Post Go back to editing

HMC6981LS6 gate bias self oscillation

Hi everyone,

I designed a K-Band Transmitter using HMC6981LS6 power amplifier.

I followed the layout guidelines from the original EVB. The substrate is RO4350B 0.254mm thichness.

All ground pins goes directly to thermal ground pad.

The decoupling capacitors on Vgg and Vdd pins are the same as EVB.

The only difference is that I put a 10ohm resistor in series with the Vgg pins (before the capacitors) to add some RC in the bias network.

I tested the amplifier alone using my Rigol lab supply, but I've a strange behaviour. I never applied RF, so everything is with no signal and 50ohm terminations.

This is my power ON sequence:

- Vgg and Vdd at 0V, 

- Set Vgg at -2V. Vdd still 0V, Idd still 0mA. (correct)

- Set Vdd to 5.5V. Idd still 0mA (correct)

- Start moving Vgg from -2V to -0.9V (typical Vgg). When I reach Vgg=-1.4V, Idd goes immediately to 1.6A and I see 105MHz oscillation at gate terminal.

- If I rise Vgg in this state, for example at -1.3V, -1.2V etc, Idd remains constant to 1.6A. At -1V, always with oscillation present, the Idd is slightly higher (1.8A).

- If i lower Vgg in the oscillation state, I get pinch-off hysteresis: Vgg must reach -1.8V to kill the oscillation and power off the device (Idd = 0).

So, the oscillation starts at Vgg = -1.5V and remains until I go below -1.8V.

I read in the "MMIC Amplifier Biasing Procedure" page 5, at point #10 that a 200ohm resistor is needed if oscillation occurs in the biasing procedure.

I try that, but there's no difference.

I've another board with the same behaviour, so I don't think the chip is defective.

I see that someone had similar behaviour but with another MMIC --> https://ez.analog.com/rf/f/q-a/71375/hmc8410-not-operating-as-expected

Best regards

-Federico



link with similar behaviour added
[edited by: ParTech at 6:15 PM (GMT 0) on 9 Jun 2020]
Parents
  • PROBLEM SOLVED

    After several tests, we found the problem.

    The major difference between our previous releases, EVB and the actual board is the capacitor layout.

    In the previous release of this board, the decoupling capacitors where all in the top layer. The values are 4.7uF+100nF+1nF+100pF.

    Now we spit the capacitors in two groups: 4.7uF+100nF on bottom (under the chip) and 1nF+100pF in top.

    So the layout is: 4.7uF --> 100nF --> via (0.6/0.3mm) --> 1nF ---> 100pF.

    The funny thing is that Vgg1 and Vgg2 had different oscillation amplitude with the shared bias! So "something" was happening from bottom to top layer.

    Finally we moved the 100nF capacitors of Vgg lines from bottom to top layer, and now the chip is working!

    I simulated the via inductance with LTSpice and the resonance with 1nF+100pF. And....the LC network resonate heavily in the 120MHz region! Ouch!

    We used the same split solution for the other chips in the same board: HMC751, ADL5545, HMC368, but they work perfectly.

    It could be a pcb manufacturing problem or really the layout, but we'll take more attention to that in the future.

    Hope this will help.

    Best regards

    -Federico

Reply
  • PROBLEM SOLVED

    After several tests, we found the problem.

    The major difference between our previous releases, EVB and the actual board is the capacitor layout.

    In the previous release of this board, the decoupling capacitors where all in the top layer. The values are 4.7uF+100nF+1nF+100pF.

    Now we spit the capacitors in two groups: 4.7uF+100nF on bottom (under the chip) and 1nF+100pF in top.

    So the layout is: 4.7uF --> 100nF --> via (0.6/0.3mm) --> 1nF ---> 100pF.

    The funny thing is that Vgg1 and Vgg2 had different oscillation amplitude with the shared bias! So "something" was happening from bottom to top layer.

    Finally we moved the 100nF capacitors of Vgg lines from bottom to top layer, and now the chip is working!

    I simulated the via inductance with LTSpice and the resonance with 1nF+100pF. And....the LC network resonate heavily in the 120MHz region! Ouch!

    We used the same split solution for the other chips in the same board: HMC751, ADL5545, HMC368, but they work perfectly.

    It could be a pcb manufacturing problem or really the layout, but we'll take more attention to that in the future.

    Hope this will help.

    Best regards

    -Federico

Children
No Data