6-gang FM stereo tuner heads

[G0HZU_JMR]

I put an aerial on the input to one of my spectrum analysers this evening and set it to capture about 20 seconds of the signal from a couple of FM broadcast stations. Reception isn't that great here as I don't have a proper aerial up for FM.

See below for the spectrum on a 200kHz span and the demodulated stereo signal from 0 to 100kHz. This shows the 19kHz pilot tone and the demodulated L+R up to 15kHz and also the L-R modulated on a suppressed 38kHz carrier. I think the strange modulation up at 57kHz is RDS.

The first image captures a quiet moment and this clearly shows the 19kHz pilot tone, the suppressed 38kHz carrier and the RDS waveform up at 57kHz. The second image shows a classic instrument and the third shows some pop music. The analyser works in real time so the screen display is very fluid in terms of update rate so it can be quite interesting to watch it whilst listening to the station at the same time.

It looks like the music has about 15kHz bandwidth although some modulation effects can be seen around the 19kHz carrier. I was expecting the demodulated signal to be a bit cleaner than this in this region.

[Radio Wrangler]

Q as we know it today, used to be called the Circuit Magnification Factor. Marconi Instruments' Q meter was called their Circuit Magnification Meter, making it sound rather grand.

What this means is that the circulating power in a tank circuit of reasonable Q is significantly larger than the input and output power.

In FM tuners, it's usual to use dual-gate MOSFETs and to match these circuit impedances are run high. This means that the signal voltage in the tank is much larger than the signal voltage coming from the antenna.

Varactors are very much voltage controlled devices. Voltage sets their capacitance, so the increased voltage from unwanted large signals is made even worse, modulating the tank's tuned frequency by a greater amount, dramatically worsening the strengths of intermod products.

To avoid overload, most people would think to narrow the input tank to try to filter them out, but this means higher Q and greater voltage magnification of unwanted signals close to the wanted channel.

Ever get the feeling you just can't win?

The root cause is that the varactor diode needs to operate n a low-signal-voltage environment, and you want to use diodes needing plenty of tuning volts to scale the problem down. So the trend to low voltage varactors so tuning voltages can come from 5v supplies is very undesirable.

Varactors can be embedded in low impedance tanks and can produce reasonable Q and not suffer magnified signal voltages.... they get magnified signal currents instead, but that's not so devastating.

An R&D engineer at HP in California, Bart McJunkin, invented what he called a 'cartwheel oscillator'. It used a printed inductor looking like a very short very fat piece of coax. say 50mm in diameter, 1.6mm long! So it was a 2 inch ring of track on each side of the PCB with hundreds of stitches connecting them. This formed the outer of the coax, and the inner was a lily-pad of copper in the middle of each side, also connected together with loads of stitches. His oscillator circuit amplifier lived on the lily pad.

So this needs a tuning capacitor... The cartwheel has spokes, each spoke being a back-to-back pair of varactors. So there is a dozen varactor pairs, all working in parallel. Our tank has only tiny inductance and very large capacitance. It is very low-Z. Signal voltages are very low, yet the varactors run with tens of volts of DC bias to tune them.

Coupling into and out of this monster tank was by coupling loops also printed in the PCB. A thin ring at half diameter distributed tuning voltage to the common cathode points of the varactor pairs.

It was patented as an oscillator, where it gave an advantage in phase noise, but I spotted that it was a route to low intermod varactor tuned agile filters. It would be superb for an FM tuner and would haul back some of the advantage that mechanical variable capacitors still have. It could be taken to extremes and beat the mechanical capacitors (unless you use butterfly variable capacitors, you wind up with a sliding contact in your tank - everyone forgets this!)

So, yes, there is still fun to be had in FM tuner design, and designers still have unused tricks up their sleeves. But FM radio seems to be in decline. The BBC doesn't broadcast my sort of music, and I don't see any commercial return from designing something a bit different. I could do it for the hell of it, but I might as well pick an area with some profit in it to spend the effort on.

Anyway, so now you know why mechanical variable capacitors currently have advantages in FM tuners.

David

[regenfreak]

Originally Posted by Radio Wrangler

Q as we know it today, used to be called the Circuit Magnification Factor. Marconi Instruments' Q meter was called their Circuit Magnification Meter, making it sound rather grand.

What this means is that the circulating power in a tank circuit of reasonable Q is significantly larger than the input and output power.

In FM tuners, it's usual to use dual-gate MOSFETs and to match these circuit impedances are run high. This means that the signal voltage in the tank is much larger than the signal voltage coming from the antenna.

Varactors are very much voltage controlled devices. Voltage sets their capacitance, so the increased voltage from unwanted large signals is made even worse, modulating the tank's tuned frequency by a greater amount, dramatically worsening the strengths of intermod products.

To avoid overload, most people would think to narrow the input tank to try to filter them out, but this means higher Q and greater voltage magnification of unwanted signals close to the wanted channel.



So this needs a tuning capacitor... The cartwheel has spokes, each spoke being a back-to-back pair of varactors. So there is a dozen varactor pairs, all working in parallel. Our tank has only tiny inductance and very large capacitance. It is very low-Z. Signal voltages are very low, yet the varactors run with tens of volts of DC bias to tune them.

Coupling into and out of this monster tank was by coupling loops also printed in the PCB. A thin ring at half diameter distributed tuning voltage to the common cathode points of the varactor pairs.

It was patented as an oscillator, where it gave an advantage in phase noise, but I spotted that it was a route to low intermod varactor tuned agile filters. It would be superb for an FM tuner and would haul back some of the advantage that mechanical variable capacitors still have. It could be taken to extremes and beat the mechanical capacitors (unless you use butterfly variable capacitors, you wind up with a sliding contact in your tank - everyone forgets this!)

So, yes, there is still fun to be had in FM tuner design, and designers still have unused tricks up their sleeves. But FM radio seems to be in decline. The BBC doesn't broadcast my sort of music, and I don't see any commercial return from designing something a bit different. I could do it for the hell of it, but I might as well pick an area with some profit in it to spend the effort on.

Anyway, so now you know why mechanical variable capacitors currently have advantages in FM tuners.

David

Interesting! The cartwheel oscillator sounds like a Frankenstein kind of looking device! I can't find the patent for this.

In term of the equivalent circuit model, the high Q varactor diode can be idealised as the a capacitor (junction capacitance of the depleted region) in series with a resistor (low resistance undepleted regions). The Q is sensitive to the reverse biased voltage. As the bias voltage increases, the depleted layer thickens, reducing the junction capacitance and the undepleted layer resistance. Hence the Q will go up.

One of the key advantages of varactor tuning is the possibility of miniaturisation of receiver circuits. However, the Q of inductors suffer as their physical dimensions, getting below a certain threshold. I attached the photo and schematic of my Mitsumi FE-352 micro tuner. It is super cute but I have not got time to build circuit to make it working.

There is a kind differential tuned varactor L-VCO that has very low phase noise used in microwave applications. I dont know much about:

https://ieeexplore.ieee.org/document/1324731

The demise of AM/FM analogue broadcast will be inevitable, like digital photography replacing film photography. The only difference is that the DAB is an epic technological failure that it is not fit for purpose,..a joke. I have a very diverse music taste; classical, pop, trance, hip hop, dance, heavy metal, blue, jazz...I listen to BBC 1, BBC 4 and London LBC most of the time.

Originally Posted by G0HZU_JMR

I put an aerial on the input to one of my spectrum analysers this evening and set it to capture about 20 seconds of the signal from a couple of FM broadcast stations. Reception isn't that great here as I don't have a proper aerial up for FM.

See below for the spectrum on a 200kHz span and the demodulated stereo signal from 0 to 100kHz. This shows the 19kHz pilot tone and the demodulated L+R up to 15kHz and also the L-R modulated on a suppressed 38kHz carrier. I think the strange modulation up at 57kHz is RDS.

The first image captures a quiet moment and this clearly shows the 19kHz pilot tone, the suppressed 38kHz carrier and the RDS waveform up at 57kHz. The second image shows a classic instrument and the third shows some pop music. The analyser works in real time so the screen display is very fluid in terms of update rate so it can be quite interesting to watch it whilst listening to the station at the same time.

It looks like the music has about 15kHz bandwidth although some modulation effects can be seen around the 19kHz carrier. I was expecting the demodulated signal to be a bit cleaner than this in this region.

Having no external antenna, I do get multi-path interference in my flat a lot.
Two years ago, I read the book "Stereo FM Radio Handbook by Harvey and Boltman. For me, it is a classic text that introduces me all the basic concepts of multiplexing. At that time, I knew nothing about FM.

[regenfreak]

I have tested the Nelson Jones's IF and detector circuit. It works OK but it does not sound that great. My homebrew valve IF strip is staggered tuned and it sounds great at the expense of poorer selectivity.

I use 10.7MHz Toko IF transformer from ebay as the quadrature detector coil for the TAA661B chip. It gets lots of distortion if the coil is not fined tuned at its sweet spot. Maybe the audio quality gets better if I change its 4.7k resistor across the tank circuit...anyway I am going to try LA1222 IF chip based on the IF stages of the Kenwood KR-9600 (1600Watts per channel, I mean who can play that level of output power before the neighbors complain and police turn up at your door step)

[regenfreak]

I have replaced the lossy cores by T50-6 for the 5-pole 10.7MHz Chebyshev with a bandwidth of 300kHz and 330ohms I/O impedance. The insertion loss is about -6bd including some mismatch losses with the L-pads. Tuning is very sharp and it is very difficult to make it perfectly flat and symmetrical in practice. So this is the best I can get so far. The S21, S11 and group delay are attached.

I will compare the overall responses between the following IF Amp and bandpass configurations:

1) 280kHz ceramic filter ---->CA3053 cascode IF amp--->280kHz ceramic filter

2) 280kHz ceramic filter ----> CA3053 cascode IF amp------> 5-pole 300kHz Cherbyshev

3) 280kHz ceramic filter --->LA1222 IF amp--->280kHz ceramic filter---->LA1222 IF AMP--->two 280kHz ceramic filter in cascade (total 4 ceramic filters like the Kenwood KR-9600)

4) 280kHz ceramic filter --->LA1222 IF amp--->280kHz ceramic filter---->LA1222 IF AMP--->5-pole 300kHz Cherbyshev

LA1222 is two IF amps combined in one package. I am building the TA7303P IF amp and discriminator IC using homemade discriminator coil with a blank Toko style transformer. I can add a signal strength meter for the TA7303P later. The MPX decode is TA7343AP.

Further experiment with the TAA661B chip has been futile. I swapped different 10.7MHz tank coils for the quadrature detector with different impedance and Qs. I can only get mono with some audio distortion, potentially due to the mismatch in impedance and Q. I do not bother to mess about with the value of the damping resistor which affects the loaded Q of the LC quadrature detector coil. There is plenty of gain and selectivity though. I strongly believe the TA7303P IF amp and discriminator IC will be much better.

[regenfreak]

As described in the previous post, here are the results from the replica of the IF amp circuit for the mighty Kenwood KR-9600 with LA1222:

Attachment 1 and 2: S21 and group delay for the four 280kHz ceramic filters in cascade,-3db bandwidth is 210KHz, shape factor = 1.96 at 6db/60db;

Attachment 3: two 280KHz ceramic filter+ 5 pole Chebyshev LC filter, -3d bandwidth = 220KHz, shape factor = 2.15 at 6db/60db

Both configurations have fairly flat phase shift linearity and they work well with the 6-gang stereo tuner, not too wide or too narrow.

In my next stage of nerdy experiment, I will build a 7-pole Butterworth 10.7MHz IF of bandwidth of 280kHz so that I will have total of 12 poles in the IF amp combined with the 5-pole Chebyshev.

The Butterworth has better phase shift linearity than Chebyshev but it also has rounder shoulders. The 7-pole Butterworth should make the roll-off as steep as the 5-pole Chebyshev filter.

Attachment 4: LA1222 IF amp, TA7303 IF amp and demodulator, and TA7343 MPX decoder. I will need to experiment with different IF amp as the TA7303 tends to get overloaded easily with MPX audio distortion. There is plenty of gain and selectivity with the LA1222 IF amp.

Attachment 5: is the IF response curves of the RIMO filter designed by Richard Moddafferi for the highly regarded Mcintosh MR-78. Note its shape has Bessel filter like rounded-shoulders. The -3db bandwidth is very narrow in both normal, narrow and super narrow configurations. According to the MR-78 service manual, the -6db bandwidth are 150kHz for normal and 130Khz for narrow which would chop off some of the MPX audio. It would be good for DX and overcrowded FM band with adjacent strong and weak stations. For stereo MPX fidelity, I think they are too narrow.

Here is his dissertation:
https://digitalcommons.njit.edu/thes...=PDFCoverPages

It seems that he moved the poles in S-plane to achieve the best flatness of the group delay response using a Fortran programme. The Rimo filter was introduced at the time that the ceramic filters started to become the norm. Since no other tuners used Rimo filter, its complexity and costs probably outweight its benefits.




Attachment,

[regenfreak]

Here are the results of two super-wide 330KHz ceramic filters in cascade with 12 pole-LC filters; a 5-pole 300kHz Cherbyshev and a 280kHz 7-pole Butterworth:

Attachment 1: set-up with two LA1222 IF amps
Attachment 2: S21 and S11 of the cascade
Attachment 3: comparison of the group delay for the 7-pole Butterworth (280kHz) and 5-pole (300kHz) Cherbyshev.

I have had lots of fun tuning 12-pole LC filters. The flatness Butterworth of the 7-pole Butterworth is not really significantly better than the 5-pole Cherbyshev.

At the moment, I am building the group delay equalizer (attachment 4) taken from the Sansui TU517. It is based on the idea that the typical group delay of ceramic filters have double humps like the back of a camel and a one-pole LC resonator has group delay like a bell shape. By adjusting the value of the shunt damping resistor for the LC resonator, it will smooth out the gap between the two humps of the group delay curve. The closest analogy I can think of is the Area Rule in the design of transonic and supesonic fighter jets.

[regenfreak]

Here is a little fun circuit for the group delay equalizer. It works but produces very subtle improvement in the flatness of the group delay graph when it has been tested with 4 ceramic filter cascade with an IF amp. The loaded Q of the LC tank can be variable between Q=35 to 45 by adjusting the variable damping potentiometer. I have replaced the fixed 12K series resistor by lower or higher values, it makes little difference to the maximum Q. I dont know if I can improve it further by replacing the LC tank with a higher Q. Attached is the S21 and group delay in blue line. Note that although the group delay looks very flat in the graph, the vertical axis can be stretched or rescaled to exaggerate the distortion.

The Sansui patent for their group delay equalizer consisted of very complex LC network and it is not simple like this.

[regenfreak]

I have not updated lately because I have been rattled by tumultuous time in the UK. I dug deep and purchased the Siglent SSA3021x vs plus a few weeks ago before all these madness started to unravel.

I have experimented with LC IF filters using traditional Toko type of transformers and ratio detector. The circuit is nothing special, very commonly found in probable FM transistor radios. The attachments show the accumulative response curves of stage 1, stage 2 and stage 3 (different coloured curves) probed at the collectors of the BJT If amp using the Ukranian FET probe from Ebay. The overall gain is about 44db.

With synchronous tuning of the LC filters, I got the overall bandwidth of 180kHz which is too narrow for my liking. With a bit of staggered tuning, i can get a flatter top response with a bandwidth of about 230KHz.

Attachment 4 shows the classic S-curve of the ratio detector using my Rigol DZ1022Z as a sweep generator.

Attachment 5 is the LA1235 IC quadrature detector with double tuned coils. Combining with 4 wide ceramic filters, the FM stereo sound is truly amazing. This is the best performing detector that I have tried. The combo is simple and effective. The double tuned quadrature coils require very critical tuning for the best audio quality and lowest noise. I replaced the normal 300khz and 280khz ceramic filters by Murata G.D.T filters with high degree of group delay flatness but higher insertion loss. I didn't notice in the difference in perceived audio quality.

The Ebay FET probe has input capacitance of at least 1.5pF and not 0.5pF as claimed even with very short ground lead. This is good enough for 10.7Mhz but no good for VHF and can load down tuned LC circuits. The brass needle is very thin, and it broke today. This is the second version i purchased from Ukraine. I am experimenting and building my own FET probes which will have true 0.5PF input capacitance.

[nemo_07]

Originally Posted by regenfreak

The Ebay FET probe has input capacitance of at least 1.5pF and not 0.5pF as claimed even with very short ground lead.

The original article claims 0.75pF, not 0.5pF.
https://www.vintage-radio.net/forum/...7&postcount=43
The remaining 0.75pF or so, are due mostly to the two red crocodile clips in the photo
https://www.vintage-radio.net/forum/...0&d=1664724744
while the broken needle had a contribution of the order of 0.1pF.
Bulky conductors not allowed here, must use "slim" conductors, as thin as practicable.

In the range 10-100 MHz the lenght of the ground lead is not ctitical here when making relative measurements (anything usually expressed in dB or the like "ratios"); this statement is the more valid the lesser the input capacitance, Cin is made.
For a probe with Cin ~ 0.1pF you don't need a ground lead as long the probe stays fixed in place (no movement) in a well defined setup when probing with hands off.
When making absolute measurements without ground lead, its attenuation will vary introducing a negative error up to few dB when moving the probe around without touching it with the hand; these will drop significantly when your hand has a contact with the probe's body (GND), and then the systematic error will not exceed -0.5dB as compared to perfectly grounded probe.
As long the tip ("needle") is thin, its length is not critical at all, be it 1cm or 5cm.

Originally Posted by regenfreak

This is good enough for 10.7Mhz but no good for VHF and can load down tuned LC circuits. ... I am experimenting and building my own FET probes which will have true 0.5PF input capacitance.

This is not good enough in either case, and primarily not because of loading ("de-Queing"), but because of significant detuning when touching a LC tank, be it IF or front end.
From the well known equation for a LC tank we can derive the value of relative detuning delta_F/F resulting from a small detuning capacitance Cin, being
(1) delta_F/F ~= -0.5*Cin/C, for Cin << C.

For example, with C = 100pF, Cin = 0.5pF we get -0.25% detuning, corresponding to -26.75 kHz at 10.7MHz IF. Too much for fine adjustments on IF strips. Also, your J310 candidate or any other silicon JFET will never beat the BF998 in such design.

As mentioned earlier, for a work in the 10-110 MHz range this Elector design
https://www.vintage-radio.net/forum/...8&d=1653256711
would be a quite usable tool, after some modifications.

The first step would be to replace the the R4 value of 47 Ohm with 150 Ohm.
This will increase the impedance at the input (G1) of the MOSFET, and the gain of stage. The total capacitance at the gate G1, Cgtot will be ca. 2.5 pF, the resistance > 60 kOhm at 110MHz, and > 3 MOhm at 10.7MHz.

For values of C1 = Cgtot/10 or below, the loading of LC tanks (IF or front end) will be negligibly low, not worth to mention for any practical Q value in this frequency range.
What remains is to find a fine balance between detuning and attenuation, the last being
(2) Att[dB] ~= 20 log[Cgtot/C1] + 6.

The simple way of reducing C1 in this design would be cutting off (and later removing) some area of the upper PCB pad (component side). The resulting Cin can be estimated by probing the LC tank of an oscillator (assuming the C value is known) and measuring the detuning, see equation (1).
You could also insert a MMIC gain chip (50 Ohm purely resistive load required) before entering the coax line to compensate for attenuation.

[regenfreak]

Originally Posted by nemo_07

The original article claims 0.75pF, not 0.5pF.
https://www.vintage-radio.net/forum/...7&postcount=43
The remaining 0.75pF or so, are due mostly to the two red crocodile clips in the photo
https://www.vintage-radio.net/forum/...0&d=1664724744
while the broken needle had a contribution of the order of 0.1pF.
Bulky conductors not allowed here, must use "slim" conductors, as thin as practicable.

In the range 10-100 MHz the length of the ground lead is not ctitical here when making relative measurements (anything usually expressed in dB or the like "ratios"); this statement is the more valid the lesser the input capacitance, Cin is made.
For a probe with Cin ~ 0.1pF you don't need a ground lead as long the probe stays fixed in place (no movement) in a well defined setup when probing with hands off.
When making absolute measurements without ground lead, its attenuation will vary introducing a negative error up to few dB when moving the probe around without touching it with the hand; these will drop significantly when your hand has a contact with the probe's body (GND), and then the systematic error will not exceed -0.5dB as compared to perfectly grounded probe.
As long the tip ("needle") is thin, its length is not critical at all, be it 1cm or 5cm.

This is not good enough in either case, and primarily not because of loading ("de-Queing"), but because of significant detuning when touching a LC tank, be it IF or front end.
From the well known equation for a LC tank we can derive the value of relative detuning delta_F/F resulting from a small detuning capacitance Cin, being
(1) delta_F/F ~= -0.5*Cin/C, for Cin << C.

For example, with C = 100pF, Cin = 0.5pF we get -0.25% detuning, corresponding to -26.75 kHz at 10.7MHz IF. Too much for fine adjustments on IF strips. Also, your J310 candidate or any other silicon JFET will never beat the BF998 in such design.

As mentioned earlier, for a work in the 10-110 MHz range this Elector design
https://www.vintage-radio.net/forum/...8&d=1653256711
would be a quite usable tool, after some modifications.

The first step would be to replace the the R4 value of 47 Ohm with 150 Ohm.
This will increase the impedance at the input (G1) of the MOSFET, and the gain of stage. The total capacitance at the gate G1, Cgtot will be ca. 2.5 pF, the resistance > 60 kOhm at 110MHz, and > 3 MOhm at 10.7MHz.

For values of C1 = Cgtot/10 or below, the loading of LC tanks (IF or front end) will be negligibly low, not worth to mention for any practical Q value in this frequency range.
What remains is to find a fine balance between detuning and attenuation, the last being
(2) Att[dB] ~= 20 log[Cgtot/C1] + 6.

The simple way of reducing C1 in this design would be cutting off (and later removing) some area of the upper PCB pad (component side). The resulting Cin can be estimated by probing the LC tank of an oscillator (assuming the C value is known) and measuring the detuning, see equation (1).
You could also insert a MMIC gain chip (50 Ohm purely resistive load required) before entering the coax line to compensate for attenuation.

The measurement of the input capacitance of the Ukrainian probe was done without crocodile clips like in the photos(it uses gimmick capacitor). Instead, I used a very short 50 ohm transmission line strip with a VNA 50 ohm load termination. With the NanoVNA v2 plus4 the input capacitance is measured to be 1.24PF for the new version of the Ukrainian probe with 2.5 inch ground lead(including crocodile clip's length). Using a calibrated DER DE-6000 RLC meter, the DC input capacitance is about 1.7pF. I have also another version of the Ukrainian with op amp, it has also input capacitance of 3pF (two 6pF capacitors are soldered in series).

I made the Poorman's probe a while ago. Here I attached the S21, S11 and Smith chart of the Poorman's probe up to 1GHz. The gain is -30db instead of -20db that I was expecting. I used rigid gold-plated probe tips which are not ideal for contact reliability. I am waiting for the shipment of pogo test tips from China. I will make an improved version of this with built-in Gali 51+ amp and 1000V smd DC blocking capacitor at the output. Like you said, I use gimmick capacitor which are just shaved copper bits from the PCB. By aggressively grinding off the copper surfaces, I can attain input capacitance as low as 0.1pF for gimmick capacitor but the insertion is too large for my liking. Now I use a 0603 0.1PF capacitor in series with the tip. The surrounding stray capacitance of tracks result in the overall input capacitance of about 0.7PF. I should not get hung up with trying to get input capacitance as low as possible. The HP and Tektronic active probes have input capacitance of 0.7pF to 2pF.

Yes I use Gali 51+ as amplifer for the ukrainian and Poorman's probes from time to time.

To ground or not to ground is an interesting question. With the floating ground, it picks up lots of noises and also give unreliable readings or absolute readings.

The ground lead not only contributes to the stray inductance but also stray capacitance. The test pin + ground lead + input capacitance form a series LC network.

The inductance of the ground lead is about 25nH per inch. With two inches of ground lead and a 3pF probe input capacitance, the series resonance frequency is 410MHz. If an input pulse with the leading-edge rise time faster than trise = 0.35/410 = 0.85nS, ringing occurs. Similarly, if the input capacitance is reduced from 3pF to 0.7PF, the resonance frequency goes up to 850MHz and rise time is 0.41nS.

Yes BF998 is much better than J310. I made Q-measurement probe pen with J310 for 10.7MHz circuits with success but FM is not good.

he first step would be to replace the the R4 value of 47 Ohm with 150 Ohm.
This will increase the impedance at the input (G1) of the MOSFET, and the gain of stage. The total capacitance at the gate G1, Cgtot will be ca. 2.5 pF, the resistance > 60 kOhm at 110MHz, and > 3 MOhm at 10.7MHz.

For values of C1 = Cgtot/10 or below, the loading of LC tanks (IF or front end) will be negligibly low, not worth to mention for any practical Q value in this frequency range.
What remains is to find a fine balance between detuning and attenuation, the last being
(2) Att[dB] ~= 20 log[Cgtot/C1] + 6.

Interesting, how did you come up with those figures? Using Spice?

[nemo_07]

Originally Posted by regenfreak

... I should not get hung up with trying to get input capacitance as low as possible.

There is nothing about "as low as possible".
From the equation (1) we get the maximal allowed Cin for any tolerated max. detuning of your choice, and the example given with a typical value C of LC tank shows the range of required Cin values.
An excessive detuning when probing the shape a single LC tank is not a big problem, but trying to align coupled tanks could be challenging, and especially multi-stage filters seem to fall into category "mission impossible".
I could be wrong now ...

Originally Posted by regenfreak

The HP and Tektronix active probes have input capacitance of 0.7pF to 2pF.

Such probes are designed for looking at logic or other circuits with moderate or low impedances, not for probing Hi-Q LC tanks. Beside detuning, some of them will introduce measurable damping at 100MHz.

Originally Posted by regenfreak

Interesting, how did you come up with those figures? Using Spice?

No need for Spice. The figures come from the BF998 data, reading the plots and calculating simplified circuit models.

[G0HZU_JMR]

One thing to watch out for with any FET based follower circuit is (unwanted) negative resistance at the gate up at VHF. The BF998 has a high transconductance so will be very prone to producing negative resistance at the input if care is not taken with the choice of components at the source and gate pins.

The original poor man's probe circuit has a 47Ω bias resistor at the source pin and having a low resistance here will help minimise (or even prevent) negative resistance at the gate. I think the probe design also relies on there always being a 50Ω load resistance at the source with minimal shunt capacitance. Even when this is present, I think the circuit could still be prone to producing some negative resistance at the input.

I think that changing the bias resistor from 47Ω to 150Ω is very risky in terms of circuit stability because I expect that it will introduce a fair bit of negative resistance at the gate 1 pin of the BF998. This negative resistance will be generated across a huge bandwidth. Probably spanning VHF and also through a fair bit of the UHF band.

[regenfreak]

Originally Posted by nemo_07[This is not good enough in either case, and primarily not because of loading ("de-Queing"), but because of significant detuning when touching a LC tank, be it IF or front end.
From the well known equation for a LC tank we can derive the value of relative detuning delta_F/F resulting from a small detuning capacitance Cin, being
(1) delta_F/F ~= -0.5*Cin/C, for Cin << C.

For example, with C = 100pF, Cin = 0.5pF we get -0.25% detuning, corresponding to -26.75 kHz at 10.7MHz IF.

We are not aligning a narrow band FM IF stages. 0.25% detuning is nothing for wideband FM alignment. I wonder if you have ever tried aligned wither mono stereo IF strips for broadcast receivers before? I can easily do alignment for a FM mono radio "blinded" by the using output voltage method in synchronous tuning. However visual alignment is the king for both synchronous and staggered tuning.

I am not sure how you define C1...if C1 is the input capacitance of the probe including stray of the ground lead and clip. C1 acts like a voltage divider. The input capacitance for BF888 is 2.1pF. If C1 = 0.7, the effective input capacitance is 0.525pF.

The signal integrity on the scope and rise time is not my concern as I am probing sine waves at broadcast FM frequencies well below the resonance frequency of the LC network for the probe tip. Beside the resonance ringing can be suppressed by adding a simple Butterworth low pass filter at the output.

If you measure the S21 and S11 of the probe without the ground, the results are a total mess.

Originally Posted by nemo_07Such probes are designed for looking at logic or other circuits with moderate or low impedances, not for probing Hi-Q LC tanks. Beside detuning, some of them will introduce measurable damping at 100MHz.

How you do know? Have you tried? This guy probed the IF output for the residual output of FM front end tuners with multi-stage tuned bandpass filters. He used a Tektronix 500MHz P6202A with input capacitance of 2pF:

http://ham-radio.com/k6sti/swept.htm

I am so intrigued by his results that I bought a Tektronix P6202A on ebay USA last week. I am waiting for it to arrive. I have a Farnell triple output TOPS 3D power supply that will enable me to use it without the Tektronix power supply. I have attached the Rp and Xp charts for P6202A and P6201.

Correct me if I am wrong, Xp is the parallel equivalent reactance of the capacitive parts of the probe tip and FET input. Looking at the trend, it looks like the reactance of the total input capacitance dominating and the stray inductance is negligible. Rp is equivalent parallel resistance of the probe..I am not sure how it is calculated. For the P6201, the reactance for the stray inductance only starts to kick in above 800MHz.

Originally Posted by G0HZU_JMR One thing to watch out for with any FET based follower circuit is (unwanted) negative resistance at the gate up at VHF. The BF998 has a high transconductance so will be very prone to producing negative resistance at the input if care is not taken with the choice of components at the source and gate pins.

Funny you wrote that. Originally i used the same component values as in that Poorman's article and it didn't work. I changed the bias voltage with different R2 and R3 voltage divider , then it works all fine and dandy.

In the third attachment, I attached Z11 measured by Hirshbuscher's degree thesis using a 4-port VNA. I have no idea how to extract Z11 from S11 and phase angle. It is beyond my technical understanding. Basically, we attach a high impedance DUT to a 50 ohm transmission line, there is a massive mismatch in VSWRs between these two devices and hence high uncertainties in the Z11 measurements.

[G0HZU_JMR]

The bias network for that poor man's probe does look to be quite aggressive as it places a positive voltage of about 2V at g1 and the source bias resistor is only 47Ω. I'd therefore expect the BF998 to operate beyond Idss so it could draw well over 20mA drain current. This is off the datasheet graphs and heading towards the 30mA max limit for the device. The magazine article does mention it can draw up to 30mA.

The 47Ω bias resistor will be in parallel with (1/gm) so it will spoil the source match into the 50Ω cable.

However, if it is increased to 150Ω then something else will also have to be done to try and offset the negative resistance this will generate at the g1 pin. A snubber network could be fitted at the g1 pin for example. This will have to be designed carefully because it could introduce excessive damping at the input if overdone.

[regenfreak]

Originally Posted by G0HZU_JMR

The bias network for that poor man's probe does look to be quite aggressive as it places a positive voltage of about 2V at g1 and the source bias resistor is only 47Ω. I'd therefore expect the BF998 to operate beyond Idss so it could draw well over 20mA drain current. This is off the datasheet graphs and heading towards the 30mA max limit for the device. The magazine article does mention it can draw up to 30mA.

The 47Ω bias resistor will be in parallel with (1/gm) so it will spoil the source match into the 50Ω cable.

However, if it is increased to 150Ω then something else will also have to be done to try and offset the negative resistance this will generate at the g1 pin. A snubber network could be fitted at the g1 pin for example. This will have to be designed carefully because it could introduce excessive damping at the input if overdone.

Sounds complicated. I currently use 1.5V bias.

I have repeated the S21 and S11 measurement using a short transmission line with open. I got slightly different input capacitance. Now the S21 is closer to -20db. However, it is a bad idea to use rigid and pointed test pins for this measurement as the contact resistance is high and pressure sensitive. It is difficult to get repeatable readings. I got the Rp is between 45k to 130k. It is better to use blunt pogo probe pins or solder the test pins to the transmission line. So the project wont be continued until I receive the pogo probe tips.


So far I have tried C1 =0.3pF, 0.56pF and 1pF. Attached photo is the measurement of the gimmick low frequency capacitance using the DER EE De-5000. It uses Kevin 4-terminal measurements and has the resolution of 0.01pF using parallel admittance Y measurement for small capacitance:

Y = G +jB = 1/Rp + 1/jXp


I have never had any success sniffing the residual RF from the output of IF stage. I know the IF can be a buffer to reduce probe loading. If i probe the double-tuned RF amp stages, there is severe probe loading...

[G0HZU_JMR]

nemo_07: Such probes are designed for looking at logic or other circuits with moderate or low impedances, not for probing Hi-Q LC tanks. Beside detuning, some of them will introduce measurable damping at 100MHz.

regenfreak: How you do know? Have you tried? This guy probed the IF output for the residual output of FM front end tuners with multi-stage tuned bandpass filters. He used a Tektronix 500MHz P6202A with input capacitance of 2pF:

I had a play on a simulator and the BF998 MOSFET probe has the potential to have an effective input Rp that is a couple or orders of magnitude higher than the P6201 at 100MHz. This is partly due to the capacitive divider at the input. However, this can only happen if the MOSFET probe has a snubber at g1 to offset the negative resistance the BF998 will generate here across VHF and well into UHF.

Achieving low input Cp for a useable probe is always going to be a challenge and a lot depends on how small and short (and fragile?) you can make the tip section. This will minimise any capacitance to free space.

[regenfreak]

Originally Posted by G0HZU_JMR

I had a play on a simulator and the BF998 MOSFET probe has the potential to have an effective input Rp that is a couple or orders of magnitude higher than the P6201 at 100MHz. This is partly due to the capacitive divider at the input. However, this can only happen if the MOSFET probe has a snubber at g1 to offset the negative resistance the BF998 will generate here across VHF and well into UHF.

Achieving low input Cp for a useable probe is always going to be a challenge and a lot depends on how small and short (and fragile?) you can make the tip section. This will minimise any capacitance to free space.

It has been 18 years since the Poorman's probe article published. I hope someone can design a snubber circuit to fix the negative resistance issue.

I have attached the schematic for P6202A. The trimmer C12, R13 and C13 form a low impedance path to the ground at 100MHz. I think it is optimized for signal waveform integrity really. I didnt pay much for my P6202A, but the shipping and import together cost as much as the probe itself, just like everything else sold in ebay from the USA.

Most expensive active probes have replaceable tips. For me, I dont mind if the poorman's probe tips are fragile. I just replace them at a little cost.

[G0HZU_JMR]

Thanks. That P6202 probe circuit has a classic shunt RC snubber circuit in it. This is R13 (570Ω) and C13 (1.8pF). There is also a series 50Ω resistor R10 that helps to offset any remaining negative resistance.

In the case of the poor man's probe, the value of the shunt RC snubber will depend on the components fitted at the source. If the 47Ω resistor was changed to a higher value, then the snubber might be something like 39Ω in series with 1.3pF. This assumes there is a terminated 50Ω cable connected at the output of the probe.

If the BF998 produces a broadband -39Ω in series with 1.3pF at the g1 pin, then the shunt RC snubber would need to be 39Ω in series with 1.3pF to cancel out the negative resistance. In reality, there will need to be a compromise here if the aim is to get rid of all negative resistance at the input with some margin. The snubber will need to have some additional damping in order to give adequate margin at all frequencies.

The penalty of the RC snubber will be increased input capacitance, but this can be offset by having a tiny series capacitance in the input divider.

[regenfreak]

Originally Posted by G0HZU_JMR

Thanks. That P6202 probe circuit has a classic shunt RC snubber circuit in it. This is R13 (570Ω) and C13 (1.8pF). There is also a series 50Ω resistor R10 that helps to offset any remaining negative resistance.

In the case of the poor man's probe, the value of the shunt RC snubber will depend on the components fitted at the source. If the 47Ω resistor was changed to a higher value, then the snubber might be something like 39Ω in series with 1.3pF. This assumes there is a terminated 50Ω cable connected at the output of the probe.

If the BF998 produces a broadband -39Ω in series with 1.3pF at the g1 pin, then the shunt RC snubber would need to be 39Ω in series with 1.3pF to cancel out the negative resistance. In reality, there will need to be a compromise here if the aim is to get rid of all negative resistance at the input with some margin. The snubber will need to have some additional damping in order to give adequate margin at all frequencies.

The penalty of the RC snubber will be increased input capacitance, but this can be offset by having a tiny series capacitance in the input divider.

oh thanks, interesting!! Shall i add this RC snubber to the Poorman's probe as you have suggested. How can i tell if the circuit is unstable with negative resistance?
I could not understand why there are the 50ohm R10, R12 and C13 there. Originally, I was thinking of building the circuit for P6202A but the service manual did not include this crucial information. I can add a 200 ohm smt trim pot in series with the 47 ohm resistor. Checking Hirchbuechler's thesis, he added an additional 33 ohm resistor to the 50 ohm output.

I have also attached the 900Mhz Tek P6201 1M FET probe. It uses a MMBF4416 FET. Note the presence of R100 and R120 =27 ohms at the probe tip input. I think R100 = 50 ohms. It is not clear in the service manual. C120 = 4.7PF , it couples the gate to the source. It is possible for me to build this circuit except that i may have to replace Q120 and Q130 by more available equivalent transistors.