Full-Range IF Selectivity; The Hammarlund Variable Crystal Filter

[Bazz4CQJ]

In previous threads we’ve discussed the issue of selectivity in vintage (valve) receivers. Most often the problem is a simple lack of selectivity, or in the case of the HRO, a choice between the eye of a needle or wide open barn door.

Continuing to delve in to the internet on this, I discovered the crystal filter system which was patented in 1938 by the Hammarlund Corporation and described in great detail in the magazine “QST” December 1938 (page 35). It’s 13.4MB so not attached. Looking at a number of the Hammarlund Rx’s made after ’38, the design appears to have become a “trade mark” feature of all their sets.

It’s quite a simple design, using a single 455kHz crystal which is subjected to various degrees of load impedance by a circuit which switches in resistors of values which set the range of the bandwidth and then a capacitor which fine tunes the bandwidth within that range. This “two knob” control was reported to be very effective and required no skill to use. The claim is that “The higher the impedance load, the wider the filter”.

Shown below are the circuit of the filter and its frequency response at different settings, as taken from the ops manual for the Hammarlund HQ129 –X receiver.

Now, it has to be said that in the QST article, it is implied that the crystal necessary for this circuit is something special, but I’m not sure to what extent any crystal made at Hammarlund in 1938 was very distinctive from a crystal made by any other company in 1938, but I could be wrong.

The only other slightly quirky feature of the circuit is that it needs a differential tuning capacitor to make it work. Such capacitors are not common in the UK. I’ve been wondering if the use of two varicap diodes, suitably configured, might be substituted for the differential.

Curiously, we rarely seem to discuss Hammarlund receivers on the forum. I wouldn’t claim to have the expertise to look at this circuit and make deeply informed comment, but Hammarlund were flogging this for some decades.
Thoughts?

B

[Bazz4CQJ]

Ooops a post appeared then disappeared, but this was my reply.

I'd need to go back to the QST article, but I seem to recall that the 2nd IFT was a low impedance arrangement.

Just a comment on the QST article, there's about three solid pages of it and then additional half-page incremental bits scattered through the rest of the mag, so a bit inconvenient to read and absorb.

B

[Cruisin Marine]

It was my post, quickly removed after I realised it lead nowhere.
I wonder if anyone has tried to reproduce this and test its viability?
I guess, it is not one of those things can be discussed in detail without practical experiment.

[Bazz4CQJ]

Originally Posted by Cruisin Marine

I wonder if anyone has tried to reproduce this and test its viability?

Let's say that it was in every model of receiver they made between 1938 and 1960; I doubt if it could have lasted that long unless there was somthing to it. It seems that Hammarlund were in trouble by the late 50's and were taken over twice before finally disappearing, but lots of good companies had the same experience.

I will be building this, but recommend that you don't hold your breath waiting for me to complete it .

B

[CarbonMike]

Didn't Eddystone Radio do something similar in their 680/888 series?

[Synchrodyne]

From a description of the Hammarlund SP-600 receiver:

"Selectivity is varied by control of the IF section band-pass. A six-position switch gives 6 dB bandwidths of 0.2, 0.5, 1.3, 3, 8 and 16 kHz. In the three narrow-band positions of the switch a crystal filter is used. A crystal phasing control operative in these switch positions provides precise control of the skirt for discrimination against very close interfering signals."

It might be said that nose selectivity is varied stepwise. Skirt selectivity is continuously variable over a range for each of the crystal positions. Quite possible the ranges abut or even overlap, giving continuously available skirt selectivity over the range covered by the three positions, even though the nose selectivity is still stepwise.

That does not seem incompatible with the circuit presented above. The crystal phasing control is there, and the crystal is evidently bypassed in some positions of the selectivity switch. Selectivity control is then done by varying the amount of resistance in series with the tertiary IF coil.


Cheers,

[Radio Wrangler]

Originally Posted by Cruisin Marine

I wonder if anyone has tried to reproduce this and test its viability?

Switched resistance loading of LC resonators, and then the same thing with quartz crystals for the narrower bandwidth is the standard method of implementing the wide range of switched IF bandwidths in spectrum analysers.

So it works, and there's a lot of them around. Usually the filter is five synchronously tuned stages and they all switch together. As sync tuned filters they are bullet-shaped on top but the skirts are not as steep as Chebyshev or Butterworth tuning would give so their shape factor is poor as communications filters go, but their pulse response and lack of ringing is more important in this application.

In later analysers, the resistors aren't switched but done with PIN diodes whose resistance is set by individual DACs. The phasing/centring is done with varactors and individual DACs. The analyser software allows you to connect the cal signal to the input and the analyser jumps around calibrating the IF filters on all the offered bandwidth (1, 3, 10 etc series). It's fun to watch. The resulting DAC settings are then saved and evoked whenever that bandwidth is selected.

Care is needed with varactors. They exhibit capacitance controlled by the immediate and instantaneous voltage acros them. They are very fast and big signals will pump the capacitance thus pumping the filter characteristic and intermodulating other signals. Not what you want in an analyser or receiver. Operating levels have to be carefully planned to keep this below noticeability. Usually this compromise means the noise figure suffers.

David

[Bazz4CQJ]

The filter being discussed is indeed used in the SP600, which I think went in to production in 1957, and according to radiomuseum.org, continued to 1971??

B

[ms660]

The same Xtal filter arrangement was also used in the Hammarlund HQ150 which also had a Q multiplier fitted which in conjunction with the filter could provide a notch both sides of the signal.

The local oscillator frequency was above the signal frequency on the top HF bands.

Lawrence.

[Cruisin Marine]

Perhaps the best thing about this post is that it demonstrates the ulmitate importance of giving filters the correct feed and termination impedance, if you don't do that you are ignoring the capabilities of what a filter is actually there for in the first place.

[Radio Wrangler]

Everyone would assume that if you had access to perfect components, you could make a perfect filter. Perfectly shaped and perfectly lossless.

To make perfectly rectangular filters, you would need an infinite number of those perfect components because of the order number required.

It's the perfectly lossless bit that gets interesting.

All those signals on frequencies not destined to get passed through.... where does their energy go?

It doesn't come out the output of the filter, it can't get dissipated in those lossless components, it can only get reflected backwards, out thi input port.

So, for a filter made out of really good components, there is a direct relationship between the transfer function to the output, and the reflection function at the input. Chebyshev and Cauer ripply filters match their passband ripples with ripples in the return loss.

If the source isn't a good match, the reflected stuff gets reflected back in and out and in and out and in.... If the filter is a bit imperfect, then some of the rubbish will get to the output. So it therefore depends a fair bit on the quality of impedance presented to the input.

A second way to view this is with coupled resonator filters. Make one with perfect components and there is nothing inside it to set the Q of each resonator, so the poles aren't spaced correctly wrt. the real axis of the s-plane. So the response is dominated by sharp large peaks.

In real life, the filter uses the source and load impedances presented to the filter to govern the Q of the first and last resonators. Intermediate resonators see these losses through the pattern of internal coupling factors. So your lowest Q poles have to be the outer resonators, and the higher Q ones live towards the middle of the structure.

Whichever way you look at it, passive filters strongly depend on getting the source and load impedances right.

David

[Bazz4CQJ]

As I think someone said earlier, in terms of "copying this circuit" the only way you're going to know is to try it. Now, even if it you don't get to work as well as Hammarlund may have, if it gives you a significant improvement, then you win the fat cigar!

The filter in the HRO is just scrap metal in terms of using with telephony. I took it out some time ago with no intention at all of ever putting it back.

I wonder if one of the LT Spice gurus could assess the circuit and make useful findings? I suspect that the L19 /85pF might need some tinkering, depending on strays?

Attached, a first draft at the HRO-Ham hybrid. It does not look like a huge amount of work to try it. Moreover, it can be done without any visible change to the interior or the front panel and that is sooo important for some people.

B

[Jon_G4MDC]

You could load the HRO filter in similar way to widen it. It's still only a single resonator so it won't be worth much. The ceramic ladder eg Murata will be much better if correctly terminated.

[Bazz4CQJ]

Originally Posted by Jon_G4MDC

The ceramic ladder eg Murata will be much better if correctly terminated.

I believed that the last time you posted regarding ceramic filters you were thinking that you'd need to use semiconductors with them and so you were leaving them alone?

When you say ceramics would be "much better", much better than what? The idea of loading the HRO filter is not on the table.


B

[G3PIJpeter]

The AR88 switched bandwidth single-crystal filter taps the crystal / C75 down L34, with additional capacitors C80 and C81 to 'tune' the shape at narrower settings - see attached simplified filter circuit. It would be interesting to compare the shapes of the AR88 and the Hammarlund filters to see which arrangement (when correctly set up) is the best (or least worst).

Peter G3PIJ

[Bazz4CQJ]

Hi Peter
that's an interesting idea and I don't recall looking at it previously, but I just looked in the my file and found that I looked at that circuit (detailed circuit attached) back in October. I wonder why I seem not to have been attracted to it?

It does not seem to be substantially more or less complex than the Hammarlund, and it does not require the differential capacitor.

I guess that the only people who could make much comment would be those familiar with both the AR88 and the various Hammarlunds.

B

[Radio Wrangler]

Each of them has only a single crystal, and so when that crystal is in-play, there will be a single-pole filter response. Changing the loading to change the Q will adjust the -3dB bandwidth, but the skirts will also change in proportion, until you get far enough out on the skirs for the bandwidths of the resonators in the IFTs to have some effect. A single pole response will always be bullet-nosed and the slopes will be asymptotic to a fall of 20dB per decade (of frequency difference from the centre)

The phasing capacitor of the HRO front panel is there on the AR88, but inside, on the AR88 deck. It's primary purpose is to correct the response for the C-nought parameter of the crystal... this is essentially the capacitance of the case and electrodes of the quartz unit.

Look at the IFT driving the crystal and you see the secondary has a tap. The tap goes to RF ground and has only the AGC voltage on it, for biasing the subsequent valve. The ends of the transformer therefore have equal and opposite signal voltages on them. One end drives the crystal, and the other drives the 'phasing' capacitor. The far ends of the crystal and of the phasing capacitor are joined.

This makes a bridge circuit, and when the phasing trimmer is adjusted to equal the crystal's C-zero capacitance, the currents through the phasing trimmer and through the C-zero component of the crystal will cancel.

Quartz crystals normally exhibit both series and parallel resonance modes, with similar pairs for all the overtone modes, and also for most spurious resonances. Cancelling C-zero this way removes the cause of the parallel resonance modes, leaving us with just the series mode resonances.

So a crystal alone as a series-pass element into some load resistance will normally give a sharp series mode peak with, slightly higher in frequency, a sharp parallel mode null.

The null distorts the skirts of the series mode peak.

When C-zero is nuilled by a phasing capacitor in proper adjustment, the skirts of the series resonance become symmetrical.

This is what goes on in the AR88. Whoever aligns the IF sets the phasing trimmer for good symmetry and leaves it.

The HRO is a bit cleverer than the AR88, the phasing trimmer is brought out to the front panel as a user control. Artful CW operators have found that a null which they can move around gives an added ability to substantially reduce nuisance signals close to the one they are trying to copy.

I've never had a Hammerlund, so I can't comment from an operating viewpoint, but it looks very similar. The Hammarlund circuit switches different loading resistors in, while the AR88 has only one loading resistor, but by switching the output end of the crystal between the top of the resonator sporting the load resistor and two different taps down the resonator inductor it performs the same function in loading the crystal into different resistive impedances. Here the AR88 could have used the voltage step-up to drive the grid of the following valve, but they stuck with the output end of the crystal. This could have increased gain a little on the narrower bandwidth settings, compensating for the increased loss in the crystal. This is something you need to do with variable bandwidth filters such as those in spectrum analysers where you also have to worry about the temperature coefficient of the loss varying with selected bandwidth. I think the AR88 was done this way to avoid still further complexity in the switching, as more trimmers would need to be switched to keep the peaks in the same place for all bandwidths.

Hammarlund's use of a differential phasing capacitor rather than a plain trimmer looks to be a way to keep the various resonators from being pulled as much when the phasing control is used. The AR88 with fixed phasing doesn't need this, the HRO just lives with it. I think it's a neat way to reduce an unwanted effect (reduce not remove) in a set with phasing as a front panel knob.

While it's interesting comparing these three approaches, there doesn't seem to be any driving need to convert an AR88 ot HRO into an SP600. We live in an age where SP600s are affordable, so we can just get one of those to restore.

If filter performance really is a problem, then these sets only deploy a single crystal response, and that only on bandwidths too narrow for telephony. What is needed is a filter of telephony sort of width, but with a flatter top and better shpe factor down into the skirts. This is a portrait of a multi-pole filter. So, if you want to listen to AM or SSB with a crystal filter, then you need multiple crystals.

Those crystals can be deployed in either a lattice architecture, or in a ladder architecture. The ladder has an advantage in needing a group of similar crystals, while for a lattice they all have to be made carefully offset to each other (You can hear the cash registers chiming!) However, ladder structures become difficult for wider filters on lower centre frequencies. 455kHz is too low for easy AM or SSB bandwidth ladders, so it's lattice territory.

The likes of ICOM, Yaesu etc flog option filters for their transceivers. Some transceivers have IFs at both 9-ish MHz and 455kHz in schemes to allow variable bandwidth by offsetting LOs and sliding filters across each other. If you opt for more filters, on the poshest models, you found the 455kHz filter was three or four times the price of the one at 9MHz.

So, if you feel hadicapped by the selectivity of an HRO, AR88, CR100. SP600 etc, etc. Then the next step that really delivers any improvement is to a set with multipole crystal filters. Multipole ceramic filters can also be better in this way, but they were mainly used on cost-competitive hobbyist receivers. Collins' mechanical filters did the job for governmental and military jobs.

Ships' main receivers were pushed into multipole filters, and there are namy which had switched block crystal filters at 2MHz or 1.4MHz. A lot of 1.4MHz filters came on the market in the 1980s as various firms and agencies dumped spares stock. I collected myself a fair selection of a few different CW bandwidths, RTTY ones, USB and LSB as well as a few different AM bandwidths. So if I want to do a homebrew receiver, I'm covered. They come in useful when I want to add LSB to a military set with USB only.

Back in the sixties, you had whatever set you could afford, and if it didn't do what you wanted, you modified it. Nowadays you may want to question your starting point and maybe choose a different receiver. It all depends if you want the end-goal, or an interesting journey.

From a point of circuitry, there is not a lot to choose between the filtering of these three receivers. Each can be set up to give responses very similar to the others. The AR88 lacks phasing as an operator control, but that is only of value for CW reception under crowded conditions.

David

[ms660]

Originally Posted by Bazz4CQJ

It does not seem to be substantially more or less complex than the Hammarlund, and it does not require the differential capacitor.

I guess that the only people who could make much comment would be those familiar with both the AR88 and the various Hammarlunds.

I found the Hammarlund that I had was better than the AR88D.

Lawrence.

[Radio Wrangler]

A different cut on the compromise cake, maybe better Q in the IFTs, and also crystals not made in a wartime hurry. A few things got learned and improved from the era of the wartime sets to the era Hammarlund were active in.

Maybe the Hammarlunds were the pinnacle of the single conversion 455kHz sets? It was an era killed off by inadequate image rejection on the higher bands, and the wrong selectivity for SSB. It was a calm before several storms on the HF receiver design front.

David

[ms660]

Here's a couple of shots of the Hammarlund I had plus a link to the manual:

https://bama.edebris.com/download/ha...q150/hq150.pdf

Lawrence.