Retuning a radio front-end

[philpem]

Another one of my trademark strange projects 

I've got my hands on a box full of scrap radios which were used on a commercial project at around 200MHz. They can be reconfigured for 175MHz with solder bridges (usually changing coil taps or adding extra capacitors). I'm curious whether one of these could be converted to act as a receiver on the 144MHz (2-metre) ham band. At the moment I'm only interested in receive (I've currently got the power amp disabled with a piece of wire).

I've had a crack at it already -- the radio is controlled by a 68HC11A1, and I've written new firmware which can set the VCOs to whatever frequency I like, and a few strategically placed capacitors have altered the VCO range to give me 144 to 146MHz, less the 21.4MHz IF. The problem is, nothing is getting past the RF filter, and I'm not sure the mixer is working at the new LO frequency either.

I don't know what the value of the "#642" inductors is. They're white coil formers with tin cans and an adjustable ferrite slug. My Peak Atlas LCR reads them as either zero or 0.1uH depending on how it's feeling at the time.

So from what I can gather, Q16 and Q17 form the mixer, but they don't form any topology I recognise. Dual-JFET mixers aren't in any of my RF books. I expect the LC circuitry is there to phase-shift the local oscillator, but by how much and why I have no idea. I'm guessing 180 degrees given how the transformer is positioned.

This leaves the RF input stage and filters -- L11, Q15 and surrounding capacitors seem to be an RF input amplifier, and I think the LC circuitry is to restrict the frequency response. I plan to simulate that on SPICE later.

L12, L13 and L14 and surrounding capacitors all form part of the filter I'm stuck with. I can tell it's a three pole filter (or it looks like one) but I don't recognise the topology at all. I simulated it in LTSPICE and found out that it had a bandpass response, and inductance values between 80 and 160 nanohenries produced a reasonable frequency response in line with the frequency ranges quoted on the schematic. I can't reasonably change the inductor values without lots of desoldering and rewinding, so I changed a few of the capacitor values in LTSPICE and managed to get something resembling the frequency response I wanted. However, I don't just want to bodge it -- I'd ideally like to calculate the correct values instead of guessing and testing.

Can anyone tell me what I should be looking up to find out more about this filter topology, and how I might modify it to lower the frequency?

I've attached the schematic; the radio hardware is on the last page.

Thanks,
Phil.

[G0HZU]

I've only glanced over it but the cheap and cheerful way to do it is to simply pad the BPF inductors with caps. eg try 3P3 as in the figure below and then retune the inductors as required..

NOTE: You also need to alter the 12p cap at the input elliptic HPF to the RF amplifier to maybe 22pF in order to get the image reject notch to shift down. At 12pF it will be nulling somewhere near to 144MHz!

You also have to fit the links as shown in red. Note that the link 5A by the mixer would ideally have a couple of turns of wire instead of a straight/short link. You could try fitting a similar 2 turn inductor to the other red A links in the BPF instead of fitting the 3P3 caps but you may run into issues with unwanted cross coupling between these and it may mess up the BPF behaviour.

Note that by fitting the 3P3 caps you will also reduce the bandwidth of the BPF to just a few MHz. But I guess this is OK for 2m ham band use.

I'll look at it in more detail tomorrow.
HTH

[Richard Head]

Phil
CC95 and L11 is a trap, probably set to the image frequency to improve image rejection.
L12, L13 and L14 are part of a three stage coupled resonator BPF. The resonant capacitors in each tuned circuit are actually 5.6pF/2. Splitting the resonant capacitor in half allows the coupling capacitors to be a more managable value, otherwise they would have to be 0.9pF which is difficult to find. It is a bog standard coupled resonantor design.
You really need a spectrum analyser with tracking gen (or VNA) to set up the BPF properly.
The mixer is a double balanced active FET mixer. It's like a double balanced diode ring mixer but uses switching FETs rather than switching diodes.
Dick

[G0HZU]

A few more things to add...
The dual FET mixer requires a 180degree phase shift of the LO signal at each of its gates. There also needs to be a reasonable voltage swing here to exploit the square law response of the JFETs.

So the network involving the #642 coil L15 is designed to feed the gates like this. It will need to be peaked up at the new LO frequency which will be about 30MHz lower for operation on the 2m band.

You can still get reasonable gain and 180deg phase performance here by just changing the value of L15. However, you could get closer to ideal performance by changing the caps as well. However, I doubt the cap changes will make much difference to the overall performance. You could also optimise the network for exactly the same amount of voltage swing on each gate but I'm not sure this aspect of the mixer performance is really that critcal.

If you want to go down the path of redesigning the BPF then you will end up changing lots of capacitors in that BPF and you will have the hassle of desoldering/swapping lots of caps.

However, you would first need to decide what bandwidth you want at 144MHz because you now have control over this and then (ideally) design and optimise it on a linear simulator and then fit the new parts and tune it up on a VNA or spectrum analyser and tracking generator as Dick says.

Because of the number of local pager signals at 138MHz and 153MHz in the UK I would recommend you try for a low bandwidth at 144MHz. eg 8MHz?

The quick/cheapo option is to just fit the additional caps I suggested in my first reply and see how that performs after you repeak the #642 coils L12, L13 and L14 at 144MHz to suit the 3P3 caps. L11 would ideally be tuned to the image frequency assuming the image lies on the low side of the band. The bandwidth of the overall front end BPF will go down to maybe half of what it was originally. But you will definitely need to change the 12pF cap at the input to shift the null in the elliptic HPF well down below 144MHz. If you leave it at 12pF it will corrupt and attenuate the overall response of the preselector at 144MHz by maybe 20-30dB making the front end very deaf. You might be able to shift it down by just adjusting L11 but I doubt it will have enough tuning range without increasing the 12pF cap CC95 as well.

[KJDS]

It would be worthwhile redesigning the filter.

I'd put a fixed inductor, or at least a couple of turns of wire in place of the three link 'A's to reduce the frequency and increase CC98, 103, 111 and reduce the ones underneath those to narrow the bandwidth.

RFSim will get you reasonably close, but I'd want a VNA to tune it. The image reject filter and mixer lumped balun have already been covered.

[philpem]

I've only glanced over it but the cheap and cheerful way to do it is to simply pad the BPF inductors with caps. eg try 3P3 as in the figure below and then retune the inductors as required..

NOTE: You also need to alter the 12p cap at the input elliptic HPF to the RF amplifier to maybe 22pF in order to get the image reject notch to shift down. At 12pF it will be nulling somewhere near to 144MHz!

Thanks for that -- I had been increasing the value of all the capacitors. Adding extras to the filter legs would be much easier, thanks to the ground plane on the PCB

How did you come up with the 3p3 value? Rough guess?

And yeah -- I figured the input filter would have to be "tweaked" too.

You also have to fit the links as shown in red. Note that the link 5A by the mixer would ideally have a couple of turns of wire instead of a straight/short link. You could try fitting a similar 2 turn inductor to the other red A links in the BPF instead of fitting the 3P3 caps but you may run into issues with unwanted cross coupling between these and it may mess up the BPF behaviour.

I think I'll stick with the caps for those. I could possibly re-wind the inductors (the screening cans and bobbins are separate) but that seems like a lot of work for not a lot of gain. Adding a 2-turn inductor to LK5A might be tricky - I think it might be a pair of solder pads on the bottom of the PCB, under the screening can!

Note that by fitting the 3P3 caps you will also reduce the bandwidth of the BPF to just a few MHz. But I guess this is OK for 2m ham band use.

That probably won't be an issue, and may actually be beneficial with all the high-power stuff (pagers etc.) on either side of the UK ham band.

If you want to go down the path of redesigning the BPF then you will end up changing lots of capacitors in that BPF and you will have the hassle of desoldering/swapping lots of caps.

However, you would first need to decide what bandwidth you want at 144MHz because you now have control over this and then (ideally) design and optimise it on a linear simulator and then fit the new parts and tune it up on a VNA or spectrum analyser and tracking generator as Dick says.

I've got a fairly nice Advantest R3361A spectrum analyser with TG on my workbench, and also a Marconi 2022E signal generator. I've also got homebrew 40dB RF taps and a fairly chunky 50-ohm dummy load (I think it's specced for 50 Watts?).

The only problem is, I don't have a high-impedance probe for the SA, which means any probing I do will load down the filter quite heavily...

The quick/cheapo option is to just fit the additional caps I suggested in my first reply and see how that performs after you repeak the #642 coils L12, L13 and L14 at 144MHz to suit the 3P3 caps. L11 would ideally be tuned to the image frequency assuming the image lies on the low side of the band.

If it helps, the tuning is "low-side"; when receiving, Fvco = Frf - Fif. So if you wanted to tune 144MHz, you'd actually tune the VCO to 144MHz - 21.4MHz, or 122.6MHz.

The bandwidth of the overall front end BPF will go down to maybe half of what it was originally. But you will definitely need to change the 12pF cap at the input to shift the null in the elliptic HPF well down below 144MHz. If you leave it at 12pF it will corrupt and attenuate the overall response of the preselector at 144MHz by maybe 20-30dB making the front end very deaf. You might be able to shift it down by just adjusting L11 but I doubt it will have enough tuning range without increasing the 12pF cap CC95 as well.

I did notice it was pretty deaf when I connected up the SG. I took it as far as -50dBm, but didn't dare go any further.

CC95 and L11 is a trap, probably set to the image frequency to improve image rejection.
L12, L13 and L14 are part of a three stage coupled resonator BPF. The resonant capacitors in each tuned circuit are actually 5.6pF/2. Splitting the resonant capacitor in half allows the coupling capacitors to be a more managable value, otherwise they would have to be 0.9pF which is difficult to find. It is a bog standard coupled resonantor design.
You really need a spectrum analyser with tracking gen (or VNA) to set up the BPF properly.
The mixer is a double balanced active FET mixer. It's like a double balanced diode ring mixer but uses switching FETs rather than switching diodes.

Ahh -- that (the capacitor trick) makes sense now you've explained it. Thanks!

It would be worthwhile redesigning the filter.

I'd put a fixed inductor, or at least a couple of turns of wire in place of the three link 'A's to reduce the frequency and increase CC98, 103, 111 and reduce the ones underneath those to narrow the bandwidth.

RFSim will get you reasonably close, but I'd want a VNA to tune it. The image reject filter and mixer lumped balun have already been covered.

I'm with you on this one, KJDS, but that's a lot of design work for the sake of a proof-of-concept...


The one remaining question is, what do I have to do to make the transmitter work on the new frequency (if I even want to), but that's a question for later


Thanks,
Phil.

[G0HZU]

How did you come up with the 3p3 value? Rough guess?

Yes, I looked at each filter section and mentally worked out that there was about 7pF in parallel with the inductor if you include the self capacitance of the inductor itself.

So I worked out that the inductor will probably tune to a ballpark 120nH to get resonance up at 175MHz. So for resonance at 145MHz the inductor would need to 'grow' to 175nH. You can see that the current method to change bands is to change the inductance with the links.

I then worked out what capacitor was needed across 120nH in order to make its equivalent inductance (at 145MHz) to be 175nH. i.e. the capacitance needed to 'grow' the 120nH inductance to 175nH at 145MHz.

This turns out to be just over 3pF.

However,  the penalty for this method is reduced bandwidth in the BPF. You might be best to try 2.7pF caps if the filter becomes too peaky or lossy with this method. Hopefully there will be enough tuning range left in the inductor to still peak up the filter at 145MHz with the 2.7pF caps.

[KJDS]

One word of caution in narrowing the filter, and that is doing so will increase the loss. Whether it's worth it will depend on how much interference is nearby.

[G0HZU]

I guess a lot depends on the unloaded Q of the #642 coils at 145MHz wrt the loaded Q of the BPF design. If the Qu is medium/high eg Qu = 180 then it will probably be OK.

However if the Qu of those coils at 145MHz is only 100 then the change to the narrower bandwidth will cause an extra loss of maybe 3dB. This is because I'd estimate the original circuit probably has a loaded Q of about 30-40 and the narrower circuit with the 3p3 padding caps will have a loaded Q of over 50.

What do the #642 coils look like? I'm assuming they are basic coils inside something like a 7mm or 10mm screening can. Obviously, the tighter the metal screening can is compared to the coil diameter inside, the lower the unloaded Q will be for the coil because the screening can will damp the Qu of the coil. Also it depends on what material the tuning slug is made of. eg powdered iron or ferrite or aluminium.

Also.. just looking at the circuit there is a 4K7 resistor R100 across L12. Assuming L12 is approx 120nH then this resistor will 'heavily' swamp the Qu of L12. So it may be worth experimenting with raising the value of R100 from 4K7. This will help a bit with bandwidth and also stage gain if you try the 3P3 mod. Don't overdo it at R100 or the RF amplifier may go unstable. I assume that the 100pF cap C108 on the base of the RF amplifier is fitted snugly here because it will be needed to prevent instability way up in the UHF region. i.e. the base of Q15 will need to have a very low impedance to ground right up into the UHF region or the inherent stability of the common base topology will be lost. You don't want much in the way of stray series inductance here!  I'd expect that the quality of this ground will dictate how far you can raise the value of R100. You could try improving the layout of the 100pF cap C108 or try padding it with something smaller. If Q15 hoots it will most likely do it way out of band way up into the UHF region and I'd expect you to hear lots of hissy noise down at 145MHz if it started hooting.

[philpem]

What do the #642 coils look like? I'm assuming they are basic coils inside something like a 7mm or 10mm screening can. Obviously, the tighter the metal screening can is compared to the coil diameter inside, the lower the unloaded Q will be for the coil because the screening can will damp the Qu of the coil. Also it depends on what material the tuning slug is made of. eg powdered iron or ferrite or aluminium.

The base is just shy of 12mm; the coil former itself is 7mm with threads cut into it to take the wire. The plastic is white, if that makes any difference, and you're quite right - the screening can is separate (and a complete pig to desolder, even with a Metcal).
The slug looks like it's made of ferrite, it's certainly brittle enough. I've broken two of them so far; had to smash them, remove the bits and swap slugs from the spare-parts breaker radio. Not fun...

Also.. just looking at the circuit there is a 4K7 resistor R100 across L12. Assuming L12 is approx 120nH then this resistor will 'heavily' swamp the Qu of L12. So it may be worth experimenting with raising the value of R100 from 4K7. This will help a bit with bandwidth and also stage gain if you try the 3P3 mod. Don't overdo it at R100 or the RF amplifier may go unstable. I assume that the 100pF cap C108 on the base of the RF amplifier is fitted snugly here because it will be needed to prevent instability way up in the UHF region. i.e. the base of Q15 will need to have a very low impedance to ground right up into the UHF region or the inherent stability of the common base topology will be lost. You don't want much in the way of stray series inductance here!  I'd expect that the quality of this ground will dictate how far you can raise the value of R100. You could try improving the layout of the 100pF cap C108 or try padding it with something smaller. If Q15 hoots it will most likely do it way out of band way up into the UHF region and I'd expect you to hear lots of hissy noise down at 145MHz if it started hooting.

Thanks for that -- I'll have a look. Really need to build that Elektor high-impedance RF probe first, though... 

[G0HZU]

I don't know what the value of the "#642" inductors is. They're white coil formers with tin cans and an adjustable ferrite slug. My Peak Atlas LCR reads them as either zero or 0.1uH depending on how it's feeling at the time.

Yes, the Peak Atlas LCR meter is going to be a waste of time for RF stuff.

I can show you how to measure the inductance and unloaded Q of those inductors at 145MHz using your spectrum analyser and tracking gen if that helps?
You would be able to measure the full inductance range of the coils up at VHF with this method as well if you adjusted the slug from end to end.

I usually measure the unloaded Q of coils like this using three methods and I hope to see similar results for all three. Two methods use a VNA but the other method can use a spectrum analyser and a tracking gen and it should give similar accuracy. But you have to build a little test jig first and you also need a decent high Q 10pF ceramic capacitor and you need to extract a coil + screening can from your parts radio. I can show you how to do this Q measuring method in another post if it helps?

I measured an old Toko MC120 coil set to 120nH and got Qu =170 @145MHz for two methods and Qu = 160 with the crudest VNA method. The datasheet for the MC120 suggests the Q will be slightly lower but this is at a lower test frequency. The Toko MC120 coil is very similar in construction to the coils in your radio.

I'd expect you to get a Qu of >180 because your coils are slightly bigger with looser screening cans. But anything over 140 would be fine IMO.

[philpem]

I can show you how to measure the inductance and unloaded Q of those inductors at 145MHz using your spectrum analyser and tracking gen if that helps?
You would be able to measure the full inductance range of the coils up at VHF with this method as well if you adjusted the slug from end to end.

That'd be great, if you could!
It sounds like something that'd be very useful to know.


I've got the radio open, added the capacitors and disabled squelch -- it's still as deaf as a post. All I'm getting from the speaker is noise. LO is dead on frequency (or what I think the frequency should be), but I probably can't check the filter inputs/outputs with the SA until I've built or acquired a high-impedance probe. That'll be a job for this weekend, assuming I can find the parts I bought to build it... (they're around here somewhere!)

Cheers,
Phil.

[G0HZU]

Here's a couple of images to show the 3dB BW method of measuring Qu of a resonator.

You need an ultra high Q 10pF cap (@145MHz) for best results and I used an ATC 800B series 10pF cap.

See below for a simulation and an image of the real jig. This jig is designed for use at VHF. It won't be much use at lower frequencies or up at UHF.

The idea is to connect the coil and the cap in series and connect one end to ground. Then feed the other end with an ultra low impedance source that gets fed from your tracking generator. Then sniff near the centre of the LC network with the receive port of the analyser. Don't touch the LC network with the sniff connection. You just want to hold it fairly close and get coupling at a tiny fraction of a picofarad.

To get an ultra low Z source you have to make a section of low impedance microstrip transmission line and the low Z point is taken at a tap point along the line. I did it with a strip of copper and used a skinny piece of card or plastic as the dielectric/insulator. You can see the white card sandwiched between the copper strip and the copper plane of the PCB underneath. Then connect the LC network to the microstrip tap point and ground as per the images.

The Q will be 145MHz/(3dB bandwidth in MHz)

So if the 3dB BW was 1MHz the Q would be 145.

In the simulation the 3dB BW is 145.420-144.563 = 0.857

The Qu = 145/0.857 = 169.2. This is very close to the Qu of 170 shown in the model for the inductor in the simulation.

I tested an old Toko MC120 coil with the jig on my Advantest analyser+ TGen. You can see in the last image below that it has a 3dB BW of about 0.885MHz. Therefore, this coil has a Qu of about 164 at 145MHz.

Because it resonates with a 10pF cap at 145MHz then it's inductance must be about 120nH at 145MHz. So with this little jig it is possible to measure both Qu and the inductance at 145MHz. The Qu of the 10pF cap must be very high in comparison. In reality, the Qu of this cap will be around 3000 at 145MHz and this is enough to cause a slight error with this method. i.e. the cap will spoil the measurement slightly and the Qu of this coil will be slightly higher than 164.
Hope this makes sense. If you need more info then let me know...

[G0HZU]

The other way to do it is to simply track through with the LC network in shunt as in the image below. You have to ground the screening can as well and I've done this at the back of the can.

The Q will be proportional to the notch depth at resonance. i.e. at resonance you have a resistive potential divider and you can work out the Q from the notch depth.

However, this method relies on the source and load impedance of your test gear to be a very accurate 50R.
But this method is quicker and it gives a ballpark estimate. You can improve it by fitting precision 10dB attenuators at the SMA connectors in the image and this will lessen the uncertainty. In reality there will be a lot more uncertainty with this method compared to the 3dB BW method in my previous post.

[G0HZU]

For completeness, here's the VNA results for the MC120 coil.

If I cal up the VNA to do a component measurement in a test fixture (with the coil's screening can fitted and grounded)  then I get the VNA plot as below.

This measures the impedance of the coil on its own. i.e. without the 10pF cap.

This method needs very careful use of the VNA to calibrate it because the reflection coefficient is very close to 1. So care is needed to get a result that is meaningful.

But it shows 120nH  X= 111.2 ohm and Rs = 0.67 ohm. So the Qu will be 111.2/0.67 = 166


The other VNA method is to resonate it in series with the cap and then measure either side of resonance for the magX = Rreal points. These correspond to the 3dB BW if it were tested as a resonator and I get 144.28MHz and 145.13MHz. This has a BW of 850kHz and the Q will be 145/0.85 = 170.5.

All the results agree quite well although some skill/experience is needed with all of them in order to get reasonably accurate results. To get within 10% accuracy is difficult with any of these methods but I'd hope to be well within 20% accuracy

[philpem]

Thanks for the details, G0HZU! I don't have an ultra-high-Q capacitor to hand (that I know of), so it might be tricky to do that, but I'll file that information away for reference

In the meantime, I built one of the Elektor 1GHz high-impedance probes and while the amplitude accuracy is nothing to write home about (probably because I built it with 5% resistors), it does the job. I've tuned the front end on the T545TR as good as I can, though sadly two of the ferrite slugs (L13 and L14) cracked while I was doing the fine tuning. It's not as good as I'd like it, but probably good enough.

I've attached some spectrum traces from the filter with the tracking generator connected to the radio's RF In -- L12p4, L13p4 and L14p4.jpg are L12, L13 and L14 pin 4 respectively (L2 pin 4 also being Q15's collector). Q15e.jpg is the emitter of Q15. I didn't expect the... somewhat odd frequency response of the intermediate stages.

Sadly while the RF filter works fine, I can't get the mixer to work at all!
Adding a few turns of 24SWG enamelled wire on LKA-5 didn't seem to rein it in at all. I tried coils from about 2mm to 7mm diameter, 2 to 8 turns (wound on my trimming tools!). The mixer is clearly working -- touching L15 pin 4 with a soldering iron seems to get things going and the shrill whistle of my Marconi SG's modulation generator becomes audible. Touching said junction with my finger does similar; I've noted similar effects with other junctions around the mixer (CC115 for one). Based on that, I tried adding a 3.3pF over C113 and C116 alternately but it didn't seem to have any effect.

Does anyone have any suggestions for other tricks which might get the mixer going?

(I'm actually slightly tempted to bodge in an SA612 if only to see what happens...)

Thanks,
Phil.

[Andy Watson]

Have you confirmed that the output of the RX-VCO is still present and at an appropriate level at the new lower frequency? There appear to be more tuning options around C191 and C192.

[G0HZU]

Thanks for the details, G0HZU! I don't have an ultra-high-Q capacitor to hand (that I know of), so it might be tricky to do that, but I'll file that information away for reference

You can use a standard SMD ceramic cap eg a common or garden COG 0805 10pF but the Q you measure will be a bit lower because the cap Q will slightly damp the Q of the coil. I just tried a cheapo cap like this and got a Q of 153 for the MC120 coil. It measured about 165 with the high Q cap.

Note: I wouldn't have used the Hi Z probe. I would have fed the analyser RF input to the tapped output of the BPF after isolating the mixer. The impedance looking into the FET mixer should be in the ballpark of 50 ohm so the BPF will be designed for a low output impedance. So you should be able to measure the gain up to the mixer input like this. I'd expect to see about 10-12dB gain to this point wrt the receiver input.

[philpem]

I finally got around to doing the measurement.

I found a scrap of double-sided FR4 in my drawer, which was sanded clean and pressed into service. 12.7mm wide copper foil tape provided the transmission lines (two pieces, one 7mm long and one 25mm long) and a piece of thin card was cut to size to form the dielectric. A right-angle SMA socket served as the input from the TG (though I had to dig around for an SMA-to-SMA cable for my SA) and a BNC panel socket with solder lug formed the output to the SA.

The capacitor was a 10pF 0805 Johanson Dielectrics thing I had sat in a cap kit on my workbench. 5% tolerance apparently. Sadly the best resistor I could find was a 47R, 5% carbon film.

I tried the inductor in a few different tuning positions -- the lowest frequency I managed to achieve was 138.56MHz and 617kHz 3dB-BW (~115nH, Q=~225).
With the slug only just engaging in the threads, I got 172.343MHz and 785kHz 3dB-BW (~85nH, Q=~220). [EDIT: Inductance calculated with the aid of http://www.pronine.ca/lcf.htm]

By my quick-hack SPICE simulation of the mixer (changing only the value of L15), the mixer would need L15 ~= 230nH to work effectively at 145MHz. In contrast, for a stock configuration, tuned for 187.5MHz (LO=166MHz), L15 would need to be ~= 110nH.

The problem was, adding a few turns of wire on the PCB didn't really help. I think I might end up adding a few extra turns to the inductors themselves. Thankfully I have a completely dead radio which I've been stealing parts from (notably an almost complete set of coils as my "experiment" radio had a bunch of cracked or broken tuning slugs, and almost the entire FM IF and demod stage for a 198kHz Radio 4 phase-mod data decoder).

Cheers,
Phil.

[philpem]

Right then, after re-winding the coil with 5.5 turns of 22SWG tinned copper and sacrificing three more 10pF chip capacitors to the cause (they're fragile and I'm clumsy, that's my story and I'm sticking by it!), I ended up with a coil which was adjustable from 207nH to 260nH.

Slug level with the top of the core out at 110.62MHz, BW=469kHz, for Q~=236 and L~=207nH.
Slug driven in most of the way gave 98.596MHz, BW=426kHz, for Q~=231 and L~=261nH.

Which means I've got enough adjustment range that my target is somewhere around half way along the track.

I think it's time to see if the SPICE simulation was anywhere near accurate... I've attached said simulation to this post. The JFET model came from the Philips model library, mirrored at http://www.gunthard-kraus.de/Spice_Model_CD/Vendor%20List/Spice-Models-collection/phil_fet.lib. After switching to that model from the one I was previously using (of unknown provenance), I found that L=95nH was sufficient in the 175-200MHz band.

EDIT: I just re-ran the simulation for F_RF=145MHz, F_LO=123.6MHz, F_IF=21.4MHz. Turns out L=160nH is more like it for a modified radio. Guess I need to take a turn off this coil.

EDIT2: Getting somewhere! The LO on this radio is a hundred or so KHz off frequency; I thought it was tuned to 145MHz, it's actually on 144.895MHz. That suggests I need to tune the reference oscillator.
Also got the coil wound for the higher inductance (L=~200 to 260nH) and the mixer seems to be mixing. Tweaking the trimmer doesn't seem to make much difference, so maybe I do need one less turn. Who knows? I'll do the mixer alignment tomorrow and see what I get. At least I'm getting somewhere!

EDIT3: A software bug was causing the tuning to be off by -100kHz. That leaves the 5kHz shift which is probably adjustable with IF trim. I'll do a full alignment tomorrow, but it's 2am now and I need some sleep.

EDIT4 29/12/14 11:13: Just done a quick LO realignment with the frequency counter. Gone from 5kHz off to within a couple of Hz (spec is +/- 300Hz). Also trimmed the mixer according to the service manual:

CHECK:
Short the gates of the two mixer FETs
Connect a voltmeter to the junction of R107/C112 via a 33k isolating resistor and measure the voltage. It should be about 2V.
Remove the short and note the increase. This should be ~0.5V at the band edges and ~2V at the band centre.

ALIGN:
Select the middle channel.
Align L15 for a maximum reading at the R107/C112 junction [via 33k isolating resistor]
Select the lowest frequency channel and note the injection voltage.
Select the highest frequency channel and adjust L15 for the same reading as the lowest frequency channel.
Repeat the two previous steps until the mixer reading is identical on the highest and lowest frequency channels.

I've had to take a turn off, as the slug was barely inside the former when I finished aligning the mixer (the adjustment was very touchy - literally, if you think it's moved, it's moved too far!). It's still touchy now, but the slug is at least inside the former. So the updated simulation (with the Philips JFET model) is probably right at 160nH ish.

I just wish I had a SINAD meter to see how the sensitivity of the radio has been affected by retuning it... and maybe see if I can improve on the RF tuning by following Tait's instructions instead of "guesstimating" it with the spectrum analyser.


Cheers,
Phil.