Exploring Switching MOSFETs for HAM RF Power Amplifier Use
Hello! Today, I want to give you an insight into my (probably inaccurate and fumbling) design considerations and estimations for a low-bands (160, 80 and 40 metres) linear amplifier. It might help you out if you want to design something similar. At least it makes me think more carefully about what I’m doing, writing it down like this! What I am looking at here is a push-pull class AB linear amplifier, in the kilo-Watt input class, with a bunch of devices in parallel.
Where to start…?
Not quite on the back of a fag packet, but this time in a notepad document, and based on previous experience of exploding MOSFETs (!) I did some rough calculations. Bear in mind I *never* use the expensive, “proper” RF devices, and I stick instead to switching MOSFETS operated in the linear region, for reasons of cost. I don’t really care if they don’t offer anything on or above 40m (even though a couple I made work OK to legal limit (400W PEP) on 40m) — but you might!
The devices I discuss today are the IRFP250N, which, if you are prepared to buy 20 or more at a time, can be picked up for around a quid each (that’s a couple of Dollars if you are reading from the Colonies). So when I test them to destruction, I am sad 🙁 — but not too much poorer!
Basic stats from the datasheet here show that we are dealing with a 200V, 30A 214 Watt device. Stop right there if you think we will achieve anything like that for a single device. Note that both the max. allowable current and the maximum power each device can dissipate depends totally on temperature above 25 Deg C.
Heat Sink Considerations
“Commercial extruded aluminium heat sinks have a thermal resistance (heat sink to ambient air) ranging from 0.4 °C/W for a large sink…”
In addition, we have to consider the device’s thermal conductivity from junction to case:
Insulation between Drain tab and the heat sink varies with the method used:
The device junction temperature must never exceed 175 Deg C, if we don’t want to destroy any devices. I want to add a safety margin to this and say, “Let’s try to keep the (estimated) junction temperature below 130 Deg C under any conditions”.
In order to continue, I have made some assumptions (which will be revised at the end) to begin with. Please read all this article before forming your own conclusions, specifically about power ratings and the junction temperature relationships — if you don’t you will nuke FETs left, right and centre!
No matter whether we use blowers, fans, heat sinks and so forth, that device will get
warm hot during linear operation. Very hot. A conservative designer might perhaps design the whole thing running at a temperature of 100 Degrees plus — but I am using an enormous heat sink in a ventilated case, so I will shoot for 80 Degrees case temp. (hotter than a central heating radiator, wet your finger before touching it!). Even when I design like this, I will check it during operation, using an infra-red thermometer. Once I factor the temperature derating in, I reckon each device can only dissipate around:
214 – [(80 – 25) x 1.4] = 214 –  x 1.4 = 214 – 77 = 164 Watts
Remember this is the MAXIMUM any device should ever be asked to dissipate, and this is at 80 Degrees C. And it doesn’t allow for thermal conductivity of the heat sink coupling and cooling. I don’t want heat related failures, so I am going to allow a 50% margin for calculation error, inadvertent overheating, station mishaps (and other fuck-ups) and so I am going to say that I want to design so that each device should never exceed 80 watts of dissipation at 80 Deg C (case temperature). We know that the device’s junction-to-case spec’ is 0.7 Deg C per Watt, so the junction will be running at around 80 / 0.7 = 114 Deg C (estimated). But again, we have not yet considered getting the heat away from the device!
Let’s take a look now at the drain current maximum and check we will not exceed it:
Let’s say 20 Amps then, to be cautious, and then derate by my safety factor of 50% again.
Each device should have no more than 10 Amps of drain current, ever.
So, with my safety factor of 50%, I’m not going to allow any device to get hotter than 80 Deg C case temperature (is this realistic, even? — will test and prove later), I’m not gonna allow the current in each device to exceed 10 Amps, and no device will dissipate more than 80 Watts of heat.
Using the old rule of thumb that any transistor used for linear RF power amplification should have a max. voltage rating of at least 4 times the supply voltage (anyone have a reference for this — especially for MOSFETs vs bipolars? — please contact me if you do) then I am going to choose a nominal supply voltage of 48 Volts. This is perhaps sailing a little close to the wind, but I have a 1000VA 40-0-40 transformer, so I will simply have to use it! I say nominal, because, for simplicity, I am not going to regulate it. More on this as the project progresses — perhaps I will regulate it at 40 Volts or so later if sag is an issue.
In a worst-case scenario, the amplifier will be driven only gently and used for linear amplification of a small signal with Amplitude Modulation. This will be mostly in the class A region — so expect efficiency at low drive levels of only around 20%. We can’t expect AB operation until the higher drive levels (with the attendant overall improvement in efficiency). AM is a good design target in my book, because its easier for me to think about than PEP — and has a duty cycle of 1. If it works for AM, it will coast on SSB!
Let’s design the thing to use all of the transformer output power — 2 x 12.5 Amps at 40V ac. By the time this is rectified, I reckon a nominal “HT” of 48 Volts will do for estimation purposes.
So, at 1kW input power (continuous) we can expect just 20% of this to be output as our AM carrier. 1kW input = 200 Watts output. This means 800 Watts of heat must be given off by the FETs. This also means the total current taken by the amplifier will be approx. 1000/48 = 21 Amps. If we were to use two devices in push-pull, our maximum drain current would be exceeded by over 100%. Worse still, each device will dissipate 400W of heat — way beyond the rating of the device even at 0 Degrees Kelvin!
So, we must run devices in parallel — and significantly so! If we are going to dissipate 800W of heat, and we have our safety-factored single device power max. at 80W each, then we need 10 devices — so 5 in each half of the push-pull arrangement. Doing so also helps the current situation, because now (assuming all devices share current equally — and they won’t!) we can divide the total drain current by the 5 devices on each side (anyone — is this correct — or should I divide by the 10 devices? Since I am unsure, I’ll err on the side of caution until anyone tells me they know better…) to give each drain taking approx. 5.2 Amps. Much more sensible and well within each device’s rating.
Thermal Considerations and Heat Sink, Revised
We know from the datasheet that the junction-to-case relationship is 0.7 Deg C per Watt. So, for a max. junction temp. of 130 Degrees, case temperature must never exceed:
That is to say, when the case temperature is 91 Degrees, the junction is as hot as we would like it ever to get (see my safety factor above).
So, if the case must never exceed 91 Degrees, first we have to estimate the thermal resistance of the material we use to insulate the drains from the grounded heat sink. See the table above: traditional mounting methods using an insulating material will give us a resistance of around 0.8 C per Watt. In my explored design, each transistor will dissipate around 80 Watts, where the case temperature reaches 80 degrees. Just getting the heat from one device’s drain to the heat sink, via the insulating medium costs us 0.8 Degrees for every Watt we need to dissipate. A very good heat sink gives us also around 0.4 Watt / Degree C.
Using this calculator, we see the following:
How can we improve this? The killer here is clearly the ‘Thermal Resistance 1’, which is the thermal resistance of the insulator between the drain and the heat sink. (Even if you buy really ‘good’ ones). Let’s pretend there is no budget, and we find the ‘best in the business’ aluminium oxide insulators:
Not cheap – just over a pound each!
We still haven’t quite reached our goal of 80 Watts per device, but we are getting close.
Remembering we have a safety factor built in (the actual max. junction temp. is 175 Degrees but we end up with 130 Degrees @ 75W), let’s see what the junction temperature actually is when each device dissipates 80 Watts:
We are just about getting away with it! Not too far away from the design figure — but we have 10 quid’s worth of insulators, plus delivery — not good for a tight git like me!
Enter the copper spreader…
If, instead of spending money on insulators, you don’t mind ‘fannying’ around a bit, its possible to use a heat spreader made of copper (or aluminium) — the thicker the better, say 3- 8mm thick. If we attach the drains for each half of the push-pull amplifier directly to the copper spreader, with nothing more than a little heat sink compound, and then insulate the whole spreader from the heat sink, surely this is better than fiddly, inefficient insulators? The question then arises: what to use to insulate the heat spreader from the heat sink? I have successfully used, in the past, greaseproof paper (believe it or not!) smeared with heat sink compound for this task. I have *no* idea what the Degrees Per C figure is for this crazy method, but since the drains are in direct contact with the spreader, with nothing between, surely its around the same figure as the aluminium oxide insulators that are expensive and hard to find? (Must measure it sometime and report back!).
Here’s a close-up of one ‘RF brick’ I made using the heat spreader method:
Wow, this gets complicated fast — and we haven’t even considered driving the devices yet! With the help of a little guesstimate and known data, we can arrive at a reasonable approximation of how switching mode MOSFETs might work in linear class AB RF operation at frequencies that are quite low (up to around 40m). The most limiting factor, by far, is getting the heat away from the MOSFET junction as quickly and as efficiently as possible.
I don’t have all the answers, and these days my brain is fucked up by pain medication much of the time.
If you see any mistakes, or know better (perhaps by littering your pathway to success with dozens of destroyed FETS (!)) then please update me right here. I’ll credit you with the correction or the new information…
Other interesting stuff: