[SOLVED] HMC 772 Power Amplifier problem

[guytoub]

Helle Guys;

I cascaded three HMC772 amplifiers with a gain of 15dB and a noise figure of 1.8dB. When I measure the final noise figure I have 7.5dB.
The final gain is indeed 45dB. I also found a power of -25dbm by having the input in open circuit.
Where does this problem come from?
Thank you for your help.

 

[G4BCH]

can you explain me your formula please: -174+10*Log10(15e9)+45+1.8 = -25.4dBm.
-174 it is the noise floor ?The

-174 is noise floor dBm/Hz (KTB)
the formula
KTB +10*log10(bandwidth Hz) + Gain dB + Noise figure dB = O/P noise power dBm

To detect a -80dBm signal you will have to include some filtering before the detector as pointed out by Biff44. The input refered noise over the bandwidth of the amplifer is about -70dBm, KTB +10*log10(bandwidth Hz), assuming 15GHz BW.
An input filter may also be needed, not knowing your application I can't say for certain.
The problem is most likely electro-magnetic feed back from input to output. It is a 3 dimensional EM problem and difficult to simulate even with a good EM simulator.
Good screening and absorber should cure it. Be aware that a filter at this frequency will radiate and needs to be screened. A gap of 10mm or less if there is dilectric involved will radiate like an antenna at 15GHz i.e. the sceening will be virtually non-existant.
Place a screen between stage 1 & 2 or stage 3 & 3, ideally screen each stage but that's over kill.

The power distribution looks to me as though it would cause probelms. Appart from the capactors there is no isolation between the stages. the track is just a low impedance transmission line. You need some high impedance in series with each feed line; a thin track should do.

it might seem at first glance to be easy to string a few ampliers together, but as the others can no doubt confirm there is a lot that can go wrong at these frequencies and even those who have many years of experience can have problems. At this frequency every part of the design has to be right or something will come out and bite.

 

[guytoub]

There isn't any issue on stability but this simulation does not affirm that unstability -if there is- does not exist due to decoupling or EM coupling or any kind of imperfection.Therefore take the suggestions here into account and re-implement your board.

Thanks a lot for your help I will take into account all the advice given here an redo my board .

--- Updated Jun 3, 2021 ---

The power distribution looks to me as though it would cause probelms. Appart from the capactors there is no isolation between the stages. the track is just a low impedance transmission line. You need some high impedance in series with each feed line; a thin track should do.

I did not understand very well, would it be possible to give me a little diagram please.

thank you for your help and your suggestions which have been very useful to me. it allowed me to increase my knowledge in this very interesting area.

 

[vfone]

Cascading HMC772 don't need any inp/output coupling capacitors, because the IC have them inside.

https://www.analog.com/media/en/technical-documentation/data-sheets/hmc772.pdf

 

[G4BCH]

The attached file shows a simple sketch of what I mean. I have not shown the low and medium frequency decoupling, the placement of that is not too critical. You should be able to remove two of the three 4.7uF tantalums on each supply, having 3 is not going to help much if at all. I would keep the HF decoupling close to the devices and within the screen.

The bias tracks both + & -V need to be thin to keep the RF impedance high, 250um to 300um would be OK. Keep the tracks for each stage close together so that the ‘mouse hole’ in the screen can be kept as small as possible. Try to keep the breaks in the ground plane for the track crossings as small as possible.

The objective of the screens is to keep the input and output currents constrained as shown. This can be tricky it is a 3 dimensional structure and the gap around the connector will leak.

 

[guytoub]

Thanks everyone for your help, everything is working perfectly now