What else can I deduce from this Smith chart?

[MisterHeadache]

I’ve finished testing seven different transistors (see my earlier thread: https://www.eevblog.com/forum/rf-microwave/modern-smt-bjts-to-replace-classic-to-39-bjts-in-amateur-hf-amplifiers/).  I’ve plotted the respective input impedances of my test amplifier from 1MHz to 50MHz on a Smith chart.  See the attached image.

I understand the basics here by looking at this data, meaning:

• The 2N5109 performed the best in this group (meaning it has the least amount of deviation from 50 ohms over the frequency range) and the 2SCR573 performed the worst.
• The 1MHz responses are in the lower right quadrant.  The data progresses clockwise as the frequency increases, moving toward the 50MHz response in the upper left quadrant.  So this means that the impedance starts out with a capacitive reactance component that gradually transitions towards inductive, and a resistive component that progressively decreases with frequency (before increasing slightly on the 2SCR573 and BCP55).

I am thinking that’s about all I can conclude from these plots, however, the arcuate shape of the responses has me curious:

• They’re not following the circles of constant resistance.
• They’re not following curves of constant gamma/reflection coefficient, which would be centered at the origin.

Question: is there any other significance to the arcuate shape of these response curves?

[tszaboo]

You can deduce two things:
1) Making a wide-band amplifier is difficult.
2) You should do input impedance matching.

[szoftveres]

I can deduce that the input impedances of the 2SC5994 and 2SCR513 behave very similarly in an unknown test setup (unknown topology, unknown circuit, unknown bias conditions, unknown output load, etc..)

[G0HZU]

I think it might be the circuit shown at 7:10 in the youtube video below?



If so, this circuit doesn't look that good to me. It seems a really curious choice. There's some odd design choices that significantly compromise the bandwidth and also compromise the large signal performance.

The smith chart plots kind of look OK but I'm surprised to see even the 2N5109 cross the real axis before 50 MHz. Maybe there is a reference plane issue.

Either way, that amplifier design is going to produce a lot of input capacitance (Cp) especially at low frequencies. The input Cp could get as high as about 1nF at 1 MHz? Cp might reduce down to about 50pF at the top of the HF band (30 MHz).

The input capacitance is what will be pulling it clockwise round the smith chart in an arc with increasing frequency. The arc seems more exaggerated than expected though? Maybe you have a slight reference plane issue with your VNA calibration.

[MisterHeadache]

I think it might be the circuit shown at 7:10 in the youtube video below?

Yessir, that's it.  I chose it because it was a known quanitity, had been around for a few years and was known to work (but not necessarily known to work optimally).  Also the choices for feedback resistors matched very closely with the formulas in EMRFD, so predicting its performance was straight forward.  I've attached the schematic.  My starting point is to just substitute the modern transistors into the exact same circuit and see how much different they perform.

Either way, that amplifier design is going to produce a lot of input capacitance (Cp) especially at low frequencies. The input Cp could get as high as about 1nF at 1 MHz? Cp might reduce down to about 50pF at the top of the HF band (30 MHz)

What part of the design is determining this input capacitance behavior?

Part #2 of my experiments with these modern transistors will build on what I learn from these experiments, and if there are ways to tweak this design to improve its broadband HF performance, I'll be including that.

[G0HZU]

What part of the design is determining this input capacitance behavior?

I think there will be a few components that contribute to the way the parallel input capacitance Cp changes vs frequency.

Through upper HF and partly into VHF the dominant contributor will be due to Miller effect. There's probably over a x20 voltage gain (Av) and the collector base capacitance of the 2N5109 will be about 2pF. So there could easily be (Av+1)*2pF =   maybe 50pF Cp at the base of the 2N5109 due to Miller effect combined with the internal Cbe capacitance.

Once you look below about 10 MHz, there will be other aspects of the design that will boost the Cp much higher as the frequency is reduced. The inductance of the transformer at lower frequencies will cause some phase shift at the collector and so the resistive feedback (1800R) will no longer arrive back at the base in anti-phase with the input. It will will be phase shifted to make the input look very capacitive and this effect will increase as the frequency is reduced. The reactance of the 100nF emitter capacitor will also become significant at lower frequencies and this will also contribute towards the shifting phase at lower frequencies (and this adds a bit more capacitance at the input at lower frequencies).

So the net result will be that you will see a changing input Cp vs frequency. It might be 50pF at 30 MHz and maybe climb to 100pF at 7 MHz and it might climb to 1000pF down at 1 MHz. This will cause the s11 arc to not neatly follow a circle on the smith chart as the frequency changes.

[G0HZU]

If it helps, the things I'm not too keen on about that test amplifier are the choice of transformer, the excessive RF gain, and (possibly) the 1uH inductor in the feedback.

The transformer looks to be a conventional (and quite lossy?) 6:3 turn transformer (4:1 Z) and I'm used to seeing a transmission line transformer (TLT) here. Maybe the lossy transformer helps with stability up at VHF or UHF but otherwise I can't see the advantage. A TLT will offer lower loss over a much wider bandwidth and this includes the low frequency end because the 6:3 turn ratio transformer will have a fairly low primary inductance and this will cause a lot of loss down towards 1 MHz.

The high gain of the BJT (>20dB without the 6dB pad?) will make it harder to maintain a good input and output match across 1-50 MHz. I'm not sure why the 1uH inductor is there but maybe it helps compensate for something up at VHF.

I looked online at the SV1AFN website  at the specs for that amplifier and I think the claims are a bit misleading. It isn't going to provide a 23dBm P1dB compression point because of the 6dB pad at the output. Also I don't see how it can work well down to 0.5 MHz unless several dB lower gain and a poor input match is acceptable. Having the higher gain will probably help with the noise figure so it might manage to produce a sub 4dB noise figure.

Because of the high gain the input IP3 will be compromised. Maybe this was all done for a reason but it seems to be a slightly odd design. Maybe you could try reducing the gain and try a TLT and this may help give a better match over a wider bandwidth. You woukld probably see the most benefit down towards 1 MHz in this case.

I'm not familiar with some of the transistor types you are checking out but some of them don't look that good on their datasheet. It will be interesting to see how they perform in your tests.

[MisterHeadache]

G0HZU, thanks for the detailed replies! You've given me a lot to think about!

The transformer looks to be a conventional (and quite lossy?) 6:3 turn transformer (4:1 Z) and I'm used to seeing a transmission line transformer (TLT) here. Maybe the lossy transformer helps with stability up at VHF or UHF but otherwise I can't see the advantage. A TLT will offer lower loss over a much wider bandwidth and this includes the low frequency end because the 6:3 turn ratio transformer will have a fairly low primary inductance and this will cause a lot of loss down towards 1 MHz.

I think you are referring to something like a FT37-43 core with 10 bifilar turns wound on it, right?  I see those frequently used in EMRFD for transforming 50 ohms to 200 ohms at the collector, with the next stage load attached at the center of the windings.

The high gain of the BJT (>20dB without the 6dB pad?) will make it harder to maintain a good input and output match across 1-50 MHz. I'm not sure why the 1uH inductor is there but maybe it helps compensate for something up at VHF.

I looked online at the SV1AFN website  at the specs for that amplifier and I think the claims are a bit misleading. It isn't going to provide a 23dBm P1dB compression point because of the 6dB pad at the output. Also I don't see how it can work well down to 0.5 MHz unless several dB lower gain and a poor input match is acceptable. Having the higher gain will probably help with the noise figure so it might manage to produce a sub 4dB noise figure.

I was scratching my head over that 1uH inductor in the feedback too.  So I ran my LTSpice gain and input impedance sims to 100MHz with and without it and saw no difference.  But I left it in on my PCB design, for package protection for no other reason in case it turned out to be necessary.  Regarding the 23dBm P1dB, my testing confirmed the 2N5109 does reach it...without the 6dB attenuator.  Maybe that's what he meant?  Just a guess.

Because of the high gain the input IP3 will be compromised. Maybe this was all done for a reason but it seems to be a slightly odd design. Maybe you could try reducing the gain and try a TLT and this may help give a better match over a wider bandwidth. You woukld probably see the most benefit down towards 1 MHz in this case.

I'm not familiar with some of the transistor types you are checking out but some of them don't look that good on their datasheet. It will be interesting to see how they perform in your tests.

I will add those to the list of things to try for round #2.  And those mods should benefit all the transistors.  Yes, I did deliberately include the 2SCR573 and BCP55, knowing that their specs didn't look so good for Ft and capacitance.  I felt I needed a few marginal parts in the mix, if for no other reason than to prove my testing was good enough to detect differences in performance.

Also looks like I need to go study the Miller effect some more.  I've seen several IF amplifiers use cascode designs to help mitigate it, but I hadn't run across a cascode design that could output power in this ballpark (~20dBm.)

[G0HZU]

It's a long time ago, but I have designed a few amplifiers like this in the past using a TLT. It's generally best to use the minimum number of turns as this helps with stability up at UHF.
10 turns seems like a lot. I would have used half this many turns? I've used twin hole bead based transformers and also toroidal. I used fairly fast BJTs back them biased at about 8Vce and 25-40mA Ic.

With a fast BJT it should be possible to design for a gain of about 16dB and get a 3dB bandwidth of about 250 MHz. With a low noise device the noise figure can be as low as about 3dB across most of the frequency range at currents up to about 40mA if a suitable BJT is selected. I've also used modest devices like the 2N3904 with good results. I recall designing for 16dB gain and a 3dB BW of about 60-70 MHz with a 2N3904. It was probably biased at 8Vce and maybe 25-30mA. The noise figure will be higher with the 2N3904 and I think I achieved about a 5dB noise figure, maybe a bit less.

[G0HZU]

The other thing to be wary of is the amount of inductance in series with your 5.6 ohm emitter resistor. Some of it will be in the leaded resistor package itself and some in your board layout. This probably doesn't matter that much at 2-30 MHz but if you do want to explore up at 50 MHz and try and improve things, the reactance of this combined inductance can begin to become significant at 50 MHz compared to the 5.6R emitter resistance. It's going to start affecting the gain up at 50 MHz. Not by much, but it will affect it slightly.

[mtwieg]

I don't see any point in measuring only S11 of a component which is only useful as an amplifier.

Should measure all S parameters, and derive stability factor and available gain. And that should be done the absolute minimal external components (no emitter degeneration, high impedance chokes on base and collector, etc).

[MisterHeadache]

I don't see any point in measuring only S11 of a component which is only useful as an amplifier.

Should measure all S parameters, and derive stability factor and available gain. And that should be done the absolute minimal external components (no emitter degeneration, high impedance chokes on base and collector, etc).

Valid points for a more thorough evaluation.  In my case, I was doing a first-pass evaluation what (if any) modern SMT transistors could be used to replace obsolete TO-39 transistors in RF amplifiers (more specifically in the Amateur HF frequency spectrum), and how much difference in performance to expect. 

I've uploaded part #2 which includes my test results:
https://youtu.be/A1avYYQc9fo