EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: rs20 on February 24, 2017, 09:25:49 am
-
Hi folks,
I've been messing around with a SPICE model (straight from NXP) of the BFU590G (http://www.nxp.com/documents/data_sheet/BFU590G.pdf), an NPN wideband silicon RF transistor. As you can see from the image at the end of my post, I've configured it in a super-basic emitter follower configuration (apologies for the box symbol instead of normal NPN symbol), following a 27MHz signal, and as you can see it's singing like a gigahertz rooster on meth.
I do get an error Unrecognized parameter "imax" -- ignored associated with the transistor, but I'm assuming that's unrelated since I don't care about testing for maximum currents and LTSpice isn't capable of that anyway (which would nicely explain why LTSpice is rejecting the paramater.)
1. Does this look like a believable outcome, or a bad/misconfigured simulation?
2. How does one apply the Barkhausen criterion to a circuit without any loops?
3. Now I appreciate that it's unusual to apply a 8.5 GHz fT transistor to a 0.027 GHz usecase, but is there anything about the transistor that makes it fundamentally unsuitable for low-frequency operation? (Beyond just not being particularly well-optimized for the task).
My reason for using such a high fT transistor is that I want to disable the follower to leave the output node in a high impedance state with very low parisitic capacitance, and I believe the BFU590G has extremely low capacitance (as a side-effect, perhaps, of being designed to operate at microwave frequencies.) But I haven't even got past making a basic follower...
(https://rs20-static.firebaseapp.com/pano/follower.png)
-
All kinds of emitter followers have a strong tendency to oscillate if no suitable base resistor is present ...
So just add a resistor in series to the base and try again
-
Thanks, adding a base resistor as small as 11 ohms killed the oscillation. Does this type of oscillation have a name? What's the mechanism? I've found a few sources online (funny how much easier it is to find the answer when you already know the answer) that say "a small base resistor is often necessary to prevent oscillation", but they don't give further comment.
-
It is, as far as I know, parasitic oscillation and can destroy transistors.
It's a problem with RF MOSFETs too, the simple cure is the same, add a resistor to the gate but that's not a guarantee, with parts that can do GHz or even VHF you need to be very careful of board layout and construction because it's not unheard of to find you're emitting significant power at some very unexpected frequencies.
It's better to choose a more suitable transistor (lower Ft) if possible but the number of Bipolar RF transistors available these days makes that less simple.
If you really only need half a watt at <30MHz and can use through hole parts then there are many CB transistors available, there's even the BD139 which is a dirt cheap part and is capable of a couple of watts if you heatsink it and drive it properly.
-
1. Does this look like a believable outcome, or a bad/misconfigured simulation?
If you look at the spice model, you'll se the package parasitics added. I think the main culprit is the line "Lb_wire 4 5 1500p ", which adds a tiny (1.5nH) base inductance due to the base lead wire. This causes gain peaking of the emitter follower near the GHz range. If you simulate a regular ltspice 2n3904 and add that tiny inductance to the base, you'll see a similar gain peaking in the AC analysis. In this case the peak in gain is curbed by the feedback capacitance of the transistor, so the faster the transistor, the greater the peak gain.
2. How does one apply the Barkhausen criterion to a circuit without any loops?
There is feedback from emitter to base. Since the transistor has all the lumped package parasitics added, there is a lot going on between the base and the emitter terminals. The UHF gain added to the feedback to the base through the parasitics could allow oscillation at some very high frequency.
My reason for using such a high fT transistor is that I want to disable the follower to leave the output node in a high impedance state with very low parisitic capacitance, and I believe the BFU590G has extremely low capacitance (as a side-effect, perhaps, of being designed to operate at microwave frequencies.) But I haven't even got past making a basic follower...
I'm not an expert in this area, but you could try with a PIN diode. They have very low capacitances and can work as RF switches. Here is a magnific youtube tutorial by w2aew about PIN diodes:
https://www.youtube.com/watch?v=XpYsCM_Wf50 (https://www.youtube.com/watch?v=XpYsCM_Wf50)
-
The "belt and suspenders" (American usage) approach to stopping parasitic oscillation with high-frequency active devices is to put a small resistor (maybe 10 to 50 ohms) in series with two of the three leads. Of course, any resistor in series with the emitter or collector (or cathode and plate for warmer devices) must be considerably smaller than that in series with the base (grid, gate) lead.
-
If you really only need half a watt at <30MHz and can use through hole parts then there are many CB transistors available, there's even the BD139 which is a dirt cheap part and is capable of a couple of watts if you heatsink it and drive it properly.
Watch out for the ST BD139s. They are a bit variable in Ft and they don't quote it on the datasheet.
-
Does the model include package parasitics?
The result may be reasonable. You've stuffed the transistor between two zero ohm impedances. Something physically preposterous, especially at GHz!
Try adding some series resistance, which in practice you will have to provide with ferrite beads or bypass cap ESR. Especially for the base terminal, which can be tens or hundreds of ohms (or a ferrite bead of the same value).
Tim
-
Thanks, adding a base resistor as small as 11 ohms killed the oscillation. Does this type of oscillation have a name? What's the mechanism? I've found a few sources online (funny how much easier it is to find the answer when you already know the answer) that say "a small base resistor is often necessary to prevent oscillation", but they don't give further comment.
The reference at the end of the 1993 Linear Technology Applications Handbook Volume II has a paper discussing the details and the math. With a suitable load on the emitter, the input resistance at the base is negative making it easy to inadvertently produce a negative resistance oscillator. MOSFETs used as source followers do the same thing. To prevent this, include lossy elements in 2 of the 3 leads.
Michael Chessman and Nathan Sokol, "Prevent emitter-follower oscillation," Electronic Design, 21 June 1976, pages 110 to 113.