Most stable inductor winding methods

[MrBungle]

Whilst awaiting parts for another project (damnation to RS lead times of two months) I'm trying to build a simple 40m DC receiver. I've built a basic working VFO which is to be honest, rubbish. It drifted up and down 1-2 kHz. Fun. However this was done in purpose! The plan is to eliminate sources of drift piecemeal and write up findings. So far the primary oscillator transistor has been replaced with a JFET, capacitors have been changed for opposite tempco, a buffer stage has been added, very stable voltage regulation has been added and now we're down to a 50Hz/minute drift which is still rubbish.

Now I wound the original tank circuit inductor on a T50-2 core which doesn't have the best reputation for temperature stability. Ergo, this was replaced with a T37-6. This is down to 20Hz/min now. The next step is air wound!

This leads to the original question: what can I use as a coil form that doesn't ruin the Q of the tuned circuit? I'm trying to use a 1:1 ratio of winding length to diameter as a start. Preferably something readily available or easy to manufacture at home.

[G8HQP Dave]

I seem to recall that the best VFO coils were wound on glass, but for DIY polystyrene was often used. A plastic toroid is something else to try, as it may have weaker coupling to nearby objects than a solenoid.

[Argus25]

One of the most stable local oscillators I have come across is, of all places, in the HMV 904 TV set. I was amazed at its temperature stability (The tube is the X41C). The coil was wound self supporting from hollow (I think) copper tube maybe 1/8 inch diameter. Its stiff, but bends well and self supports.This copper tube is easy to get, used in a number of applications in automotive & air con/refrigeration. As I recall it was about maybe 1/2 to 3/4" diameter and 6 or 7 turns. That was for about 37MHz, but, if it was 1" dia and maybe 8 or 12 turns, depending on the tuning capacitance, I'm sure you could get that to 7.5MHz and then you'd not have to worry about a former, or its losses either. That is if you have the physical space to fit such a coil.
I have also seen some very nice silver plated copper tube coils, with ceramic bases to support each end, they have these things at Surplussales in Nebraska USA.

[Skywave]

Originally Posted by G8HQP Dave

I seem to recall that the best VFO coils were wound on glass . . . .

An extract from an article found on the 'Net:
A characteristic of all Philips Superinductance radios are the coils. Each coil was wound using Litz wire on large 2-inch glass formers. These are housed in substantial copper screening cans in order to reduce the losses, and hence improve the sensitivity (and selectivity) of the receiver.

Al.

[G0HZU_JMR]

I guess a lot depends on what stability you want to achieve but for a receiver application I would have thought that a T50-6 core would have been OK in a VFO. This is because the temperature should stabilise quite soon after power up. By contrast, if this was for a transceiver, then there would be temperature gradients caused by the duty cycle of transmit/receive operation and that's when it gets really hard to make a stable VFO.

There are a few traps that can cause confusion even on a stable VFO. I've seen VFOs that drift much more when the counter is connected. I'm not sure if this is some kind of subtle integration effect but I'd recommend coupling the counter as lightly as possible. Many years ago I designed a 'trainer' oscillator at about 10MHz that was designed to demonstrate Leeson's equation for phase noise and also to demonstrate basic oscillator theory. It was designed such that the oscillator could be measured in open loop with a VNA so the loaded Q and gain margin and the oscillator frequency could be predicted. From a few fag packet equations and a basic simulator (eg the freebie RFSIM99?)it is possible to predict the frequency and the phase noise profile very accurately.

I think this oscillator used a T50-6 core and as long as it was placed in an enclosure and allowed time to settle it barely drifted more than 50Hz in about half an hour. From memory the phase noise was about -152dBc/Hz at 10kHz offset and -173dBc/Hz at 100kHz offset which agreed with Leeson's equation.

[Radio Wrangler]

Probably one of the most stable VFOs is the G3PDM version of the Vackar. He also included construction notes with it about how to get best stability. It's in the G3PDM receiver article in the 'bull' and it went in late 1960s and 1970s RSGB handbooks. I nobbled it and put it in the ARRL handbook (editions after 1995) it's a circuit worth keeping.

David

[kalee20]

Originally Posted by Skywave

An extract from an article found on the 'Net:
A characteristic of all Philips Superinductance radios are the coils. Each coil was wound using Litz wire on large 2-inch glass formers. These are housed in substantial copper screening cans in order to reduce the losses, and hence improve the sensitivity (and selectivity) of the receiver.

Al.

That's true, really low losses on a glass former... But the OP wanted low drift not necessarily low losses. You can have low losses and still a lousy tempco if the former expands with heating.

[Argus25]

Originally Posted by G0HZU_JMR

I've seen VFOs that drift much more when the counter is connected. I'm not sure if this is some kind of subtle integration effect but I'd recommend coupling the counter as lightly as possible.

This is so true. I spent some time designing buffer amplifiers, so as to buffer the output from a radio's VFO to a 50R source to drive frequency counters with IF frequency offsets. One thing I found was that no matter how light I made the coupling (got down to about one pF into a Fet's gate and 3 or 4 additional stages) if the buffer's output was loaded variably, for example connecting or disconnecting the output cable to a 50R load, it still affected the VFO frequency measurably to one extent or another. The acid test was shorting out the buffer output. Its interesting how the impedance changes get reflected right back to the input despite many stages and loose coupling.
In the end the only way I could make the buffer behave independently and have no measurable effect was to make an optical coupler !

It doesn't matter as much of course if there is a constant load on the buffer, but its a problem if it changes. Likewise attaching or coupling a frequency counter to a VFO can have a bigger effect on VFO stability than one might imagine. One trick I use is to attach the scope probe to the pvc insulation of a wire in the oscillator, so its only a fraction of a pF coupling, then connect the probe through a high gain broadband amplifier to the counter or scope.

[MrBungle]

Thanks for the replies everyone - much appreciated. I have written a mind map of everything here on paper and will sit and work it all out later. I've got a couple of things on fire this week by the looks already so it doesn't look like I'm going to get to actually do anything for a week.

I found that the counter can load it quite badly as well. I was just using an RG58 patch and some pomona clip leads. This is bad. I moved to a 10x scope probe on a dummy load (two 100 ohm resistors) after the buffer stages to try and eliminate this. Seems to work reasonably well but it turns into schrodinger's cat paradox pretty quickly! My counter has a 50 ohm input so I will try this directly when I get some more BNC sockets arrive (hurry up RS!) as this accurately(ish) approximates the load of the double-balanced mixer that will come after it.

[Andrew2]

I think ceramic or glass are the recommended materials for the former, and IIRC ferrite or dust-iron slugs are best avoided.
The best one I ever built was the old 'synthetic rock' cct which turned up in many magazines in the 60's & 70's. Across the tuned cct is a string of three good quality (SM or polystyrene) caps. The top one is a lowish value, the bottom ones are about ten times this value. The emitter of the transistor goes to the top of the bottom cap and the base to the bottom of the top cap.
I found this arrangement to be very stable, moving no more than about 10 Hz all day. A two-stage buffer followed.
The other good one was a standard Colpitts, but with a dual-gate mosfet in place of the usual bipolar.
I've also found that attaching any test gear directly to a VFO will always b&&&er things up.

[G6Tanuki]

My experience is that a ceramic former is agood place to start. Ideally one of the kind cast with a helical groove to contain the winding.

Ideally you wind it with silver-plated wire, under tension [heat the wire, freeze the former, wind quickly and make sure the ends are tied-off before it warms up!] but that's not always possible in practice.

Avoid where possible any kind of ferrite or brass cores, and if you must coat the finished coil with something use Polystyrene solution not wax [wax can change chemically over time, becoming horribly lossy as plenty of owners of 1970s/1980s Japanese radios have discovered: they slathered the FM local-oscillator components with wax to reduce vibration-effects but after 40-50 years it becomes lossy enough to stall the LO!]

[MrBungle]

I remember that waxy gunk from my jumble sale digging days - bit annoyed with myself really as I threw out a box of IFT's many years ago I could do with now!

Assuming polystyrene cement is suitable? I've seen "Q-dope" mentioned in old literature but can't find a single supplier of it.

[G6Tanuki]

Polystyrene cement will do, yes.

Back in the days we made our own by dissolving polystyrene ceiling-tiles in Trichlorethylene or Chloroform, but these days the H&S types would have a fit at such a suggestion since "Trike has been deemed a health-hazard and Chloroform's interestingly explosive.

[John M0GLN]

Back in the days we made our own by dissolving polystyrene ceiling-tiles in Trichlorethylene or Chloroform,
Chloroform's interestingly explosive.

I just had to look and yes, you can buy it on eBay, it's expensive though.

John

[Radio Wrangler]

Cellulose thinners works on old expanded polystyrene packaging. It's quite convenient that lots of plastic things are mrked with their basic type for recycling.

David

[turretslug]

Originally Posted by Argus25

This is so true. I spent some time designing buffer amplifiers, so as to buffer the output from a radio's VFO to a 50R source to drive frequency counters with IF frequency offsets. One thing I found was that no matter how light I made the coupling (got down to about one pF into a Fet's gate and 3 or 4 additional stages) if the buffer's output was loaded variably, for example connecting or disconnecting the output cable to a 50R load, it still affected the VFO frequency measurably to one extent or another. The acid test was shorting out the buffer output. Its interesting how the impedance changes get reflected right back to the input despite many stages and loose coupling.
In the end the only way I could make the buffer behave independently and have no measurable effect was to make an optical coupler !

It doesn't matter as much of course if there is a constant load on the buffer, but its a problem if it changes. Likewise attaching or coupling a frequency counter to a VFO can have a bigger effect on VFO stability than one might imagine. One trick I use is to attach the scope probe to the pvc insulation of a wire in the oscillator, so its only a fraction of a pF coupling, then connect the probe through a high gain broadband amplifier to the counter or scope.

Agreed! Maybe I'm fussy, but I hate doing things to original and existing designs that detract from their effectiveness and it's quite difficult to sample an oscillator without adding capacitance, impedance, reverse isolation or tempco headaches. It's straying from the thread a bit in that presumably narrow-span semiconductor VFOs are in mind but the classical free-running, wide-span, high-impedance typically thermionic VFO needs a very gentle sniff indeed. After sporadic years of tinkering, I ended up using a DF60 (minimal heat and size) with input via 0.5pF and low 1k anode load feeding a conventional semiconductor buffer. The HT restriction being overcome by supplying from the existing mixer screen potentiometer, only 1mA or so taken.

[MrBungle]

Did some work on this very late last night. I would an air core inductor, on a bog roll tube for lack of anything better and covered it with polystyrene cement. I also tacked together a box around it with FR4 stock. This appears to be quite stable now. It was on overnight (8 hours) and drifted up 300Hz but I don't know if that was up, down, all over the place, nor whether this was in line with temperature or not.

Now my counter doesn't have any interface so I can connect it to a computer and actually measure stuff periodically, so I spent a few minutes between gaps at work today playing around with pytesseract ( https://pypi.python.org/pypi/pytesseract ) so I can point a webcam at my frequency counter and record the frequency drift. The counter has an OCXO in it so is stable longer than the VFO is at least. I already have a meter with temperature and RS232 so I'm intending on collection stats on both in parallel

I'm going to measure the absolute drift over time i.e. +2Hz/-2Hz in a 5 minute period means a total drift of 2Hz appended to the total.

Have been looking at huff-puff stabilisers as well: http://www.hanssummers.com/huffpuff.html

It's amazing how small problems turn into large interesting ones!

[Andrew2]

Mr Bungle, I can recommend the Huff-puff stabilisers. My homebrew 160m transceiver uses a Colpitts local osc and although it was easily stable enough (a slow, steady drift to a max of about 200 Hz) for AM, obviously it was a bit too wayward for SSB. Rather than spend days trying to improve it, I added a Huff-puff stabiliser.
The cct I felt most at home with was G3DXZ's 'Huff & Puff in Practice', adapted to suit my particular oscillator. With a 28 MHz reference xtal and a suitable division ratio, it provides lock points every 50 Hz-ish and listening to its 'note' on another receiver it sounds like a crystal osc. Never goes anywhere!

[MrBungle]

I built another VFO, well VXO as an experiment quickly before i take a look at huff puff stabilisers. It has been on all day and sampling hourly and it has drifted a total of 24Hz. Turns out the humble super VXO is pretty solid. I can pull a 7030 crystal pair reliably between 7025 and 7035 so that gives 10KHz spread per pair. I am now looking at a crystal switching scheme as well. Probably fine for the CW section of 40m.



This side steps the original question which I will return to shortly.

[Al (astral highway)]

Originally Posted by MrBungle

This leads to the original question: what can I use as a coil form that doesn't ruin the Q of the tuned circuit? I'm trying to use a 1:1 ratio of winding length to diameter as a start. Preferably something readily available or easy to manufacture at home.

I recently got hold of an All Wave Signal Generator, 65B, from Taylor Instruments, as an easy renovation project.

When I opened it up I was stunned how few passive components it contains. Discounting those in an RF bypass role or mains suppression role, it has just four or five capacitors and resistors combined.

I know these sets were partly built with whatever was to hand and the circuit details vary, but, relevant to this post, the thing is rock solid stable, to four decimal places.

From memory now, the inductors are wave-wound using Litz wire on I don't know exactly what former, but they are obviously at the heart of the stability.
If I get time to open it up later this week, I'll photograph the inductors.

For larger, high-Q coils, lexan is excellent in my experience.