BFU590G Troubles

[jamfletch]

Hi all,

I've made a common base amplifier test board (You can read more about it here : https://github.com/jamesfletcher22/NortonAmps/tree/main), which I'm using to compare different topologies of amplifiers for use in low noise HF applications.

One thing I've discovered is that the BFU590G transistors I'm using seem to stop functioning very easily. They don't blow up in an extroadinary fashion, but instead will draw maybe 4x the current (from about 5mA ->20mA) in their failure condition. They also exhibit low gain whenever they reach this condition.

These are high Ft devices, and I'm just wondering why they're so sensitive? It seems I will get random failures for some reason or another. Has anyone had experiences similar to this?

[G0HZU]

Have you ruled out instability? These devices can go unstable and oscillate up at many GHz, possibly beyond the limits of your test gear.

The common base topology is very prone to instability up at UHF even with fairly mundane BJTs.

Do the devices stay in the failed bias condition if you test them with just resistive biasing and no other RF related components in the circuit?

[G0HZU]

Your PCB layout isn't really tight enough for fast BJTs like this. There are some fairly long connections to the BJT in some cases. This may cause stability issues.

I've got a basic RF amplifier here somewhere that uses a BFU590 but this is a common emitter amplifier with feedback. I don't remember having any issues with failure modes or instability even though I used non SMD parts and built it ugly style on a sheet of bare copper.

If used in common base, it's usually wise to include a parallel RL suppression network somewhere when using fast BJTs like this. Often this can be placed at the collector. Sometimes this isn't needed though.

[KE5FX]

I've got a basic RF amplifier here somewhere that uses a BFU590 but this is a common emitter amplifier with feedback. I don't remember having any issues with failure modes or instability even though I used non SMD parts and  because I built it ugly style on a sheet of bare copper.

There, fixed it for you. 

[G0HZU]

I've got a basic RF amplifier here somewhere that uses a BFU590 but this is a common emitter amplifier with feedback. I don't remember having any issues with failure modes or instability even though I used non SMD parts and  because I built it ugly style on a sheet of bare copper.

There, fixed it for you. 

Building it ugly style on a copper plane does make it much quicker and easier to model (for stability). I used basic models for the components and the manufacturer's s-parameter data for the BFU590 for this.

This method of construction obviously doesn't guarantee unconditional stability though... it's easy to (deliberately) design an unstable amp using this construction method

[RFDx]

One thing I've discovered is that the BFU590G transistors I'm using seem to stop functioning very easily. They don't blow up in an extroadinary fashion, but instead will draw maybe 4x the current (from about 5mA ->20mA) in their failure condition. They also exhibit low gain whenever they reach this condition.

The transistors didn't fail, they are oscillating somewhere in the UHF region. A spectrum analyzer would show you where that is happening. You have very long PCB traces  around the amp, at least the base is somewhat short and tight. The single ferrite bead in the collector isn't effective enough to keep the amp stable. As you have access to ADS, include all the traces as they are on the PCB and find out the frequency range where S11 is outside the Smith chart.

Common measures to prevent instability are Zobel/Boucherot networks from emitter to ground, from collector to emitter, ferrite beads or resistors in series with base and collector.

[G0HZU]

To get some idea as to high in frequency a small signal BJT can go unstable, it's best to measure one on a 2 port VNA at various operating points. I've done this with my VNA many times with BJTs, JFETs, MOSFETs and some ICs up to many GHz. The VNA data can be processed to calculate the k factor (and B1) and this can be plotted against frequency.

This can give some surprising results...
A genuine 2N3904 at 10V 10mA can go unstable and oscillate up to about 1000MHz.
An MMBTH10 (SMD version of the MPSH10) at 10V 5mA can go unstable and oscillate up to about 2.2-2.4GHz.

I've not measured a BFU590 on my VNA but looking at the manufacturer's data you can expect it to be able to go unstable beyond 5GHz, maybe even beyond 6GHz.

[jamfletch]

Thanks all, I'll try and have a look on my Tiny SA at the amplifiers that are causing me trouble at the moment and see if there's any oscillations higher up in frequency.

[profdc9]

For common-base/common-gate (CB/CG) and common-collector/common-drain amplifiers (CC/CD), you have to be careful about preventing oscillations.

The junction capacitance of the base/gate and emitter/source and lead inductance can form a series LC circuit that causes the voltage on the base/gate to oscillate. 

A good way to minimize this is to ensure that the capacitor that grounds the base/gate (CB/CG) or collector/drain (CC/CD) (depending on configuration) has as short of leads or connection as possible between the ground and base/gate (CB/CG) or collector/drain (CC/CD)  Also, a small damping resistor, say 10 ohms or so, can be placed in series with the emitter/source (CB/CG) or the base/gate (CC/CD) with as short of a lead as possible as well, as stray capacitance from the lead to the ground plane can also be a source of oscillation.   The components on the side of the base/gate and collector/drain should be as separated as possible (180 degrees apart) as to not couple together if possible.

Similarly, for common emitter or common source, the emitter/source lead should be as short as possible and connected to the ground plane as closely as possible.

In some cases, even vias, which can have an inductance of 1 nH or even more, can be too much.  Using many vias in parallel and a larger area of the top of the board for ground helps reduce the inductance.  In extreme cases, you might have to think about using structures such as radial stubs around the leads to create a ground point at the lead as well.

[G0HZU]

Also, a small damping resistor, say 10 ohms or so, can be placed in series with the emitter/source (CB/CG)

Can you please explain why? It's just that I'm not sure that's going to achieve much for a common base or common gate amplifier in terms of improving stability.
Normally, the best place for a series resistor would be to put it in series with the collector for a common base amplifier. This should improve the stability and it can often achieve unconditional stability at all RF frequencies.

[mike449]

To quench the oscillations, I've used 0805 SMD ferrite beads placed in series with either base or collector (NOT emitter) of BFU550AR (same process/series as BFU590, lower power), right next to it. The layout is critical. 10mm trace is too long. A solid ground plane on the bottom side, with lots of vias to the top ground pieces is a good idea.
TinySA doesn't show the oscillation as a peak, but as a greatly increased noise all over the spectrum, down to DC.

[profdc9]

I'll explain my reasoning for putting the resistor on the emitter of the common base.

A common base typically already has a large resistance on the collector to provide voltage gain.  So it seems to me that adding another isn't likely to change things much.  One might add a series resistor to the output of the amplifier to minimize the capacitive loading of the collector though.

One can add a resistor to the base, which is effective at eliminating oscillation, but with enough resistance one reduces gain and no longer has a common base amplifier.  The added resistance adds to the base resistance rbb and so adds noise as well.  So this is not ideal, and it is better to just get the base bypassed to ground as effectively as possible rather than resort to this.

The emitter is part of the LC junction that oscillates and itself has a lead inductance and a capacitance to ground.  A resistor close to the emitter lead can isolate the emitter lead from the trace that goes to the signal source, reducing the effect of stray capacitance at the emitter.  It is a tradeoff, however, as it increases the input impedance.  It probably won't help much if the layout is already compact.

I found these papers on the subject that might be helpful as well

http://www.iceamplifiers.co.uk/people/papers/analytical.pdf
http://www.iceamplifiers.co.uk/people/papers/transistorisedtia.pdf

Of possible interest, I made a project on github which is several RF circuits using a UHF transistor, the 2SC3356 which is similar to the BFU590G

https://github.com/profdc9/RFUtilityKnife/

[G0HZU]

Your explanation doesn't make a lot of sense to me. I'd also recommend that you analyse a common base amplifier using the proven and accepted techniques that date back many decades.

The best place to start is to analyse the s-parameter data for the BJT and compute the k and B1 factor for all frequencies and then use stability circles on the smith chart to look at what external circuit elements can cause oscillation at the problem frequency if a given circuit is showing instability.

You should find that the common base amplifier generates a LOT of negative resistance at the collector due to the internal feedback capacitances and the package inductances. It really doesn't matter much what you do at the emitter connection, as the common base amplifier (that uses a reasonably fast BJT) will typically generate negative resistance at the collector over a huge bandwidth. This can easily cover 50MHz to 2GHz or more. This negative resistance is usually in series with a small capacitance so you just need some stray inductance in the collector to create a UHF oscillator.

In other words, if you want to achieve unconditional stability across a huge bandwidth, the best place for a stopper resistor is usually at the collector. I don't think you can achieve this with a series resistor at the emitter. You can tame an amp with an emitter resistor at a known problem frequency but you won't be able to achieve unconditional stability over a huge bandwidth with a series emitter resistor.

If you want to make an amplifier for general bench use where you might want to connect it to various circuits it's best to try for unconditional stability over a huge bandwidth.

[G0HZU]

In the case of the Norton/Lankford amplifier, things are a little more complex because it uses a transformer to create feedback. This makes it harder to analyse for stability as an accurate model of the transformer would be needed up to many GHz. This would typically require a 4 port VNA to model the transformer.

I still think that an RL network at the collector ought to be effective at suppressing any oscillations up in the GHz region.

I've got a 2SC3355 based Norton amplifier somewhere here but I can't remember where it is. I didn't have any stability issues with it but I really only made it to see what noise figure and IP3 performance I could achieve.

[profdc9]

All I can suggest is that you might want to read the articles I referred to.  They say that a rough guide to when oscillation occurs is when the time constant of the base lead inductance and base spread resistance is greater than 1/fT of the transistor.

They don't specifically mention adding emitter resistance.  That is only something I have found by using the technique of seeing how the circuit oscillates with nearby capacitance (waving my hand near the circuit and seeing the changes) and trying to quench it.  But it may not work for you.  The main thing they said about it is:

"Once the oscillation frequency has stabilized, it is sometimes possible to lower the frequency of oscillation by bringing an object close to the circuit. This principle underpins several musical instruments including the Theramin [41]. The emitter node is particularly sensitive in the common base amplifier configuration. The magnitude of the frequency variation is dependent on the magnitude of the additional capacitance to ground formed by the object, with respect to the capacitances already present in the circuit."

They don't mention adding resistance at the collector at all (section VI of the first paper):  "Stabilization of an amplifier by increasing the collector – base capacitance or load resistance is unlikely to be practically helpful."  The main technique they mention is to add resistance at the base (a base stopper), but also advise that has its own problems which I mentioned.

I think there's a lot of layout issues that contribute to oscillation, which they do go into depth about, especially as the prime contributor to oscillation is inductance at the base lead.  That the loop inductance of the loop that includes the base lead to the ground, the emitter-base junction, and the path of the signal from the source (the trace and ground plane)  is the prime contributor to oscillation as its that junction that oscillates.  While feedback capacitance from the collector can contribute to this, it's mostly outside this loop.  The paper derives conditions for negative resistance looking into the emitter, not at the collector, as the negative resistance of that junction is what causes oscillation.

I must admit that I didn't understand this nearly as well until I read this paper, so it might be worth a look.

[G0HZU]

They don't mention adding resistance at the collector at all (section VI of the first paper):  "Stabilization of an amplifier by increasing the collector – base capacitance or load resistance is unlikely to be practically helpful."  The main technique they mention is to add resistance at the base (a base stopper), but also advise that has its own problems which I mentioned.

The usual technique involves adding a series resistor at the collector. However, to get unconditional stability over a the full operating range of the transistor, a shunt resistor of several thousand ohms can also be added at the collector followed by the series (33R-100R) resistor at the collector. A suitable inductor can often be placed across the series resistor and this effectively shorts out the resistor for in band signals. Up at UHF the reactance of the inductor is much greater than the resistance so the series stopper resistor will still be effective up at UHF.

This technique has been used with common/grounded grid tube amplifiers for over 70 years and it is also used with common base BJTs and common gate FETs. I'm not proposing anything new here...

[G0HZU]

To give a practical demonstration, see the circuit below. This probably doesn't look like a UHF oscillator to most people but the circuit below demonstrates the inherent stability issue that the CB amp suffers from.

Note that the emitter connection effectively already has a 1k series resistor in it. Yet the circuit will still easily oscillate up at UHF because the CB amp produces negative resistance at the collector across a huge bandwidth. So all you need to complete the oscillator is to add a suitable inductance at the collector and this is L1 in the diagram below. For oscillaion up around 1GHz L1 can have a value of maybe 10nH up to 20nH or higher. If you keep increasing L1 the oscillation frequency will reduce until the resonance occurs where the BJT is no longer able to produce adequate negative resistance. It will then stop oscillating. This will usually be somewhere in the VHF region for a fast BJT.

If you were to AC short the input (at the emitter) the output would no longer be negative, it would show something similar to the admittance seen on the datasheet. However, if you were to slowly release the short at the input and explore input resistances of 2 ohms, 5 ohms, 10 ohms and higher, you would see the output impedance rise dramatically, and eventually it would go negative and it would probably do it from maybe 50MHz to several GHz for transistors like the BFU590.

There are equations for this stuff.

Don't get distracted by the 1nF caps in the diagram below. Treat them as a perfect AC short if you like. The circuit will oscillate because the BJT will produce a negative resistance in series with some internal capacitance and this resonates with L1. Note also that the PSU at the emitter resistor is a negative supply, -5V in this case.

With a fast BJT like the BFU590 you can probably solder the base pin direct to a copper ground plane and it will still oscillate because of the internal inductance and capacitances of the BFU590.

To stop it oscillating, you can put a small series resistor at the collector. You can also experiment with a resistor in the base or try other things, but the collector resistor is very effective at cancelling the negative resistance generated by the BJT at the collector and it can usually do this over a huge bandwidth.

[profdc9]

I'm trying to simulate what you have there to understand what you're getting at.

I understand with say, no resistance whatsoever, like you might use with a plate choke for a common-grid amplifier, you need some resistance to damp the oscillation.  Here's an example with a BFU590G I did simulating a common base with an inductor to the supply.  Clearly there is oscillation, as one has the oscillation of the collector-base capacitance and the inductor.  The transient analysis shows the rise of the oscillation.



But it doesn't really make sense to do this, I think.  A common-base usually has a resistor at the collector, and then you don't get this sort of oscillation.  You could use a combination of the two, which is perhaps what you're proposing?



But if I add inductance to the base lead, we can clearly see the oscillation starting, which is the prime issue with trying to do common-base amplifiers.  I chose the value based on the information in the SPICE model file.



You can see the resonance peak after the base inductance is added.  This is the primary problem with common base I think, that the base grounding is imperfect.  An effective way to get rid of this is to add a base stopper resistor:



However, this has a noise and gain penalty.  Adding an emitter resistor lowers the gain, adds noise, and raises the resonance frequency.



So you can decide what kind of situation you think is best.  I included a zip file of the Qucs-S simulation if you want to play with it, including the BFU950 spice model.

[G0HZU]

This is the primary problem with common base I think, that the base grounding is imperfect.  An effective way to get rid of this is to add a base stopper resistor:

Yes, it's usually the package inductance in the base. Often, this inductance isn't included in a Spice model.

It's generally best to explore this stuff using s-parameter data from a real device. Try plotting the K (and B1) factor for a common base amplifier with just the BJT in the circuit. This is much easier if you use s-parameters.

Then try and get the K factor above 1 by adding a series emitter resistor. It won't really help much. Then try adding a base stopper resistor. This can help but it will reduce the GMAX gain you can expect from the device up at VHF and it can degrade the noise figure.

Then remove these resistors and try a shunt (1k?) resistor at the collector followed by a series (100R?) resistor at the collector. This should improve the K factor and you might achieve unconditional stability at all frequencies. If not, you can then try adding a really small base stopper resistor to get it over the line where K is >1 at all frequencies. You can also try a series emitter resistor to try and get it over the line but I usually try and do it all at the collector if I want good gain and noise figure from the amplifier.

This is much better than trying the suck it and see approach by adding resistors to cure an existing instability. It's best to try and make it unconditionally stable over as wide a bandwidth as possible.

[G0HZU]

For example, see the simulation below of a little MMBTH10 BJT. I've added resistors at the collector to get K above 1 at all frequencies. So this amp should be unconditionally stable regardless of what matching elements I then choose to add at the output ports.

The GMAX plot shows I can expect just over 16dB gain at 300MHz if I put a (low loss) tuned network at the input and output at 300MHz.

If I try and do this using a base stopper resistor and a series emitter resistor I can't get anywhere close to unconditional stability at all frequencies and the GMAX number plummets as well.

Of course, you don't 'have' to achieve unconditional stability at all frequencies and if the amp does go unstable when built, you can add stopper resistors at the base or the collector or the emitter and they might be successful at taming the instability.

The problem with adding them at the base or emitter is that you might then want to connect the amp to different circuits or you might want the amp to remain stable when the ports are left unconnected or with shorted inputs or outputs. That's when it's good to design the amp to be unconditionally stable at all frequencies

Pin 1 of (two port s-parameter model) SP1 in the simulation below is the emitter and pin 2 is the collector of the BJT. The ground pin is the base. This is directly grounded in the schematic but the model includes the internal package inductance of the BJT. So the base isn't perfectly grounded even though it appears so in the simulation below.

[G0HZU]

To further show what I mean, look at the alternative simulation below. Here I've tried to achieve unconditional stability using a base stopper resistor and a series resistor at the emitter. Not only will the noise figure be degraded, you can see that the GMAX number at 300MHz is already lower and the amp is still nowhere near unconditional stability as K crashes to way less than 1 below about 1500MHz.

If I increase the resistances to 100R then the GMAX number dips below 10dB at 300MHz and K still crashes below 1 at frequencies below about 700MHz. So not much improvement despite the GMAX now dipping below 10dB at 300MHz.

[G0HZU]

The other thing to bear in mind is that your linked article about common base amplifiers uses fairly crude models and these probably fall apart at frequencies way below where a BJT can actually oscillate.

The little MMBTH10 can oscillate at frequencies much higher than its quoted Ft frequency. The plot below is of the K factor of the MMBTH10 (in common base) at 7Vce and 5mA Ic taken using my VNA and bias tees up to 3GHz.

This shows that K doesn't recover above 1 until about 2.5GHz. This is about three times higher than the typical Ft (unity current gain) frequency stamped on the datasheet.

So in theory at least, a common base amplifier based on the MMBTH10 can oscillate up to about 2.4GHz or so. Realistically, the only way to model and analyse this is to use VNA derived small signal s-parameter models as below.