VNA Measurement Issues of an amplifier

Hello all,

I designed an amplifier recently, and I have some issues while trying to measure it.

I did calibration and put in DC-Blocks and some external attenuators to prevent excessive power to the port of VNA.

Then I tried measuring the gain(S21) of my device, and I saw a "jumping" s21 behavior. First thing, which came into my mind were
that I biased the device in a wrong manner, hence changed my dc-block capacitors to a coupler and a filter on each side, and looked at it again.

S21 of my amplifier appears to be at the number I desired then "jumps" down like 5-db. Then back-up, so confusing.

If any of you have seen this before, can you enlighten me on what to do?

Regards,

ktr

If there is a jump on S21 , your amplifier probably oscillates..

Is the source power from the VNA small enough to not overload the amplifier?

@BigBoss

That's what I thought first, then I looked back to my design in ADS, K >1 at all times. My S11 and S22 are quite fine at -15/-20 db levels. How can I be sure it oscillates? Is there a setup you use to be sure for it?

@volker

hmm, thats something I did not looked at. I will check on it. What u say is different from putting an attenuator after vna, is that correct?

In many VNA, you can set the source level from the menu.

Being of K>1 in ADS does not mean that the amplifier is unconditionally stable.Practical measurement is the right reference.
Collect s-parameter data up to fmax. of the transistor ( if it's available) then check the stability with these s-parameters in- for instance- ADS or MW Office.
VNA start-up power level is generally 0dBm ( some VNAs keep the last Output Power ) and this might be excess power for your LNA. Decrease the output level for instance -10/-20dBm and don't forget to redo the calibration at that level.( calibration will be more difficult due to low level !)

I did what u said, and I found out that, amplifier was in fact working correctly. Jumping behavior was stemning from wrong calibration. But, collecting the s-parameter made it quite revealing.

And yes, my VnA had a 0db start-up. Did not know that previously. P1-db is around 17.5dbm for output and I have a 16db gain, 1.5dbm is my leeway. I am connecting a 20db attenuator at Port2 to decrease power.

What would be your choice on while collecting data?

A-) Putting an attenuator to port 2
B-) Decreasing Port power of VNA ( but u said this would be hard to calibrate, can you elaborate on this)

The 20dB attenuator on port 2 is "safe" because you can't damage the VNA with high power. But with the 20dB attenuator at port 2, you get 2*20dB less reflection for DUT port 2, so even with calibration it will be difficult to do accurate S22 measurements. If that matters, you better decrease the source level, or use a combination of both (lower source level and less output attenuation).

Hmm, mb I should do that, but the "jumping" s21 behavior is still what confuses me. How can a s21 jump because of wrong calibration?

I tested the same amplifiers in a Agilent VNA yesterday at a friends' workplace, and they worked quite fine with respect to my simulation results. But today, in my Tektronix TTR500 series VNA S21 still jumps like 5-6db.

Can you show the measurement, so that we see the jumps? What calibration method have you used?

Initially, I was thinking that your VNA does internal switching of the frequency source, and might have different levels in different band segments. Combined with a DUT driven into saturation, that might show up as jumps.

But if it's calibration, it might be as simple as a loose (varying) contact during calibration. According to the photo, you have a lot of adaptors in your setup...

I used Spinner Cal-kit, and did SOLT 2-port calibration. Calibration is done in a 1Ghz band.

I reduced the number of adaptors and added the extra adaptors which were after cal-kit as port extensions.

However, I am making a simple measurement, they should not even be necessary at this level(port extensions and stuff)

Strange, the jump is at different frequencies? I assume you have repeated the calibration and the effect is still the same? Do you also see this with other DUT like an attenuator?

I tried passive elements(which I designed on the same substrate, but on a different pcb) namely a coupler, filter and a divider. They all work fine in terms of s11 and s21.

Active elements seems to be the problem, I tried with different devices(namely a Driver amp, a PA and this LNA).


Tried different calibration bandwidths and also adjusting the port power as you suggested. At some point, I thought amp's must be faulty.

However after seeing the expected results in a Agilent's VNA, I tried to measure with preset of VNA. In preset, although skewed I am getting similar results. It all goes bad after calibration. Which kind of mistake I should make for this behavior, that I can't figure out.

To close the topic, these are the things that I tried, and now have a sense of whats going on.

1- I made 10-12 calibrations trading off between attenuation and port power of the VNA varying from 0dbm port power to -40dbm; while using 6,10 and 20db attenuators on either side depending on configuration.

2- Tried connecting a Spectrum Analyzer after an VNA, to be sure that incident port power is actually the said value, and tested the attenuators same way, measuring power.

3- The jumping issue kind of solved by giving -40dbm from input port(alas, calibration of thru was a nightmare), but im expecting a return from Tektronix as well.

:)

So you think the amplifiers generates some harmonics that the VNA receiver detects, depending on frequency? It will be interesting to see the answer from Tektronix.

Solution: Tektronix VNA software and Windows itself(what a surprise) were causing the bugs. As sad and simple as it is, I got the latest version of software, and changed the decimal sign in Windows, and its done.

Interesting, thanks for the feedback!

Blaming Windows sounds like an excuse to me. The decimal sign issue is something that caused trouble for our own RF software products some decades ago, because we forgot to set that internally in our code and used the system setting. Amazing to see developers today still make the same mistake.