s2p file from VNA

[regenfreak]

To start with you have to ensure the input and output impedance are both 50 ohms or connected to matching networks or transformers. For non-50 ohms DUT, impedance mismatch of input/output can give completely wrong results. Garbage in = garbage out

For non-50 ohms DUT, sometimes it is a tall-order to figure out the unknown input and output impedance of the device, it is a bit of chicken and egg situation full of uncertainties in the measurements.

W2AEW illustrated how he measured a 50 ohm bi-directional amp with a VNA:

https://youtu.be/7TtKE39TWpI

He can be 100% confident of his results because of the certainty in using 50 ohms input and output impedance.

[G0HZU_JMR]

To start with you have to ensure the input and output impedance are both 50 ohms or connected to matching networks or transformers. For non-50 ohms DUT, impedance mismatch of input/output can give completely wrong results. Garbage in = garbage out
For non-50 ohms DUT, sometimes it is a tall-order to figure out the unknown input and output impedance of the device, it is a bit of chicken and egg situation full of uncertainties in the measurements.

I might be wrong but in this case I think dmowziz is trying to measure the s-parameters of the transistor itself rather than measure a complete amplifier. In this case it does mean the VNA has to measure a device where the input impedance and output impedance is a long way from 50 ohms. This does require good hardware, a good calibration and an experienced operator or the results will be very poor.

It's fairly rare to see a transistor manufacturer publish s2p data for a BJT below about 10MHz because it becomes too difficult for the VNA to make accurate measurements when the VSWR climbs up above 100:1 and into the thousands. The reverse isolation can also be very high and this introduces some noise and uncertainty onto the s12 measurement. I generally still try and measure s2p data for a BJT down to about 1MHz anyway and often the data is still useful/educational although it can become very noisy.

[dmowziz]

Originally Posted by G0HZU_JMR

I might be wrong but in this case I think dmowziz is trying to measure the s-parameters of the transistor itself rather than measure a complete amplifier. In this case it does mean the VNA has to measure a device where the input impedance and output impedance is a long way from 50 ohms. This does require good hardware, a good calibration and an experienced operator or the results will be very poor.

You are right, not an amplifier. The TIA in the video from w7zoi, i've done that before and measured.

Jeremy, i think you can assist me in many ways and make me a better engineer (almost graduate). Including David in many ways
Don't hold back.

I have a SVA1015X. Trying to get SMA cables asap.

God bless you sir

[dmowziz]

@David @Jeremy

Abit off. Please how can one learn the most to become an expert in Electronics...

I usually find myself spending so much time and the math before soldering.
Thinking maybe should solder alot of circuits to gain mastery

Please what method to become an expert someday

Thanksss... and sorry, off topic

[G0HZU_JMR]

To gain experience with s-parameters you can always practise the export of s2p files by doing it all within Genesys. If you draw up your BJT circuit and bias it up for 10Vce and 5mA Ic and then simulate it over 1MHz to 3GHz you can then click on the linear1_data icon in the workspace tree and then move the mouse up to the File menu and select File>Export>S-Parameters.

This will give you an s2p file containing the s-parameters of the circuit. You can then add some microstrip to the circuit (to mimic a basic test fixture) and then experiment with ways to de-embed the microstrip to shift the reference plane back to the pins of the transistor.

[dmowziz]

@David @Jeremy

Please how can one learn the most to become an expert in Electronics...

I usually find myself spending so much time on the math before soldering.
Thinking maybe should solder alot of circuits to gain mastery.

Please what method to become an expert someday

Thanksss... and sorry, abit off topic.


Thank you sirs

[G0HZU_JMR]

Things are very different today compared to when I was a student so maybe I'm not the one to give the best advice in 2022. However, I'd always recommend to study the basics about how EM waves propagate through networks. This doesn't mean you have to master Maxwell's equations but just be aware of how this happens. I'd also recommend to spend a lot of time studying lumped components to master how they behave (and misbehave) at various RF frequencies. I'd always recommend using a good RF simulator as a learning tool. I wanted the (Circuit Busters) Eagleware simulation tools more than I wanted the exotic RF test gear at work. Today there is a lot of choice for free simulation tools and lots of decent new/used RF test equipment at low prices.

See below for a GMAX and K plot of a 2N3904 BJT that I measured recently with a VNA. This is a transistor we are all familiar with. This shows some of the information that can be taken from a single VNA derived plot.

It shows that at this bias point of 10Vce and 10mA Ic this jellybean BJT can produce about 25dB power gain at 50MHz (with K>1) it can produce 15dB power gain at 145MHz and about 8dB gain at 433MHz. The K curve dips below 1 at about 1GHz and so this transistor can go unstable and oscillate up at 1GHz . That's a lot of useful info about this common transistor from just one plot.

If you look closely, the strange kinks in the GMAX plot occur each time K passes through 1 and this is normal. GMAX can be thought of as the maximum power gain that can be achieved from the transistor whilst maintaining K>1.

[dmowziz]

I've "seen" what you do. Learning from you will be great

Thanks for the advice

The K and GMAX, is the equation to get them the one on page 196 of ITRFD ?

Using S Parameters.

Please can you share the s2p file? Will like to compare with what I'll measure

Thankss

[G0HZU_JMR]

The GMAX trace in my version of Genesys plots Max Available Gain when K>1 and it plots the theoretical Maximum Stable Gain (for K=1) if K is less than 1 on the graph. That's why it's always best to plot GMAX and K together in Genesys because each time K passes through 1 Genesys will produce the kinks in the GMAX curve as it hops between GMAX and MSG.

The equations are given in Genesys for both MAG and MSG.

I still think you should either experiment with JFET parts or maybe try a BJT in common base. Otherwise the BJT will produce too much power gain in common emitter and your VNA will overload the BJT and also port 2 of the VNA. See below for an s2p file of the SMD version of the MPSH10 BJT.

Despite what the authors of many RF theory books claim, a VHF BJT configured as a common base amplifier is very prone to instability. This instability can happen well above 1GHz.

In common emitter the MMBTH10 can go unstable up to about 2.7GHz at 10vce and 10mA according to my VNA data although this assumes lossless resonator components. In reality 2.4GHz would be a more realistic upper limit for oscillation assuming real components were used to make up the oscillator circuit.

A 2N3904 at a similar bias point can perform a similar oscillation trick up to about 1GHz. The GMAX and K data from my VNA show that the 2N3904 can achieve about 15dB power gain at 145MHz with K>1 at 10Vce and 10mA Ic.
This is based on my VNA data from an On-Semi 2N3904 bought from Farnell recently. Ideally you should buy from Farnell (or similar) if you want genuine On-Semi parts to compare against.

Have a play with the common base s2p model below using Genesys. You should be able to plot GMAX and K.


! Filename: MBTH10CA.S2P Version: 3.0
! Philips part #: PMBTH10 Date: Dec 1990
! Bias condition: Vce=10V, Ic=5mA
! CB CONFIGURATION
# MHz S MA R 50
! Freq S11 S21 S12 S22
40 .775 176.8 1.757 -3.6 .002 85.3 .988 -1.0
100 .769 172.4 1.750 -8.7 .006 95.1 .992 -2.5
200 .751 165.3 1.727 -17.4 .012 99.7 1.004 -4.7
300 .729 158.7 1.689 -25.9 .019 107.0 1.017 -6.8
400 .703 152.6 1.642 -34.4 .027 108.7 1.031 -8.8
500 .675 147.2 1.591 -42.6 .035 108.2 1.040 -10.8
600 .650 142.4 1.535 -50.8 .045 107.7 1.045 -13.1
700 .626 138.1 1.473 -58.8 .055 106.8 1.046 -15.7
800 .607 134.2 1.403 -66.9 .065 105.5 1.045 -18.7
900 .587 130.7 1.338 -74.8 .077 104.8 1.044 -22.0
1000 .571 127.4 1.274 -82.8 .088 103.8 1.045 -25.5

[G0HZU_JMR]

To give you some encouragement I built up the 145MHz (15dB gain) amplifier using a 2N3904 from the same Farnell bag. I carefully set the collector voltage to 10.0V and the Ic to exactly 10.0mA and then replicated the AC equivalent circuit below.

If you go back and look at the GMAX curve in post #27 it shows that just under 15.7dB gain is possible at 145MHz with the 2N3904. In reality there will be some losses in any lumped matching components and the simulation below predicts the matched 145MHz amplifier will have about 15.2dB gain with realistic losses included for the matching. The simulation doesn't include a 10V power supply or any bias resistors because the biasing is already locked into the s2p model. In the plot below you can see the agreement between the built amplifier (real 2N3904 and real matching components) and the simulation based on my own s2p model of the 2N3904.

You can see that the agreement is very close for s21 (both are close to 15.2dB at 145MHz) and also for the s11 and s22 traces.

To get results like this does require a fairly decent VNA and a lot of experience. However, it may be possible to get close to these results using one of the newer nanovna models. At some point I'm going to buy another nanovna to see how much they have improved.

[G0HZU_JMR]

I also dug out a SMD MMBTH10 BJT to do some further testing. These state 650MHz Ft on the bag but this is for unity current gain rather than power gain.

If you look back at marker 4b on the plot in post #20 my VNA data suggests that this device can still go unstable up at around 2.4GHz when biased at 10Vce and 5mA Ic. This is nearly four times higher than the Ft frequency stamped on the bag by Farnell and on the On Semi datasheet. The BJT can still oscillate well past the Ft frequency because an oscillator doesn't need >= unity current gain to meet the requirements for oscillation. It can still produce power gain at much less than unity current gain.

I loaded my VNA derived s2p data model of the MMBTH10 into Genesys and I added a simulated transmission line resonator at the collector and with this resonator added Genesys predicts that this BJT can still deliver a whiff of negative resistance at the base at about 2.4GHz. By adding a small tuning cap at the base there is resonance and a net negative resistance at about 2.4GHz as in the plot below.

I built the oscillator circuit with the resonator and just tested it on a spectrum analyser. Sure enough it oscillated close to the expected 2.4GHz as in the analyser plot below. This shows the power of measuring these devices using a modern VNA. This technology is now coming within reach of hobbyists as the low cost nanovna technology improves year on year.

I suspect that I could get it to oscillate beyond 2.5GHz by optimising the resonator components but I doubt I would get close to 2.7GHz unless I also changed the bias point of the MMBTH10. At the moment it dies just before 2.5GHz if I try and tune it higher in frequency and this is to be expected as my resonator parts introduce ESR losses and the simulation plot below shows there is only a few ohms of negative resistance available at the base of the transistor up at 2.4GHz.

[dmowziz]

Thank you very much Jeremy!

The Gmax and K I studied briefly before (it was "dry") now seems fun ... It's something I can measure! Thank you

Which will you get? A right angle SMA plug or straight?

I'm getting funds this week so getting components.
First thing being to try and measure the S parameters of a transistor with my big VNA and compare with what you have.

"
I still think you should either experiment with JFET parts or maybe try a BJT in common base. Otherwise the BJT will produce too much power gain in common emitter and your VNA will overload the BJT and also port 2 of the VNA."

Thank you, I'm doing this today with a NanoVna


Btw, post #29, will this be the GMAX and K for that? I don't think so

[G0HZU_JMR]

The K plot looks OK but the GMAX plot looks different to what I'd expect from Genesys. In Genesys it should be possible to just enter GMAX in the 'variable' box (after you double click on a graph) and it does it all for you automatically. You don't need to enter any equations.

The other way to do it is via the equations engine but you would get a different result unless you put in all the same conditional equations that Genesys uses for GMAX. In parts of the graph where K>1 the built in GMAX function plots the max available gain. However, if K dips below 1 then Genesys computes the maximum stable gain that is possible with K=1. This is a really powerful way of presenting the data as it can quickly show how much gain can be extracted from the device at each frequency if the design aim is for K>1 at that frequency.

It is possible that your 2020 version of Genesys is different. Maybe GMAX isn't plotted this way on the new version?

I generally don't like using right angle SMA connectors up at UHF but they should be fine for your initial experiments. However, I'd prefer to use straight connectors.

[dmowziz]

I made a mistake.

This is correct I think :


Thankss..Getting straight

#30, the inductors, please what did you use?

[G0HZU_JMR]

I don't think you are quite there yet, your GMAX plot still doesn't look right to me. The numbers are too high down at 50MHz for example. However, there may be different interpretations as to what GMAX means.

There's some info about GMAX here in the Keysight knowledge centre. Scroll down to find the stuff about GMAX.

https://edadocs.software.keysight.co...0/S-Parameters

I think the GMAX plot below follows the equations given in the link above.

Now that you have that common base MMBTH10 s2p file loaded into Genesys have a go at plotting the complex impedance looking back into port 2 (the collector). You should see loads of negative resistance up at VHF and into UHF. This should be a clue as to how prone VHF common base amplifiers are to instability if care is not taken with the design. Some authors of RF theory books will often tell you that with the base grounded the device should be quite stable. That s2p file from Philips for the MMBTH10 will tell you the opposite is true. In common base, this transistor is very keen to be part of a UHF oscillator.

I just used regular air wound solenoid coils for the 145MHz amplifier.

[G0HZU_JMR]

If you do have a go at making the 145MHz amplifier make sure you use the same 2N3904 manufacturer as me. My parts are OnSemi and were bought from Farnell. I don't recommend buying 2N3904 parts from ebay or from small surplus shops if you want to do critical experiments like this. For most general purpose stuff it's OK to use the smaller suppliers for a 2N3904 but you might not be buying a real 2N3904. To get the same results as me you need a genuine OnSemi part.

The 2N3904 is quite an old part and it is different to other jellybean BJTs even though the datasheet description and specifications appear very similar to other jellybean transistors. The internal die is (was?) made with a very different process (process 23?) and this gives it superior RF properties to process 4 transistors like the BC107, BC547 etc.

See below for an old spectrum analyser plot of a 2N3904 configured as a 1GHz oscillator. I don't think this could be achieved with the BC107 type of jellybean BJT. I suspect I could get closer to an oscillation at 1200MHz with the 2N3904 if I tried again.

[dmowziz]

"Now that you have that common base MMBTH10 s2p file loaded into Genesys have a go at plotting the complex impedance looking back into port 2 (the collector). You should see loads of negative resistance up at VHF and into UHF."

Thanksss...Trying this

Meanwhile, is my plot similar to yours now?
Unit dB in my version does 20*log but GMAX is power

If my plot is similar then I think we plotted for a CE, not CB.

Thankss, I'm ordering parts tonight finally!
But I might not be able to buy $500 calibration set for my VNA (I saw a thread by you describing a homemade calibration set, I'll check for the thread)


Before making the 145 MHz amplifier, I will want to test and measure my setup for post #20 using a MMBTH10

Thanks Jeremy

[G0HZU_JMR]

Your GMAX plot looks much better now, a good result!

If my plot is similar then I think we plotted for a CE, not CB.

I'm not sure what you mean but if you look at the first line of the sdata for the MMBTH10 at 40MHz it shows this:

40 .775 176.8 1.757 -3.6 .002 85.3 .988 -1.0

The input s11 data shown in green shows mag 0.775 and an angle of about 177 degrees. On a smith chart this would indicate a low resistance. The magnitude of the reflection coefficient can be converted to a VSWR as (1+0.775)/(1-0.775) = VSWR = 7.89:1 VSWR. As the angle of 177degrees is close to 180degrees this is mainly resistive and the input resistance at the emitter can be crudely approximated to be about 50/7.89 = 6.33 ohms according to the s2p data at 40MHz.

The MMBTH10 was biased at 5mA when this data was taken. Classic BJT theory states that in common base the input impedance at the emitter can be approximated as 26/Ic = 26/5 = 5.2 ohms.

So this does indicate the s2p file is for the MMBTH10 in common base at 5mA Ic as the 5.2 ohms figure is close to the 6.33 ohms indicated by the s2p data.


If you look further down the s2p data to the line for 400MHz you can see there is negative resistance at port 2 (collector) because the magnitude of s22 is greater than 1 as highlighted in blue below.

400 .703 152.6 1.642 -34.4 .027 108.7 1.031 -8.8

This transistor is so keen to go unstable in common base that the only thing it needs to complete an oscillator at 400MHz is a collector inductance with a reflection coefficient that has an angle of 8.8degrees. This equates to an inductance of about 260nH at 400MHz and this would need to be fitted at the collector to complete the 400MHz oscillator.

Of course, this is quite a large inductance and the really scary stuff happens up at >= 1GHz where a stray layout inductance of maybe 12-30nH at the collector can create the conditions for a parasitic oscillation at over 1GHz.

[dmowziz]

Thanks Jeremy,

Since I joined you've made me understand some things required to work at RF. Grateful.. I'm taking notes

so at 800 MHz

800 .607 134.2 1.403 -66.9 .065 105.5 1.045 -18.7

an inductance of about 59.3 nH ?



I'm waiting for parts so can do the experiments.

Please the copper clad board you used here #35
https://www.vintage-radio.net/forum/...=187177&page=2

What's the thickness please

Thanks Jeremy

[G0HZU_JMR]

I think I used 0.02" thick Rogers 4003C PCB material for the little strip section of PCB but I don't think this is critical for that 75MHz filter design.

Yes, according to the Philips s2p data 59nH would make it oscillate at 800MHz. However, I do have some old s2p data of my own that suggested only 22nH was needed for 800MHz oscillation. I'll have a go at measuring a MMBTH10 and MPSH10 again in common base on the VNA.

My old data suggested the oscillation frequency would be much lower than 800MHz with 56nH so I need to look at this again. Also that s2p data is for a Philips part and my MPSH10 devices are Nat Semi and my MMBTH10 devices are all made by OnSemi. I have several bags of them here.