CBBN PA failure - other options?


Gareth G4XAT
 

After initial success and seeing 0.5W out of this board I continued testing beyond a dummy load.
Knowing the PA had no output circulator I fitted one (Marked 10GHz and measuring 52ohms in and out) before any other part. First testing was with a commercial dipole feed WG16, that seemed fine. I then progressed to a home-made G3PHO feed (built to the recommended dimensions found here, http://www.w1ghz.org/QEX/circular_wg.pdf  p45 0.050" probe diameter [sma pin]). At some point during these tests the signals across the shack reduced, so I reverted to the dummy load and found that it was not longer warming after some minutes of TX. Connecting a suitable power meter confirmed that I was down from 0.5W to around 100mW. :-(

From which I surmise the PA has failed in part, but not totally. Looking at the data sheet https://docs.broadcom.com/doc/AV02-0243EN the internal diagram might support that theory.

I have measured voltages under TX condition (but with no drive). The Drain had 4.5V, the negative bias showed as -0.8V (pin 7) so I think all is well in terms of suitable voltages. There is no evidence of smoke egress.

I'm unsure what might have caused this failure, all the 'control circuitry is as the makers intended' AFAIK. I used a circulator which I hoped would provide a safe place for the output if what I connected was a poor match.

My options are various:
1. Use as-is and see how I get on with 100mW. 2. Lift C79 and re-direct the pre-driver output to a suitable SMA, this is likely to be much less than 100mW though as the final PA has around 18dB of gain. 3. I have a second identical transverter so I could just plug that in....but I don't want to kill a second PA....

4. I have a 10W TWT that I can use, but would need to tap the driver chain before the pre-driver for a lower level drive.

From the same source of the CBBN units I also bought a '3 Watt 11GHz' board. The PA strip in that has been hacked by others and used with success at 10GHz. It does have an output circulator on-board (Marked 10.7-11.8GHz) so it might be OK at 10.3GHz? 3W would be a useful increase.
But ultimately, I'm wondering how it failed, so I don't do it again.
Suggestions and ways forward welcome...learning curve being climbed here...
Thanks
Gareth


Mark GM4ISM
 

Gareth

Sorry I don't know there boards, but something to check would be the SMA Elbow on the output.

 I have come across far to many that don't work well at 10GHz, presenting a poor match.  Perhaps 1 in 10 of the ones I have tested have been really bad with rapid match changes at or around 10GHz


Other than that,  is the PA device well enough heat-sinked and is it rated for 100% duty cycle CW?    That board was probably mounted originally using the case as a heat spreader.   I have come across devices (most for the cellular market) that look really great from the headlie figures but their internal thermal resistance junction to case is high enough that they cannot dissipate the CW rated excess heat for more than 10% of the time, even on a perfect heat sink.  What is the device?  a close look at the datasheet may tell you a lot.

Mark GM4ISM

On 11/05/2022 08:22, Gareth G4XAT via groups.io wrote:
After initial success and seeing 0.5W out of this board I continued testing beyond a dummy load.
Knowing the PA had no output circulator I fitted one (Marked 10GHz and measuring 52ohms in and out) before any other part. First testing was with a commercial dipole feed WG16, that seemed fine. I then progressed to a home-made G3PHO feed (built to the recommended dimensions found here, http://www.w1ghz.org/QEX/circular_wg.pdf  p45 0.050" probe diameter [sma pin]). At some point during these tests the signals across the shack reduced, so I reverted to the dummy load and found that it was not longer warming after some minutes of TX. Connecting a suitable power meter confirmed that I was down from 0.5W to around 100mW. :-(

From which I surmise the PA has failed in part, but not totally. Looking at the data sheet https://docs.broadcom.com/doc/AV02-0243EN the internal diagram might support that theory.

I have measured voltages under TX condition (but with no drive). The Drain had 4.5V, the negative bias showed as -0.8V (pin 7) so I think all is well in terms of suitable voltages. There is no evidence of smoke egress.

I'm unsure what might have caused this failure, all the 'control circuitry is as the makers intended' AFAIK. I used a circulator which I hoped would provide a safe place for the output if what I connected was a poor match.

My options are various:
1. Use as-is and see how I get on with 100mW. 2. Lift C79 and re-direct the pre-driver output to a suitable SMA, this is likely to be much less than 100mW though as the final PA has around 18dB of gain. 3. I have a second identical transverter so I could just plug that in....but I don't want to kill a second PA....

4. I have a 10W TWT that I can use, but would need to tap the driver chain before the pre-driver for a lower level drive.

From the same source of the CBBN units I also bought a '3 Watt 11GHz' board. The PA strip in that has been hacked by others and used with success at 10GHz. It does have an output circulator on-board (Marked 10.7-11.8GHz) so it might be OK at 10.3GHz? 3W would be a useful increase.
But ultimately, I'm wondering how it failed, so I don't do it again.
Suggestions and ways forward welcome...learning curve being climbed here...
Thanks
Gareth
-- 
Mark  GM4ISM
gm4ism@...
dc2light.co.uk



AVG logo

This email has been checked for viruses by AVG antivirus software.
www.avg.com



geoffrey pike
 

Hi Gareth,
One of the best things i ever bought was a SMA 8-12 GHz directional coupler and an HP crystal detector and actually measure
the forward and reflected voltage and derived the SWR into the load.
I have done this just recently into the horn for my 10 GHz setup before firing 13 watts into it
cheers
Geoff
GI0GDP

On Wednesday, 11 May 2022, 08:21:45 BST, Gareth G4XAT via groups.io <g4xat@...> wrote:


After initial success and seeing 0.5W out of this board I continued testing beyond a dummy load.
Knowing the PA had no output circulator I fitted one (Marked 10GHz and measuring 52ohms in and out) before any other part. First testing was with a commercial dipole feed WG16, that seemed fine. I then progressed to a home-made G3PHO feed (built to the recommended dimensions found here, http://www.w1ghz.org/QEX/circular_wg.pdf  p45 0.050" probe diameter [sma pin]). At some point during these tests the signals across the shack reduced, so I reverted to the dummy load and found that it was not longer warming after some minutes of TX. Connecting a suitable power meter confirmed that I was down from 0.5W to around 100mW. :-(

From which I surmise the PA has failed in part, but not totally. Looking at the data sheet https://docs.broadcom.com/doc/AV02-0243EN the internal diagram might support that theory.

I have measured voltages under TX condition (but with no drive). The Drain had 4.5V, the negative bias showed as -0.8V (pin 7) so I think all is well in terms of suitable voltages. There is no evidence of smoke egress.

I'm unsure what might have caused this failure, all the 'control circuitry is as the makers intended' AFAIK. I used a circulator which I hoped would provide a safe place for the output if what I connected was a poor match.

My options are various:
1. Use as-is and see how I get on with 100mW. 2. Lift C79 and re-direct the pre-driver output to a suitable SMA, this is likely to be much less than 100mW though as the final PA has around 18dB of gain. 3. I have a second identical transverter so I could just plug that in....but I don't want to kill a second PA....

4. I have a 10W TWT that I can use, but would need to tap the driver chain before the pre-driver for a lower level drive.

From the same source of the CBBN units I also bought a '3 Watt 11GHz' board. The PA strip in that has been hacked by others and used with success at 10GHz. It does have an output circulator on-board (Marked 10.7-11.8GHz) so it might be OK at 10.3GHz? 3W would be a useful increase.
But ultimately, I'm wondering how it failed, so I don't do it again.
Suggestions and ways forward welcome...learning curve being climbed here...
Thanks
Gareth


Mark GM4ISM
 

Gareth,

Just spotted the device data in the original post.

It should be OK for CW operating at 100% duty cycle so long as the case does not exceed 85C  that would require the heatsink to have about 16 °C/W thermal resistance.. Not too hard.. So i suspect you have not cooked it :)

The datasheet does not mention max Mismatch the device can stand, but at this power level that's not uncommon. If the mismatch causes excess drain current it could damage the device. I note that the Abs max drain current is 900mA and at 500mW out it should be drawing maybe 720mA , reasonable headroom.



-- 
Mark  GM4ISM
gm4ism@...
dc2light.co.uk



AVG logo

This email has been checked for viruses by AVG antivirus software.
www.avg.com


-- 
Mark  GM4ISM
gm4ism@...
dc2light.co.uk


Dave_G0WBX
 

Re:-

Knowing the PA had no output circulator I fitted one (Marked 10GHz and measuring 52ohms in and out) before any other part.


What was the spec' of the circulator?   And did it have a load on the third port?

(I don't recognise anything in that image that could be a circulator+load aka "Isolator".  There again, I'm used to higher power level systems.)

A "Circulator" on it's own wont help protect a PA, even a relatively low powered one, unless there is a capable load for at least that power level on it's third port, creating an "Isolator".

How did you "measure" 52 Ohms in to the circulator?   Has that value changed?  (Presumably the DC resistance of a built in dump load.)

73.

Dave G8KBV.


--
Created on and sent from a Unix like PC running and using free and open source software:


Gareth G4XAT
 

Correct, its an external unit - see picture. I'd be amazed if it wasn't capable of absorbing 1/2 a watt and yes, I measured its DC resistance which remains at 52ohms. Noted on the 90 degree connector, that is a high quality part, the other SMA parts I have are largely Chinese sourced.
Yes its heat-sunk, at this point by its large aluminium box. At worst it got pleasantly warm to the touch. No over-current seen, it draws the same as it always did on TX.
Although the photo shows an elbow, the power meter was a straight SMA-N-Power meter head (an old Marconi 6061 with a 3 watt head :-)) and that reading, along twith the little SMA dummy load not even geting warm, confirms the power drop.
Working on comparing the RX path with a Bullseye LNB this afternoon, whilst pondering the PA issue.
Thanks for all comments :-)
Gareth


Gareth G4XAT
 

Well this is interesting....The receive side is only 1-3dB down on a bulls-eye LNB in terms of received MER across the shack (really basic testing....). Using a Pluto on 3.460GHz then a 10.3GHz Pipecap to a little Pluto Whip.
So the RX side is good enough for testing.
And equally ODD, the PA has come back to its previous power level. I can but assume operator error in entered frequencies or something.
So I compared the two circulators I have. The earlier pictured one lost about 2/5th of the available power, its replacement only around 1/10. I'd consider that tolerable for the protection it may offer.
Testing on Saturday with DVBS2. Will probably have a quick go with the Langstone too, since its just a screen-press away.
Thanks for all the earlier suggestions, more handy bits to look for......
Gareth