RF Amplifier-Converting a pretty-close schematic to a PCB layout

[Xnke]

Moving forward in the TRX project, I'm looking at the RF amplifier. I need to get the PCB layout worked out and be fairly confident in it before too long, as the PCB will be sent out for production on a metal PCB substrate as the RF output devices are ST's PD55015E/PD55025E LDMOS devices-and are available in SMT only. I *could* mash the leads out and solder them to a copper block like other LDMOS transistors, but there really isn't much need to at these kinds of power levels.

Since I'm new to power RF amplifiers (well, solid state ones...the quad of plate-modulated GU50's keeps the shack warm) I was pleased to find that DL2EWN has already used the bigger version of these devices (same package) in pretty much exactly the configuration I am hoping to use, and left a schematic on the web for us.



Now, the bad-I speak and read very poor german, but I almost think I know what most of this is. He's using the 35W output devices, but I have 15 and 25W versions here. I plan to build two amps, using the same PCB, (since the mosfets are in the same package, they'll fit all three output ratings) and over the same range (1.8 to 52MHz) that this amp was built for.

I don't understand what that BD677 is doing in the bottom, other than acting as a strange zener diode of sorts, and I can only assume the two 1N5711's are used for temperature compensation of the LDMOS bias voltage.

I don't know, but I assume, that the doubled up resistor pairs at R1/2, and R3/4, then again at R15/16, and R17/18, are for power handling, and not a special RF technique.

He calls out 43 material cores for the input and output transformers, and gives turns counts on the input transformer. This is nice-although I have BN61-202 cores here, I am familiar with how to adjust for that. He does not give much info on the output transformer-just the 1000-16-43, which may be a core size? Not sure. I'm thinking it's likely a single turn  primary, and three to four turns on the secondary, when running at 30ish watts and 12-14V main supply voltage.

Before I make the first attempt at PCB layout, can anyone here help me out with the german notes, and what's going on in the bias circuit with the oddly connected transistor (I've only ever used them that way to create noise.)


EDIT: I found a photo of the actual PCB that this schematic belongs to, and hopefully it will help shed some insight on things.

[David Hess]

I don't understand what that BD677 is doing in the bottom, other than acting as a strange zener diode of sorts, and I can only assume the two 1N5711's are used for temperature compensation of the LDMOS bias voltage.

The BD677 base-emitter junction is for temperature compensation of the power MOSFET Vgs.  I think the idea is to use a tabbed transistor for this so that it can be easily thermally coupled to the power MOSFETs.

[Xnke]

That does make sense, but then are the diodes just there for another PN voltage drop?

[Twoflower]

Probably I can help a bit to translate it:
"von Steuerbaugruppe" -> from control unit
 "zur Tiefpassbaugruppe" -> to low-pass unit
"zur Anzeigebaugruppe" -> to display unit
"+8V (beim Senden)" -> +8V during transmit

I think the PCB you found is different. The part numbers don't match and the transformer has 2x two resistors in series.

[David Hess]

That does make sense, but then are the diodes just there for another PN voltage drop?

I do not really know but I was thinking that at high power levels, the schottky diodes rectify the gate drive to increase the idle current reducing distortion.

[KJDS]

The low value resistors are to prevent oscillations. I'd start with higher values and if it's stable but low gain then reduce the resistors.

[chrisl]

I don't know, but I assume, that the doubled up resistor pairs at R1/2, and R3/4, then again at R15/16, and R17/18, are for power handling, and not a special RF technique.

The pairs are used to reduce the stray inductance.  These resistors are to stabilize the amp.

[Xnke]

After going through and making sure that the pictured PC board is in fact the board that matches that circuit diagram, I can totally see why he doubled up resistors. They're 1/4W through-hole components, and thus have some lead inductance more than a surface-mount resistor would.

I've always built amplifiers working from the output stage to the input stage, and it's worked pretty good for me so far. In this case, the schematic is sound, and it *does* match that board in the first post. Stepping my way through one component in the schematic at a time, I am able to locate all the components in the schematic on that board. Of course there are a lot of extra bits on that board, but they're related to the power input/filtering/regulation and input/output switching.

Still, it doesn't hurt to work through the design and make sure I am not about to layout and order a dozen metal-clad PCB's, only to find out I'm a dummy and made a huge mistake!

In my case, the desired output is 30W for the PD55015E devices. This is pretty nominal output power for a push-pull pair, according to data I can find on these. Next, the supply voltage-I will be supplying 12.6 to 13.8V power. The voltage swing I estimate will be about 11.5 volts and That points me to a drain-to-source impedance of roughly 4.4 ohms. Each device exhibits about 50pF of drain-source capacitance. So, the output of the RF amplifier, as viewed from the primary of the RF output transformer, will look like 8.8 ohms in parallel with 25pF of capacitance-Am I correct so far?

Now, normally at this point I'd figure the output transformer. I'm going to use the methods I'd normally use-but I KNOW they are not correct for best performance in this kind of application. I'd like to learn how to do it correctly, so here goes.

The primary of the output transformer will need to match 8.8 ohms to 50 ohms. IIRC, that means a 1 to 5.68 impedance ratio. That would, IIRC, mean I need a turns ratio of 1:2.38, which is a bit interesting to wind so I figure that I'll just have to get close. Almost invariably, these output transformers are built with either a binocular core, or a pair of toroids glued together to form a binocular core. These just can't have a good centertap with just the single U winding, so I would have to have 2 turns on the primary (center-tapped at 1 turn) and 4.76 turns on the secondary. That's just as irritating to wind...

How should this be done? What's the "proper" way to figure this?

[dmills]

Ditch the centre tap and use a DC injection transformer wound on a small ferrite (Not iron powder) core instead, much better behaved in most designs.

Have a look at what 2:5 gets you, 8 ohms on the drains, not horrible, and the extra voltage headroom should make for lower distortion.

Po = 2*(Vd^2)/R = 33W which is near enough.

Your voltage swing calc is I think wrong, but factored into the above eqn, in that Vd should swing from somewhere near ground to somewhere near twice the supply voltage.

30W PEP means about 60W of DC input and around 30W of waste heat, do calculate the junction temperature rise due to the thermal resistance of the package, it can be an issue. 

73, Dan.

[in3otd]

Your calculations end up with the correct theoretical impedance ratio but you'll need to leave some headroom as the swing on the transformer may be even less due to the devices and the parasitic inductances.
So you can try with a 2-turns center tapped primary and a 5-turns secondary and see how the circuit behaves. I always try to have a 2-turn primary to avoid the additional DC injection core, it may have more leakage (not sure) but at this power level it could still be acceptable.
The PA in your initial post seems to be designed with quite some margins to me, e.g. the output transformer RF1000-16-43 is rated for 250 W PEP, https://www.communication-concepts.com/rf-1000-transformer/. You should be able to use something like a BN43-3312 for that, the 61 material has lower losses but I didn't check if you'll get enough inductance.

Interestingly, I was recently playing with a push-pull of AFT09MS015N which seem similar to the  PD55015, maybe with a somewhat lower maximum drain current. With a 2 turns/5 turns output transformer on a BN43-202 core at present I can easily get 30 W at 1.8 MHz, which go down to 25 W at 10 MHz and to 15 W at 30 MHz. The core does not seem to get too hot and interestingly its temperature is higher at higher frequencies, I think that's actually the losses in the transformer wire that increase due also to the compensation capacitance. Getting a non-transmission-line transformer with a good frequency response at high frequency, so not needing much compensation capacitance, is not too easy especially if one cannot use the usual brass-tube primary as here.
I also initially thought to do a metal-core PCB but for this prototype I've used a 0.8 mm FR4 with a full ground plane on the bottom side which is screwed to a big heatsink and it appears to be good enough, especially for SSB service. At 10 W out, so near the maximum PA dissipation, the devices case top temperature stays below 65 C so the junction temperature should still be ok.

[Xnke]

Well that demonstrates part of my problem with the conventional style output transformer-this amplifier will cover 3.5 to 54Mhz. I had planned on using 61 material for the core, and 43 material if I really had to get the inductance up. It is looking more and more like I'll have to use a transmission-line style transformer and a separate DC feed balun. You might try winding your output transformer on a 2861000302 core, if the BN43-202 sized core works but isn't so good at higher frequencies. It is slightly larger than the BN43-202, and is a 61-material.

Knowing that the transformer in the original schematic is good for 250W, I can use a much smaller transformer for the 30 and 50W design here. (just change the output device and bias voltage, to change the power output) The only reason I'm working with the PD55015 is because I have two of those, and two more PD55025. The same basic schematic and circuit board should work for both of these devices AND the PD55035 in the original schematic-they're all the same device family and same physical package. (Power SO-10RF)

I believe with a little optimization, the only significant change I need to make is in the thermal compensation transistor and it's placement. It's a darlington transistor to get two Vb-e drops, and twice as much tempco, in the same thermal package. I am sure there is a good surface-mount part that can be had.

[dmills]

Change bias, device AND Transformer ratio to change the output power.... It is really the drain impedance that sets the output power level, not the device (Which just needs to be butch enough to survive the drain current and voltage swing).

Personally I would go for a larger 61 mix pigs nose, something like a 2861010002, more likely to get a sane amount of magnetising inductance at low frequency with the larger part.

73 Dan.

[Xnke]

I have thought about that exact core, Dan. At 5 bucks, it's not too expensive, but it's also a 250W sized core. Roughly the same size as the core in the original schematic, the only difference it's a binocular core instead of two ferrite tubes side by side.

I guess I could just use the 2861010002, I mean even the price is the same between the 15W part and the 25W part!

[dmills]

It is not always about power handling.

The bigger part gets you more inductance per turn, lower flux density and better linearity even at low power levels, and unless you are space or very cost constrained (Or have gone so far that winding length is hurting high frequency performance) bigger ferrites seldom have any real downside in this application.

That core might be stupidly large for a 10W stage but it is available and gives you options when it comes to winding the thing, a 50W stage at 12V will be handling ~8A so wire sizes are not negligible. 

Lay the board out for the big core and buy a few smaller variants, a two tone test will reveal what you can get away with.

73 Dan.

[Xnke]

That's the plan now. I actually found a pair of 43 tubular ferrites, the same size as what's used at the transformer in the original design. It'll be good enough to start with for testing, and will make a good comparison for the multi-hole ferrite blocks.

Having the dimensions of the transformer available lets me finish up the PCB layout. Once I have a PCB, I'll wind up the first output transformer and start testing.