Roller Inductors, Center Taps, Air Core Transformers, and Associated Craziness

[ikrase]

I've been studying more on how to design a quadrupole mass spectrometer power supply (summary: drive 100+pf load with 2MHz sine wave split phase having DC component and DC offset, w/ good voltage stability).

I've been solidifying my schemes around a sine wave oscillator that drives either a tuned resonant inductor or a resonant-secondary transformer (when combined with a tuning capacitance and/or just the capacitance of the electrodes and cables).


In order to produce the proper split-phase DC offset signal, one needs a transformer that is not merely center-tapped, but which has a split (or dual) secondary.

At the frequency I'm dealing with, an air core transformer is viable and possibly the best option -- easier to design and has higher Q (and it's easier to reduce Q than increase it).

It occurred to me that one could add a few-turns primary coil to a roller inductor to produce a tunable isolated air-core transformer -- is this actually a viable idea?

Also, is it practical or desirable to make the "split-tapped transformer" actually be a pair of independent transformers?

[rhb]

I can't see any reason you can't use two transformers.  You'll need to account for the series primaries.

Do you have suitable roller inductors?  If not,  I may have some.  My Dad left me a few dozen (!!!) WW II surplus tuning units.  There may be some suitable inductors in some of them.  As a rough guess these will take several hundred watts of RF.  The stuff is all silver plated with steatite insulators etc.  Cost was no object.  PM me if you want me to go looking through the pile. 

[ikrase]

Interesting...


Couldn't the primaries be in parallel?

[rhb]

They could.  If you're making it up yourself it can be whichever you find more convenient.  You just need to account for the effect of parallel or series connection on the circuit.

Making a dual secondary is pretty easy.  Wind the primary on a suitable form and then wind the secondaries over the primary with an  insulating sheet (aka fish paper) between the primary and the secondary.

Have you calculated inductances, turns ratios, power level and approximate dimensions for an air core transformer yet?  I *know* I have fixed inductors on very high grade ceramic forms.  But I need an idea of what I'm looking for.  It's a midsize U-Haul truck pile of stuff.

[T3sl4co1l]

I would worry about imbalance. You want that thing pretty well centered, no?

Tim

[rhb]

Working with inductors wound on 2" cores with #12 wire I'd expect that tuning  two transformers would not be too difficult.  Fiddly perhaps.  I'd actually worry more about the field interaction between the two.  They would need to be very solidly mounted mechanically or there would be all sorts of headaches.

Personally I'd pick a plain air core coil with the proper inductance and power rating, wrap suitable insulation around it and wind a pair of secondaries over it.  I'd only resort to using two transformers if I couldn't find a suitable primary for making a dual secondary transformer.

That said one might find a large toroid from a SMPS that would be suitable.  That would avoid fringing field issues. The OP needs to provide a bit more information. This is an oddball RF PA.  There are lots of issues one has to get right for it to work as desired.

[ikrase]

Yes, I need a lot more information.

I'm trying to build something based on this paper: https://web.stanford.edu/group/Zarelab/publinks/798.pdf  with some changes.

The important things to understand are that the output is meant to be in peak-peak Voltage, not in Power-at-some-impedance and that both the isolation from ground and of the two "ends" of the amplifier from each other are important.

I attached a sketch I made.



The changes I would make:


- Trying to push the working frequency up to 2 MHz (though lower is probably tolerable, and it's intended to be changeable)

- Using a low tuning capacitance, or none at all, to achieve resonance (other than the actual load)

- No automated stepper motor tuning

- Using variable inductor-transformers to tune to resonance.

- Replacing the expensive and probably unneccesary Apex power amp with a much cheaper boosted op-amp

- Possibly boosting the primary voltage to reduce the required transformer ratio

- Crystal-controlled frequency source




Anyway, the linked work uses a cylindrical air-core transformer consisting of 3 turns primary over a secondary that isn't fully specified but has an inductance of 300 uH and looks like it probably has a secondary of around 300 turns on a 20mm former (this gives the proper numbers and dimensions that agree with the picture, as well as providing a transformer ratio of about 10 which agrees with my simulations).

Since I want to go to a higher frequency, the combination of inductance and capacitance needs to be reduced. Getting rid of the tuning capacitor (which only makes things worse) is a first step.




(I was somewhat resisting using ferrites because as far as I could tell, the need to baby the flux in the ferrite leads to adding another constraint -> even more headaches.



Another possible idea: Minimal-inductance air core transformer (probably very small diameter to allow the combination of high turns ratio and low inductance) with the center tap / split, and the ends leading off to roller inductors (in series) that are not used as transformers.


Anyway the target inductance for roller inductors would be in the 20 to 100 uH range.

[rhb]

I've read the paper and your comments.  Your requirements are self contradictory and the paper was clearly written by people who had little understanding of RF amplifiers.  Litz wire in a power amplifier is nuts.

Some suggestions:

For the oscillator an AD9850 DDS will provide better spectral purity than the approach in the paper.  They're cheap and easy to use. EBay is awash with them.  They are also referenced to a crystal oscillator so the  frequency accuracy is  10-20 ppm or better.  a GPSDO will give you  several orders of magnitude better accuracy though at 10x the cost.

Feed the DDS to a wideband amp capable of giving you a watt or two of drive.  MiniCircuits probably sells something that will serve well.  The more power the better, as it will reduce the gain required in the output stage.

The only distinction between the mass spec PS and an ordinary radio transmitter is the output LC tank.  The output is intended to drive two antennae rather than one.  This is done with phased array antennae where directivity is required.

The load has a characteristic impedance. The power is E**2/Z, so for a 100 Vrms output into 50 ohms you need a 200 watt amplifier.  That's a good bit of power, but not difficult.  If your poles are higher impedance the amount of power is reduced.  But you still have to match the impedance of your poles.  If you don't, the power is reflected back towards the DDS which will not work well.

Getting a tuning range from 300 KHz to 2 MHz is quite easy.  Use a tapped inductor for the tank circuit in combination with the variable capacitor and use a multipole relay to switch taps.  Step the DDS and alternate  the sweep direction when you switch taps at the end of the capacitor rotation.  The AD9850 can change frequency in less than a microsecond.

Because your tank circuit has two outputs, you'll have to tap both sides of the secondary.  As you want to tune over nearly 4 octaves, you'll  need to tune the primary side of the output transformer.  Your output impedance is constantly changing with frequency.

To tune the output with a roller inductor you would need to wind the two sides in the opposite direction.  That is perhaps possible, but keeping both halves of the coil balanced will be very difficult.  With a tapped inductor, you can insert a trimmer capacitor in each tap to correct for slight differences in inductance on either side of the center tap.

I'd like to strongly urge you to read the ham radio literature on RF amplifiers.  This is only difficult if you follow the paper too closely.

[ikrase]

Well, it appears that the paper writer and myself have "not being very competent with RF" in common. And I'm not very good at explaining things.

I thank you very much for your criticism and advice, but I need to ask more questions. How bad actually is the Stanford paper, and what in particular strikes you as wrong?


(Didn't they just use Litz wire because it was what was on the shelf?)

I was however always planning to use a crystal-controlled DDS (though not necessarily specifically the AD9850) to provide the original source for the sine wave. I doubt a GPSDO is worth it, maybe just a good TCXO.

Why only a watt? Most of my (admittedly rather ridiculous) simulations (and the paper) suggested that at least ~10W would be needed, I was imagining either power-transistor-boosted opamps or (as suggested in my earlier thread that was less focused on my attempt to follow the Stanford paper) one of the many ~50W QRP ham power amps targeting the 160 meter range using IRF510s. One design for a quadruopole driver used tubes, 6146B's, but it was flatly unsuitable for a QMS.

I would say that there are two further important distinctions between a QMS power supply and a radio power amp. First, the QMS needs the DC offsets, and second the QMS in the final reckoning needs to tightly control voltage, not power. Of course, that just means accurately (and stably) knowing your load impedance (frequency dependent since its purely capacitative)


For your tuning scheme, am I correct that you're suggesting an output transformer with (numerous) taps in the secondary to adjust the L, and then a (small) variable capacitor for analog, fine tuning? Or would this multi-tapped output tank inductor be separate from the transformer?



Finally, I would also be interested in what you think of the paper's power monitoring/feedback detector circuit (and the use of the VGA that it controls.)

[rhb]

The way the writers wrote up their work.  That is not how someone who had mastered "The Art of Electronics" would describe the build or the way they'd build it.  The concept is sound, but the execution is quite inept. 

The watt was so you'd have something you could buy off the shelf to follow a DIY DDS  to drive more stages of amplification.  That will let you work out the impedance matching problem without having to deal with high voltages and power levels.  The impedance matching over almost 4 octaves will be very difficult.  It's doable, but you should expect to make many iterations.

A TCXO certainly should do fine.  I doubt that you can maintain amplitude tightly enough to even need that.

Calculate the load impedance vs frequency and determine what power level you need at each frequency to provide the desired output voltage.  That will give you a rough idea of what sort of gain range you need to maintain a constant voltage.  If you want to vary the voltage  that will require additional gain.

What I'm suggesting is a tapped secondary with each tap having its own trimmer cap to adjust for variations in the coils and the tap positions.  You'll probably need to do the same thing on the primary to get a match over 4 octaves.  Life will be much easier if you have a vector network analyzer available.  The DDS and a 4 channel DSO will work with some MATLAB and SCPI programming.  But it can all be done by pencil and paper.

The central concept to what I'm suggesting is that the air variable is swept from max to min position, then you jump the DDS to the maximum frequency of the next tap and sweep down, then step to the next tap and sweep up.  The trimmers in both tap legs will let you match the min frequency on one tap to the maximum on the adjacent tap.  The number of taps required will be determined by the VSWR of the output transformer.  It's really  just a weird automatic antenna tuner.

The DC offset really doesn't effect the RF design other than requiring an output transformer and the opposite phase outputs require a center tap transformer.  Aside from that it's a normal RF amplifier with the exception of the tight output level control.

Disclaimer:  I am not an RF expert.  I recently read a couple of upper undergraduate or early graduate level texts on RF design.  I messed around a bit with RF 30 years ago and am only now getting back into it.  But I di know the difference between what I do know and I don't know and I know how to learn.  In the end that's what matters.  I happen to think this is a very cool project.  The sort of thing I'd like to do if I am granted the time.

To get this done I suggest doing this:

Get a DDS you can sweep under computer control.  The obvious choices are the FeelTech FY-6600 and the JDS-6600.  I think the latter probably has a better output stage than the former.  There are extensive threads about both.  You can build one, but when all is said an done you'll have spent almost as much money and a bunch of your time.  You're better off spending your time on the matching problem.  IIRC the JDS will deliver a 20 Vpp signal with good fidelity.  The FeelTech fades around 15 Vpp unless you modify it.  Both units have ungrounded switching mode power supplies and will require correcting that issue either by grounding the chassis or replacing the PS.

Design and build a prototype of the transformer and air variable but skip the automation and use slide switches and manual tuning with a dummy load capacitor.  Then measure the VSWR seen by the DDS at a series of frequencies.  I'd do this with just the maximum inductance taps and trimmers, manually tuning the air variable each time.  That will get you started on the impedance matching problem.  I think you would be well advised to calculate the impedance curves for prospective tap positions before construction.  It really is just a weird automatic antenna tuner with a weird antenna.  There is almost certainly ham oriented software available online for calculating the impedance of a tuned circuit.  I think Elsie will do this, but am not sure as I've not played with it much.

The level control that they used looks OK.  But I suspect they were mostly using it to compensate for poor matching.  The better the matching, the easier the gain control becomes.

[ikrase]

Oooooooohhhh, thank you, that makes more sense.


I was never planning on frequency adjustment being that automated - would be open to having manual tuning and even plug in module with transformer for each octave.

I'm imagining sine wave source -> VGA -> power amp ->  primary (ferrite toroid?) -> tapped centertapped secondary -> tuning capacitance -> load.

Any comments on the capacitative dividers for level measurement? The Standard paper didn't specify them, I think silicon or mica may be appropriate for low tempco.

[rhb]

The FeelTech and JDS  AWGs can be controlled over RS-232 or USB, so if you use one of those for the DDS you already have the variable gain amplifier. They appear to adjust in 1 mV steps.  But I don't think anyone has checked them for accuracy yet.

Ferrite toroids will reduce the Q thus requiring fewer taps.  The materials for both air core solenoids and ferrite core toroids are cheap.  I'd try both.

The capacitive dividers are just to scale the amplitude of the RF without being influenced by the DC bias.  If you put an RF diode bridge and capacitor feeding the ADC of an Arduino  or similar you can adjust the DDS output as needed.   A good design is actually easier than what the Stanford folks did.  Maybe that's why it's a Junior University  ;-)

I'm glad you found that useful.  I spent a good bit of time on it.  I think a mass spec is a really cool project and would love to build one myself.  I bought a VNWA 3E, but I've not actually fired it up yet.  Never having used one I'm expecting it will take a few days to get acquainted.  But I'm in a position to check impedances of the coupling.

 My parts for the Zynq project came and the USB hubs want a microUSB for power  so I ordered some hubs that take 3.5 mm barrel connectors.

Have Fun!
Reg

[ikrase]

I did some research and some simulations today and I think I *finally* get it.


I'm kinda getting the idea that the Stanford design is clumsily trying to do DC-style "zero source impedance, infinite load impedance" with the addition of the resonant-secondary transformer to avoid needing to waste massive amounts of power charging the capacitive load backwards and forwards. And this is not working well for them as you say.


So it sounds like what you're suggesting is that the load capacitance be placed in a series loop with (tunable?) inductor(s) so as to resonate at the proper frequency. Then this resonant loop should be driven (also in series) by a much-lower-inductance impedance matching transformer with multiple taps for tuning the frequency match (as well as the caps you mentioned) and the center tap to provide the DC offset.

(Falstad link)

The transformer will mostly be matching the power amp output impedance to the resistance (parasitic or deliberate ballast) of the tank circuit since the capacitative reactance and the inductive reactance will be cancelling out when tuned to resonance. It will likely be a *step down* transformer.


It's allll finallllly making sense. I think.


What about air core toroids (technically foam or plastic core probably)?  I'm considering putting interchangable transformer-blocks on an Octal base. And of course airgaps. I guess I better get some rings.

(Quick questions: Are the trimmer caps in series between the tap and the switch contact?)




I'm intrigued by what you're doing-- you have a manufactured mass spec that you're bringing up? Would love to see pictures and/or hear commentary on construction techniques.

[T3sl4co1l]

That's a thing, but notice you have to match six components(!) to ensure balance.

How important is balance, anyway?  I've been assuming it's important but do you have any data on common mode AC ("VDC" term varying in step with "VAC")?

Tim

[rhb]

It's probably best to put the trimmers in parallel with the active section of the secondary and use a double pole switch. to select the desired tap and trimmer.  The other end of the trimmers tie to the center tap.  I had been thinking of  the caps in series when I mentioned them, but that results in needing excessively large capacitors.

 I suggest using a DP switch but leaving out the trimmer caps until you know you need them.  It might be a good idea to use a 4 pole switch so that if it is necessary to select taps on both sides of the transformer that's easy to do.  Mechanical fabrication can take a depressing amount of time for very simple circuit changes.  And things become a nightmare if you didn't allow enough space.  I'll go see if I can find you a good switch.  I know what you need now and I'm sure I have some really high grade, silver plated steatite wafer switches good for several hundred volts.

Aside from a VCO rather than a DDS, the conceptual design in Fig 1 of the Stanford paper looks fine.  It's the actual implementation that is duff.

I think that automation is important, but only when the instrument is running.  Then you want to automate it so that everything is repeatable, measure the errors and correct them in the signal processing.  I'd build the PSU so that it was easy to add stepper motors of appropriate size at the very end.  I don't see any reason that you can't achieve a research grade instrument result. 

A friend of mine built the magnet mass spec at the Geosciences department at UT Austin.  Sadly he was hit from behind and killed by a careless driver while waiting to make a left hand turn.  His wheels were cocked to make the turn and the impact drove him into an oncoming car pretty much ripping his Sentra in two.

I can't speak to the matter of air toroids.  You'll just have to try it.  I suspect that there are geometric constraints.

I'm working on FOSS firmware for Zynq based DSOs.  I *wish* I was bringing up a mass spec.  My lifelong dream retirement job was as an instrument tech and builder at a university.  Unfortunately things didn't work out that way.  But I would like to build my own mass spec which is why I'm so interested in what you're doing.  I'd greatly appreciate if you would post some of your simulation results and other information about the project.  Perhaps also a documentary thread in the Projects section?

As I commented in another thread about an electrokinesis transducer, I was hopelessly corrupted at an early age by Clair Stong's collection of Amateur Scientist columns from Scientific American.  I was probably OK until I got to the chapter that said, "For less than the price of an average set of golf clubs you too can smash atoms in your backyard."  At 10 or 12 I was enthralled.

[ikrase]

There are three documented (har har) DIY mass specs that I know of.

Two are made by the German experimenter Thomas Rapp,

http://www.rapp-instruments.de/Radioaktivitaet/div/quadropol/quadro.htm a quadrupole mass spec. Note his low-capacitance design and less-than-successful power supply... and beautifully machined poles. This machine is only really suitable as an RGA.

http://www.rapp-instruments.de/Radioaktivitaet/div/sector/sector.htm for his cheap pipe tee mass spec, which is really no better than a toy but which probably presents a good platform to experiment cheaply with basic ion optics on.

The other is being designed by some people in Hong Kong called M-lab -- they're mostly working on quantum computing, not mass-spectrometry, and their docu isn't great. Apparently their device is meant to be used with Ebay-garbage-can quadrupole mechanicals that have no power supplies.




My (initial) intent is to make the quadrupole very much a bench-stack design -- there will be the filament controller, the multiple HV supply for different DC electrodes, the AC quadrupole supply, and my trusty Keithley electrometer.

[rhb]

Sadly I can't read German.  But the workmanship is beautiful.  I'll see how google translate does later.

I don't know what an "RGA" is.

I was mostly asking about your power supply simulations.  I'd like to see what your results.

Unfortunately, building a mass spec is far down on my list.  I've got a scrap water tank to turn into a cupola for melting iron, and a scrap SS pool filter to build a crucible furnace.

For a completely different project, I got my  linear ball bearing rails today along with my steppers.  I'd been planning to put them in an old 14 gauge steel chassis, but after looking at it I think I'm going to order a 12" x 12" x 2" grade B (0.0002") surface plate to mount the rails. Unfortunately shipping is more than the cost of the surface plate.

[ikrase]

An RGA is a Residual Gas Analyzer. It's a simple mass spectrometer that measures the bulk contents of a high vacuum system and identifies the... residual gases. Very useful for finding leaks or determining why you aren't getting a very good vacuum.

Recently some extremely compact RGAs such as the MKS VQM have appeared on the market, these being advertised as an upgrade for ion gauges.

I included a link to Falstad in a recent post, it should bring up my simulation.

Are you building your own CMM?

[rhb]

What's a CMM?

I'm building a CNC XY stage and am goofy enough to scrape the extruded rails flat and true.  The initial goal is just to learn how to control the thing.  But I'd like to have something that could be used for precision work.  I bought continuously supported rails which can't be adjusted to as tight a tolerance as unsupported rails, but they take more load.  It's an experiment.  I can always fit more accurate rails if needed.  I am going to need a 12" machinist's level.  I've got an 8", but I bought four 12" rails rather than tow eichts and two twelves..  The ballscrews are only 10", but I should be able to position a 4" platform anywhere.  I've got an HeNe laser which should be sufficiently coherent to give me interferometric positioning.

There's a great anecdote about this.  In 1948, Scientific American did a series of articles about John Strong's construction of a ruling engine at Johns Hopkins.  The article mentions that many amateurs had come to grief and financial ruin  attempting to build a ruling engine.  In 1958 the Amateur Scientist column featured a ruling engine built as a hobby project in the UK.  The builder was employed at a university as an instrument technician and designer.  So I'm not sure if "amateur" really is appropriate.  What he did was use a DIY cadmium lamp interferometer to position the carriage.  Because the errors were random the gratings did not suffer form Rowland's ghosts.

From what I've read, I should be able to get 8-10" coherence length from the unit I assembled.  I've stripped old CD & DVD drives of splitters and mirrors.  I should note that the unit is not finished.  I still need to mount a switch and power connector, but was terrified of botching the mechanicals.

I've wanted to do this stuff for a long time, but wouldn't write the check.  But having had several close friends die, I'm more willing to sign the checks.  I might not get a chance if I don't get to work on it.

[ikrase]

CMM is an ultra-accurate motion platform (usually cartesian, only sometimes motorized) equipped with a mechanical stylus, which is used to measure the shape and dimensions of arbitrarily shaped objects.

The bread and butter of modern machine shop metrology -- and oh so expensive.

[rhb]

It would have understood immediately if you had said coordinate measuring machine.  I'm quite aware of those. CMMs will do for most tasks, but there are things for which they will not meet the specs and things such as  toolmakers buttons made specifically for the job are required.  Then it gets *really* expensive.  You're making tooling to better precision than the object being made.  Things like the gravity B gyros are certainly beyond CMM capabilities.

Acronyms are very context sensitive even between different groups working in the same field which is why I never use an acronym without defining it on first use.

Shars.com has a 12" x 12" x 2" grade B surface plate for $33.  Unfortunately UPS ground will be slightly over $33.  But I'll have over $200 in the mechanicals, so another $67 for the base is not disproportionate.  I think I'll be able to get 0.001" accuracy and repeatability, but probably will have to switch to full round ball slides and rail to get to anything better.  So I may just build on the steel chassis as originally planned using the supported rails and  do a larger, higher precision table later using a larger surface plate.  The supported rails are quite stiff so shimming them to fit the steel chassis will probably suffice for 0.001" tolerance in 3 axes.

[rhb]

I have been playing with the steppers for my XY stage.  Which brought to mind the last couple of sentences of the Stanford paper in the "Variable Capacitor and Frequency Tuning" section.

There are no forces on the variable capacitor when the system is deenergized.  So the loss of holding torque has no effect.  In fact, no holding torque is ever needed with an air variable capacitor unless it is subjected to mechanical vibration.  The authors are mistakenly attributing to the stepper motor the effects of heating in their ridiculous choice of a coil wound with Litz wire.  Very likely a close examination would reveal other heat related issues, but that is the most likely suspect.

I have a lot of friends with PhDs from Stanford.  Some of whom are a bit strange, but they are very capable.  The authors of this paper are appalling.  Their supervisor should have never allowed such a construction and the reviewers should never have allowed its publication. However, I know someone who as a PhD candidate in geophysics did not know Snell's law.  And he got his degree.

[ikrase]

I'm sorry, I should have guessed it was the acronym, not the idea.

One other question: What about RF chokes -- do you think they are needed or desirable on the connections to the DC power source?

[rhb]

Oh, yes!  Leaving them out would play havoc with the voltage regulation and make your amplifier oscillate on its own.

To paraphrase a remark by Ken Thompson, the primary creator of Unix, "Never underestimate the value of brute force."

I'd very much like to see your calculated impedance curves for a 300 uH inductor based transformer and the variable cap tuned across the full range.

I'd offer to do them except I'm trying to build a CNC XY stage and learn FPGA programming.  So I've got other hills to climb at the moment.  I just had someone point out that I was overlooking thermal expansion effects in the XY stage.

[ikrase]

So for my power amp, I'm looking at using a variant of this circuit, a Class AB amplifier using IRF510s. : http://www.g0kla.com/scpa/SimpleCheapPA.php

(This is a form of the popular WA2EBY ham radio amp, which in its original form is able to go down to 160m (1.8 MHz)


How would I adapt it for the range a bit below that? Few more turns on the inductors/transformers?