VHF Amplifier-Can I get it to behave across 400Mhz???

[Xnke]

Working on a three-band handy-talkie project, and everything's good until I get to the RF Power amplifier. Modulation, encoding, receiver, all is good there. But, the Transmit Power Amplifier is not. The VHF/UHF PA has to handle three bands: 144-148Mhz, 220-225Mhz, and 420-450Mhz. I don't have to do it all in a single device-I am OK with using as many as three amps, but obviously it'd be better if I could do it with less. If I was doing it in tubes...just bandswitch the tanks and be done. But, there's better ways with the new LDMOS devices!

I'm using a BFU520WA as the driver, and AFT05MS004NT1 as the PA. Target output is 4W, and input is 6mW from the FM modulator. Before I order silicon, I'd like to get at least close with the passives so I can place one order with Digikey and at least get something going. I haven't had any luck with simulations for RF stuff...LTspice apparently doesn't do a good job.

Here's what I've got so far:



And a description of what I think I have:

Gain should be around 15-16dB, and with 8dBm input I should have 24dBm output to drive the PA.

VBAT is 7.4V, from a Li-Ion battery pack. It will range from 8.4V on a fully charged pack, to 6.8V where the batteries are officially "dead".

C20/C22/C21 are filtering and decoupling for the stage. R8 is part of the filtering/decoupling.

L1 should be >100 ohms impedance across the band, 10nH is only 10 ohms at 144, 13 ohms at 220, and 27 ohms at 440...so it's not big enough. Because L1's reactance change over frequency will mess with the next stage's input impedance, R9 has been inserted to help with broadband matching into the next stage.

C17 is a DC block.

C18 and L4 are intended to be part of the impedance match into the next stage.

R10 and R5 set the bias for the BFU520WF at about 7.5mA Iq, C19 and L2 are there to add a bit of bias stability.

C16 is a DC block.

If I am understanding how to calculate input impedance, the Zin of this stage should be about 219 ohms. Zout would vary from 10 to 20 ohms, if terminated into a 50R load.

What am I missing here? So far, I think I need to increase L1 and R9 a bit to improve matching, but I'm worried a bit about stability given this transistor's F(t) is 10Ghz...My test gear goes to 400Mhz here at the house and I'll have to go into the lab at my cousin's to do spectral analysis higher than 1gHz.

What do I need to do to promote stability and gain flatness over the extremely wide bandwidth I need to work with?

[TheMG]

Virtually every multi-band VHF and above transceiver design I've ever seen uses independent RF power stages for each band.

One very good reason for this is that you would have to switch in different harmonic filters at the output of your final amp depending on the band. Since the 70cm band is right around the third harmonic of the 2m band, they obviously can not share the same low pass filter to get rid of harmonics.

Also, of particular importance for battery powered transceivers, it is much easier to design an efficient narrowband power amplifier than it is to design one that is very broadband. Such designs typically rely on resonant tank circuits, which pretty much rules out the same circuit operating on more than one band.

[Xnke]

I figured that was the case-I know that MOST two-band HT's that run 2M and 1.25m are a single power amp, with switchable LPF's on the output. Unfortunately, not many HT's are dual-band with 1.25M, they're 2M/70CM units and about half I've looked at are single PA's, half are separate PA's. I haven't seen the guts of any 3-band HT's. Basically all low/moderate cost HT's use a single amp and switch in the filters, and anything more than mid-range switches the PA completely.

The LPF's I am using are good up to about 8W output power, and it's not like the amps are all on at the same time, so heat won't be a huge issue. Space...maybe. Depends on how tight I can pack parts together.

[vk6zgo]

Working on a three-band handy-talkie project, and everything's good until I get to the RF Power amplifier. Modulation, encoding, receiver, all is good there. But, the Transmit Power Amplifier is not. The VHF/UHF PA has to handle three bands: 144-148Mhz, 220-225Mhz, and 420-450Mhz. I don't have to do it all in a single device-I am OK with using as many as three amps, but obviously it'd be better if I could do it with less. If I was doing it in tubes...just bandswitch the tanks and be done. But, there's better ways with the new LDMOS devices!

I'm using a BFU520WA as the driver, and AFT05MS004NT1 as the PA. Target output is 4W, and input is 6mW from the FM modulator. Before I order silicon, I'd like to get at least close with the passives so I can place one order with Digikey and at least get something going. I haven't had any luck with simulations for RF
stuff...LTspice apparently doesn't do a good job.

Here's what I've got so far:



And a description of what I think I have:

Gain should be around 15-16dB, and with 8dBm input I should have 24dBm output to drive the PA.

VBAT is 7.4V, from a Li-Ion battery pack. It will range from 8.4V on a fully charged pack, to 6.8V where the batteries are officially "dead".

C20/C22/C21 are filtering and decoupling for the stage. R8 is part of the filtering/decoupling.

L1 should be >100 ohms impedance across the band, 10nH is only 10 ohms at 144, 13 ohms at 220, and 27 ohms at 440...so it's not big enough. Because L1's reactance change over frequency will mess with the next stage's input impedance, R9 has been inserted to help with broadband matching into the next stage.

C17 is a DC block.

C18 and L4 are intended to be part of the impedance match into the next stage.

R10 and R5 set the bias for the BFU520WF at about 7.5mA Iq, C19 and L2 are there to add a bit of bias stability.

C16 is a DC block.

If I am understanding how to calculate input impedance, the Zin of this stage should be about 219 ohms. Zout would vary from 10 to 20 ohms, if terminated into a 50R load.

What am I missing here? So far, I think I need to increase L1 and R9 a bit to improve matching, but I'm worried a bit about stability given this transistor's F(t) is 10Ghz...My test gear goes to 400Mhz here at the house and I'll have to go into the lab at my cousin's to do spectral analysis higher than 1gHz.

What do I need to do to promote stability and gain flatness over the extremely wide bandwidth I need to work with?

LTspice thinks L1 is a perfect inductor----- in the real, horrible, nasty world, it is a parallel LC network, so even if it gave the results you were looking for, it would still be meaningless.

[BigBoss]

This circuit does not work..see below.
-Biasing is completely wrong
-0.1uF (C18) cannot be used as a matching element because it's too high..
-R5(680 Ohm) is too low, it should be around 5K
-33 nH (L2) is too low for 144MHz.
etc.. There are possibly other wrongly estimated elements ..

[vk6zgo]

I can't believe I looked at it, worried about L1's parasitic capacitance, but didn't notice the bloody great RF short circuit across the output!

[M0HZH]

The circuit is completely wrong, both in DC and AC.

It can be fixed, but why not just use a MMIC like BGA 616 ?

[Xnke]

MMIC would work-sure. But that's not the point of building it this way. Even if I ultimately use an MMIC for this, I clearly don't know what I'm doing with a regular transistor anymore.

BGA616 has a bit much gain to fit where I need to put this, but other than that yes, it would work fine. It's hard to find an inexpensive, widely available, MMIC with only 12-15db of gain at the frequency being used here.

I'll go through this again and make another attempt. I don't know how the math got so screwed up-the bias resistors are an order of magnitude is off. 11.5nA of base current is what the bias resistors I chose above yield, but I aimed for 105uA. I already knew L1 was too small, and L2 was just "put an inductor here" placed. C18 was the same-Just a "put this here" that I forgot to put a value in.

For reference, NXP AN11427 is an LNA in the middle of the 440Mhz band, using this device, however the supply voltage is 3.6v. It would work fine, at the lower supply voltage, but would not give the 75-100mW output needed.

[M0HZH]

BFU520WA's specified P1dB is +6dBm @ 433Mhz / 10mA / 50ohm Zout, that's lower than the signal you drive it with. You need a different device, capable of the required output.

Keep in mind you need some headroom in gain and output level to compensate for different frequency response in each band. Also, the final stage would have to have at least 3dB of headroom in output at the end for ruggedness purposes (antenna mismatches, etc).

Edit: +11dBm @ 20mA, possibly close to +14dBm at 30mA but not specified. Definitely very far from +24dBm you are aiming for.

[Xnke]

Yep. I managed 2vp-p output at 433Mhz on the breadboard, but that's about it. Changed the bias resistors to 5.2K and 1.5K, added a 2 ohm emitter resistor, and that was all I got.

I don't think I have any other transistors that are suitable, but I do have some GALI-51+ MMIC's that would do. I need to figure out how many other circuits in this project I've jacklegged together and fix them. The receive side works fine, minimum recoverable signal is about -119dBm.

Next one to work on, I guess, is the PA.

[M0HZH]

The 1K drain resistor is seriously limiting power output; you also need to test with output load connected.

GALI-51+ has the same problem as my previous suggestion (BGA616), output limited to around +18dBm. Not sure how easy it is to find MMICs that achieve over +24dBm P1dB. One device I am considering for a design is AFIC901N, it has two amplifier stages in one package, you would only need to use the second one here but it doesn't seem to be available though the usual channels. Another option would be back to transistors, something like BFQ790 or RD01MUS2 are easy to get, have good performance and plenty of headroom.

[Xnke]

Well, moving on to the next stage, I need to see if I need the full 100mW or if I can overcome it with biasing. The AFT05MS004NT1 datasheet is a bit fuzzy on that-they call for 100mW but in the charts later on in the sheet, as low as 40mW is enough to drive the device to full output.

The breadboard testing on BFU520WA was into a 100R load resistor, but even as is i think a little retuning would make a fine LNA for the three bands on receive. I did swap the 1K drain resistor for a 22 ohm.

I've been working through the examples given in the datasheet for the 136-174Mhz amplifier and 350-520Mhz amplifiers. They rely on microstrips for a LOT of the impedance matching it feels like-and I know that VHF/UHF is the prime time to start using them, but I am not skilled with even lumped elements. The board layouts and microstrips are based on a 0.020" thick FR4 PCB, I had planned on a 0.040 or 0.060" thick material.

I think lumped constants can work for these, because I don't need the 100+Mhz bandwidth the UHF amp delivers, I only need it flat over about 40Mhz. My ARRL handbooks that are heavy on theory and examples are too old to include LDMOS, or even mosfets in general, and the few bipolar transistor examples are all 100W tuned class C units with transformer coupled input and output stages.

Class C operation is fine-I'm only doing FM and FSK. Is there a good resource for Class-C LDMOS design out there on the web? I suspect it will be extremely similar to a triode Class-C amplifier with higher capacitances and lower impedances, but given my track record with silicon I'd like to start at the beginning. I'm having a hard time sifting through some of the EE design courses I have found to locate the information I need.

[Xnke]

The reference design in the datasheet for AFT05MS004NT1 calls out needing 60mW of drive at 155Mhz, gain is 20dB, and output power is 6.0W. That's basically ideal-but it's optimized for a very wide band of operation. I think by narrowing the band of operation I can get equal performance with the same or less drive power, and a more compact design. (more importantly, one that I understand and can adapt to 220Mhz as well!) Either I can narrow it up and build three separate PA's, or I have to widen this one even more, which probably isn't going to work well and I'm not confident in my ability to make it work at all.

I've drawn up the below, using the values from the example design and adding a non-temperature-compensated bias solution. It won't work as is-the input and output matching relies on the stripline elements that aren't visible in the schematic. But, it's a place for me to start.



Starting from this, I want to determine the input match network first. I need to understand the unmatched gate impedance at 146Mhz (center of the band) and work from a 50 ohm impedance into that. I'll work on that, and post back what I can figure.

Don't laugh too hard T3sl4C0IL, it's been 16 years since you taught me this stuff and I haven't used it nearly enough to remember it!

[Xnke]

In NXP's datasheet and engineering bulletin EB212, they claim an "MSTEP" dimension is listed for each device, and it is the dimensions of the stripline that the actual device terminal is connected to. I had to dig for it, but for the part I'm using, that block is 0.070" long, by 0.140" wide, on a 0.020" thick FR4 substrate. They do not mention the copper thickness, but it changes the impedance only very slightly. The same size microstrip is used at both 146 and 430Mhz, as the "lead in" and "lead out" directly connected to the transistor. From that block, the actual impedance transformation happens. It's impedance is 18.8 ohms-Do I need to maintain the impedance of the stripline as I change the substrate thickness, or maintain the stripline dimensions? NXP doesn't mention, and so I assume the impedance is what matters.

If the impedance is what matters, and the MSTEP impedance is what is important (seems like it since it's the same for both sets of frequencies, for multiple different devices) then I have enough info to get the stripline PCB layouts to work. This is handy-although dangerous since I don't fully understand it. Guess it's time to layout a board and find out?

[Xnke]

This one works.



Adjusting the input match a bit, down to 6.5nh and 39pF, it works and is a fairly good (1.8:1SWR IN) from 121 to 390Mhz, with the output terminated into a 50R dummy instead of the ceramic filter. No luck yet in getting across that UHF line, but I'm getting closer.