Some thoughts on the design of RF linear amplifiers for amateur radio use. Chris GM3WOJ - July 2022
I built my first HF amp in about 1982 - using a pair of 4CX250b valves - in a 19" cabinet that was surplus from a UK nuclear facility (I always wondered if the cabinet was slightly radioactive, hi) This amp is still in use today, but only on 70MHz.
Since then I've built another six
HF/VHF amps, mainly using 4CX1000a/4CX1500b valves. I used a QY5-500 valve in
one amp - it lit up the shack but did not deliver much RF really. I also built a
50MHz amp with an NOS Eimac 8877 - it worked well for only a few months then the
8877 died - it was one from the year when Eimac thought an internal 'heat dam' was a good idea - in other
words, a bad version of
the usually good 8877 series. These dubious 8877s were first manufactued in August
1986 approximately, without Eimac telling anyone immediately about the modification (if my
info is correct). I have always liked Eimac products.
Most of my homebrew valve amps have only rudimentary protection - mostly just fuses - but have proved surprisingly reliable over the years.
In 2011 I decided to build/assemble my first solid-state HF amp. I purchased a 'ready made' 600W o/p HF PA module from Israel, an 1800W 50V DC 'blade' SM mode PSU (ex cellphone site), protection and control boards from Australia, and a PCB for a six-band BPF from the USA. Despite hours of very careful construction, and the end result looking like an amplifier, this project was a complete disaster. The PA module was utterly useless, as was the so-called protection circuitry, which seemed to need 'thinking time' before reacting. After burning up several hundred pounds in cash replacing MRF150 MOSFETs, I abandoned this project.
This bad experience did not put me off solid-state amps - just off trying to build my own one! So I purchased a Tokyo Hy-Power HL1.2kfx from a friend - what a great amplifier! Compact, reliable and extremely well protected against most scenarios. This THP amp uses four of the 2933 devices as used in the KPA500, which only uses two of the devices. This link from 2010 is interesting : https://elecraft.mailman.qth.narkive.com/JGXGdxjY/kpa500-vs-tokyo-hy-power-1-1kfx Sad that Tokyo Hy-Power closed down.
In our pre-COVID DXpeditions (ZK2V, V6Z, A35V, VK9CZ) with my friend Keith GM4YXI, we initially used these THP amps and in more recent years we have borrowed Expert 1.3kFA amps. We have never had any issues with them - they worked well even when the air temperature in the shack reached +35C. On one DXpedition an AC voltage surge lit up every LED on the Tokyo HP amp, but it was undamaged - that's good protection circuitry! However we *always* go for reliability over power output - never more than 1000W out of the Expert 1.3 and never more than 700W out of the THP amps.
Some sympathy for manufacturers
Designing and manufacturing *any* product which is then released into the 'wild' is a difficult, almost impossible, task. Despite months of pre-production testing, there will almost always be some real-world situation which causes problems - whether it's operator error, environment, bad luck, etc. The amateur radio market is a prime example of a 'wild' environment, especially for an item like an LDMOS amplifier. However, that does not excuse any manufacturer who is marketing a product that has deficiencies, whether they know about them or not.
Firstly, there are expectations which stem from any manufacturer's advertising. In my view, anyone attempting to run an early Expert 1.5 at 1500W o/p all the time is asking for trouble. The clue is in the NXP LDMOS device name i.e. MRF1K50H - in the NXP lab under tightly-controlled conditions it is capable of 1500W o/p, but they have an endless supply of them, hi. Here in the UK an MRF1K50H currently costs around £250 = US$300 approx so you don't want to blow them up too often (or ever). Later - Series 2 onwards I think, Expert 1.5s are fitted with the MRFX1K80N 1800W device. Here is some info from David G3YYD about the Ampleon BLF series devices :
Secondly, some purchasers of LDMOS amps will be unable to provide them with the stable RF drive and/or antenna conditions that are essential to minimise the probability of damage to the amp. I am lucky and have all resonant single-band antennas, so never use an ATU. Many other amateurs are faced with matching random lengths of wire, etc. - not so easy.
Allan GM4ZUK/GM4Z made an interesting comment about LDMOS devices i.e. they are designed to run continuously 24/7/365 in commercial service. This implies that the device is held at a stable temperature e.g. +60C and does not undergo the 'temperature cycling' that it does in an amateur amplifier - especially on FT8. Allan thinks this temperature cycling may be a contributory factor in the longer-term failure of the LDMOS device, due to internal expansion and contraction.
Power density comparison - LDMOS vs Ceramic tetrode
I am not sure sure why these LDMOS devices need to be so physically small - maybe for the relatively new generation of tower-mounted cellphone equipment, etc - I don't know? Obviously the smaller the device, the greater the difficulty in removing heat, which in turn also might lead to lower long-term reliability.
Let's compare an NXP LDMOS chip with an Eimac 4CX1000a valve, making the simple assumption that they each have to dissapate 1000W of heat energy within their own structure (ignoring heatsinking, forced air cooling, etc.)
[Cuboid body of LDMOS (L x W x H) 32.2mm x 9.8mm x 3.8mm -> volume = 1199.1 cubic mm. Cylindrical body of 4CX1000a anode (Pi x radius squared x H) = 3.14 x (42.5)^2 x 31 -> volume = 175820 cubic mm]
Device | LDMOS | Tetrode |
Body shape | Cuboid | Cylindrical |
Volume | 1199.1 mm^3 | 175820 mm^3 |
Volume ratio | 1 | 147.6 |
Power per mm^3 | 0.83 W/mm^3 | 0.0057 W/mm^3 |
Power density ratio | 145.6 | 1 |
This is NOT an entirely valid comparison - for example I can't work out the actual volume of the metal used in the 4CX1000a anode structure which is actually two concentric rings with multiple fins between them etc, but it does show roughly how much higher the power density is in an NXP LDMOS device.
Some specific comments about SPE Expert LDMOS amplifiers
SPE amps, in the early days, had a poor reputation - expensive fuses. However I think they really got their act together and the 1.3, 1.5 and 2k amps are, generally speaking, good products.
The SPE designers have designed a popular product - however there do seem to be a number of unhappy purchasers. It is important to separate out purchasers who (a) probably have damaged their amps through mis-use (overdriving, drive spikes, overheating, RF feedback in shack, bad antennas, etc) and (b) those who think that the design of the amp. has caused it to fail through no fault of the operator or operating conditions. Any online forum is likely to have a high number of complainers who like a platform for their complaints. Obviously, some complaints are justified and should be listened to by the designers and manufacturers.
Sorry SPE, but I do not think that your idea of running any transceiver at full power (100W or even 200W o/p) and relying on a hard-wired ALC link to the Expert amp is good engineering. Any phono to phono cable can go open circuit > LDMOS destroyed. I appreciate that a lot of clever design has gone into this, but this Expert 'AGC' system may be an exasperated attempt to stop careless (or just unlucky) users from damaging the amp with excessive drive, even momentarily. As someone said on the reflector, this is almost 'daring the ALC link not to fail'.
Some comments about ALC
[Completely off-topic, my February 1959 copy of the Collins Radio Company's iconic publication 'Fundamentals of Single Side Band' - page 7.10 - outlines an 'ALC' system - however in this context it means 'Automatic Load Control']
Any amplifier/transceiver ALC system is a 'closed loop' system i.e. the output level is fed back from the amplifier to the radio to limit the drive level. Quite a few early valve amplifier/ transceiver ALC systems actually made the transmitted signal quality worse - they caused 'flat topping' by limiting the peak SSB signal incorrectly. By definition any ALC system will have a finite response time - even if this response time is a few microseconds, that may be too long to peak limit the drive to an LDMOS device. The transmitter part of most (all?) current transceivers have their own internal ALC system, which is designed to minimise internal overdriving and distortion.
Amplifier protection in modern amplifiers is usually achieved by a combination of good hardware design and well-written firmware. Some transceivers can 'handshake' with the amp. to help prevent overdriving. That's all very well for e.g. Flex radio owners, but for any amplifier manufacturer of course, they cannot assume that any particular make or model of transceiver will be used with their amp, so any protection circuitry has to be independent of the transceiver overall. Protection should work effectively with any transceiver - on condition that the user maintains a suitable low drive level, constant antenna load etc.
Finally, this quote from Art K6XT on the reflector really sums things up ... "Mind boggling how some can blow LDMOS after LDMOS connecting to the same old stuff, blame the amp, and seemingly never contemplate something else as the cause." Spot on, Art!
-----------------------------------------------------------------------
These are just my personal opinions - feel free to disagree. 73 Chris GM3WOJ / GM2V near Inverness, Scotland. Updated 11th July 2022