I also view it as one of a long list of things I want to play with some day, but there are more interesting engineering problems....
Some day!

Regards, Dan.

Yep, and it is worth noting that even if you have narrow selectivity filters immediately behind the roofing filter, these often sacrifice IMD3 performance for shape factor so a good roofer helps even then.

All of this does of course only show under contest conditions,  and even then everything else needs to be spot on to make the IF filters the weak point.

Regards, Dan.

Back in the day when the throttle and clutch were controlled mechanically by wires run in spiral metal tubes, you sometimes used to see these tubes heavily burned where the link between the engine block and chassis had come adrift and the starter current had found an alternative ground path!

In the UK there is actually a standard relating to this, http://www.fcs.org.uk/my%20files/fcs_pdfs/codesofpractice/fcs1362_2010.pdf aimed mostly at business radio rather then ham but the principles are the same.

Regards, Dan.

These amps seldom run anywhere near the optimum drain load point anyway, for the simple reason that the match will only really be right at ONE power level (PEP), everywhere else the thing will be matching the drain at a lower then optimum impedance, so no with a power level 20db down, the optimum match for the harmonic energy would be to a drain impedance 10 times what the network will get you.

On 'mixer' behavior, I would not that the devices clearly are non linear as we need output filters, which we wouldn't if the stage was truly linear, and with a third harmonic typically being somewhere around -15dBc or so unfiltered, we even have a reasonable indication of how non linear.
S21 obviously != S12, but still clearly gain and Vds are related so there is a mixer mechanism there.

Further,  in VHF PMR service where multiple transmitters often share a single hilltop tower, spurs produced by PA intermod (with received RF from colocated transmitters) is a well known issue and it is common practice to store a before and after spectrum to be able to demonstrate that nothing has gone pearshaped when a new transmitter is installed.

I think some experiments are in order.

Regards, Dan.

Are you sure about that?

Most solid state HF amps use broadband matching networks (which after all is why we need the output filter) which are typically designed to provide something approaching a match from 1.8 - 30Mhz or so. Certainly if operating on a low band the harmonics will fall within the designed range of the broadband match.

The win (if there is one) with diplexers is probably more to do with reduced circulating currents (or peak drain voltage) and getting some heat away from the devices then IMD (at least on aggregate, there will be cases where significant improvement is measured).

Regards, Dan.

For us legal limit, I personally would be looking at multiple devices with power splitters and combiners, MUCH easier to cool and you can test each amplifier separately at maybe 4 or 500W output.
I do this even for UK legal limit (400W at the feedpoint, so maybe 500W TPO), 4 * BLW96 (Two, two transistor amps) makes a nice 500W set that is robust (800W nominal transistor rating). I have a bucket full of the things but would not otherwise recommend them for a new design.

The lower per device dissipation means the heatsinks can run hotter (because the drop from the die to the heatsink is smaller), and the lower power per device makes the output match a lower ratio, so easier to build and more efficient.

4 * 600W PEP modules will run at the required ~400W reliably into a much bigger mismatch and will be much easier to cool (As well as having lower IMD and giving you a fall back to half power if one module goes south). Design the output match for the intended PEP power, not the maximum the sand is capable of, it will help the efficiency. I never like to exceed 75% of a devices data sheet rating, and 60% is happymaking.
My homebrew also tends to be designed assuming CCS not ICAS service, so it tends to be a bit big......

Do pick a 50V device, the matching tends to be less painful and you need rather more reasonable supply currents.

A LOT of the RF manufacturers have taken to rating devices at 25 degrees C case temperature, the derating to something you might actually sustain is painful (IXUS looking at you!).

Regards, Dan.

For SSB and non keydown modes I can think of a few good ways to shrink the power dissipation as well (A variation on polar loop modulation), but that does not help when you put the thing into CW or rtty mode.

Personally I like my homebrew CCS rather then ICAS, it makes for a big radio that runs wonderfully cool (And the higher rated magnetics and less stressed finals help the IMD figures too).

Regards, Dan.

You might want to consider going to a bipolar Norton feedback design like the Clifton Labs product, 10db gain and IIP3 somewhere in the +40dBm region.

For the most part preamps before the bandpass filters are IMHO the kiss of death on HF unless the aerial is electrically very short or inherently very narrowband.

regards, Dan.

Yea multi MW inverter drives are cool technology!

More worryingly then poor heatsink flatness I have anecdotal reports of non flat transistors (or even parts with inclusions in that casting!), NOT happy making.
Of course the tendency to design for the minimum possible heatsink area and to try to design out heat spreaders probably does not help.
Interestingly I have seen CPU heatsinks with a layer of soft metal foil (indium?) pre fixed to the surface which presumably makes the surface flatness somewhat less critical.

And yea, amateur service transmitters are notoriously crap, in all sorts of ways, I am not that shocked to add melting finals to their list of infelicities, you want it done right, design and build it yourself!

Regards, Dan.

Hi all,
I was hacking around with some signal processing maths and a thought occurred....

If you were to tune an AFSK signal such that the receiver tuned frequency was centered exactly between the two tones, then you would effectively see a single tone that switched between LSB and USB, rather then AFSK in the normal sense.

Now if the RX is a quadrature phasing type then (I+Q) is one sideband while (I-Q) is the other so this phase modulation can be readily detected.

Because only a single tone is now involved, it should be possible to apply synchronous demodulation to the thing together with raised cosine filters (assuming RTTY transmitters use them also) only the width required for the symbol rate (so 45hz or so) which should improve the ability to winkle weak RTTY or related signals out of the noise.

Comments?
Is it worth trying to implement?

Regards, Dan.

There have also been some reports of some MRF parts turning up with poor surface flatness on the mounting flange which is of course lethal to thermal conductivity across the device/heatsink interface.

I very much doubt that many commercial manufacturers lap the heatsink/device interface and when combined with evidently inexpert assembly this will lead to reliability issues especially if the heatsink is marginal to start with.

There may actually be something to be said for soldering the damned power devices to the heatsink or heat spreader, at least that way the easily screwed up manual assembly stage goes away, but repair becomes more then a bit painful.

I can think of several power efficiency tricks that would help significantly when running SSB mode, but all anything like that does is force a further reduction in allowable power for key down modes like RTTY and some data modes, not a good tradeoff.

I would have thought that the rig designers would have taken a leaf from the PC thermal design crowd, the heat (and power density) are similar.

Regards, Dan 2E0CHE.

Not really, 100W is only 10dB more then 10W  and propagation can easily make 50 or 60+ dB worth of difference to received signal strength at your QTH.
A good first mixer, narrow roofing and selectivity filters and use the attenuator to put the band noise just above your receivers noise floor (and thus maximize the systems dynamic range) will solve most problems. Wide waterfall displays are fun when RF conditions permit them, but they are not a requirement and should be sacrificed to make a more reliable contact when dynamic range matters without a second thought. 

1 KW is only +20dB on 10W, hardly a very significant difference at the receiver in this context.

A 500Hz or narrower first IF filter will really help as the poor phase noise and linearity performance of most transmit chains make working closer  then that problematic, never mind the tendency to use audio injection into an SSB rig with all the infelicities that offers... If you have a 250Hz filter and it has an acceptably constant group delay over a few tens of Hz that can improve things further, but just going to 500Hz from 2.4K and using the input attenuator will solve most problems.

Personally while I can make a contact on QRP power out of the transmitter, I do not consider a 10dB link margin excessive, and in fact do consider a reasonable amount of link margin to be a good thing for reliable communication. 

Particularly on the lower frequency bands the input attenuator is a much maligned control as band noise usually exceeds receiver self noise by a wide margin.

Regards, Dan.

Granted the expectation is a lot in a small box (Rather too small for my fingers in some cases), but when you look at the actual feature set on something like a FTdx5K (Sorry Vertex Standard for picking on your radio, other high end sets are similar), what you discover is that ok, it needs a DSP core and possibly also a microcontroller (You can combine the two), but once you have those anyway things like split, multiple modes, memories, audio processing on transmit and CAT become "A trivial matter of software".

I still say the major custom parts consumer is the front panel, and the major repair risk is the LCD, followed by the digital bits due to inability to obtain the image files needed to flash a stock part and as ever the power supply (Custom switcher transformers are a bigger pain then custom line frequency iron ever was).

I note that they make a marketing feature out of a custom DSP part, which I really do not see the need for, there just cannot be that much going on to warrant it.

I would note that anyone buying such a radio is going to ship it back to the manufacturer for repair and that they are going to keep a reasonable stock of spares (Kind of embarrassing to sell something like that then be unable to fix it a few years later), the real risk is probably far more in the midrange and cheap sets that are going the same way.

Of course some of us consider a broken high end rig at a rally for a few percent of its new price to be a gift from the gods, so bring it on.......

Of course commercial production has a different set of marketing imperatives, fixed costs and per unit costs then my building a radio in my basement does, that is obvious, but the architecture is very similar in both cases in all probability.

Regards, Dan.

 

That sounds painfully familiar.

I have had a few similar experiences (One of which convinced me to figure out how to build high Q notches from coupled stub lines, not hard if you have a VNA for tuning), got to hate it when a upstream suppliers supplier drops something unique (National Semi looking at YOU!).

I admit to still being surprised by seriously custom parts going into ham grade kit, guess I don't see the point, even something like a FTdx5000 does not actually push the receive performance limits that hard (I can beat the IP3DR by 10db with a homebrew radio (not easily, and not small, but it can be done using off the shelf sand)). I mean at the end of the day filters, mixers & local oscillators (which are the performance critical parts) are fairly well understood and are for the most part things you can build with parts second sourced from multiple suppliers.

I would be very surprised to see any more then a minimal amount of single sourced components going into a amateur service radio (Some of the Analogue devices stuff is a potential problem in this respect (DDS and VGA parts mainly), but push comes to shove there are other ways to do those bits), the risks of single sourced parts are just too great for the manufacturer to design them in without good reason. 

I still say the biggest liability when it comes to being able to get the things fixed is the custom buttons, knobs and (particularly) LCD  displays. These tend to be both custom parts and the ones that take damage, a custom BGA package is (Provided you are not dumping massive amounts of heat in it ("Playstation"/"X-Box" video chip style), actually really **very** reliable.

Custom OLED is actually worse as that has a very finite life expectancy due to the organic polymer breaking down with use.

BTW: I am very much an amateur at this stuff, having not worked in electronics for a long time and having no formal training in the subject whatsoever.  If anyone in the UK is looking to hire a self taught radio hacker, please talk to me!

Not modular? At that price? Ouch!

Got to admit that the modular thing does not make for a cheap (or small) radio, mine is a ~10U rack of card cages and rack boxes, with a remote operator control panel so I don't have to be in the same room as the PA blower, and yes, it is a bit like your hammer, 3 versions of the LO (Probably soon to be four), 4 revisions to the first mixer, a couple of revised IF strips... And git tells me I am up to version 179 of the firmware (Which is not PIC based (horrible things), ARM is much nicer).

Plenty of modular parts for radios out there, minicircuits have made a business out of selling building blocks terminated on SMA connectors at 50 ohms, some of them are not at all bad.

It is of course up to the purchaser of any equipment (of any type) to decide how deeply they wish to understand its operation and thus how repairable the gear will be for them (not very in the case of me and some of my Agilent kit) , and both ways have costs.

If you make the decision to buy a very heavily custom DSP based rig (and there are circumstances where this makes sense), you need to be aware of the risk cost in so doing, same if you roll a fully modular where the opportunity cost in learning to do it is very much non zero, but the chances are better that you can repair pretty much anything that goes wrong with it.

Guess for my particular interests buying a $10,000 radio takes a back seat to a better network analyzer or whatever is shiny from the used test gear dealers (It would only get taken apart so I could see how it worked anyway).

Ham radio is a broad church and whatever you buy, you should consider all the benefits, risks and opportunities it presents.

73 Dan.