Push pull HF amplifier using UHF bipolar devices

[W4AMV]

Good day,

I have a number of old but good 80 to 100 W UHF bipolar transistors. Many of the devices use internal matching to assist in the external matching and increase the operating bandwidth. At HF this internal matching is of little value.  Nevertheless, the devices have significant power gain at HF and may be difficult to stabilize, but I am willing to try ! Has anyone on the FORUM looked at applying these broadband UHF bipolar devices to an HF push pull design ? The devices in the attached schematic are rated at 80W at 400 MHz. Simplified operation in push pull is shown. The input and output 16:1 and 1:16 transformers are not an ideal way to do push pull but they are a simple start. I have added feedback and loading in anticipation of fighting some instability. The design will run near class C, CW only. The bias control is regulated with temp sense feedback to the 723 voltage regulator. This is all standard items from the Moto app notes and bulletins. Any comments on this approach, pro or con, are appreciated.

73, Alan

[AB2EZ]

Alan

Hi! Looks like a fun project.

Something looks strange (to me):

There is a 22 ohm resistive load across the 1 turn (output) side of TLT-1. With the 16:1 turns ratio, the resistive impedance looking into the input side would be 16 x 16 x 22 = 5632 ohms. The transformer provides a 16:1 voltage step down input-to-output

But there is a network that presents a 67 ohm load directly across the RF input, and also provides a small voltage step down of 120/152 = 0.79

I suspect (perhaps incorrectly) that you wanted to 3-resistor input network to work in conjunction with TLT-1 to present a 50 load to the RF input signal... and I suspect that you didn't want to attenuate the RF input signal by more than 20dB.

Best regards
Stu

[W4AMV]

Hello Stu. Thanks for the response.

The 11 ohm each side of base is a first attempt to provide some RF input stabization (shunt loading). These values are subject to change ! The 16:1 is really operating on the base to emitter input R and on these big devices is usually an ohm or two or even SUB OHM. So for all intent and purpose, I suspect that the device base to emitter R will dominate. The R-pi network on the input is a 6 dB attenuator if I recall. All of the C values on the input, 18 pF, 1100 pF, are present to provide a near 50 ohm transform from that low B-E resistance of the MRF327 operating push-pull. Again, all subject to change as the design is tuned.

[AB2EZ]

Alan

Okay... I understand now. I forgot that the impedance from base to emitter would be very low for these high power RF devices.

Thanks!

Stu

[AB2EZ]

Alan

Since each of the transistors will be operating in class C, its base-emitter impedance will be mostly resistive in the "on state" (a low-value of R... as you pointed out), and mostly capacitive in the "off state". It is not clear that you can depend upon the base-emitter capacitance of the "off state" transistor to provide a low-enough impedance path for the base-emitter current of the "on state" transistor to return to the secondary of TLT-1.

Is 1000pF (to the right of L1) enough capacitance to provide a low enough impedance return path from ground to the center of the secondary of TLT-1?

Stu

[W4AMV]

Good catch Stu. No, it is not nearly enough. I probably will select a C value whose capacitance with its lead length and any PC trace length provides a "good" short circuit impedance over the HF band.

Thanks !

[W4AMV]

As a follow up, here is the card and updated schematic. The feedback and shunt loading remained the same. The shunt feedback R at 470 ohms. Not including the R-pad (~ 6 dB) at the front of the PA, the large signal gain is 19 dB and all seems stable. I have not tried to get ugly with the terminations, yet. At 0.5 W input the output is 55 W at 20 V and ~ 5 A. One interesting outcome is the output transformer. Was a 16:1 and now settled on a 4:1. The efficiency at 16:1 was quite poor. The load line at the collector way off base and the desired collector pp Z needed to be higher! Also, switched from a DC center tapped output transformer to a floating center tap. Hence, DC feed to the collectors is now accomplished with a 1:1 Balun choke, see photo, sitting atop the output xmfr. The card is single sided FR4, the pc connections use copper tape, and the grounds provided by small copper rivets. Well, its a breadboard. So its easy to re route and tear up the traces. from top to bottom in the pix, driver card, PA, bias card. The PA devices in center of card. The entire area shown is the heat sink and eventually the assembly flipped.

Alan

[W4AMV]

I received some informative emails on the collector transformer selection. Thank you. The final turns ratio will be determined by the available supply power capability. The selection of 4:1, 9:1, 16:1 and so forth is very dependent on the final collector voltage, Vce sat and the device collector current capability. Conservative estimates for these devices and with no fan on the sink probably places this PA output power at 120W. The use of the bifilar collector Vdd bias feed choke permits better push pull balance, the IMD is reported to be improved and the 2nd harmonic rejection appears improved. As I assemble some measured details, I will document.  Again, thanks for the informative email information.

[W4AMV]

I have updated the schematic of the original post. The current results obtained verify the use of UHF bipolar devices pp at HF is viable, but stability is a real concern. It is addressed in this schematic. As it turns out, the schematic is not to different than found in the original Motorola notes where the MRF454 and similar devices are used. The original schmatic provided for 0.5 W to 55 W output and including the 6 dB input attenuator has ~ 25 dB power gain. When this PA was terminated in a 40 meter LPF oscillation occurs. The base shunt load (11 ohms) and feedback was inadequate. At low supply voltage where the collector base C is larger, the oscillation is quite severe ! This updated schematic operates with no instability noted from zero Vcc up through ~ 24 V. I suspect it will be fine through 28, but need to check and get real ugly with the load termination to be sure. The gain now is 17 dB and 5W provides 70W. There is 6 dB input pad. Key is the use of SERIES R on each device base and the feedback R dropping to 6.8 ohms. Low impedance (voltage feedback) is now provided by a single turn through the collector bias choke ( as it appears in the original notes). The output xmfr is still 4:1 but eventually will probably go to 9:1. It is easy to modify these transformers if you leave adequate room to remove the turns and re thread. The copper foil stick-on landings to FR4 is proving to be quite handy and makes breadboard and modifications a snap. The supply current for the unit is 6A at 20 V so the efficiency is running about 58 %.

[AB2EZ]

Please clarify

How does the base to emitter current ... that is flowing from one end of the secondary of the input transformer into the transistor that is "on" ... get back via ground to either the other end of the secondary or the center of the secondary?

Stu

[AB2EZ]

Please clarify:

You show a 0.1uF capacitor from each collector to ground. This would provide essentially a short circuit for RF across the output of the amplifier.

Stu

[Opcom]

The RF application engineer warned be about excessive gain at low frequencies. Looks like you have dealt with that here. Please clarify more, this would be a great advantage, to use UHF rated FETS, sometimes cheaper and surplus.

There are some 1KW RF FETs (STAC4932B), but they too are UHF and the A.E. warned me the UHF ones would have too much gain at HF and oscillate. Funny the datasheet for the latest one says HF/VHF/UHF - someone must have figured it out, or it's a new type. The SOA says 20A@30V, so it's kind of a larger unit. The cost is way too expensive.

[W4AMV]

I apologize for the schematic errors. I will repost. The last schematic is revised. The 0.1uF on the collectors are NOT present. Please delete. The base bias is provided by a connection from the "Bias" point, the 270uF to the CENTER TAP of TLT1. Some photos here for clarification. The driver amp is updated and a LPF off card is attached. This filter for 40 meters and to be revised. Although the IL is under 0.2 dB, it could be improved with a different choice of filter network.

On the stability issue. If device data is available, at least S parameters, a good initial estimate of the degree of difficulty is easily found. Otherwise, you are correct, the low frequency gain has to be controlled. The feedback and addition of series R//C into the base lead of the device if appropriate values selected is one approach. The emitter inductance and via connections from the lead to the back side ground plane are of equal importance. I used a number, 3-to-4, of small copper rivets around the emitter area.

[N4LTA]

In the past I did some bipolar amp designs using Motorola devices. I had terrible problems with instability in amplifiers when the devices were not matched to the design frequency. This was a problem using 150 MHz devices at 6 meters. The only device characterized for 6 meters was the MRF 492 and they were harder to find than the 150 MHz devices.

Helge Granberg at Motorola warned me not to try doing it and he was correct. I never could tame them. Hope that you have better luck.

I have had some opportunities to get surplus UHF devices in the past and always turned then down do to my past experience.

Pat

[W4AMV]

Hi! Great to hear of your experience. Yes, well I am making progress. Your point is well taken and the jury is still out on this one. In the end, one has to get real UGLY with the load terminations, for example a screwball antenna. The out of band impedance, for example this low pass filter, creates a nice headache! Lots of instability issues until I addressed the current feedback now in use. There may be more added and one thought was to use a diplexer filter for the load filter termination. In any case, the price of these devices and their capability was just to good to pass up. Next on the list if this plays out is a pair of 125W LDMOS S band devices in PP. Oh my head hurts.

There were some questions on the TLT xmfr turn ratio selection and a pdf attached to address.

[W4AMV]

Here is a short note addressing the stability issue. Hope it sheds some light on the topic.

[AB2EZ]

Interesting!

With regard to the feedback:

The two ferrite beads between the 5T windings of T3 and the B+ would prevent differential mode current from flowing through the windings in series... and therefore prevent coupling to the 1T feedback winding. Please clarify whether there is an error in the latest schematic.

If I connect the two 5T windings on the left side of the ferrite beads, I estimate that the gain around the loop is (the transistor current gain) x (3.3 ohms / 6.8 ohms) x (1T/10T) = approximately 5. Do you agree?

Stu

[W4AMV]

Well, thank you again for catching my horrible goofs! Please see updated again, now called v7. Let me know if there are any others.

[AB2EZ]

Oops

Revised schematic still has ferrite beads blocking differential mode current

Stu

[W4AMV]

Ah yes. Point in fact, I am currently not using C1 and C2, now added to the schematic and called out as V8. The reason at this point is layout and I should address their effect on PP balance. I suspect this would have come out as I evaluated the amplifier performance. Again, thank you and flame on me if I have other mishaps.

[AB2EZ]

I think the 1T winding of T3 is also going to need a center tap with a suitable capacitor to ground. All of those center tap / capacitive ground returns are a consequence of the amplifier being "first push-then-pull", rather than "push-pull".

Stu

[W4AMV]

Hi Stu.

Under the assumption the devices are "matched" and if current drive appled at the base, push pull operation without CT is provided. Look at Motorola AN758 which provides a good reference point. It is useful to go through the math and fun to run a simulation and see if the "forced CT" vs. "floating CT" has a marked effect on the PP performance. This would include even harmonic levels and 2-tone IMD.

[AB2EZ]

Alan

I read the Motorola AN758 application note. It is very informative and thought provoking... and I plan to read it again, more carefully.

I am still skeptical about using the path from emitter-to-base in the "off" transistor to complete the drive circuit; rather than an RF grounded center tap on the input transformer's secondary winding... but if it works, it works (-:

It would be interesting to compare the behavior/performance of the two approaches at 3.885MHz. This would include the addition of an RF grounded center tap on the 1T feedback winding of T3, as well as the secondary of T1 (which is already shown in the latest schematic).

The author states that, in class C operation,  the emitter-to-base path in the "off" transistor can be modeled as having a resistance that is about 3x that of the base-to-emitter path through the "on" transistor. This is, of course, a very rough approximation, based on some time-averaging of the time-varying behavior of that path over the portion of each cycle that the "on" transistor is conducting current from base-to-emitter. It would depend upon such things as: the biasing, the time-varying emitter-to-base capacitance, various sources of stray capacitance-to-ground, etc. But, in the end, all that matters is that you get a pulse of collector current on each half cycle whose area is sufficient.

Anyway... I have found this very interesting and educational.

Stu

[W4AMV]

One of the items which came up in earlier posts was stability due to out of band impedance termination of the PA. I mentioned the use of a diplexed low pass. I also ran across an article by Bill Sabin, W0IYH, in QEX and he discusses the same application. Good read. The explanation of the diplexer design was a little void of detail, but there are references that provide the solution and if there is enough interest I can put an Xcell SS together. The end result is the PA sees a constant 50 ohm termination despite the screwball out of band Z of the filter as well as the antenna. The PA running here on 20 meters at 50W CW backed off from the nearly 100 W as I have no FAN on this thing yet. The 2nd harmonic from the PP PA alone is 30 dBc and after the LPF at -63 dBc. The 3rd harmonic is at -70 dBc. The return loss looking into the filter is ~ 20 dB or better expect at the cross over frequency of the HPF/LPF and is ~ 15 dB. Still very acceptable. The insertion loss was ~ 0.3 dB. Quite good for small T60 toriods. The use of a low pass filter configured with a high pass section or diplexer is a key design element for achieving stability especially when applying high gain devices.

Alan