Q-meter questions

[G0HZU_JMR]

If it helps, see below for a comparison of a 50uH coil wound on a Siemens pot core. This is based on s1p data of the coil I took with the VNA some time ago.

The brown trace is the inductance vs frequency of the real 50uH coil as seen on a VNA. The blue trace is a best fit model that uses a fixed value for Cdist.

You can see the creeping error that starts to widen at 4MHz which is half the Fo frequency.

If I try and adjust Cdist such that Fo is the same (8.1MHz) for both the coil and the model then the curve fit for inductance doesn't track as well at lower frequencies. So it's poor again for a different reason. All this is because the basic lumped model for C dist in all the basic RF text books is a bit flawed.

[G0HZU_JMR]

Note that the above applies mainly to solenoid type inductors. For toroidal inductors (eg wound on a typical powdered iron toroid) the wideband physical model is quite different compared to a conventional solenoid model and in my experience the inductance vs frequency for the toroid will be much better agreement with the fixed Cdist model right up to Fo.

[Argus25]

Originally Posted by regenfreak

Woww awesome they are beautifully built. Thanks for the ideas on how to measure the actual tracking errors..I am sure I will spend a lot of time figuring it out.

Thanks.

One of those links was incorrect, this was the one I should have posted:

http://worldphaco.com/uploads/THE_AF...OC16_RADIO.pdf

Unfortunately, on my first attempt I had an incorrect value of padder. When I corrected it to fix the tracking, then the dial calibration markings were off and I had to re-do the dial markings, so what you see there is the second attempt.

I'm a bit fanatical about accurate dial markings, but because my radios have a buffered L/O out and I have frequency counters with IF offset options I can see the tuned frequency that way, which of course would draw my attention to an inaccurate dial.

The dial on that radio is a epicyclic ball type, and I added lamps to a light box at the rear of it. This radio has all germanium transistors, all but one, which is for the L/O buffer output. It also contains two of the now rare OA31 power germanium rectifiers which are used to "mix" or OR gate either the internal battery supply or an external DC supply without switching.

[David G4EBT]

There was the PW 'Sarum' Q Meter in November 1978 Practical Wireless, which can be found here:

https://www.americanradiohistory.com...PW-1978-11.pdf

Also, the Raymond Haigh 'Q' Meter Adaptor, which featured in Electronics Today International, Vol 27, Issue 9 and was a much simpler device, no doubt with its limitations. Unfortunately, that's a missing issue of ETI in the American Radio History Archive, but I've attached scans of the pages of the magazine, (not very good quality).

If nothing else, the articles might be of interest, albeit the original poster's request for a 'Q' meter has yet to be met, and presumably he's looking for a commercial instrument - not a home-brew project. I got all the bits together to build the Sarum, etched the PCB and wound the six coils, but I never got round to building it. I guess the bits have long since been scattered to the fours winds - quite scary to realise that the article was written 4 decades ago.

[Radio Wrangler]

Q meters live between two problems... measuring devices whose self resonant frequency isn't far removed from the test frequency makes it difficult to determine the value of the inductor, and as Jeremy has shown, models can be iffy in this area.

The second problem is that measuring high Q components places severe requirements on the losses of the test jig built into the Q meter. THe HP and Marconi G meters are solid machined structures integrated with the variable capacitors. Ordinary radio variable capacitors aren't in the same league, and the business of producing a very low injection resistance without bad uncertainty or stray L is difficult. To a point, the Q meter's Q can be accounted for, but it makes for growing errors at higher Qs.

Home built ones can be fun, but there are plenty of Marconi ones around and they are restorable to a decent extent and a much better starting point than an ordinary variable C.

The HP Q meter is FET based and avoids a difficult valve, but they are rare beasties.

A dead Marconi even the very old version may be a good starting point for a home brew version, just for it's hefty tank circuit

David

[G0HZU_JMR]

One thing to be wary of with a JFET based solution will be the risk of negative resistance appearing at the gate. This is very likely to happen with a JFET arranged as a source follower with a diode detector at the output, especially if two diodes are used in the detector. The small signal capacitance of the diodes at the source is probably going to be enough to let the JFET generate significant negative resistance at its input. So instead of looking like a passive load it can resemble a load with a positive reflection coefficient.

In terms of measuring Q this would begin to affect the accuracy of the instrument when measuring inductors with fairly high Q. The JFET would act as a very mild Q multiplier and it might be possible for it to artificially boost a Q reading of 300 by maybe 10% for example. However, this depends on the coil inductance and the test frequency and so the impact will vary depending on this.

This is no big deal if you only want to measure low Q inductors but this issue alone would be enough to take up the entire accuracy spec of a commercial Q meter unless it was countered in some way.

The HP Q meter is FET based and avoids a difficult valve, but they are rare beasties.

Thanks. I had a quick look at the HP detector circuit and it looks like they use a complicated circuit involving a JFET and several BJTs. It also runs off a huge split supply of +/-25V to allow it to cope with big voltages. It looks like they use bootstrapping to boost the input impedance of the JFET .

I have to assume the other BJT based gubbins around the JFET will help prevent any significant amounts of negative resistance. There are some ferrite beads in there to keep it all tame so this probably took some dev work to get right.

On the Haigh Q meter circuit posted up by David, you can see that the JFET has a series 100R resistor at the gate and the JFET source feeds to a diode detector that will probably look like a small capacitor to small signals. This 100R resistor is about right for taking away the worst of the risk of negative resistance as I'd expect the 2N3819 JFET gate to look like -100R in series with maybe 2 or 3pF. This model would hold true over a few MHz up into VHF. However, a lot depends on the capacitance of the detector diode so these numbers are just an estimate.

So my guess is that this series 100R was added to cancel the negative series resistance at the gate and tame the circuit when certain test networks were presented to it.

[regenfreak]

Thank you all for the contributions. This thread would be better placed in the Test Gear section. There are so many useful attachments here. It is immaterial whether I can buy one or homebrew one, the articles and the information here are all very useful for understanding the limitations and challenges of measuring Qu.

[Skywave]

Originally Posted by David G4EBT

There was the PW 'Sarum' Q Meter in November 1978 Practical Wireless, which can be found here:
https://www.americanradiohistory.com...PW-1978-11.pdf

That Q-meter is the one I was referring to in my post #8, q.v.

Al.

[G0HZU_JMR]

All of the magazine based Q meters shown so far are quite poor in my opinion.

Whilst dinner was in the oven I knocked up an excel sheet with the Rowe correction equation in it. This is the equation that corrects for the finite source impedance. The equation does work well but I thought I'd show a screenshot of what can go wrong if you don't use this equation after using the meter for some inductor values at certain frequencies.

If you had a 4.7uH inductor with a Q of 200 at about 7MHz and you put it on the Rowe meter then this is what I think will happen...

If we assume there is no other source of error in the meter other than the source impedance then on the X1 range (Rs = 4.6R, Q = 20-100) I think the meter will read a Q of just 36.5 at 7MHz with this inductor. On the X5 range (Rs = 0.92R, Q up to 500) I think the meter will read a Q of about 105. So this is a long way away from the correct Q value of 200.

Applying the Rowe equation (10) corrects the Q to 200 in both cases but I'd question how many people would actually do this? Would someone in 1969 really have sat down and worked this out with a pen and paper each time they used the meter?

Also, when the error gets this big the correction relies on the accuracy of the meter to get the first reading right. In this case where the initial Q error is so large, a 10% error in the initial reading of 105 would cause equation 10 to 'correct' with about a 20% error compared to the correct Q of 200.

It's not all bad news, the meter won't always give errors this high and for many inductor types the results will be quite reasonable. However, I think you would have to become quite savvy and be alert when the X1 and X5 range readings are a long way apart as in the case below. Equation 10 would definitely be needed here!

[Philips210]

Hi

There was also a Q meter project in Elektor Electronics, April 1990 issue and may be found here https://www.americanradiohistory.com...or-1990-04.pdf

Regards,
Symon.

[regenfreak]

thanks. This is more complicated schematic.

$9000 today price? Must you be loaded in order to afford the measurement of the unloaded Q?

https://www.youtube.com/watch?v=7et0pJzrQJU

[regenfreak]

This are the mathematical derivations of the equations:

http://www.docente.unicas.it/userupl...les/qmeter.pdf

[G0HZU_JMR]

Originally Posted by Philips210

Hi
There was also a Q meter project in Elektor Electronics, April 1990 issue and may be found here
https://www.americanradiohistory.com...or-1990-04.pdf
Regards,
Symon.

Interesting transformer in that one!

The author suggests that this meter is accurate to within about 10%. That's quite a claim. I don't know how much harmonic distortion there will be from the oscillator but this will contribute to uncertainty and this alone will knock a dent in that accuracy claim. Also the BF981 buffer amplifiers are a bit scary from a stability point of view. Even a crude model of the BF981 predicts a LOT of negative resistance over a huge bandwidth. My model predicts the negative resistance will still hold up even with 15-20pF shunted across the BF981 input. Note that this model assumes there is a small capacitance shunted at the BF981 source leg and this will be due to the detector diodes.

So the big air variable 500pF capacitor would need to have a minimum capacitance of over 20pF in order to prevent a net negative resistance at the Gate 1 terminal that could cause oscillation. Otherwise I think the detector port could oscillate up at VHF with certain high Q inductors at low capacitance settings. This is based on a crude VCCS model derived from the BF981 datasheet.

It's an interesting circuit but I don't believe the global 10% accuracy claim and I'd expect it to exhibit some strange behaviour with certain settings especially if the main tuning capacitor can get down to below about 20pF with certain test inductors and at certain test frequencies. My most basic Q jig here has a 500pF variable cap and it gets down to 15pF on the minimum setting.

I very crudely modelled the BF981 based on the datasheet and then put a 1.5uH test coil at the input that has a high Q of 200 or so. With a dial capacitance of 18pF the model predicts the detector will oscillate at 30MHz as in the simulation below. Even if it doesn't oscillate like this in reality, I'd expect the negative resistance issue to spoil the 10% accuracy claim.

I really must be missing something here because I'd expect to see some obvious problems with grossly elevated Q readings with certain highish Q inductors at certain test frequencies. To me there is a fundamental design problem there but maybe this doesn't matter if you never ever use it under these conditions. I'll look at it again tomorrow but I really want to walk away from this one too...

[mickm3for]

Hi the circuit has to have a very low .2 ohm feed resistance this gives added problems of drive current. the pw sarum gives a low reading less than half value of Q
i tried lots of circuit ideas wide band transformers with solid wire through center of toriod etc but settled on this as a resonable circuit (not mine)
It gives resonable readings on known Q coils.
feed resistance is .2 ohms later changed to wb transformer if details required will find details as worth doing . circuit is coppyright Lloyd Butler VK5BR ok for non commercial use. Tony Naylor of Sprectum comms was going to produce a kit i dont know if he has hope this helps

[G0HZU_JMR]

Yes, you could try improving the Sarum circuit in various ways. I can see that the series 100R has now been added at the gate of the JFET in that circuit. This will be to offset the typical -100R negative resistance in the JFET when it is fed to a capacitive load at the source. There only has to be 1 or 2pF here to cause problems with negative resistance.

Here's a couple of classic articles about Q measurement by Wes Hayward W7ZOI. He uses a couple of the methods I use but I also have others including the swept method with E and H field probes.

But the attached pdfs by Hayward are worth a read. His Q jigs are as basic and scruffy as mine!

Anyone with an untested Q meter could use the methods described by Hayward as a means to cross check the accuracy of their meter.

I also dug out a genuine old 2N3819 and configured it as per the original Sarum circuit with a 1Meg gate resistor and 1k in the source. Instead of using diodes I placed a 2p2 cap in shunt at the source of the JFET. This is a more controlled test as the capacitance of the detector diodes will vary with drive level. But even 2p2 is enough to give plenty of negative resistance.

It is compared to a model I derived from the 2N3819 datasheet and the curves for resistance and capacitance agree very well.

You can see there is about -50 to -100R at the gate over quite a bandwidth. This is a very difficult measurement for a VNA once below 30MHz. You can see the trace data getting quite noisy on the resistance trace. A very good VNA is needed here!

[G0HZU_JMR]

Negative resistance is a bad thing to have in a Q meter detector because it can cause instability but it can also cause the indicated Q to appear higher than it really is.

You can think of this along the same lines as a regenerative detector in a radio. As the feedback is adjusted with the regen control, the gain and selectivity improve. These are nice features in a regen receiver but this is bad news in a Q meter. Even a small amount of negative resistance can affect the Q meter accuracy. However, it's most likely to happen at higher test frequencies with the capacitor dial near its lowest settings. So this would be inductors of a few uH with moderate to high Q tested up at maybe 12-30MHz with the capacitance dial down at 25pF or so.

[G0HZU_JMR]

Hope I'm not boring too many people with this stuff but here is a simulation of a comparison between an almost perfect Q meter (the top circuit) and the same Q meter but with the impact of connecting the 2N3819 detector to it. Note that this detector does not have the series 100R gate resistor so it is presenting all of its negative resistance to the circuit under test.

The equivalent negative resistance model of the 2N3819 is -90R in series with 2pF and this gives exactly the same 'negative' loading effect as the 2N3819 data file from the VNA.

The 4.7uH test inductor has a Q of 160 at 16MHz and the top meter gets quite close. The -3dB markers on the red trace show a Q of about 156.

However, because of the regenerative effect when the 2N3819 is added, the blue trace shows a Q of 210. Quite a difference. So I think it's a good idea to minimise the negative resistance and also try and minimise the capacitance of the diode detector. Maybe choose diode(s) that have low capacitance when they are detecting large signals. The diode capacitance usually gets a lot less once it is reverse biased by its own (detected) DC output.

In the example below, the total diode capacitance at the 2N3819 source was modelled with a fixed 2.2pF cap. This assumed there would be two low capacitance diodes fitted here.

[Radio Wrangler]

Originally Posted by G0HZU_JMR

You can think of this along the same lines as a regenerative detector in a radio.

In the past it was not unusual for shortwave radios to have a valve with a bit of regeneration coupled into a tuned circuit to "sharpen it up a bit" the gain needed careful adjustment to have it on the side of stability. As a clue to what negative resistance would do in a Q-meter, the circuit was well known as a "Q-multiplier" You can find a lot by searching that name

Life on the outside of the Smith chart...

David

[regenfreak]

Originally Posted by G0HZU_JMR

Here's a couple of classic articles about Q measurement by Wes Hayward W7ZOI. He uses a couple of the methods I use but I also have others including the swept method with E and H field probes.

But the attached pdfs by Hayward are worth a read. His Q jigs are as basic and scruffy as mine!

In the Q-twofaces file, part 1, I have seen similar methods used in another article that I cannot find now. Cc are de-coupling capacitors with very small values like 1-2pF. The bigger the Cc value is, the more Q is loaded down. There are variations of this method using weak inductive and capacitance coupling without the need of a VNA.
I don't have a vector network analyser or other fancy equipment. The best thing in life is free or wont cost a fortune. I am only interested in poor man's measurement techniques using a basic scope and a signal generator. The biggest problem with 10:1 probe is that it will still load down the Q significantly. I have ordered myself a 100:1 probe and a cheap active scope probe from Ukraine(I know people don't trust it but hey I am measuring Q up 1MHz not 1GHz)...I think there is a way working around it and I will just have to figure it out.

[VT FUSE]

How does Red Pitaya by STEMlab stand up as a tool for analysing Q of inductors? Seems like this board could form the basis of a pretty comprehensive bit of kit.

Attended a short presentation by a well respected Radio Engineer where he presented his usage and the versatility of the board.