Hi,

So, I think I have probably screwed up my NanoVNAv2Plus4 (by supplying too much input power to port #1), but I'd like some input on:
1. Did I actually brick it?
2. Is there any way to fix it (I am reasonably good at SMD soldering)/rework

So, here are two screenshots from NanoVNA-QT

As you can see, the uncalibrated plot is way on the left side.

I can calibrate it, but now the measurements are extremely noisy.

How this happened?
1. I have a two stage FM amplifier who's output impedance I was trying to measure. I terminated the input of the RF PA, and connected the NanoVNA port 1 through a 15dB attenuator to the output of the amplifier. In hindsight, I should have gone for greater attenuation, but what is done is done :-(

2. So, the NanoVNAv2's measurements suddenly changed as I was adjusting the bias point of the amplifier. I turned everything off as soon as I could, but it was already probably too late.

So, my question:
1. Is this nanovna now bricked?
2. Is this fixable?

Thank you!

P.S. The NanoVNAv2 has been a super helpful product in my RF PA and Antenna design, so rather than refraining from using it, I'd learn to use it properly and safely.

You will not likely to get usable results measuring through at 15dB attenuator; an ideal 15 dB attenuator will have a 30dB return loss with either an open or short at the far end and higher  return loss if terminated.  Difficult to calibrate through this.   Also, the output impedance of a power amplifier will vary with drive and output power.  The impedance seen with no output is likely significantly different than the working impedance with full drive. See this applications note:

file:///C:/Users/morri/Downloads/application_note_output_return_loss_of_high_power_class_ab_amplifiers_an056.pdf

FWIW, asking on Tindie, I have been suggested to "replace U801 (MXD8641). It is the QFN IC closest to the port 1 connector, labeled U801 on the board."

Yes i also think that the first ic is broken..

Should be easy to fix...

Dg9bfc sigi



Am 07.04.2022 17:01 schrieb Anish Mangal <anishmg@umich.edu>:

> FWIW, asking on Tindie, I have been suggested to "replace U801 (MXD8641). It
is the QFN IC closest to the port 1 connector, labeled U801 on the board."



_._,_._,_

* * *

Looks like I sent a bad link: I had the app note stored on my laptop. Just
Google " RFMD Application Note AN-056" and it will download from the Qorvo
site.

Here is a clip:

Measuring “hot S22” of an amplifier is usually quite difficult. A power
device operating at or near rated power levels produces enough power to
overdrive and damage most network analyzers. Additionally, the “hot S22”
measurement must be made with a frequency that differs slightly from the
input frequency so discrimination can occur between the reflected output
power and the amplified input power. Additional information on “hot S22”
measurements is available from Anritsu. (
http://www.us.anritsu.com/downloads/files/11410-00295.pdf).


One way I have seen this measurement done involved using a high directivity
directional coupler and filters while having the amp driven to full output
into a dummy load.

On Thu, Apr 7, 2022 at 2:58 PM Bob W0EG via groups.io <morrisnc7=
gmail.com@groups.io> wrote:

> You will not likely to get usable results measuring through at 15dB
> attenuator; an ideal 15 dB attenuator will have a 30dB return loss with
> either an open or short at the far end and higher return loss if
> terminated. Difficult to calibrate through this. Also, the output
> impedance of a power amplifier will vary with drive and output power. The
> impedance seen with no output is likely significantly different than the
> working impedance with full drive. See this applications note:
>
> file:///C:/Users/morri/Downloads/application_note_output_return_loss_of_high_power_class_ab_amplifiers_an056.pdf
>
>
>
>

When you get your NanoVNA fixed or replaced, you will want to look into load pull measurements.

Or maybe you have a working PA and are attempting to  determine if it is optimized by measuring its output impedance. Let me note that RF PAs and for that matter audio PAs are not 50 ohm (for RF) nor 8 ohm for driving an 8 ohm speaker. In the case of audio, modern amplifiers are fairly low impedance output making them more of a voltage source like an AC power line. RF PAs are somewhat similar, but probably not as low impedance. The output network in the RF PA transforms (usually down) the 50 ohm load of the antenna to the impedance that the power device needs to see as a load to produce the power output desired. So measuring the output impedance with a VNA isn't the thing to do.

Hi,

Thanks for the responses. I understand the logic behind load pull measurements, and I do not have the equipment (two directional couplers capable of handling 100W) necessary to perform load pull measurements.

I am trying to design a class-A P.A. capable of ~100-20W RF Power out in the 88-108MHz band.

The transistor I'm using is the MRF300AN. Though I don't need it's max 300W drive power, it has a lower case to junction thermal resistance than its lower power sibling - MRF101AN. I also needed a driver stage (because the source of the signal is a low power LimeSDR).

So, I first set about making the reference designs from the datasheet for MRF101AN and MRF300AN. They provide reference designs for 88-108MHz in case of MRF101AN and 81.36MHz in case of MRF300AN. They also provide starter kits which are quite reasonably priced.

Consulting with NXP, the reference circuits are for when the transistor is biased to be Class-AB. Adequately modifying the load line impedance for class-A operation, and hence modifying the matching networks for these reference designs, I was able to get reasonably close. I also used simsmith to do a very crude form of load-pull (sweeping the complex output impedance of the transistor and observing what its equivalent 50-ohm transformed impedance looked like). Then, as a final step, I connected the NanoVNA along with attenuators and fine-tuned some trimmers to get a very nice match.

This approach worked rather well. *I even varied the drain current significantly to still see that the output matching was decent for different values, telling me that I've a good matching network.*

The problem appeared when I tried to tweak another of their reference design which combines two transistor stages

Here, these two are also presumably meant to run in class-AB, and I want them to run in class-A. One challenge here is that instead of matching between the stages (i/p of stage#2 and o/p of stage #1) is just a 1nF capacitor. There isn't a matching-to-50-ohm-and-back network as it would just make things inefficient.

I know I should much rather learn load-pull but for this iteration, the time to get a working amplifier is limited for me.

Anyway, I tweaked the output matching circuit, for what I "think" would work, but there were two problems. I'm dealing with a lot more gain here, and needed perhaps stronger attenuators. Also, the matching network, is perhaps poorly designed. When I'm at the Idq that I target - 4A, the match is good, but the match deteriorates much more significantly with different Idq values, than in the case of the original MRF300AN-reference design.

Anyway, it looks like for now, I will go with the two-independent-amp-stages for now, learn more about advance techniques for rf pa design (load pull and simulating in ADS) and reserve this effort for a future iteration.

I've tried to best capture the somewhat chaotic process I followed, but I do have two working amplifier stages giving be decent class-A performance which I was looking to consolidate the circuit of. If someone has more thoughts on how I could refine the design for a future setup, I will definitely keep in mind when coming back to this circuit.

73 // VU2TVE

You may want to try this: connect a manual tuner and power meter between
the amp an a good fifty ohm power load. Drive the amp fully and adjust the
tuner for maximum power into the load. Disconnect the amp and use the
nanoVNA to measure the S11 looking into the tuner. The output impedance of
the amp at full power will be the complex conjugate of the measured S11.

On Thu, Apr 7, 2022 at 8:46 PM Dave W6OQ via groups.io <david.hostetler=
ieee.org@groups.io> wrote:

> When you get your NanoVNA fixed or replaced, you will want to look into
> load pull measurements.
>
> Or maybe you have a working PA and are attempting to determine if it is
> optimized by measuring its output impedance. Let me note that RF PAs and
> for that matter audio PAs are not 50 ohm (for RF) nor 8 ohm for driving an
> 8 ohm speaker. In the case of audio, modern amplifiers are fairly low
> impedance output making them more of a voltage source like an AC power
> line. RF PAs are somewhat similar, but probably not as low impedance. The
> output network in the RF PA transforms (usually down) the 50 ohm load of
> the antenna to the impedance that the power device needs to see as a load
> to produce the power output desired. So measuring the output impedance with
> a VNA isn't the thing to do.
>
>
>

I guess, here when you imply 'amp' you meant the output of the transistor itself (perhaps fed through a DC blocking cap). That way I get the output impedance of the transistor. Accordingly, I can then measure the tuner's impedance and figure out the amp's impedance and design a matching network. This approach would allow me to measure the impedance across the range im interested in and then design a suitable broadband matching network.

One question though. Would an "input matching network" affect the output impedance in a significant way? In my limited testing, it seemed like it was the case. If I changed the input matching network, it seemed to have an affect on the output impedance, or it could be that I was measuring it wrong.

Thoughts?

Personally I would use a different device for 100W class A. You will get a maximum efficiency of 50% if the rest of your circuit is ideal, in a real life scenario, you end up with an much lower One of our amplifiers at work has the MRF300 as the driver running 30w out  class A water cooled and it still runs quite hot at 88Mhz.

You might want to watch this and play with the software. https://www.youtube.com/watch?v=GhPqPVlDRPY

John

You're right. I think I may have overstated the 100W figure. The current setup that is already working to our desired performance level is putting out roughly ~65W in class-A (I think).

But I did explore NXP devices with lower case-to-junction thermal resistance and the next step up would be something like this - https://www.nxp.com/products/radio-frequency/rf-industrial-scientific-and-medical/1-600-mhz-broadcast-and-ism/1-8-600-mhz-300-w-50-v-lateral-n-channel-broadband-rf-power-mosfets:MRFE6VP6300H

But allow me to restate the problem. I already have a working setup with two amplifier stages. The first one is an MRF100AN driven at just 12V/0.6A Idq, and the second one a MRF300AN at 50V/3A, and it works. But I was looking to instead build off of the reference design which includes the AFT05MS004N as stage #1 and MRF300AN as stage #2. The idea behind was also to not match the stages to 50 ohm but match them to one another. However, while fiddling with the matching, I caused the problem with the NANOVNA, so I (a) need to fix it (b) try a better approach than the one that i've been following for impedance matching for a future re-design.

The actual application here is that rather than 64W going into one FM carrier, there will be 8 carriers with 1W each. This setup was already trialed in the real world, for the application it was meant for (in this event - https://www.dalailama.com/videos/teachings-in-dharamsala-jataka-tales ), so I was just looking to refine the design a bit further.

To help with some calculations, I designed this spreadsheet - https://docs.google.com/spreadsheets/d/1sBoRY1teJVLkM9s9JaP73GgkqS_izmh1ziyh5p3Rt4o/edit#gid=2126483046 which is meant to help with finding the right operating point for the two transistor stages.

73 // VU2TVE