Broadcast Warehouse design problem

[XXL]

Does anyone know why the Broadcast warehouse design, most recently used by zozo, only does around half a watt on the bottom of the band ? The 4427 runs RED HOT, which is why you need that huge chimney heat sink on it, and will only do around half a watt on the bottom of the band.

Is there a design issue with it ? no other designs using the 4427 get that hot.

[jvok]

I had the opposite issue with the zozo board. I'd get almost a full watt on 87.5 but less than half a watt at 108. My guess is the output matching could do with a bit of work.

[teckniqs]

It's not like that on the BW version, does around 1 to 1.5w across the whole band.

[XXL]

It’s the same with the NRG board on eBay. It’s suppose to do 1w but I can barely get 800mw out of it, and the 4427 is red hot. There’s no need for a heat sink that big. Iv had 4427s doing nearly 2w and they don’t run that hot.

[Albert H]

There are several issues:

1. The collector choke needs to be a higher value - we always used something like 10µH, rather than the 1µH or 2µ2H that most people use.

2. The match needs to be re-worked, too. Vcc²/2Po gives an output impedance of 112.5Ω. If you're going to match into 50Ω, you have to do an impedance conversion..... The "design" typically uses a couple of variable inductors, and some fixed capacitors. Typically, there's a significant mismatch over much of the band, unless the inductors are specifically adjusted at each frequency - there's no way to make a simple network like that truly "broadband". This mismatch will reduce the available output power and will cause reflection back into the transistor, causing it to heat up a lot. Note that the tuning that gives maximal power output is VERY unlikely to be the optimum match! Originally, that stage was tuned with trimmer capacitors and variable inductors, but the extortionate prices for trimmers and the unavailability of Toko S18 inductors make "simplification" of that stage inevitable.

3. Most of the 2N4427 transistors available today are either fakes or are sub-standard. It was never a particularly good device - particularly because the case is connected to the collector. For that stage, I always preferred a transistor with its emitter connected to the case. This allowed for trivially easy heatsinking. I used to use 2SC730, MRF237, SD1143, MRF229 and many others.
These days we always design for FET devices in that position, and usually go for pairs of devices and a broadband transformer to do the impedance match - the push-pull transistors tend to cancel even harmonics, and both gain and efficiency are enhanced. We use ceramic bandpass filters on the way into the exciter final stage - the Soshin parts (for example) have low insertion loss, but virtually eliminate out-of-band products.

Remember - all electronic design is a set of compromises!

[teckniqs]

I am just wondering what manufacturer the said 4427 is ?

[Albert H]

Every semiconductor manufacturer seems to have made 4427s! The "originals" came from (amongst others) TRW (now part of Motorola). I have some Philips ones, some branded "SEM", some from APT and a few others.

[nrg-uk]

Zozo board works fine and does not run hot it's your clone Chinese transistor get some real ones very hard to get but if you want one I can put on on eBay £10 each they should have gold legs and base!

[XXL]

Iv got real ones, new old stock from the 90s. All got ST on them. They’re not shiny like the fakes on eBay.

Take the big chimney off the 4427 and replace with a regular size one and you’ll see it gets hot very quickly. They shouldn’t get that hot. A regular sized heat sink should be enough and still be able to hold your finger on it at full temperature.

[radium98]

RCA are only gold base , like the mitsubishi4w that have gold base and very long legs. i have few from about 10 years and they were in that era expensive , how then nowadays .

[jvok]

Just tested a couple of zozo boards I have, both give almost bang on 1w at 87.5mhz at least when cool but drop off a bit when they warm up. Tests were done with 2x 100r resistors soldered flush to the board so probably as close to a perfect 50r load as you'll get.

One I built myself with a (probably fake - shiny and the writings wearing off) motorola part pulled 1.8w dc at 15v just for the pa, which works out to 56% efficient. Not great when class c should be 70% odd in theory but not too bad for an old school bjt at vhf.

The other is one I bought fully built on ebay with a gold base part marked "2N4427 41 SSS". This pulled 3.3w at 15v, only 31% efficient. This one came with a much bigger heatsink than what I use but no surprises still runs very hot.

Current draw was very sensitive to output tuning but even then the one board consistently pulled much more than the other.

[sinus trouble]

Maybe the design could be 'Tweaked' Yet i doubt it would have much effect?

We seem to be missing a point here which Jvok eluded to!

I would like to have a magical solution, But the fact is the 4427 is pretty poor on efficiency! (Compared to todays standards) Even when operating in optimal conditions!

We also need to consider that it is a 20V device (Absolute Max) and was never intended to run @ 15V

An original Motorola will give more gain! However, it will still get hot regardless!

XXL, Some of the other devices Albert mentioned may give better result?

CPC also have 2N3866s in stock at a reasonable price? May be worth a try?

Good luck Geez!

[XXL]

If I run the driver board at 18v I get 1w. The mpsh10 can also run at that voltage so it’s fine. I was just questioning the heat problem.

Also I’m sure 3866 will only do about 300mw at 15v. It’s a 28v device.

[jvok]

I agree with sinus, the real answer is to switch to more modern parts. Old bjts like the 4427 or 3866 are still low gain and inefficient even if you find a legit one.

I've been playing with a design using an rd01mus2, tiny little smd fet with no heatsinking besides the board itself. But it barely gets warm doing 1w at 12v and it takes less than 20mw drive. I even got 2w out of it without pushing it too hard.

[sinus trouble]

As Albert always says! Generally think of your Driver Board as the 'Signal source'

Try not to push it too hard! Let it do its job in a stable manner!

Leave the Gain aspect to the following Power stages!

You will find that you get better performance from your 'Rig' overall!

[Albert H]

We've been using the Mitsubishi "12V" FETs since they became available. I can remember that I was astonished at how much gain they gave, and just how efficient they were, compared to the old (and newer) bipolar transistors I'd used for years.

My latest exciter uses a dual-gate FET oscillator driving into a BFG135 for about 300mW, then a Soshin GFW ceramic bandpass filter, and then into RD01 FET. The output of the FET has a level-sensing circuit, that "wraps back" to the second gate of the oscillator FET, so that we get exactly flat power output right across the band.

The FET oscillator uses a couple of varicaps for tuning, and two for the modulation - one with voltage-variable bias to linearise the deviation sensitivity - the same peak-to-peak voltage will give 75kHz deviation right across the band. The exciter is essential a Voltage-tuned power oscillator block that requires no adjustments whatsoever. It's a little more expensive than the usual "prate" exciter, but has a spectral purity and accuracy that really simplifies the construction of the rest of a transmitter, thereby saving on calibration costs.

[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.....

[XXL]

I agree with sinus, the real answer is to switch to more modern parts. Old bjts like the 4427 or 3866 are still low gain and inefficient even if you find a legit one.

I've been playing with a design using an rd01mus2, tiny little smd fet with no heatsinking besides the board itself. But it barely gets warm doing 1w at 12v and it takes less than 20mw drive. I even got 2w out of it without pushing it too hard.

Would it be a straight replacement for the 4427 on the BW design ? Obviously with a different pcb pad or by bodging it on.

[jvok]

Would it be a straight replacement for the 4427 on the BW design ? Obviously with a different pcb pad or by bodging it on.

No, biasing circuit is different with fets.

I'll post a circuit when its finished but right now I'm struggling to get the input matching right

[Bton-FM]

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.