The regulatory regimes between the EU and US are actually very different in that CE is (essentially) a self certification that your product complies with all applicable regulations, while you may take the view for a certain product that you need to get it tested at an independent test house, this is not actually a requirement. 

While VDE, UK, CSA are all essentially insurance industry test houses they have never really obtained the level of being almost mandatory approvals in the EU that they have for many things in the US, the regime is just totally different but I am far from certain that one is meaningfully easier then the other (You guys have FCC class B testing for example, with paperwork with large fines attached for getting it wrong, nothing quite equivalent exists here), six of one half a dozen of the other.   

I know why wall warts and line lumps are popular, that was not what I was bitching about, it is the annoying tendency to make equipment gratuitously difficult to get into that pisses me off.

Regards, Dan.

It is particularly annoying when seen on things like professional audio mixers where a significant amount of end user configuration is done by moving jumpers on pin headers inside the equipment!

I think it has something to do with that whole UL thing that the us market is so enamored of.

The other major annoyance is the common use of 'security' screws on things like power supplies, WHY? I mean sure if I screw up in a big power supply I might die, but a simple standard screw will stop me getting in there by accident and if I go in anyway that is clearly me accepting the risk, why make access harder then it needs to be.

Regards, Dan.

I never quite saw the virtue in that approach, why let more get to the first mixer on receive then you have to, and if you are going to filter on receive before the first mixer then surely using the same filter on TX only makes sense (You might want to switch in a buffer between the mixer and filter to terminate the mixer out of band)?

Agreed that LO phase noise can only really be dealt with at source, and that some of the very wideband VCOs and synthesizers leave a lot to be desired, the saving grace here is the move to DDS which (if you use the good stuff) makes the dominant close in PN component that of the crystal/SAW oscillator. I will take a few low level spurs over PN any day as the spur requires you to get unlucky with a strong carrier, while PN will mix with everything out there.

Still seems to me that doing transmit BPF 40dB earlier in the chain makes more sense then doing it at the 100W level.

73 Dan.

Actually NFB does still help, for all that one usually cannot apply as much of it as would really be desirable.

The pain is that most RF stages can only run a few dB of NFB because the open loop gain just is not there and feedback around multiple stages raises stability issues.
That modern LDMOS has in the region of 20dB open loop gain at HF makes NFB a far more useful technique then it was with the old 10dB bipolars, but far too often this is seen as an excuse to reduce the number of gain stages rather than increase the linearity of the whole chain. 

73 Dan.

This actually begs a question:

How much of the PN is down to the low power stages in a transceiver as opposed to the stages running at more then say 20dbm?

If most of the noise is down to the LO and low power stages could not a pair of relays be added to switch the receive bandpass network in between the mixer buffer and the pre driver on transmit?
You want to end up with the filters as late in the chain as possible, consistent with the drive power not being excessive.

This would effectively eliminate the phase noise contribution from the low power stages leaving only the noise inherent in the medium and high power amplifier stages, which if they are starting at +20dbm or so is not likely to be overly significant.

Seems to me that doing the work at low power usually trumps doing it at high power, particularly as the rig likely has suitable filters already.

Of course if the PN is coming from an overly noisy PA strip then to do any good filtering must be at the high power output, but I have a hunch that is not usually the case.

73 Dan.

If I was to try a 20 year repairable amp, I guess I would probably use 50V LDMOS, because much of it is more or less interchangeable and there is a ton of it out there, MRF151/VRF151/BLF177/SD29... All of it will work in more or less the same board with at most minimal changes.

Go SMT for the lot, because availability is frankly going to be better, and anyone doing repair work will have to get familiar with it.

I am far from convinced that actually it much matters what you use, the money is in the thermal management and output filters, not the RF gain parts so much, so if in ten years someone needs to cook a new PA module for the thing using whatever is new and shiny then that is not a big deal. 

Similarly, power supplies are not that big a deal anymore, the advent of switchmode power for telecoms applications has seen to that and provided a wide range of off the shelf power supplies entirely suited to our needs.

Specify the interfaces between the controls and the RF deck, use standard connectors and provide a drill diagram for the heatsink (so if I have to do a new amplifier board sometime I can make it fit the heatsink) and I don't see anything to make me think a 20 year life is impossible or even all that unlikely, but the thing will be at least as expensive as (inflation adjusted) something solid from any of the old players.

Forgo the micro (Possibly difficult with the US market and the 11M sillyness you guys have), most everything can be handled in trivial discreet logic (Or use a micro, but provide full documentation so I can sub something else). 

The key thing to making a repairable radio actually is not what technology you use, but is that you supply ample technical documentation, ideally including a BOM, schematics, commented VHDL or Verilog sources, mechanical drawings, G code, assembly drawings, setup instructions, JTAG boundary scan sequences, the lot. Sure it means I can duplicate your radio, so what, by the time I get one you should already be working on the next big thing.

73 Dan. 

I am not really convinced by the 'modern kit is too complex to fix' argument, I mean sure the techniques have changed, VHDL toolchains instead of bench drills and VTVMs for example, but ultimately if you can figure out how the thing works (And radios really are not that complicated) you can usually bodge up a replacement for an unavailable part.

Sure if you loose a custom LCD panel or big ASIC the repair can be a pain and can end up with a rig that looks a little Frankenstein, but end of the day, a DDS is in some ways simpler then a good VFO, a mixer is a mixer and most SSPA strips are pretty similar. 

Move the DSP to a small embedded board if the custom DSP is unavailable, given the rate of progress in computer power your little embedded board probably stomps the rigs original DSP anyway.

The big pain is not the parts, it is the diagrams, and the source code (Including the HDL source code), really the two should be included in the documentation if the rig is to be considered properly fixable.

In reality I have to question the extent to which self repair was practiced on any kit more modern then the 1960s, my instinct is that most folks took rigs to a repair shop, and this has not changed significantly. 

The cost in any product is generally NOT the small signal electronics, you can design a lot of that in with only a very small effect on the per unit cost, it is the metalwork and the power stuff that really costs, and that never gets much cheaper, so I can see an argument for throwing electronics at the problem if it adds features without significantly impacting the overall unit cost.

KK3AN: You might want to think about a set of relay contacts brought out for 'alarm' monitoring outputs, or if the thing talks TCP/IP anyway maybe SNMP? For the rack of amplifiers in a lights out rack room somewhere, both are extremely useful.

73 Dan (Who considers 'No user serviceable parts inside' to be a challenge).

Note that those are switching parts and not designed for linear service, note also that the 730W thermal rating of the DE475 is given at 25 degrees C case temperature and with a derating factor of 4W/C (The others are worse)......

For RF plasma generators and such in the 13.56Mhz ISM band (clearly what these are designed for) these are a solid choice, but I doubt you would get good performance in a linear amplifier.

TANSTAAFL applies to RF power devices.

73 Dan.

And there are a useful range of smt mount connectors available as well, so even for things you are going to keep around interconnection is not a big issue (See the samtec catalog for the sort of thing I have in mind).

Sometimes you need a BGA however simply because you need something like a large FPGA and the large ones only come in packages with lots of IO, irrespective of need for it, and that means BGA, still there are ways to get even these (And large QFNs and CSPs that can also be a pain) mounted, you just cannot do it yourself.

Anyone else out there in Hamburg for the CCC meeting?

73 Dan.

Sounds like a job for EB104 to me (Or one of its more modern variants).

This is not a trivial project by any means, and you will, especially if you have no experience, blow up at least some of the expensive sand before you get the protection circuits right, I may well be tempted to bite the bullet and just buy a KPA500 as a part built 'kit', at least that is known to work well and will have support. 

Note that 400W is only 6db up on 100W, and you can easily make more difference then that with a better aerial system (Which also helps on RX), so make sure that is all it can be first.

73 Dan.

No chance mate!

In a push pull stage the drain voltage peaks at twice the supply, due to transformer action in the DC injection bifilar.
I don't see the RD16 surviving 100V somehow.

Regards, Dan.

Ouch!
Karma is a bitch sometimes!

I am guessing someone now owned by Thales group?

Regards, Dan.

I thought Cdg usually dropped over the first part of the operating region then became more or less stable as Vds increased?

73 Dan.

Never tried it but note that the miller capacitance will become an increasing pain in the arse at higher drain voltages, still 20V might be Ok if the reduction in SWR tolerance is an acceptable trade off, not sure I would want to go higher than that.

I have wondered about the possibility of isolating the tab and doing source degeneration on those parts, maybe a handful of 1 ohm resistors from the tab to the ground plane? Gain will come down, gate impedance will go up linearity will improve, question is will it be stable? 

Regards, Dan.

From the data sheet, carefully, paying attention to operating mode, and specified cooling conditions (There is a common trick to rate sand at 25C case temperature which can never be maintained in practice....IXYS Looking at YOU!).

Most RF power semis specify various limits (Pd, Tj, current, voltage), so it is not really a one number sort of game.

For typical class AB service you can assume somewhere around 50% efficiency in a typical amateur service design (It is possibly to do much better in a more specialized architecture), so a 100W Pd part is probably good for about that much output if properly cooled and matched. 
In a purely SSB service application it may be permissible to go a bit higher on PEP by lowering the drain match impedance at the cost of reliability (You get away with this because SSB has a very high power back off ratio), I wouldn't do it, but you can if smoked power fets are not a big issue to you.

73 Dan.