6-gang FM stereo tuner heads

[regenfreak]

I have attached the comparison of different VNA measurement configurations and their relative error from 0.001ohm to 100K ohms. Clearly the transmission measurement is the winner compared with S11 measurement in DUT with high impedance. I cant remember where I got theses pictures from...

The third attachment shows the frequency sweep of my 6-gang FM tuner using the 0. 7pF Poorman's probe measured at the air variable capacitor of the last LC bandpass filter. Note the presence of the oscillator signal and noises from a powerful local London pirate station.

The 4th attachment shows my first attempt to measure the higher order harmonics of the 6-gang FM tuner oscillator with the poorman's probe measured at the frequency counter indicator is coupled to the BJT oscillator via a 1pF capacitor so there is little load pulling...

[regenfreak]

I have made my 2nd attempt to do series through measurement of a 1206 SMD 10K resistor with my homebrew calibration standards (see attachments) from 1MHz to 4GHz. I use the tips of the SMA connectors as measurement planes. This time I have got much closer results to the graph in the web link in my previous post through careful calibration. My measurement shows the asymptotic stray capacitance of a 10K resistor is about 0.054pF or 54fF at 2.3GHz. I attached 3db attenuators for both port 1 and port 2 to improve the accuracy of the measurement. A small bit of excess solder would have noticeable impact on the stray capacitance measurement.


PS for my sanity check, I have repeated the measurement of the helical filter using the Tek P6202 probe, I can see the identical dip at the resonance peak.

[regenfreak]

The explanation for the strange dip of S21 in the resonance point of the helical cavity filter is simple. When i measured the S21 with an active probe sampled on an open tranmission line, there is a bigger impedance mismatch and energy is reflected from the port 2. One end of filter was not terminated. Once i repeated the measurement with the transmission line terminated with a 50-ohm load, then the dip disappeared. However, there was a certain degree of mismatch. As i mentioned before, the helical cavity filter may have non-50ohm I/O impedance as the S11 is close to -10db no matter how hard i tried to fine tune it. In any case, i didnt see obvious evidence of negative resistance of both Poorman's and Tek probes when i probed both the high- and low-impedance nodes of moderately high Q filters. The effect is not easily quantifiable.

[regenfreak]

One of the most unpalatable aspects of learning to use Smith chart is to find the matching network of a black box device. Taking the previous measurement of the 465MHz helical cavity filter, the complex impedance reads Z = 81.37-j29.52 at the resonance frequency of f = 465.7MHz from the Smith chart(attachment 1). Last night I stumbled across this online Smith chart software:

https://www.will-kelsey.com/smith_chart/

Using the paper-based Smith chart "Yang-Ying" method described by W2AE2W in his video:

https://youtu.be/TsXd6GktlYQ

I have found the matching network for the helical filter to a 50-ohm system at the resonance frequency is a parallel inductor of 24.5nH and series capacitor of 7.5pF.

I guess the most effective strategy to overcome phobia of snakes is through exposure therapy. I am learning to get more comfortable with Smith chart in spite of the past phobia...it makes more sense if it is intuitive to use.

The design of broadband impedance matching network seems very complex as illustrated by this video with the use of Keysight ADS software simulation:

https://youtu.be/XpR6uoKYfF4

[Radio Wrangler]

Spot frequency matching of something isn't too difficult, until you introduce some additional requirements or limitations.

Going back to FM tuners, 88 to 108MHz is a 20MHz band centred near enough on 100MHz, so that's a 20% relative bandwidth. Matching networks trying to hit a reasonable compromise over that sort of bandwidth are difficult.

There is one trick with the Smith chart which can be used. Instead of going from one impedance to the other in one arc pair, go in two pairs. Have an intermediate point at the geometric mean of the two impedances you are coupling. Make one arc pair be a deviation upwards, and the other a deviation downwards. What now happens is that as you swing frequency around, the impedance you make comes close to the target point and does a loop around near the target and then departs to mismatch territory.

This is more a sort of chebyshev technique where the impedance achieved is close to the target, but has a 2-dimensional ripple around it.

Take a step further and allow a lot of intermediary points, carefully spaced in equal geometric frequency increments and you network approaches a lumped element version of an impedance transforming horn structure. It's quite amazing once you see what were all rather different looking methods all converging and you see that they are different views of the same thing. The deviations from the ideal line for each of these arc pairs is smaller, the more of them you have, so Qs are lower.

I had to design a couple of transmitters for the air band, 118-137MHz which is almost as wide (in octaves) as the FM broadcast band. The difference is that broadcast transmitters get assigned a single frequency and get tweaked-up for it. The airband one has to work anywhere in the band without tweaking every time ATC tells you to change frequency. It's and awkwardly wide band, especially with efficiency targets to meet. So you have to get inventive.

Worse than this is when you have to march a device with impedance varying across your band. The only way I've been able to do this is to create a lumped element network which models the device Z, and to treat that as part of my matching network, a part I don't have to put on the board, at least.

You're growing familiarity, it's slow, but things get easier as you get more experienced. There are often several routes you can take, experience gives you a flying start on picking suitable ones to begin with

David

[G0HZU_JMR]

I think that trying to measure high impedances like this can be very challenging. Also, to get the very best from a VNA the operator has to use a good cal kit that is properly corrected and a test fixture that is properly calibrated out.

If you look at the W0qe link below you can see that the test fixture isn't great, and the results aren't that great either.

https://www.w0qe.com/Measuring_High_Z_with_VNA.html

The Jeroen Belleman s21 plots in your post #202 are relatively poor as well. There's lots of uncertainty ripple in the plots and there appears to be a creeping uncorrected error below about 100kHz. This stuff is fine for casual experimentation, but the plots should look a lot better than that.

See below for an old s21 plot I made of a series 100k 1206 resistor. I did this twice. Once with an expensive ecal unit and once with my homebrew mechanical cal kit. The results are overlaid in the plot below and agree very well up to about 6GHz. You can see there is no wobbly ripple anywhere.

I think you are asking too much of a VNA when you try and measure the input impedance of your probes using an s11 measurement and I can only suggest (again) that you make up a single resonator BPF at about 100MHz.

I'd recommend using decent RF caps (eg ATC or PPI) and a very tight point to point layout over a PCB groundplane.

Maybe start with 1pF coupling caps and a 5.6pF resonator cap and a 300nH inductor. A 7 turn inductor wound with 0.71mm diameter wire with about a 1cm diameter and about 1.5cm length should give an unloaded Q of about 400 at 100MHz.

The second image below is a simulation of this resonator. You can see the insertion loss is just under 2.5dB and the bulk of this loss is due to the 300nH inductor with its unloaded Q of about 400. If the simulated probe with 500k Rp and 0.7pF capacitance is connected (resistor R1 in the image below models the probe) then the resonance goes down by 4MHz and the insertion loss drops 0.5dB to about 3dB. If the probe Rp was 100k then the insertion loss would drop to about 4.25dB. This is a big difference in insertion loss and it sould be easy for the VNA to measure..

This resonator method is far from perfect, but it should give results in the correct ballpark. In the past I've used this method at 10.7MHz and at 70MHz and I have some old test jigs already built up for this purpose.

The idea is to measure the insertion loss with the 50R VNA and leave this measurement in place all the time. Then touch the probe at the resonator as in the image below. The far end of the probe cable can be terminated in a 50R load. This test only measures the loading effect of the probe. No actual measurements or plots are made with the probe.

[regenfreak]

Originally Posted by Radio Wrangler

Spot frequency matching of something isn't too difficult, until you introduce some additional requirements or limitations.

There is one trick with the Smith chart which can be used. Instead of going from one impedance to the other in one arc pair, go in two pairs. Have an intermediate point at the geometric mean of the two impedances you are coupling. Make one arc pair be a deviation upwards, and the other a deviation downwards. What now happens is that as you swing frequency around, the impedance you make comes close to the target point and does a loop around near the target and then departs to mismatch territory.

Take a step further and allow a lot of intermediary points, carefully spaced in equal geometric frequency increments and you network approaches a lumped element version of an impedance transforming horn structure. It's quite amazing once you see what were all rather different looking methods all converging and you see that they are different views of the same thing. The deviations from the ideal line for each of these arc pairs is smaller, the more of them you have, so Qs are lower.

I had to design a couple of transmitters for the air band, 118-137MHz which is almost as wide (in octaves) as the FM broadcast band. The difference is that broadcast transmitters get assigned a single frequency and get tweaked-up for it. The airband one has to work anywhere in the band without tweaking every time ATC tells you to change frequency. It's and awkwardly wide band, especially with efficiency targets to meet. So you have to get inventive.

Worse than this is when you have to march a device with impedance varying across your band. The only way I've been able to do this is to create a lumped element network which models the device Z, and to treat that as part of my matching network, a part I don't have to put on the board, at least.

You're growing familiarity, it's slow, but things get easier as you get more experienced. There are often several routes you can take, experience gives you a flying start on picking suitable ones to begin with

David

Thanks. I have only made a baby step forward. I downloaded the SimSmith software a few months ago but never tried to use it until today. It has taken me a few minutes to find the matching network for the helical filter using SimSmith. I didnt know i have a hidden gem installed in my PC!

Today it is my first time I have heard about the constant Q circles in broadband matching, its all advanced stuff:

http://www.circuitmason.com/work-boo...nstant-q-lines

I really like how W2AEW suggested the use of the sentences to help memorizing which direction the arcs should be moving:

"Adding inductors eLevate thru real axis"
"Adding capacitors Crash down thru real axis"

G0HZU_JMR[I think that trying to measure high impedances like this can be very challenging. Also, to get the very best from a VNA the operator has to use a good cal kit that is properly corrected and a test fixture that is properly calibrated out.

If you look at the W0qe link below you can see that the test fixture isn't great, and the results aren't that great either.

https://www.w0qe.com/Measuring_High_Z_with_VNA.html

The Jeroen Belleman s21 plots in your post #202 are relatively poor as well. There's lots of uncertainty ripple in the plots and there appears to be a creeping uncorrected error below about 100kHz. This stuff is fine for casual experimentation, but the plots should look a lot better than that.

See below for an old s21 plot I made of a series 100k 1206 resistor. I did this twice. Once with an expensive ecal unit and once with my homebrew mechanical cal kit. The results are overlaid in the plot below and agree very well up to about 6GHz. You can see there is no wobbly ripple anywhere.

I think you are asking too much of a VNA when you try and measure the input impedance of your probes using an s11 measurement and I can only suggest (again) that you make up a single resonator BPF at about 100MHz.

I'd recommend using decent RF caps (eg ATC or PPI) and a very tight point to point layout over a PCB groundplane.

Maybe start with 1pF coupling caps and a 5.6pF resonator cap and a 300nH inductor. A 7 turn inductor wound with 0.71mm diameter wire with about a 1cm diameter and about 1.5cm length should give an unloaded Q of about 400 at 100MHz.

The second image below is a simulation of this resonator. You can see the insertion loss is just under 2.5dB and the bulk of this loss is due to the 300nH inductor with its unloaded Q of about 400. If the simulated probe with 500k Rp and 0.7pF capacitance is connected (resistor R1 in the image below models the probe) then the resonance goes down by 4MHz and the insertion loss drops 0.5dB to about 3dB. If the probe Rp was 100k then the insertion loss would drop to about 4.25dB. This is a big difference in insertion loss and it sould be easy for the VNA to measure..

This resonator method is far from perfect, but it should give results in the correct ballpark. In the past I've used this method at 10.7MHz and at 70MHz and I have some old test jigs already built up for this purpose.

The idea is to measure the insertion loss with the 50R VNA and leave this measurement in place all the time. Then touch the probe at the resonator as in the image below. The far end of the probe cable can be terminated in a 50R load. This test only measures the loading effect of the probe. No actual measurements or plots are made with the probe.

Thanks for the suggestions. Your 100MHz resonator will be my next weekend project. I will wind the coil with silver plated wire and see if I have a small-value silver plated piston trimmer for the cap. I have some NPO 1pF ceramic capacitors lying around. I have not forgotten the diplexer you designed a while ago but the more I am digging into the measurements of IP3 for amplifiers and mixers, the more I realize how challenging it is to do the measurements properly.

To be fair, I am doing the 10K resistor measurement as a casual exercise for helping me to understand the nuance of measuring high impedance DUT better. My NanoVNA V2 plus 4 is many times cheaper than WOQE's or Jeroen Belleman's VNA. I dont expect too much of its accuracy, particularly at low frequencies. The V2 plus 4's accuracy below 10MHz is not great but still much better than the Siglent sva1015x which I owned for two weeks before i sent it back.

[G0HZU_JMR]

I'd recommend making the resonator jig on a thin piece of copper sheet. It would have to be thin enough that you could still solder to it and also thin enough to bend one large corner up. Maybe choose something 0.4mm thick?
Any thicker than this and it will be difficult to solder to unless you have a powerful iron and a suitably large tip on it. Also wear safety glasses when soldering stuff like this as excess solder can splash up really easily.

I'd recommend putting the connectors and the caps and the inductor on one side of the sheet (at the corner of the bottom side) and have a drilled peep hole in the metal shim to allow a test point to your probe on the other side of the metal sheet.

The idea would be to be able to bring the probe close to the top of the resonator without causing any proximity effects to the high Q inductor. I'll see if I can put something together to show you what I mean.

[G0HZU_JMR]

The highest impedance (commercial) probes I have here are my Marconi 2388 1GHz FET probe and a Marconi TK2374 RF probe.

I've also got my trusty old diode detector probe. This uses a classic HP 2800 Schottky diode and converts RF to DC. This is the classic probe that connects to a DVM. This probe works well but it is level dependent. The input impedance improves with higher RF drive levels. Both Rp and Cp improve. This probe can often be measured with a VNA. It's not going to perform well compared to the poor man's probe.

The Mi TK2374 is going to look poor as well.

The Mi 2388 probe has a variable (rotary) tip capacitor that acts as a variable 0-40dB attenuator so it may perform fairly well in a test at 100MHz.

Somewhere, I've also got a crude lashup of the Bob Pease probe although it isn't going to work well up at 100MHz. It also generates a lot of negative resistance across HF and into VHF. I'll see if I can find it.

[G0HZU_JMR]

I had a rummage and found some of my old impedance plots for my Mi2388. The plot below was taken with 40dB tip attenuation dialled in and this was done using my current lab VNA.

This shows that at 100MHz the Rp was about 75kΩ and the tip Cp was about 1.7pF.

You can see by the noise of the plots that the VNA was struggling a bit to make this measurement. Below 60MHz the VNA is really struggling. I don't think I used a low RBW or any averaging, so the Rp plot is very noisy even at 100MHz.

I can normally measure an Rp of 100kΩ fairly well up to about 100MHz with this VNA so I think the 75kΩ Rp measurement at 100MHz is going to be close to reality.

The user manual shows a graph of input Cp at 40dB tip attenuation and this shows just over 1pF. I would have made this measurement with a grounding sleeve and a ground spike fitted to the probe so maybe this is why I measured 1.7pF and not just over 1pF as in the manual.

[regenfreak]

Originally Posted by G0HZU_JMR

I'd recommend making the resonator jig on a thin piece of copper sheet. It would have to be thin enough that you could still solder to it and also thin enough to bend one large corner up. Maybe choose something 0.4mm thick?
Any thicker than this and it will be difficult to solder to unless you have a powerful iron and a suitably large tip on it. Also wear safety glasses when soldering stuff like this as excess solder can splash up really easily.

I'd recommend putting the connectors and the caps and the inductor on one side of the sheet (at the corner of the bottom side) and have a drilled peep hole in the metal shim to allow a test point to your probe on the other side of the metal sheet.

The idea would be to be able to bring the probe close to the top of the resonator without causing any proximity effects to the high Q inductor. I'll see if I can put something together to show you what I mean.

Thanks. I will have to order 0.4mm copper sheet. Can't i use double-sided copper clad PCB boards? I used up all my 0.4mm copper sheets for the ground solder points of my valve 6-gang FM tuner with its aluminum chassis. I have a 500C temp controlled solder station and some fat tips but I must say it is a little tricky to solder thick copper sheets even I have some special solder flux. I have a micro solder blow torch, but the flame is so hot that it just vaporizes everything!

I vaguely get the idea but not sure about the exact layout in your description. The other thing i have to watch out for is the clearance distance of the coil from the ground plate and the side wall, if they are too close, it will damp the Q of the LC tank.

The Mi 2388 probe has a variable (rotary) tip capacitor that acts as a variable 0-40dB attenuator so it may perform fairly well in a test at 100MHz.

Somewhere, I've also got a crude lashup of the Bob Pease probe although it isn't going to work well up at 100MHz. It also generates a lot of negative resistance across HF and into VHF. I'll see if I can find it.

Thats interesting. I will try to find the schematic for the Mi 2388 probe. I looked at Bob Pease probe a few weeks ago but he didnt explain anything about his design philosophy in his article and he called it "low frequency probe" so I didnt bother building it.

I had a rummage and found some of my old impedance plots for my Mi2388. The plot below was taken with 40dB tip attenuation dialled in and this was done using my current lab VNA.

This shows that at 100MHz the Rp was about 75kΩ and the tip Cp was about 1.7pF.

You can see by the noise of the plots that the VNA was struggling a bit to make this measurement. Below 60MHz the VNA is really struggling. I don't think I used a low RBW or any averaging, so the Rp plot is very noisy even at 100MHz.

I can normally measure an Rp of 100kΩ fairly well up to about 100MHz with this VNA so I think the 75kΩ Rp measurement at 100MHz is going to be close to reality.

The user manual shows a graph of input Cp at 40dB tip attenuation and this shows just over 1pF. I would have made this measurement with a grounding sleeve and a ground spike fitted to the probe so maybe this is why I measured 1.7pF and not just over 1pF as in the manual.

It is useful to see your old measurements. My Poorman's probe has Rp = 79.8K and Cp = 0.723pF at 100MHz. So the Rp is quite close to the Mi2388 at 100Mhz. 40db attenuation is quite a lot, equivalent to 100:1 voltage ratio.

[G0HZU_JMR]

I've had a go at making a first stab at the 100MHz resonator jig. I think the metal sheet I've used is 0.5mm thick so it was really difficult to solder to. I had to use a 4mm bit in my Weller WSP80 iron and a flux pen.

I've just tested the Mi 2388 with the resonator jig and the results agree with the old VNA measurement quite well. This isn't a very well controlled experiment as the attenuator is a continuous rotary type and the 40dB setting is set by eye. Therefore, I really should do an s11 and a jig measurement side by side with exactly the same rotary attenuation selected.

However, with no probe connected the insertion loss was 2.48dB at 101.6MHz. This was exaclty as predicted based on the predicted Q of the resonator inductor. With the Mi2388 touching the top of the resonator (via the tiny peep hole) the BPF response dipped to 93.17MHz with an insertion loss of 5.00dB now showing on the VNA.

The best way to reverse engineer the loading effect of the probe is to see which combination of Rp and Cp gives the same pulling/loading result and this delivered values of about Cp =1.55pF and Rp =83kΩ.

If I look at the Rp of the Mi 2388 down at 93MHz on my old VNA s11 measurement it is about 83kΩ. So this result agrees very closely.

I used an ATC 800B cap for the 5.6pF resonator cap and PPI (1111N series) for the 1pF coupling caps. I also fitted a 10dB attenuator at port 1 and port 2 of the jig before calibrating it. This improves the port match. I didn't bother with a full 2 port cal using the ecal module. A through cal with the 10dB attenuators should be OK for stuff like this.

I think it will be difficult to check the cal using high value SMD chip resistors up at 100MHz. Usually, my VNA tests show the Rp of a 100kΩ resistor falls with increasing frequency. I may have to try various thin and thick film types to try and find a resistor that behaves well up at VHF.

My test jig looks really crude and ugly, and it is only a temporary lashup. It's OK to prove a concept but should really be built into a proper and stable jig.

See below. This was made very quickly and crudely (and it shows ) You can see that the idea is to use the probe on the back of the jig where it's possible to probe a tiny test point that pokes through the peep hole.

Obviously, I'd recommend that you make something more robust and more presentable than this...

[G0HZU_JMR]

I just tried a 470kΩ 0805 chip resistor in parallel with a 0.7pF PPI 1111N cap in place of the probe. This was an attempt to check the accuracy of the jig at 100MHz.

When this was connected at the peep hole test point the jig predicted a resistance of 380kΩ in parallel with just over 0.8pF.

I also tried a 220kΩ in parallel with the 0.7pF cap and the jig predicted 178kΩ in parallel with just under 0.9pF.

I'm not sure why the jig reads a bit low for Rp but it might actually be correct. The effective Rp of these resistors may well fall up at VHF. I'm using cheapo thick film 0805 resistors for this test. it would be nice to know what the true Rp of these chip resistors is up at 100MHz.

[regenfreak]

Originally Posted by G0HZU_JMR

I've had a go at making a first stab at the 100MHz resonator jig. I think the metal sheet I've used is 0.5mm thick so it was really difficult to solder to. I had to use a 4mm bit in my Weller WSP80 iron and a flux pen.

I've just tested the Mi 2388 with the resonator jig and the results agree with the old VNA measurement quite well. This isn't a very well controlled experiment as the attenuator is a continuous rotary type and the 40dB setting is set by eye. Therefore, I really should do an s11 and a jig measurement side by side with exactly the same rotary attenuation selected.

However, with no probe connected the insertion loss was 2.48dB at 101.6MHz. This was exaclty as predicted based on the predicted Q of the resonator inductor. With the Mi2388 touching the top of the resonator (via the tiny peep hole) the BPF response dipped to 93.17MHz with an insertion loss of 5.00dB now showing on the VNA.

The best way to reverse engineer the loading effect of the probe is to see which combination of Rp and Cp gives the same pulling/loading result and this delivered values of about Cp =1.55pF and Rp =83kΩ.

If I look at the Rp of the Mi 2388 down at 93MHz on my old VNA s11 measurement it is about 83kΩ. So this result agrees very closely.

I used an ATC 800B cap for the 5.6pF resonator cap and PPI (1111N series) for the 1pF coupling caps. I also fitted a 10dB attenuator at port 1 and port 2 of the jig before calibrating it. This improves the port match. I didn't bother with a full 2 port cal using the ecal module. A through cal with the 10dB attenuators should be OK for stuff like this.

I think it will be difficult to check the cal using high value SMD chip resistors up at 100MHz. Usually, my VNA tests show the Rp of a 100kΩ resistor falls with increasing frequency. I may have to try various thin and thick film types to try and find a resistor that behaves well up at VHF.

My test jig looks really crude and ugly, and it is only a temporary lashup. It's OK to prove a concept but should really be built into a proper and stable jig.

See below. This was made very quickly and crudely (and it shows ) You can see that the idea is to use the probe on the back of the jig where it's possible to probe a tiny test point that pokes through the peep hole.

Obviously, I'd recommend that you make something more robust and more presentable than this...

Thanks. I am very surprised by the frequency drift of a massive 8.43MHz! I dont really understand.

Before I go to bed, I have spent 10 minutes to put together a "quick and dirty" test jig that I used previously. It is used for finding the unloaded Q of a VHF LC tank. The coil and cap specs are based on your suggestion. I used 1mm silvered plate wire and a 1-8pF piston ceramic trimmer. The wires are capacitive coupling "cat whiskers" so that there is at least -30db insertion losses both ports. The QL approaches asymptotically to Qu in a graph if the insertion loss of the external coupling to the resonator is greater than -30db. I have used my cheapo NWT200 to find unload Q which reads 357 which is somewhat lower than your simulated value of 400. I didnt bother with calibration and normalization with the NWT200 as the software is buggy and flaky, so ignore the absolute value of the vertical axis in db.

[regenfreak]

The attached file is the sweep of the homebrew 6 gang FM tuner with the 0.7pF Poorman's probe touching the first air variable capacitor gang (1st single tuned RF stage). I noticed that the measured resonance frequency is less than 700kHz deviation from what it shows on the tuner's frequency counter so the poorman's probe does not cause great frequency pulling like in your case. The 6-gang capacitor and the coils are all silver plated.

[G0HZU_JMR]

Thanks. I am very surprised by the frequency drift of a massive 8.43MHz! I dont really understand.

I chose the resonator components such that a small change in Cp across the resonator would show a large shift in frequency. Also, I designed it for a high loaded Q and used an inductor with a Q of about 400. The effective Rp of the resonator (when fitted in the jig) is about 20kΩ at 100MHz so this makes it sensitive to the Rp loading of any probe that gets connected here. This is exactly what is required if the jig is going to be able to detect the loading effect of a high impedance probe with low capacitance.

The total capacitance of the jig is 1pF + 1pF + 5.6pF. If the resonator L is 320nH then the BPF will be centred somewhere around 102MHz.

If 1.55pF is added across the resonator (by the 2388 probe) then the capacitance is now just over 9pF. This means the BPF centre frequency will drop by about 8MHz or so.

[G0HZU_JMR]

I'm not sure why the jig reads a bit low for Rp but it might actually be correct. The effective Rp of these resistors may well fall up at VHF. I'm using cheapo thick film 0805 resistors for this test. it would be nice to know what the true Rp of these chip resistors is up at 100MHz.

I'm off to bed now but I tried a quick experiment where I measured ten 220kΩ resistors (SMD 0805) in parallel. I stacked them in four groups to keep the footprint size down.

I then compared this to a single 22kΩ 0805 chip resistor on the VNA up to 150MHz.

The result was that the Rp of the single 22kΩ resistor only fell to 20.6kΩ by 93MHz. At the same test frequency, the ten 220kΩ in parallel had fallen to just 15.5kΩ.

I think this shows that the high value resistors in my cheapo (Farnell) 0805 chip resistor kit don't show a constant Rp from LF through into VHF. I don't know how much spread there will be in this behaviour but the average across ten resistors was 220kΩ reducing to 155kΩ by 93MHz.

This experiment is a bit quick and basic, but it does help to explain why I measured less than 220kΩ when I tried to measure a 220kΩ resistor at about 93MHz with my resonator jig.

[G0HZU_JMR]

Thanks. I will have to order 0.4mm copper sheet. Can't i use double-sided copper clad PCB boards? I used up all my 0.4mm copper sheets for the ground solder points of my valve 6-gang FM tuner with its aluminum chassis. I have a 500C temp controlled solder station and some fat tips but I must say it is a little tricky to solder thick copper sheets even I have some special solder flux. I have a micro solder blow torch, but the flame is so hot that it just vaporizes everything!

Yes, you could try having a vertical PCB screen between the resonator and the probe and then fit this onto another bit of spare PCB. Just have some sort of mouse hole or peep hole at a suitable height in the vertical PCB wall. This way you don't need to use copper sheet and it will be much easier to solder it all.

It would look a lot neater as well.

My version was just a lashup and it took a few minutes to bend the metal and solder it all together.

I can put the 1pF and 5.6pF ATC/PPI caps in the post to you FOC if that helps? I have a few spare ones. These SMD caps are physically very rugged, and you don't have to worry about the end caps cracking off unless they are really stressed. They have ultra-low ESR up at VHF.

[regenfreak]

Thanks. I am at work now.

I chose the resonator components such that a small change in Cp across the resonator would show a large shift in frequency.

I see. You have picked a high-Q LC tank to exaggerate the probe-loading effect. For a typical FM tuner, say L = 70nH , C = 36pF, f = 100.258mHz. A poorman's input capacitance of deltaC = 0.7PF, the frequency becomes 99.297MHz. The frequency shift is about 0.961MHz. Just a correction of last night post, when I probed the first RF air variable capacitor of the homebrew 6-gang FM tuner, the frequency shift was just below 1MHz compared with the calibrated frequency counter. The tuner was properly aligned, including the tracking of the oscillator and all 6 gangs. The frequency display automatically subtracts 10.7mHz in the display. So it is consistent with the calculation.

I'm off to bed now but I tried a quick experiment where I measured ten 220kΩ resistors (SMD 0805) in parallel. I stacked them in four groups to keep the footprint size down.

I then compared this to a single 22kΩ 0805 chip resistor on the VNA up to 150MHz.

The result was that the Rp of the single 22kΩ resistor only fell to 20.6kΩ by 93MHz. At the same test frequency, the ten 220kΩ in parallel had fallen to just 15.5kΩ.

If you put 10 in parallel, the parasitic capacitance would be multiplied by 10. So it is not surprising that the Rp for the ten 22K resistors would be lower than a single 220K.

can put the 1pF and 5.6pF ATC/PPI caps in the post to you FOC if that helps? I have a few spare ones. These SMD caps are physically very rugged, and you don't have to worry about the end caps cracking off unless they are really stressed. They have ultra-low ESR up at VHF.

Thanks that's very kind of you. It is difficult to get hold of 1pF ATC100 caps on ebay. I can send you a PM later. In the meantime, I will try to use through-hole 1pF caps

[regenfreak]

I chose the resonator components such that a small change in Cp across the resonator would show a large shift in frequency. Also, I designed it for a high loaded Q and used an inductor with a Q of about 400. The effective Rp of the resonator (when fitted in the jig) is about 20kΩ at 100MHz so this makes it sensitive to the Rp loading of any probe that gets connected here. This is exactly what is required if the jig is going to be able to detect the loading effect of a high impedance probe with low capacitance.

The total capacitance of the jig is 1pF + 1pF + 5.6pF. If the resonator L is 320nH then the BPF will be centred somewhere around 102MHz.

If 1.55pF is added across the resonator (by the 2388 probe) then the capacitance is now just over 9pF. This means the BPF centre frequency will drop by about 8MHz or so.

I couldnt wait until this weekend....I have managed to make a test jig using double sided copper clad boards(see photo). It has a test hole to shield the hand effect. Without SMT 1pF caps, I just use thru-hole 1pF ceramic NPO caps and a 1-8pF ceramic piston trimmer cap.

In the first measurement without probe (2nd attachment), the insertion loss is -1.8db which is a bit less than your prediction of -2.5db. My resonance frequency is 99.89MHz, coil L = 300nH, C =8.6pF.

In the second measurement with the 0.7pF poorman's probe, the frequency has been shifted down to 95.49MHz. The insertion loss has gone up to -3.1db (versus -5db of your case). The poorman's probe is powered up but is not attached to anything. I hope that how you did it? I didnt bother to use two 3db attenuators at both ports this time. Looking at the Smith charts, the LC is not perfectly matched to 50 ohm at resonance frequency with 1pF coupling caps but its not bad.

My previous measurement of the Poorman's input capacitance = 0.723pF. Plug the numbers into a LC calculator, it is consistent with the above measurement; assume probe cap = 0.7pF. 8.6 + 0.7 = 9.3pF, f = 95.284MHz so fairly close to the above measured frequency.