Racal RA17 - 40Mc/s Alignment

[Radio Wrangler]

Ah, but you must let the victim/trainee struggle with diverging twiddlers for what feels like two lifetimes before you show them, otherwise they won't realise just how powerful these techniques are.

The s11 phase trick bogs down on very narrow filters with high insertion loss, but it's great on low-loss filters (tried it on Yagi antennae?) The couple-to-one-resonator-at-a-time version can handle them.

There was plenty of work done on filters before Zverev, but what he did was to collect, organise and unify it, and he took the approach of going for specific shapes as achievable approximations to brick walls. He made things a lot clearer.

Then there's Seymour B Cohn's work on minimum insertion loss filters for a given skirt selectivity... you just suffer whatever shape it gives you.

So you cut your teeth 'setting the tappets'? For me ir was PLLs

David

[G0HZU_JMR]

In my case as a spotty ex student I was posted in the microwave lab for a while to learn how to manage measurement uncertainty using various bits of RF test gear up to 35GHz. Due to my obvious lack of experience I was expected to make lots of naïve mistakes (which I did) but I did learn a lot very quickly from the older members of staff.

I got to play with lots of classic microwave gear including waveguide, slotted lines, waveguide frequency meters and various network analysers. I also got my first experiences with the classic HP8405A vector voltmeter. Even in those days it was becoming a bit obsolete but it was a fantastic 'hands on' learning tool. I have wanted one of these VVMs for ages and finally bought one several years ago. I've also got a classic HP778D coupler to go with it.
Even though I have a fairly modern VNA here at home the elegance and versatility of the HP8405A keeps me wanting to play with it, even if it is just to try and stay sharp on the basics

[Radio Wrangler]

The vector voltmeter really comes into its own when looking at anything which naturally runs with its own signals and it would be a pain to insert one... VERY useful on oscillators.
The digital displayed version was designed at the 'ferry by some folk I know quite well.

David

[G0HZU_JMR]

Whilst watching/listening to the World Cup Final today I had a go at designing and building a 7 resonator BPF at 21.4MHz. This used seven Toko 10K 820nH (10mm metal screening can) inductors to make up the resonators. The centre resonators have 56pF caps across them and this filter was designed to be a classic top cap (TC) coupled BPF. This type of filter isn't very symmetrical but it will be fine for this type of demo.

The design aim was to make a filter that was about 1.5MHz wide but also I wanted very good VSWR. The aim was to get close to 1.1:1 across as much of the 1.5MHz passband as possible. With 7 resonators this means that every resonator must be tuned spot on to avoid spoiling that wonderful VSWR potential. Even a tiny nudge off tune will spoil things. This means the target for the return loss has to be better than about 26dB.

Once built and tested the idea was to try and tune it up using the Dishal S11/phase method on my VNA. This is the method I have the most experience with and it usually works really well and each resonator only has to be tuned once.

See below for the result after the first attempt to tune it. You can see that even though I only tuned each resonator once I managed to get the filter to align almost spot on. If it was perfect, the S11 and S22 plots would have been identical but this also requires that all of the components that mirror each other in the filter have exactly the same capacitance or inductance. I did try and match the caps 'a bit' before I built the filter and the results below speak for themselves.

This took about 1 minute to tune up using copper tape to briefly short each resonator in turn as its neighbour was tuned. I did try the other Dishal method using a high Z probe looking for alternate peak and null for each resonator but it wasn't quite as good. But it was still quite close and maybe if I'd tuned the peaks and nulls more accurately it would have been better. But the S11/phase method is great if your phase meter has fine resolution and can set the phase flips to within a degree or so.

Back in the early 1990s I designed dozens of BPFs like this at HF and VHF for use in wideband receiver IFs. These receivers tune up to several GHz and the receiver had a digital IF just like today's modern spectrum analysers only in those days the digital IF was a huge and power hungry stack in a 19" rack and probably cost more than a 4 bed house in Surrey. Today's equivalent of that old digital IF would run at very low power and would easily fit inside a packet of cigarettes and would cost about the same as a few packets of cigarettes

I think biggest lumped BPF I designed had about 16 inductors (to try for a brickwall alias filter) but the best one had 13 inductors and each filter had to be phase matched to several others (across a passband of several MHz) to use in a multi channel receiver used for direction finding. So today, this BPF design was a little exercise in nostalgia after I managed to find an old sample bag of these classic old Toko 10K coils.

To tune this fairly basic filter up this well manually would take an experienced operator several minutes at least and they would have to tweak each resonator quite a few times to try and get the return loss this good over the passband.

[Radio Wrangler]

You can sort the asymmetry by substituting top inductor couples for some of the top C couples. Treat the coupling inductors as negative capacitors so the resonator Cs need increasing by double the value of the coupling C replaced. remember to do the resonators at each end of each coupling L.

David

[G0HZU_JMR]

Yes, usually it's best to swap the outer caps for inductors otherwise the required inductance value can get quite big for the inner sections. Often this means the effective inductance is much higher than expected (at the filter frequency) because of the capacitance within the inductor. This can quickly make things very fussy in terms of component selection. At work I preferred to use the Shunt C Coupled transform for filters like this. With a couple of zeros added it can make a symmetrical filter and the SCC type of filter can be used for narrow bandwidth lumped BPFs up into the GHz region. But it may appear confusing to some people. The TC coupled filter looks more traditional with the array of parallel resonators.

The simulation for the filter along with the circuit is shown below:

You can see from the schematic that I practised the Dishal method on the simulation with the shorting grounds at each resonator so I was quite confident the real filter would tune up well with this method

All the caps are single caps with regular values apart from the 4.9pF caps which consist of 2.2pF and 2.7pF in parallel. Note that the Toko 10K 820nH inductors can tune up to just above 1000nH if required as they have quite a bit of tuning range.

[BackToTheLab]

Hello one and all,

Thank you for all the responses...

Apologies for the "radio silence" - I was initially busy reading the Dishal and Zverev papers, then I got distracted with an electronics project at work, and then had to take a break to get married...

Before all of this, I had chance to attempt an alignment of the 40MC/s filter. The Dishal technique is amazingly effective, and I have applied this technique to both the 37.5 and 40MC/s filters. My filter responses are now much closer to the ideal response... The tricky part is determining how to avoid the single coupling loop on the first resonator affecting the tuning of the filter...

Thank you again for your responses...

Peter