Broadcast Warehouse design problem

[Shedbuilt]

If we're designing for 1W at 12v, and we're using the "traditional / textbook" (Vcc-2)²/2Po, we end up with exactly 50 ohm output impedance. In the real world, however.....

…you’ve got to take into account the reactive part of the collector (or drain) impedance since it’s not going to be purely resistive.

Absolutely, and in a lot of cases, that’s going to be largely unknown at the design stage. Lots of variables, hence the “If we’re using”, and the “in the real world” comment at the end, because those methods are largely a finger in the air, a start point, which will hopefully put us in the ballpark of the apparent impedance (the net effect of the resistive and reactive, seen by the device), for a given expected power, at a given voltage. I was partly making the point, that the change from (say), 12v to 15v, represents a large change in impedance, at a given output power. In reality, of course, the power almost always changes as we vary the voltage.

[Albert H]

If we're designing for 1W at 12v, and we're using the "traditional / textbook" (Vcc-2)²/2Po, we end up with exactly 50 ohm output impedance. In the real world, however.....

Nope. Vcc²/2Po gives 144/2 = 72Ω I don't know where you get your "-2" from. I reality, the match is very close to 72Ω

My old 1 Watt Band 1 Link transmitters were always designed to use disused Band 1 TV aerials, so were made to match into 75Ω, so the impedance conversion was minimal. It also meant you could get away with using the brown "TV" coax, which is really cheap!

You're entirely right about the "ballpark" nature of the calculation, and the rapid change in match impedance with change in supply voltage. I always made sure that the final was run from a regulated supply, so the impedances were not changing, but as things warm up, their characteristics change.....

As ever, electronic design is based on compromises!

[Shedbuilt]

If we're designing for 1W at 12v, and we're using the "traditional / textbook" (Vcc-2)²/2Po, we end up with exactly 50 ohm output impedance. In the real world, however.....

Nope. Vcc²/2Po gives 144/2 = 72Ω I don't know where you get your "-2" from. I reality, the match is very close to 72Ω

Vcc-2 is used in a lot of reference books and other reference works. I agree though, in Class C, Vpp should be very close to Vcc. I’ve always tended to suspect, that those who designed for, what amounts to a 50 in, 50 Out, output filter, probably used Vcc-2, and assumed a 12v supply.

[Shedbuilt]

One example quoted from NXP (albeit, for this example, this would give VCC-1.5, rather than VCC-2):
“ The load line resistance is given approximately by:
Table 1. Comparison of Input Impedance for Different Operating Modes
where
VCC - VCE(sat)RF)2
RL = 2 * Pout (1)
VCC is the collector supply voltage, Pout is the required peak power,
and VCE(sat)RF is the collector-emitter saturation voltage under the frequency of operation. The value of this parameter is particularly difficult to measure, but the normal range is 1.0 to 2.5 Volts depending on the geometry, epitaxial doping and thickness. A good approximate value for 12.5 Volt devices is 1.5 Volts and for 24 Volt transistors is 2 Volts.”

[Shedbuilt]

The formatting of the equation, became garbled in the copy and paste. It (hopefully) clearly, should read
RL=(VCC-VCE(SAT)RF)²/2*Po
Description as above by NXP

[jvok]

2v vcesat sounds about right for older bjts. The newer mosfets can pull their drains almost all the way down to ground though so you can use the full 12/24/48/whatever volts when working with fets.

[Shedbuilt]

2v vcesat sounds about right for older bjts. The newer mosfets can pull their drains almost all the way down to ground though so you can use the full 12/24/48/whatever volts when working with fets.

Yes. Modern power FETs in general, have very low RDS on.

[Albert H]

The other thing that I really like about the modern power FETS is their negative temperature coefficient - as they get hotter, they draw less current, so they don't tend to suffer from thermal runaway (like the old bipolar transistors did). They also have insane amounts of gain compared to the old transistors...

Did anyone else here try to use a BLW77 on Band II? I used to get up to 180 Watts out of them, but needed close to 50 Watts of drive - hardly efficient!

The latest generation of power FETs need just a Watt or so for full (>300 Watt) output. The current crop of PA designs are astonishing, and the next ones will have even more gain!

[Shedbuilt]

Agreed all round Albert. I started using the 2SJ50 / 2SK135 pairings for audio, quite early on. As the higher voltage version of the 48,49 / 133, 134, implemented in multiples of parallel pairs, they were good for quite a lot of power, and the negative thermal coefficient, made them extremely robust in that application. The voltage dropped across the device, even cold, was a bit higher than bjts, and increased with increasing heat, but all round, they were a very good option. VHF power FETs didn’t exist then, but the early ones weren’t too far behind.
I don’t remember using BLW77s, but used a few BLW96s; more power out than the 77, but definitely needed a good kick to get there. Used quite a few different VHF (and repurposed for VHF) transistors, which didn’t want to give more than 6dB gain on Band II, and many of them were fragile at anywhere near their rated power. Things have moved on hugely !!

[Albert H]

I built a whole lot of Medium Wave rigs using the 2SK135 for the output device! It was convenient in that the body of the device was the "Source" terminal, and so could be bolted straight on to the heatsink without insulators. One of those driven by a BFY51 or a BD131 would give an easy 25 Watts of carrier, and four times that on peaks.

I used to modulate them with the LM12CL power op-amp, and the modulator had a quad op-amp for audio filtering and peak limiting. They were great little rigs, and using my "3-chip synthesiser" (CD40103, 4060 and 4046) and a 4608 kHz crystal (35p from Farnell - used in Baud-rate generators), it was easy to generate any medium wave frequency.

The most expensive parts of those MW rigs were the case, mains transformer, and the heatsink!

[reverend]

Going back to the VCC squared over 2Po discussion, I have found that 4427's aren't really capable of driving a 50 Ohm load and if you use them without any output matching, they struggle to output 1 Watt. Albert's point that the correct match is nearer 72 Ohms tallies with my experiences. Of course other transistors (such as the MRF237) play lovely into 50 Ohms, and even with just a low pass filter (i.e. no matching at all) will put out nearly 2 Watts (implying a voltage swing that goes all the way from VCC to the deck).

[XXL]

Yeh definitely needs a rework. The board is lovely and locks solid as a rock but the output let’s it down. Unfortunately I’m not that clued up on how to rework it without causing spurs. I can make a nice pll driver on a copper board that does 1w but iv no idea if it’s clean so I just use tried and tested designs to make sure it’s clean.