Biff44,
When we change the Gate bias, the drain current is getting varied accordingly. But the RF output is 22 dB below the required.
When we operated the device at about 6x the actual operating current, we had found that touching the heat sink (not the device) used to burn our fingers.

If the transistor is actually working, what else could be the possible reasons for not getting the output RF?
Could it happen due to the input/output matching network?

Added after 6 minutes:

One more point:
When we had operated the device at high current, there is a good chance that the junction would have exceeded the Tmax=225 deg C.
Thetajc = 8 degC/w
Dissipation = 21W (At high current)
Thus the heat sink temperature should not have exceeded 225-21*8 = 57 deg C. The temperature could have easily exceeded 57 deg C on the heat sink from how it had felt to us.

You can try to see the RF at various points (also input and output of the transistor) by means of a spectrum analyzer. Just use a piece of uncoated cable not connectorized at one end. Touch with the central tip of the cable the point you want measure keeping the cable with your hand close to the tip. Be careful not to touch the screen of the cable to the ground, otherwise some DC can reach the spectrum damaging it (if it is not AC coupled). Of course in this way your measuement is not accurate, but you can have rough information about gain or losses of various circuits, including the transistor, measuring before and after the circuit you want to test.
Very roughly, if you keep tight the cable screen with your hand that is in contact with the ground of your equipment, the measurement is 6 dB less than the actual value.
This method is simply to apply but quite difficult to explain.

Well, if you touched the heat sink and burned your finger...my guess is that it is time to solder in a new device! The junction temperature inside the device is a lot higher than the heat sink temperature.

You might have degraded the device, but not fully destroyed it.

Check some obvious things, though, like re-touching the solder joints at the device leads and the connectors.

Hi albbg,
We are presently using the same method suggested by you to probe the various points on the board. We are doing this by soldering the cable shield to the GND of the PCB.

Added after 4 minutes:

Yes Biff, I am also now doubting the device.
Let me check the input output matching discretes before I change the device.

Hi all,
It looks like the device is damaged. There is a reason for this. Due to a wrong choice of an output DC blocking capacitor, the circuit was getting into oscillations. This was revealed during simulations. Could such oscillations prove harmful to transistors?
Any thoughts?

Any thoughts, gentlemen?

Hello,

Because of the large difference in expected behavior and actual behavior, checking the DC behavior will tell you whether the device is broken or not.

Oscillation can easily damage a device when there are no current and/or power limiting circuits present. It is not the oscillation itself, but secondary effects like voltage breakdown, too high current and/ or dissipation.

Especially in power circuits, oscillation may occur just outside or at the edge of the intended pass band. So you don't see lots of output, but the transistor is running hot.

If you have the possibility to gradually increase bias and supply conditions, start from a low value. Slowly increase the supply conditions and monitor drain current. As soon as you see an sudden increase or decrease in current, your circuit oscillates parasitically. A jump in current can also be noted when you reduce your supply. It is like a hysteric effect.
,

Hi all,
I feel oscillation is the problem. We had designed the operating band to be from 5.7 GHz to 5.9 GHz. It so happened during tuning the circuit that the lower end frequencies are oscillating. We could observe positive S11 in the lower frequency bands over a very wide band.
When considering stability, what is the typical frequency range over which stability is to be ensured? Any advise?

Hello,

"low frequency" instability in FET amplifiers occurs frequently. When you check datasheets this phenomenon is mentioned frequently. As far as I know, there is no typical frequency range. So best is to check over a large frequency span.

One solution to avoid instability is to make sure you have some real part (seen from the FET) in your circuit to avoid oscillation. Before smoking other devices, you may add a current limiting circuit or timed shut down on over current circuit.