Relaxation oscillation in LC resonance/ Toko testers

[regenfreak]

The chokes are for B+ power line and filament to get rid of the RF. I need to go out now and be back later thanks

[G0HZU_JMR]

Thanks.
I had a go at doing a swept measurement of the filter with the nanovna. This is via the lossy 220k ohm and 68k ohm resistive matching so the response is fairly near the noise floor of the nanovna. However, it did a reasonable job of showing the filter passband despite the low signal level.

The second image shows what happens if I detune one of the resonators slightly. Note that this is a very sensitive adjustment and you won't see much unless you get both resonators tuned fairly close to each other.

I think this means you could actually do a swept plot of the filter with a nanovna and align it in a basic test fixture. I have the early cheapo nanovna £35 (delivered) so hopefully the later versions will have more dynamic range than mine.

The S11 test does not work, the Img S11 is a flat line; no sign change throughout.

You must be configuring something incorrectly. If configured correctly the nanovna should be able to make a good measurement here.

[regenfreak]

had a go at doing a swept measurement of the filter with the nanovna. This is via the lossy 220k ohm and 68k ohm resistive matching so the response is fairly near the noise floor of the nanovna. However, it did a reasonable job of showing the filter passband despite the low signal level.

Thanks. Can you repeat the test by removing the series 220k and 68k matching resistors and connect it to the 50 ohms ports directly? I want to see how much frequency shift when there is a large impedance mismatch.

In real world situations, the input and output impedance of a LC filter is often unknown, so the question remains the same: is there any value of sticking the IF transformers directly to the 50 ohms ports?

You must be configuring something incorrectly. If configured correctly the nanovna should be able to make a good measurement here.

There isn't really anything to configure. I just do a very careful calibration using the test board with an improved short.

Here is the larger scan of the Trio FM transformer user manual response graph.

[regenfreak]

I have found a 465kHz (yellow) AM transistor IF transformer. Since the commercial coiler tester (FET) does not suffer from the relaxation oscillation at this low frequency. Its reading is accurate. The VNA thru test gives the incorrect resonance frequency of 485KHz which is 20kHz above the actual frequency.

[G0HZU_JMR]

Thanks. Can you repeat the test by removing the series 220k and 68k matching resistors and connect it to the 50 ohms ports directly? I want to see how much frequency shift when there is a large impedance mismatch.

If I change the ports all the way down to 50 ohms then the filter response on s21 will be lost because the loaded Q will be very low as I will have severely damped each resonator with 50 ohms in parallel. S21 will just look like a flat line through 10.7MHz on a VNA because of this. However, the frequency where the angle of s11 passes through zero should stay more or less in the same place because the two transformers are coupled only very lightly. So I think the resonance frequency should stay about the same.

[G0HZU_JMR]

To try and show this I set up the filter on the VNA using the 220k and 68k series resistors and I added the padding caps to align it on 10.7MHz. I then removed the 220k and 68k resistors and measured the filter directly with the VNA with 50ohm ports and exported an s2p model.

I uploaded a youtube video showing the effect of changing the port impedances for both the simple model of the IF filter and also the VNA derived s2p data model. You can see they agree quite well even when the port 1 and 2 impedances are reduced from 220k to 50 ohms.


https://www.youtube.com/watch?v=wYk2UdGZZDw


Note that the 1 port angle measurement I use a the start of the video is ang[zin1]. This gives the nicest indication of resonance. In the last part of the video I change this to ang[s11] and you can see that the angle of s11 is very shallow when the filter is loaded at 50 ohms. Unless you scale for this you might think it is a flat line?

Hopefully this video is useful. It does show the value of a 1 port (and 2 port) measurement using a VNA.

[regenfreak]

If I change the ports all the way down to 50 ohms then the filter response on s21 will be lost because the loaded Q will be very low as I will have severely damped each resonator with 50 ohms in parallel. S21 will just look like a flat line through 10.7MHz on a VNA because of this.

Thanks. Very strange. I dont have this problem My S21 is well-defined and sharp when they are connected to 50 ohms ports directly, but the displayed resonance frequency is questionable. On the contrary my S11 and imag s11 are flat.

[Note that the 1 port angle measurement I use a the start of the video is ang[zin1]. This gives the nicest indication of resonance. In the last part of the video I change this to ang[s11] and you can see that the angle of s11 is very shallow when the filter is loaded at 50 ohms. Unless you scale for this you might think it is a flat line?

I will need to investigate this further. I will repeat, dump the data and sniff around the sign change of img R.

At the moment only my 4.3" NanoVNA F. can be connected to NanoVNA Saver. My NanoVNA V2 does not talk to the software like a couple who been arguing.

[G0HZU_JMR]

I think you are doing the s21 sweep the way I originally suggested in my first post on the thread. You are sweeping each resonator like a trap circuit. This gives a very well defined null and it is a good method to spot resonance although you would have to pad the resonator with 13pF or 9pF to get it resonant at 10.7MHz. Otherwise you will see the sharp S21 null up at about 14MHz.

My S21 measurement is through the whole filter with the connections shown in the video. If I do this with the VNA directly with 50 ohm ports the filter response is damped away to a flat line.

I've had a few issues with nanovna saver as well. It has probably been improved a lot since I last used it though.

[regenfreak]

Thanks I am so excited. I think I got it. The NanoVNA F can show Img S11 and S11 phase. See the zero crossing at 11MHz.This is the reflection measurement of Toko FM IF 10.7MHz. The phase angle is very flat and easy to mix. I have tried to dump the data file but the NanoVNA Saver has freezed up a few times.

[regenfreak]

I think you are doing the s21 sweep the way I originally suggested in my first post on the thread. You are sweeping each resonator like a trap circuit. This gives a very well defined null and it is a good method to spot resonance although you would have to pad the resonator with 13pF or 9pF to get it resonant at 10.7MHz.

It still does not explain my earlier experiment with a 465KHz IF transformer that the resonance tester shows frequency at 465kHz and the S12 shows 485KHz.

[regenfreak]

Referring the diagram in the link below, I have not wired the same way as the first parallel LC. Note that his port 1 ground is "floating" while my ground is connected across the LC tank in Toko FM IF measurement. I will try his approach to see any difference.

In my S21 measurement of toko FM transformer, I wire in the way as his second throu measurement:


https://www.nonstopsystems.com/radio...io_coax-sw.htm

[regenfreak]

Here I use the S11 measurement with a non-floating ground for port one and banana plug. The coil has super high Q =1400, Silver mica capacitor 100pF. I think the stray capacitance of the coil maybe roughly about 10pF. It read 1.28MHz at zero crossing of Img S11. So it is not bad!

The banana plug is not good for high frequency as you cannot really eliminate the BNC and plug out of the measurement plane in the calibration.

PS I have not tried floating ground like in the example above link. It looks almost as if he forgot to draw the ground wire!

[regenfreak]

I think any measurement cannot be validated without an alternative measurement using different instrument and technique. There is a little trick called two-frequency measurement method that allow us to find unknown L and C inside a IF transformer. I have derived the equations in the attached scan. First you measure the resonance frequency of the unknown IF transformer. Then you stick a known silver mica parallel to the transformer and measure its new resonance frequency.

I have used three methods to find the resonance frequency of Toko 468 yellow AM IF coil:

1. VNA reflection method, f = 485kHz

2. coiler tester, f = 465kHz

3. Q meter, f = 478kHz, Q = 84.

In Toko datasheet: yellow, AM 1st stage If transformer, f = 468kHz, Q =90, C = 180pF, L is not given.

Using two frequeny method with a known Cs=100pF smd capacitor and the coil tester: i got: f1 = 465kHz, f2=380kHz,
C =200pF L = 586microH for the Toko yellow AM if transformer

Now none of the methods can work out the stray capacitance of the inductor inside the Toko which explains why LC equation does not agree exactly of the expected resonance frequency with those L and C values. There is an old technique that stray capacitance of a coil can be measured using graphical method.

In conclusion, none of the three measurements agree exactly, great caution must be taken how we can obtain the resonance frequency considering the effects of impedance matching, calibration, the test leads and distributed capacitance of the coil and high frequency effect. It is a non-trial problem.

[G0HZU_JMR]

Yes, for some measurements it's worth cross checking against other methods. A decent VNA is usually king for stuff like this but it can struggle to measure things like component Q especially up at VHF and UHF. For important measurements of Q I cross check using two or three methods that are slower, but have much less uncertainty compared to a VNA measurement of (say) inductor Q using an s11 measurement.

However, even the very best VNA is only as good as the cal kit, the cal kit correction factors and the test fixture and the way the operator sets up the VNA. The vast majority of VNA users I see online manage to mess up every single one of these things. They then assume they have hit the limits of the VNA itself.

[G0HZU_JMR]

I can suggest another well controlled experiment to see how good your cal kit is across 5-50MHz. I did this a while back to compare the nanovna against the lab VNA. The nanovna did very well, not as well as the lab VNA but impressively close.

The idea was to put a precision 39pF ATC porcelain cap in series with a 1 ohm 1% chip resistor and see how well each VNA could measure the capacitance and ESR across the 5-50MHz range. Because the ESR of the ATC cap is probably less than 0.02 ohm (from the datasheet) the ESR can be assumed to be defined by the 1R resistor.

Below 5MHz this test becomes very difficult for the VNA because the reactance of the cap gets very high. However from 5-50MHz the Q of this network should gracefully drop from about 800 at 5MHz to about 80 at 50MHz. Across the reduced range of 15-25MHz the lab VNA and the nanovna should really be able to measure the 1R ESR and capacitance (and the Q ) quite well.

The chip resistor measured 1.01 ohm on a 4 wire test and this can be expected to creep up with increasing frequency to maybe 1.025R by 50MHz.

You can see the plot below for the lab VNA. It struggles to measure the ESR below about 10MHz as the Q starts to climb towards 800. However, you can see it manages to track the Q curve of the 39pF + 1.01R model very well if you visually average out the noise.

The other graph shows the ESR of the resistor measured on its own and also the capacitance and ESR when the cap and resistor are combined. Tests like this can help increase confidence that the VNA is calibrated quite well.

Below 5MHz the results from the VNA for ESR become very noisy and it is asking a lot of a VNA to measure a resistance of 1 ohm that is in series with a capacitive reactance of -800 ohms.

The 1% chip resistors cost a few pence each and the ATC 800B 39pF caps probably cost about £1 each. The ATC100B series are almost as good in terms of ESR and these can be found on ebay for about £1 each.

The second plot below shows the nanovna result. This was taken with very old firmware. It was the firmware it was delivered with. I think it will be possible to improve on this with the later versions of nanovna firmware as this allows a narrower resolution bandwidth and the performance is improved in other ways too.

[regenfreak]

The idea was to put a precision 39pF ATC porcelain cap in series with a 1 ohm 1% chip resistor and see how well each VNA could measure the capacitance and ESR across the 5-50MHz range. Because the ESR of the ATC cap is probably less than 0.02 ohm (from the datasheet) the ESR can be assumed to be defined by the 1R resistor.

Below 5MHz this test becomes very difficult for the VNA because the reactance of the cap gets very high. However from 5-50MHz the Q of this network should gracefully drop from about 800 at 5MHz to about 80 at 50MHz. Across the reduced range of 15-25MHz the lab VNA and the nanovna should really be able to measure the 1R ESR and capacitance (and the Q ) quite well.

The chip resistor measured 1.01 ohm on a 4 wire test and this can be expected to creep up with increasing frequency to maybe 1.025R by 50MHz.

You can see the plot below for the lab VNA. It struggles to measure the ESR below about 10MHz as the Q starts to climb towards 800. However, you can see it manages to track the Q curve of the 39pF + 1.01R model very well if you visually average out the noise.

The other graph shows the ESR of the resistor measured on its own and also the capacitance and ESR when the cap and resistor are combined. Tests like this can help increase confidence that the VNA is calibrated quite well.

Below 5MHz the results from the VNA for ESR become very noisy and it is asking a lot of a VNA to measure a resistance of 1 ohm that is in series with a capacitive reactance of -800 ohms.

The 1% chip resistors cost a few pence each and the ATC 800B 39pF caps probably cost about £1 each. The ATC100B series are almost as good in terms of ESR and these can be found on ebay for about £1 each.

This is interesting. I don't have the right tools to build surface mounted RF test boards yet. I am looking to buy a SMD microscope before I venture into experimentation with low capacitive and inductive stray surface-mounted components. My eyesight is bad.

I never have doubt about the calibration kits that came with my V2 and 4.3" F. Really I dont have any other alternative. Most of the time the main application of of my is VNA to check SWR for a mobile telescopic 5ft 6" HF whip antenna and experimentation of ground radials.

I can suggest another well controlled experiment to see how good your cal kit is across 5-50MHz.

Here is a 29-40MHz test of my NanoVNA v2 tested against a £1000 Rigol DSA815-TG Spectrum analyzer for a 8-pole Butterworth filter from a IF transformer stripped from a TV transmitter. It has both 50 ohms input and output impdence using tapped transformers inside. The Rigol spectrum plot was given by the ebay seller. I paid £5 for this thing.

The sweep from the NanoVNA matched very closely with the Rigol spectrum analyzer:

In the response graph of Rigol the markers are: 1=29.4MHz, 2=31.4MHz, 3=39.4MHz, 4=32.6MHz
Centre frequency 35.4MHz, passband loss 1dB.

So the nanoVNA is very impressive and I am confident with the calibration kit came with it for frequency below 40MHz

The last photo is the sweep with a cheap white noise generator and the TinySA spectrum analyzer. Just for fun

PS I have a rather expensive £35 SMA torque wrench. I watched some videos claiming that it is essential. But now I think it is overkill and a bit waste of money.

[G0HZU_JMR]

Thanks, that's an interesting looking filter! I'm tempted to buy one of the TinySA spectrum analysers but I think I'm going to wait to see if a newer model gets released. I'd mainly use it for noise/interference hunting around the house because it is so portable.

I have a nice SMA torque wrench here but I only use it for critical measurements up in the GHz region where the torque of the connector can (very slightly) affect the phase of the measurement. Otherwise I rely on a simple SMA spanner and my own judgement for general purpose stuff.

A VNA can also be used for oscillator (and amplifier instability) analysis and hopefully the V2 nanovna would work well here up into the GHz region. I'm not sure how low the source power can be adjusted in the V2 but my old nanovna is fairly limited here. I think it only goes down to about -13dBm. This is often too high and can cause incorrect measurements of active circuits. I have used it a few times to look at active circuits but the -13dBm source power is often too high.

[regenfreak]

Thanks, that's an interesting looking filter! I'm tempted to buy one of the TinySA spectrum analysers but I think I'm going to wait to see if a newer model gets released. I'd mainly use it for noise/interference hunting around the house because it is so portable.

I have a nice SMA torque wrench here but I only use it for critical measurements up in the GHz region where the torque of the connector can (very slightly) affect the phase of the measurement. Otherwise I rely on a simple SMA spanner and my own judgement for general purpose stuff.

A VNA can also be used for oscillator (and amplifier instability) analysis and hopefully the V2 nanovna would work well here up into the GHz region. I'm not sure how low the source power can be adjusted in the V2 but my old nanovna is fairly limited here. I think it only goes down to about -13dBm. This is often too high and can cause incorrect measurements of active circuits. I have used it a few times to look at active circuits but the -13dBm source power is often too high.

The TinySA is great for aligning the oscillators of FM radios and VCO/BFO of homebrew QRP receivers. Its little antenna has very little loading effect on a tuned oscillator. It can be used as a signal generator with or without AM/ wide and narrow FM modulation.

You can buy a 30db, 20db and 10 db 4GMz SMA attenuator for a fiver.

[Radio1950]

Relaxation oscillation in LC resonance/ Toko testers.

With respect to the first 15 or so posts of this thread.

Thanks to “regenfreak” for his info and reference material.

In 2010, I built up a GDO and RF Source for MF and HF use.
It was for general HF test and to drive my old DIY RF Impedance Bridge, but not to replace a sig gen.

It had coverage of 375 KHz to 35 MHz by my design, and with a coax RF output of about +10 dBm50.

The original design was based on a GDO article in QST May 2003.
It was similar to the general design of the Toko Coil Tester.

A roughie prototype of the QST variant proved the concept and a device was built and calibrated.
In later use however, every now and then, at some frequencies, it had quirks that were not obvious with the prototype.

This is the reason for this post.

Just before going on a two month holiday in July this year, I was testing some small inductors with my GDO, and sometimes the frequency would jump, or the “change in frequency” would vary across the dial.

Then I read “regenfreak’s” posts and articles, and resolved to fix my oscillator, which I have now done.

The following may assist others who might build up similar devices.

Enticingly Simple, maybe, but ...

The original QST design, like most of the others, offers the attraction of grounding of one side of coil and variable capacitor, and just a few components.

I used “rat’s nest” assembly, and this didn’t help in the end.

I could not get rid of all the quirks in the original QST design, and tried something similar to that in the ZL2PD article, but with similar results. These included low and varying output, freq jumps and funnies, poor wave shape, and variable freq counter triggering.

I then tried the circuit in the 20MHz example of the Koster-Waldow-Wolff-J-FET-VCO article.
This had low output but was stable, had reasonably constant level across the bands, with no observable spurious signals.

In the end I used a combination of the ZL2 and KWW circuits, but with emitter resistor of 47R and a Q2 drain resistor of 4K7 (with no parallel cap).
Feedback capacitor was 6p8.
All this gave better results.

The waveform across the emitter resistor, for normal circuit values, is not sine, so I buffered the signal across the tuned circuit using very light capacitive coupling (4p7) and an emitter follower, and used a SPF5189 RF Amp to obtain a nice sine output at +15 dBm in 50 ohms.

I did use the emitter signal for GDO meter dip level, and buffered it for freq. counter drive.

I now at last have a functional device. It isn’t perfect but very functional.

Hints For Others

Have another think about whether this type of oscillator really suits your specific application.

Build a prototype for proof of operation in your specific application.

This oscillator may suit operation at single frequencies, but if I were to build another GDO and Variable Freq Source, I probably would choose a Hartley, even with the “tap”.

Don’t use a large variable capacitor and few coil turns as I did for my highest range of 11-40 MHz, as the LC ratio may not suit and oscillation may stop.

I wanted a range just below 455 KHz, but it backfired on me on the higher frequencies. I suggest no more than 150 pFd max.

A GDO for operation below 1MHz is better served by a separate device.

Stray capacities are the killer for this circuit as the feedback capacitance is very low in value.
It may be better to “constrain” the strays by building the circuit on double sided PCB. Circuit layout is critical, even for higher HF.

Resist using RF chokes for the Q2 drain if you want spurious free operation with a variable oscillator; they may work well for single freq applications. It does maximise the signal output though.

Be prepared to experiment with the resistors and feedback capacitance. Use low capacitance FETS and transistors.

Read all the articles referred to in the first “part” of the thread; very interesting, but be prepared to modify.

Attached is a block diagram of my device. FETS are MPF102, buffer transistors are 2N2369.

Thanks to “regenfreak” for his info and for spurring me to understand and fix my device.

[regenfreak]

Thanks to “regenfreak” for his info and reference material.

I am glad my info was useful. I thought nobody was interested in this.

I am moving into a slightly different direction; I am looking to build the high impedance Q probe designed by a Russian Ham US5MSQ using J310 and 2N3906. I have been following closely US5MSQ's works with great interest lately. I use Google translation lurking around Russian Ham forums a lot. For poor amateur like me who cannot afford a £1000 spectrum analyzer, his works are fascinating. I cannot share his web page link because for unknown reason my home PC Norton anti-virus software blocks his russian blog.

Don’t use a large variable capacitor and few coil turns as I did for my highest range of 11-40 MHz, as the LC ratio may not suit and oscillation may stop.

There is a trick that you can change the tuning range Cmin, Cmax and Cmax/Cmin of a regular MW band air variable capacitor to suit the optimal L to C ratio for SW/HF bands. There are two ways:

EITHER:
by adding a padder trimmer capacitor in series with the air variable capacitor plus a fixed capacitor in parallel with LC tank

OR:

by adding a fixed padder capacitor in series with the air variable capacitor plus a trimmer capacitor in parallel with the LC tank.

Then you can write a little Excel calculator to solve the unkown values of the padder capacitor and the parallel capacitor by iterations..I have tried it.... It works like charm.