Building A VHF Amplifier For Valve Sets

[Herald1360]

I've heard the upside down DIL pack technique referred to as "dead bug" construction. For isolated mounting pins you could use upside down DIL sockets too.

[G0HZU_JMR]

I've taken a quick photo of the LNA PCB. See below. Unfortunately, it still has the SMD MMBTH10 transistor fitted rather than a TO-92 packaged device but you can still get the idea of the construction. I'm afraid it is a bit scruffy and wobbly because I quickly tacked it together. Ignore the Coilcraft metal screening can as this was part of the L match (not fitted).

You can see that the connections are shorter than the connections on your board and the whole thing is soldered on top of a piece of unetched PCB material and this acts as the 0V ground. The connections are still a fair bit longer than they could/should be but this is just about adequate. I also used old school ceramic disk caps and leaded metal film resistors. Normally I would do all this using tiny SMD parts on a milled/etched PCB.

Note, however that I have deliberately added a piece of wire in the emitter connection of the SMD transistor to ground. This is there to add some inductance and this introduces degeneration at the emitter and this helps improve the input match.

Also included is the result of a stability analysis from 5MHz to 4GHz. The dark blue trace is the computer model of the BF199 when in a simulated LNA circuit and the paler blue is the data imported from the VNA after it measured the MMBTH10 LNA PCB pictured below. The agreement is very good!

As well as finding some 2SC2026 transistors today I also managed to find a very old Philips non-linear model of the BF199 and this is used in the simulation of the LNA circuit to create the model data below. The main goal is for K to be above 1 (and B1 to be above 0) for unconditional stability.

You can see that both the model and the real LNA fail to meet this requirement across about 10MHz to 70MHz. However, by fitting a shunt 390R resistor across the output terminals this issue is resolved and K goes above 1 across all frequencies. I was hoping this would happen as I'd seen similar results with the 2SC2026 model.

The agreement between the BF199 model and the MMBTH10 LNA PCB seems remarkably close but these are different transistors so this is just a bit of a coincidence I think. I think the reason the stability margin for K is better on the MMBTH10 up near 4GHz is because the ceramic disk caps used in the real LNA PCB will have quite a high ESR up at a few GHz and this will be a bit like having small series snubber resistors at the base and the collector. I haven't modelled this high ESR trend in the computer model so the simulation shows a dip below 1 for K near 3GHz.

[paulsherwin]

I would expect the BF199 and MPSH10 to produce near identical results. They are essentially different manufacturers' interpretation of the same thing, cheap mass market VHF transistors for general use.

[G0HZU_JMR]

See below for a comparison of gain and input return loss. They do seem to be very similar! The simulator predicts a noise figure of just below 2dB across 88-108MHz for the old Philips BF199 model when driven from a 50R source at 5mA Ic and this does seem a bit optimistic. The datasheet suggests 2.5dB NF at 35MHz at 4mA Ic when driven from a 50R source. But these are small differences I suppose.

Note that if I change my estimate (for the computer model) of the inductance of the input wire connection from 23nH to 19nH the S21 and S11 plot agrees even better. See the second plot below.

You can glue DIL ICs on, upside down

I find that confusing and I can't see the number, bending the legs to point upwards and then sticking the IC down solves both problems.

[G0HZU_JMR]

To start with it would be worth trying to keep the antenna element a reasonable distance away from the amplifier and also the receiver. It's unlikely that there would be any regenerative feedback if they were all too close together but the connection from the amp to the receiver should be screened and the antenna should maybe be a metre or so away to start with.


I also managed to find an old BF199 to play with. It's very hot here this evening so I'm unlikely to have a play straight away. However, it would be nice to formally test the simplified version of this amplifier with and without the 390R shunt resistor at the output. I can test it for gain and match and noise figure and output 3rd order intercept point. Maybe this is all a bit OTT for such a simple amplifier but it will give some idea as to what to expect from it. It isn't going to handle large signals very well and if there are large signals nearby there could be all kinds of strange overload effects including strange noises and distorted reception.

[Radio Wrangler]

Originally Posted by merlinmaxwell

I find that confusing and I can't see the number, bending the legs to point upwards and then sticking the IC down solves both problems.

Ah, but the good ones had the type also written on the bottom!

They must have seen me coming.

David

[Thyristor]

I built something very similar many years ago; sadly, both it, and any notes are long gone; although I'm pretty certain it was configured as grounded base.

Seeing Restorastion73's post (no20) I would suggest the OP considers the PW design; anything designed by Ian Hickman is guaranteed to work. And NB the construction layout.

[paulsherwin]

Originally Posted by G0HZU_JMR

I also managed to find an old BF199 to play with. It's very hot here this evening so I'm unlikely to have a play straight away. However, it would be nice to formally test the simplified version of this amplifier with and without the 390R shunt resistor at the output. I can test it for gain and match and noise figure and output 3rd order intercept point. Maybe this is all a bit OTT for such a simple amplifier but it will give some idea as to what to expect from it. It isn't going to handle large signals very well and if there are large signals nearby there could be all kinds of strange overload effects including strange noises and distorted reception.

There's no harm in doing lots of analysis if you have the appropriate expertise and test gear. As you say, it may not help the OP much, but it's always interesting to hear these simple circuits given a comprehensive analysis by someone who knows what they're talking about. Me, I just build 'em and see if they work or not

I don't think the strong signal issue is going to be a problem in this case.

[G0HZU_JMR]

OK thanks, I'll probably have a look tomorrow. The forecast predicts it will be slightly cooler. My workroom gets very hot in the warmer months if I have test gear on.

Seeing Restorastion73's post (no20) I would suggest the OP considers the PW design; anything designed by Ian Hickman is guaranteed to work. And NB the construction layout.

Yes, that PW article demonstrates the groundplane construction technique really well. Much better than my fuzzy photo.

After a quick look at the schematic I don't think that common base amplifier will be unconditionally stable because there doesn't seem to be any resistance in the output network apart from the resistive losses in the lumped components. I would have expected to see a resistor across L1.

The in band signal handling will be a bit better than the common emitter amplifier described in this thread but not by much. The common base amplifier does offer some selectivity against out of band overload though. It's arguably an upgrade compared to the simple little CE amplifier but I think there should be a damping resistor across L1...

[G0HZU_JMR]

I had a go at testing a BF199 in the circuit earlier this evening and the results weren't quite as expected. The BF199 I have here isn't made by Philips or Motorola so this might be why it seems a bit below par.

See below for the gain and noise figure test screenshot. It managed just over 15dB gain and a noise figure of about 3.8dB across 88-108MHz. To show that test gear drift was not the reason for the extra noise I fitted a 2SC2026 after the BF199 and the 2SC2026 showed >19dB gain and a 1.8dB noise figure which was in full agreement with the model I have here.

I also tested these transistors and the MMBTH10 (MPSH10) amplifier for two tone IMD. The two tone plot below is for the MMBTH10 but they were all essentially the same. This agrees very well with the non linear models in Genesys and it shows the output 3rd order intercept point is about +10dBm. The reason the IMD performance is relatively poor is because the collector current is only about 5mA and the collector is fed directly into a 50R/75R load impedance that represents the input impedance of a typical VHF FM receiver.

[G0HZU_JMR]

To give some confidence in the test gear I measured a MMIC amplifier module from MiniCircuits for gain and noise figure. This is an official evaluation amplifier module supplied by MiniCircuits built into a PCB and metalwork complete with SMA RF connectors.

The typical data for gain and noise figure from MiniCircuits is given below in a table and I measured this amplifier again this evening. I've been using this little gain block amplifier for many years as a reference for my various noise figure test setups as it has fairly flat gain and noise figure across 10-1000MHz.

It usually measures 18.1dB gain at 50MHz and this dips to 17.5dB by 1GHz. The noise figure gradually changes from 3.6dB to about 3.8dB by 1GHz and the agreement on this particular noise figure test setup is very good indeed across 10MHz to 1GHz.

One thing I did spot when looking at the PW common base amplifier is that the parts list doesn't agree with the schematic. I think it was C2 that is 12pF on the schematic but it is listed as 1nF in the parts list. It is obviously meant to be 12pF and the parts list is incorrect.

[Radio Wrangler]

346 or N4000 series noise source?

David

[G0HZU_JMR]

It's an Agilent 346A noise source with an ENR of about 5.7dB. The ENR data vs frequency gets manually loaded into the analyser memory (by the user) and can then be called up as a cal file. I also have a homemade noise source using a Noisecom diode that I designed to work with this type of analyser and it too has an ENR cal file. This works extremely well and usually agrees within about 0.2dB for noise figure. Often the agreement is much better than this and (from memory) the hot/cold VSWR of the homebrew source is something like 1.03:1 up into UHF.

When I tested the 2SC2026 version of this little VHF preamp I also did a wide plot from 10MHz out to 1GHz. I also changed the analyser display to full screen with a combined plot. This is a bit easier on the eye and the result below may be interesting. It does show that it can still deliver about a 2dB noise figure and useful gain up to about 400MHz. A tuned design would probably give reasonable performance at 432MHz although modern PHEMT devices are much better of course.

[Radio Wrangler]

Originally Posted by G0HZU_JMR

It's an Agilent 346A noise source with an ENR of about 5.7dB.

Ah that's good. The 5.7dB ENR ones have more attenuation leading to the output than the 15dB ones, so the impedance presented to the amplifier is more stable between on and off conditions.

Some of the 346 units are very different to the others.... they have a different diode which needs the noise source to contain a little inverter to reverse the polarity of the 28v supply!

Not unsurprisingly they have different turn on-off timing.

I designed the later N4000 family, and to keep the proven RF assemblies for some of them, I wound up with the inverter in some of them. BUT because they were to go with a new series of instruments, I kept the 28v on and used a switch in the noise source to turn the diode current on and off cleanly. The unit has an EEPROM for cal data and contains a DAC so the diode current can be changed if necessary to get a better fit of CAL data. Oh, and there's a thermometer in it. Get a 6 degree error in assumed temperature of a noise source pod and you have 0.1dB error in noise figure result.

The noise fig set reads the current from the EEPROM and writes it to the DAC.
I don't believe we ever had to vary the setting as it turned out.

David

[G0HZU_JMR]

Sadly my analyser is too old to support the 4000 series as it only has the BNC output port that drives the noise source with a gated 28V. It can't do any data transfer from the noise source.

I've just done a quick experiment with the analyser and the 2SC2026 preamp and an antenna. The antenna is an indoor whip so hardly ideal.

The analyser has an internal preamp that gives the analyser a 4dB noise figure with zero attenuation. In the image below I've set the attenuator to 4dB so the analyser noise floor with a 50R termination at the analyser input should be -174 + 4 + 4 = -166dBm/Hz. You can see on the yellow trace that this is correct.

The turquoise trace shows what happens when the preamp is connected but with its input terminated in a 50R load. Note that this is the 2SC2026 preamp without the L match. This gives 19dB gain with a 1.8dB noise figure.

Therefore the noise level should go up to roughly -174 +1.8 +19 = -153dBm/Hz for the turquoise trace. In other words it should be about 13dB higher than the yellow trace. It's probably more like 14-15dB higher but some of that extra noise will be summed noise from the analyser.

The top trace is what I see when I add an antenna. It looks like the general noise level jumps up even higher. Probably 10dB higher. So it may well be the case that there isn't much point in trying for an ultra low noise preamp for use in my workroom unless the (valve FM?) receiver has a noise figure greater than about 12dB.

Maybe If I did the experiment outside in a rural location the result would be better. I think galactic noise is still an issue at 100MHz so there wouldn't be that much of an improvement. Maybe 6 or 7dB at a vague guess?

[Bazz4CQJ]

Originally Posted by Radio Wrangler

RF engineering for some reason seems to have acquired a reputation as the darkest of dark arts. Odd, that. It merely means that you have to learn a number of things that explain why non-RF people have a lot of trouble

David

But it's all got sooo complex! As the Age Poll showed, many on here (like me) started messing in the 1960's and things were so much simpler then. But the progress in electronics over the next 60 years was amazing, and if you went away for a while and came back, it was like starting from fresh. When the conversations started about FET buffer amps showing "negative resistance", I knew I that I should stay at the 'shallow end of the pool' and wear some water wings.

B

[G0HZU_JMR]

But it's all got sooo complex!

But for the discretely designed circuits that most of us (on a vintage radio forum) are interested in, not much has changed really. However, the analysis tools (eg network analysers and RF simulation tools and computers) have improved a lot and this aids understanding so I would argue that basic RF amplifier circuits are now a lot easier to understand than they were back then. This makes it much easier to appreciate how they can misbehave.

Go back >30 years and many radio amateurs would blindly tinker with an RF amplifier project until it stopped oscillating. It would be a suck it and see approach with damping resistors, screening walls and a few swapped components. It would then be declared stable, conclusions would be made and they may even pass on hints and tips to others.

Since computers became cheaply available it has been possible to quickly model and predict how and why a circuit will go unstable. Sadly, the basic skills required for this are rarely being taught (especially with practical examples) and a lot of RF teaching techniques are decades out of date. Pick up a typical book on amplifier instability and you will be presented with pages of equations and not much text. Practical examples are rarely given.

[Radio Wrangler]

It's a common fault in teaching. Throw the mathematical proofs at 'em before showing them how something can be used and what it will do for them. This will get their eyes glazed over and minds wandering elsewhere so if any do hear the what it's for bit, they've missed too much to pick it back up.

I'm afraid that all the people I know who can make the theoretical side of RF engineering sit up and do tricks were all self-taught. They learned it despite the educational system rather than via it. They found that some things made life easier and some good tricks possible, and they had to work backwards to find the maths, then it all started to make sense There are a few exceptional courses, but they are rare and very valuable.

I had a lecture to give at Dundee uni, pretty much a full morning to the gathered engineering department, different areas of engineering and various years all mixed. The subject was designing as a profession.

At one point I was talking about the Quad 303 power amplifier as an example. It was designed around the 2N3055-variant output devices contributing a dominant pole in the feedback loop of the 'Quad Triplets' and how the Ft of these parts had been progressively improved, while the Fts of the TO5 and TO18 transistors grouped with them had not kept pace in Ft improvement ratio, so the 2N3055 was no longer so dominant. Phase shift was now an issue and the amplifiers hooted at a few MHz after repair, but only with signal on.

So you could no longer buy replacement transistors BAD enough to be sure to work.

I illustrated this with a pole-zero diagram, sketched on a root-locus and a bode plot to show what was happening and how the simple little triplet circuit made a perfect trap. How you couldn't fix it with an added capacitor, that created another pole, it did not move an existing one. Life can be quite unfair.

This had Brian, the Head of Department pointing at the pole-zero sketch, bouncing up and down in delight and declaiming loudly "See! People do use this stuff!"

I was helpless, of course.

The point I'd been making was the danger of using parts with open-ended specifications. You get the minimum gain and the maximum Ft, but then the maxima can bite you. You get this in most parts one way or another. You can't avoid it, but you need to be aware and make some margins to get some safety.

David

[Skywave]

Originally posted by Bazz4CQJ:
"But it's all got sooo complex!"

Originally Posted by G0HZU_JMR

The analysis tools (eg network analysers, RF simulation tools and computers) have improved a lot and this aids understanding, so I would argue that basic RF amplifier circuits are now a lot easier to understand than they were back then.

Yes, such circuits are indeed now easier to understand how they work by using the tools you have listed. But that's the rub: those 'tools'. Learning how to use all those tools and becoming proficient with them is a large new chapter for many and can be a decidedly daunting prospect, even if one's 'hobby budget' stretches far enough to afford their purchase. Of course, if such a user is professionally employed in this field as is using such tools every day, then that 'learning curve' for hobby applications simply ceases to exist.

Al.