Q-meter questions

[Radio Wrangler]

Originally Posted by G0HZU_JMR

Sorry to be a neghead (again) but that probe design is very strange and I don't recommend anyone makes it unless they want to look at really small signals with a limited dynamic range. The overall design is not good. Best to leave it at that.

Put me down for 100% agreement with Jeremy!

AC coupling on the input and output, but then there's that pot backwards in the DC feedback path of the opamp - very strange.

An emitter follower outputting straight into a length of coax could prove rather foxy, too.

The overall structure is not good, and details of sections are risky.

David

[regenfreak]

The cheapo Ukranian probe is going to be a better bet I think.

Indeed the ukrainian seller does all the work for me. What am I going to lose paying £9 than less a meal at Nando?

For the budget conscience hobbyist, here is a cheapo way of showing Bode plot using an ordinary scope. Austerity rules :

https://www.youtube.com/watch?v=uMH2hGvqhlE

[regenfreak]

Presumably your new scope will have an FFT mode and if you select it you should be able to configure it to see quite small signals just like a spectrum analyser. You should be able to get down into the uV region like this although you would be best to select the 'flat top' window function if you want the best amplitude response behaviour as you sweep a signal with a sig gen and watch it on the scope.

Yes the siglent sds1202x-e has quite fast FFT. I will try that

[G0HZU_JMR]

An emitter follower outputting straight into a length of coax could prove rather foxy, too.
The overall structure is not good, and details of sections are risky.
David

Yes, it's a weird one alright. One aspect of it that worries me is that it looks like he's used the input bias current and the 1Meg resistor to set the DC operating point.

I've been stung so many times with opamps over the years. Sometimes it isn't enough to build a one off circuit or to simulate with a SPICE model because things like device to device variations of the input bias current (and its polarity) won't be captured by the model.

[Radio Wrangler]

Devices with input bias cancellation schemes can give surprising variations over temperature.

Look good on paper, though.

When reading a data sheet, it's vitally important to pay a lot of attention to what isn't said.

David

[G0HZU_JMR]

A few other issues with the Carlson RF probe that might affect its suitability as a Q meter buffer include:

With the circuit as drawn and assuming there is a typical/favourable input bias current at the opamp, I think the output BJT will probably degrade the large signal linearity (into a 50R load) rather than enhance it. I think this particular opamp would actually do a better job driving the cable directly although I'd do it via a series resistor. With the circuit as drawn with a BJT I can't see how the DC operating point will be controlled. It seems to be at the mercy of the input bias current of the opamp. I've never used this opamp type before so I don't know how much this will vary from device to device in terms of level or polarity. Normally, you would try to design a circuit to make it immune to a wide variation in input bias current.

As David says, the feedback pot pinout looks very strange.

I'm not sure of the significance of the 50R at the BJT emitter. This doesn't set the output impedance at 50R because there will already be a very low impedance here. But the presenter implies this sets a 50R resistance here.

The back to back 1N5711 diodes at the input will begin to look resistive once any input signals reach the tens of mV level. So that would spoil the 10Meg input resistance by maybe two orders of magnitude. Hardly anyone would notice this but it would be significant if used as a Q meter detector looking for a tiny ringdown signal.

The diodes would also introduce intermodulation and would begin to act as a weak clamp once in the many hundreds of mV region. I think the BJT would also introduce significant distortion at similar signal levels if it is fed to a 50R load via coax. So only small signals can realistically be probed with it.

Having to put a compensation capacitor at the far end of the 50R coax is a very strange requirement for something meant to drive a 50R spectrum analyser.

It needs a dual supply of +/- 5V which isn't attractive for a simple RF probe.

The 0.01uF cap at the output will spoil the LF response.

With the opamp input looking like a capacitive load at the MOSFET source follower output, the MOSFET will probably generate a fair bit of negative resistance at its input across a wide bandwidth. However, I doubt many people would
encounter any ill effects from this unless they went looking for ways to exploit the negative resistance. But it might give falsely high Q readings on certain inductors at certain test frequencies in the HF band.

[G0HZU_JMR]

If anyone is thinking of making the Cymar Q meter then they might be interested in the plot below. A few posts ago I warned that the input impedance will tumble very quickly with this buffer amplifier because it uses a common source JFET amplifier as the first stage. Presumably the design aim here was to get more drive level into the diode detector.

However, the JFET will probably have just over 1pF of drain-gate feedback capacitance and this will cause problems. 1 -1.5pF doesn't sound like much, especially down at a few MHz but it doesn't have to do much to degrade the input impedance. It's possible to predict the impedance vs frequency based on the Gm and the Miller capacitance with a basic model.

However, in the simulation plot below I used a Genesys library model of a JFET from the same process group (50) as the 2N3819. The circuit with the JFET and BJT was entered on the simulator as per the Cymar schematic and I also built it for real using a 2N3819 and a 2N3904 BJT. I tested the real circuit on my VNA after an Ecal that set the reference plane right at the input pin of the JFET. So the measurements should be correct and I've converted Rs and Cs to Rp and Cp in the plot. I also included a couple of traces that show Rp and Cp for a simulated 82k resistor in parallel with 40pF. This is to show that I really am displaying Rp and Cp correctly in the graph.

If you look at the plot below you can see that the shunt resistance Rp drops to about 35k ohm by 2MHz for both the library model JFET and also for my real circuit measured with the VNA. The agreement is very close. Much closer than I was expecting. By 10MHz Rp drops to just under 4k ohm.

The brown trace is the Rp for the real circuit and the red trace is Rp for the model. The dark blue trace is Cp for the real circuit and the light blue trace is Cp for the model.

This amount of damping is totally unsuitable for a buffer amplifier used in an LF-HF Q meter in my opinion.

[Bazz4CQJ]

I'm absolutely sure you've got it right Jeremy - I've built that circuit and it didn't work, but I wish I had £1 for every time a 3819 or an MPF102 or similar has appeared in similar designs in magazines or on the net!

B

[regenfreak]

In the BSc thesis by the student, he used VNA to sweep the S parameter for the poor man's Fetprobe. He wrote:

it is advised against using RF-probes below 1MHz, as at some point the AC-coupling may distort lower frequency signals



What does it mean "AC-coupling"? Is it to do with the AC-coupling DC blocking cap inside the scope or he is talking about something else? Why distortion occurs below 1MHz?

[G0HZU_JMR]

I'm absolutely sure you've got it right Jeremy - I've built that circuit and it didn't work, but I wish I had £1 for every time a 3819 or an MPF102 or similar has appeared in similar designs in magazines or on the net!

Thanks! I think what doesn't help is that many (most?) design papers (or theory books) that cover JFETs will treat them as having high input impedance no matter how they are configured. So someone could easily hold a classic theory book in my face, show me the design equations in the book for the input impedance and tell me I must be wrong.

However, not many books will model the JFET as an RF device and then go on to describe the consequences of the model. I tried looking online and struggled to find anything decent. The classic RF design books from 30-40 years ago are really dated in terms of presentation and they often don't fully press home the implications of feedback capacitance for example. There will be equations of course but it is up to the reader to study them and master what they really mean when impedance is plotted against frequency for example.

[G0HZU_JMR]

What does it mean "AC-coupling"? Is it to do with the AC-coupling DC blocking cap inside the scope or he is talking about something else? Why distortion occurs below 1MHz?

Yes, I think he just means that the frequency response won't be flat down to very low frequencies because of the input coupling. So the response at LF will be distorted.

However, it depends on the design. Many RF probes can be used down below 1kHz. I think my old Marconi TK2374 110MHz active RF probe works down to just 500Hz.

My Marconi 2388 1GHz active RF probe isn't quite as impressive as it is spec'd to work down to 100kHz at -3dB.

[Al (astral highway)]

Originally Posted by G0HZU_JMR

.

But the Tesla coil is a parallel resonant system.. A revised method would be to look at the right side of the smith chart. In theory at least you could find the points where the +/- imaginary value matches the parallel resistance Rp (at resonance). This would give the 3dB BW. But I don't think a regular VNA can do this because the impedances are so high. So I think I've missed what the research group means here.

Hi Jeremy, thanks for your detailed input as usual.

The research group started by noting that best practice in helical filter design could improve the performance of a Tesla transformer. They meant by this not the whole tank circuit, just the secondary coil.

They are focussed on possibilities for spectral purity, probably because a Tesla transformer, unless designed in the way they illustrate and confined to a Faraday cage, is potentially a jamming system, intentionally or not.

For the OP, they note another expression for Q...

Q= wL/RLoss

Also interesting is the table of values of Q plotted against self-resonant frequency. (See second photo).

This is taken from scientific research, but useful for home experimenters as accurately predictive for well-design/built secondary coils.

[Al (astral highway)]

Originally Posted by regenfreak

It is harder to measure Q with tesla coils because any human body or walls near the coil will affect the results. The same is true with the measurement of resonance frequency that I have to model the spark loading with a wire during physical measurement and software modelling with Javatc.

Hi there,

So true. Once the Tesla coil is powered up, the self-resonant frequency varies continuously, at an unpredictable rate. This is because the plasma at the antinode also has capacitance and is produced in a chaotic pattern, even if you are controlling the coil's on-time with regular pulses.

The best way to measure the self-resonant frequency of the unloaded coil is here, by David W Knight, G3YNH

First pic shows the helical coil next to a loosely coupled loop. Second pic shows the measurement set up.

The coil can be completely open-circuit. I did this experiment a while ago and posted the results in this thread:https://www.vintage-radio.net/forum/...d.php?t=141491

Although the secondary coil is resting on a book, the measurement is reasonably accurate. However, I later isolated it from contact with anything for greater precision.

In the background you can see on the 'scope trace the square wave in the driver circuit (loosely coupled out of the picture), and then the sine wave, which is the secondary coil at resonance. The coil is completely open-circuit.

Enjoy!

[G0HZU_JMR]

Interesting stuff!
I'm way out of my comfort zone when it comes to giving advice about those huge Tesla coils but I can offer up a few things that may be of interest.

The best way to measure the self-resonant frequency of the unloaded coil is here, by David W Knight, G3YNH

Yes, that's a neat way to do things and he is effectively using huge E and H field probes. However, that test is with the coil in free space. If it is mounted vertically and grounded to a base there will be a different type of discontinuity at the base end of the coil. So this will obviously change the frequencies of any resonances quite a bit.

It is possible to deliberately mess with the higher order resonance modes (on a typical helical resonator) if you flip the winding direction at key spots. I think Peter Vizmuller has an old patent on this. He is the chap who wrote this classic (short form) reference book and we have a copy at work.

https://www.amazon.co.uk/Systems-Cir.../dp/0890067546

However, I don't think he covers this topic in his book.

Although the higher resonance modes are referred to as harmonic modes, they rarely occur at exact multiples of the fundamental resonance. It all has to do with the dimensions of the coil. I did a lot of work trying to model all this stuff many years ago up to several GHz. At work, my main interest is with much smaller coils up at higher frequencies but I think a lot of the issues are the same. Just everything is smaller!

The other thing to be wary of is that if you lay the coil on its side and it is over a ground area and one end of the coil is grounded you can see the Q of the resonator go up a lot compared to the case when the coil is vertical. I think this is because it sets up a transmission line mode with the ground that runs parallel with it. This sharpens up the bandwidth at resonance.

So I would suggest that the Q needs to be measured with the coil in its natural environment. Otherwise the results could vary a lot.

Note:
With a huge long (transmission line?) structure like a Tesla coil I'm not sure how much uncertainty the higher order resonance modes can add to the ringdown method when viewed in the time domain on a scope. But presumably, what you see on the scope will be a waveform that has F and ~3F and ~5F components.

[Al (astral highway)]

Originally Posted by G0HZU_JMR

Although the higher resonance modes are referred to as harmonic modes, they rarely occur at exact multiples of the fundamental resonance. It all has to do with the dimensions of the coil.

That's really interesting insight, Jeremy. It's great to have your experience on board from a professional research environment, brilliant.

Originally Posted by G0HZU_JMR

o I would suggest that the Q needs to be measured with the coil in its natural environment. Otherwise the results could vary a lot.

Absolutely. And when I look at the efforts that are put into isolating air-cored coils from coupling into others in the few bits of military spec equipment I've examined out of curiosity, even small-signal stuff, that really comes through.

Originally Posted by G0HZU_JMR

with a huge long (transmission line?) structure like a Tesla coil I'm not sure how much uncertainty the higher order resonance modes can add to the ringdown method when viewed in the time domain on a scope. But presumably, what you see on the scope will be a waveform that has F and ~3F and ~5F components.

Absolutely!

The authors of the same research paper I quoted from before say that a helical resonator (or a Tesla coil, for that matter,) exhibits a minimum electric field strength at its grounded end and a maximum at the other end, whilst for the magnetic field it is a maximum at the grounded end and a minimum at the opposite end. I knew this intuitively but I hadn't made it explicit, to be honest.

Also that a standing wave is set up along the resonant winding [at the resonant frequency] and a series of related wavelengths when the electrical length of the helix is equivalent to:

f, 3f, 5f,... (2n+1)f, n=1,2...

exactly as you point out!

They go on to say:

'The standing waves terms represent a series of modes which describe the frequencies of currents flowing in the windings.'

They point out that this is true for the cavity filter's shortened helix, (something I've never knowingly seen - a cavity filter I mean)...

[regenfreak]

Yes, I think he just means that the frequency response won't be flat down to very low frequencies because of the input coupling. So the response at LF will be distorted.

Thanks Jeremy for sharing your insights.

The best way to measure the self-resonant frequency of the unloaded coil is here, by David W Knight, G3YNH

Thanks Al. It is very interesting article. The method for the measurement of self capacitance described by David Knight is the same linear correlation graphical method that I was describing in earlier posts (as seen in Tehman's textbook 1943). I can see your very interesting Russian GU81 VTTC project.

The field probes described in his scattering experiment instantly reminds me of my very first homebrew transmitter and receiver built 6 months ago; a replication of the set-up used in the famous Hertz wireless experiment performed in 1887. Instead of using a loop antenna, I used a half wavelength dipole with silver-nickel coherer detector similar to this set-up (without the parabolic mirrors):

https://www.youtube.com/watch?v=xNTHbiKmwNQ

The nodal points in David Knight's paper also remind me of the racing sparks when the coupling between primary and secondary coils is too tight or you put too much power into too a small resonator. I came across" racing sparks" that they burnt damaging holes at certain points on the secondary (very fascinating to watch)..I am not sure if the harmonics have anything to do with racing sparks. It seems no one knows why it occurs..

The other thing to be wary of is that if you lay the coil on its side and it is over a ground area and one end of the coil is grounded you can see the Q of the resonator go up a lot compared to the case when the coil is vertical.

There exists a type of tesla coil called bipolar tesla coil mounted horizontally without one end being grounded:

https://www.youtube.com/watch?v=Yj5UPufwGRw

However as I mentioned before the Q does not have to be very high in tesla coils to in order to be "bad ass".

[G0HZU_JMR]

Here's a quick Youtube video showing a vertical coil being probed for resonances. You can see that in this case the resonances aren't really at 3F and 5F etc. You can also see a few nulls in the resonances as I move the probe down the structure.

I also produced a physical model of this setup based on the 'simple' version of the works model. This fairly complicated (distributed) model captures the frequencies of the resonances fairly well and it is also possible to probe the model at various points and it correctly predicts where the nulls will be for each resonance mode. However, the errors creep in towards 500MHz and the model gets things a bit wrong up here as it gets the next resonance in the wrong place.

See the video below and also the simulation of the model at the corresponding distances along the structure.

https://youtu.be/i7hhkUPgfFQ

It's not exactly a tesla coil but it does show the expected resonance modes predicted by the model. If I get time over the weekend I'll make something with thicker wire and more turns per cm.

[regenfreak]

Very cool man

I must say I don't really understand the theory in the paper. I can forgive myself for that as I am a newbie and not an electronics engineer by trade.

Could use a string of mini neon bulbs or LED to visualise the nodal pattern if you use a high power transmitter. Something like this:

https://www.youtube.com/watch?v=adJp1zO9qfo

For those cash-strapped wireless experimenter, this is a $10 H-probe with a cheap ebay RF amplifier:

https://www.youtube.com/watch?v=2xy3Hm1_ZqI

Use it with a normal scope as a spectrum analyser:

https://www.youtube.com/watch?v=nImoQcoqkuQ

[regenfreak]

Move each probe closer and closer to the coil until you get a peak reading on the swept display and try and keep the probes far away enough that you get a steady reading but the bandwidth doesn't widen (through remote loading of the Tesla coil). Then just measure the 3dB bandwidth of the peak on the analyser and you can work out the Q. I use this method a lot for LC resonators or helical resonators. The H field probe couples a tiny amount of energy into the coil and the energy then sloshes back and forth inside the resonator between E and H fields. The E field probe can then pick up the E field part of all this. As I said earlier, I wrote some software that runs on my little netbook and this controls the VNA and it measures the Q in real time using active markers on the VNA. So once connected up, it can measure Q in real time. It will also measure the loaded Q if you bring anything near the coil that loads it. eg another smaller (primary?) coil.

This method does away with source impedance error and also the detector error because the coupling to the system is remote and very light. But you do need a sensitive analyser as the signals returning will be in uV. With a coil as big as a Tesla coil the E and H field probes could probably be a couple of feet away and still get a solid measurement.

Jeremy this H-field method is doable if I bother to build my homebrew H-probe and use the FFT of my new scope.. Thanks for the ideas.

I played with the antenna probe 3dB sweep method using the secondary of my solid state tesla coil, it worked very well but it was a rough quantitative measurement because I am such a lazy arse to set the rig properly isolating the interferences of nearby objects...

I tried the ring down method with 10 M resistor coupling, it wasn't a great method because I use a crappy cheap Chinese signal generator...

[G0HZU_JMR]

For this stuff you can probably just use a loop of bare wire for the H field probe. Maybe experiment with giving it a few turns and experiment with diameter etc. As long as the loop is connected across the coax (so it shorts it at DC) it should be good enough and it will take about a minute to make it.