Hello,

开云体育

Great work, Jacques, thanks.

John? KK6IL

Hi Jacques,
? I've been beating this dead horse for a while, but your figure on page 23 of the Rs curve over frequency of the drive transformer, makes my point.
The drive level will change with frequency and it will also change with load.
So, why not monitor the 30mV, feed it back to the oscillator to adjust the level to keep drive on the DUT at 30mv to overcome drive voltage changes that occur with frequency and load. Then Rs doesn't matter.
You will need a ganged switch at the osc output and to set the Q meter FS to set different Q levels.
?The design is over my abilities, but if we get a design, I can build it.

???????????????????????????????????? Thanks for all your input, Mikek

Hi Mikek,

I just did some simple simulations and yes it would certainly help to stabilize the 30 mV drive voltage.
Then the Rs value doesn't matter, as you wrote.
In the case of the HP4342A, at worst case, the required? voltage drive will have to be increased? by a factor of 12 at 50 MHz.
This is because the source resistance increases at 130 milliohm at 50 MHz as well as the source reactance which increases to approx. 1 ohm.
If the source resistance can be kept low, like 20 milliohm and the source inductance at 0.5 nH, then the required? voltage drive
will have to be increased? by a factor of only 2 at 50 MHz, to keep the effective drive voltage constant.
In my document I have shown how to build such a resistor.
You may verify that the source voltage dips down at resonance by monitoring the drive voltage at higher frequencies and high tuning cap values.
The attached document shows the two cases, at 50 MHz.
In all cases a feedback circuit could be designed and would add complexity.

The second option requires monitoring the source drive voltage with an RF voltmeter (Boonton or Fluke 8920A - max 20 MHz), and adjusting
the drive to keep a constant voltage after peaking at resonance. The Boonton 4200 has a DC out which could be used for feedback to
a voltage controlled attenuator.

The third option requires calculating the error generated by the source Rs and Ls.? My Excel document only compensates for Rs but it
should be possible to add the effect of the source Ls.

The fourth option would be to measure both voltages and compute the ratio, so that you don't need to stabilize the drive voltage.
I have an RF voltmeter/dB meter in development that does ratio, but it is very slow in this mode. Too slow for finding a maximum.

Food for thinking !

Jacques,

On Sat, Jan 21, 2023 at 11:45 AM, Jacques Audet wrote:

I just did some simple simulations and yes it would certainly help to stabilize the 30 mV drive voltage.
Then the Rs value doesn't matter, as you wrote.

Even a blind squirrel can find a nut once in a while. :-)

>In all cases a feedback circuit could be designed and would add complexity.

The complexity is over my ability, I sometimes see the big picture and even a block diagram, but can't design.

Yes, but once designed, it's done, I don't know if the complexity is all that much, but again, I have no design ability.

Re: the negative effect of reactance of the drive transformer. I haven't understood, how the reactance affects the measurement
as long as the Drive voltage remains at 30mV. Maybe the problem is in the nature of the feedback required to keep it at 30mV,
as Frequency changes, when reactance is involved.

I'm sure you're aware, but a change in Rs also will cause the drive voltage to sag with High Q tests and rise with low Q tests because of the Rsource resistance. This is where I started to see the problem I thought could be corrected with feedback. I wanted to use the transformer output voltage to correct the non-linearity issues of the Transformer.
? I'm still not seeing how the Reactance prevents this, but that's my problem, just a lack of knowledge. (if I was 18 again, I would do education differently)
??????????????????????????????????????????????????? Keep up the work, I enjoy the time trying to understand, Mikek

Hi Jacques and Mikek, As you can see, I have joined the group.

Jacques, I enjoyed reading your report on this stuff. You really dug deep and figured out a lot. Nice work. One thing I wonder is that if you already have a 4342A, are you investigating it to improve it, to build your own version of a Q-meter, or just for fun?

I have some suggestions on improving the transformer performance, involving the primary only at this point. The secondary gets to be tricky due to the large impedance ratio, I'll talk about that another time. The first thing is to set up the primary winding structure to be one clean, uniform layer that completely fills (no gaps between the passes through the center) the toroid surface. The beginning and end of the winding should be touching. This can be done in a number of ways, like with different core sizes, wire gauge, and number of turns. For instance, with either experimental transformer, it looks like there's maybe 10-15 percent unfilled. Keeping the same wire gauge, you could say just put on more turns until it's filled. This would change the ratio, of course, but as I understand, the 50 turns is a numerical convenience, so it should be possible to use any number, within reason, and adjust the scaling to accommodate it. You could also go with bigger wire instead, chosen to land near the desired number of turns, and fully filled. Bigger wire would reduce losses a bit, while achieving full-fill helps to minimize the leakage inductance.

Now that you have both ends of the winding at the same spot, the next thing is to twist them tightly together over their whole length, and let them exit the transformer as a transmission line. Some sleeving will help keep them together and protected. The trip to the driver amplifier should be as short as possible, where the ends are allowed to separate for making connections. This form of wiring further reduces the leakage inductance by keeping everything close-in to the core, and minimizing the loop area of the interconnect lines.

Ed

开云体育

Hi Ed,

Thanks for your good comments.? I wanted to try making my own transformer but I realized that it was not so easy !
I should try your suggestions on my transformers.
Yes, I have a 4342A? Q Meter, bought from my employer about 25 years ago.
Back in the early seventies I did some research to develop a custom Q meter for my employer.
It had to cover from 1 KHz to 5 MHz. And yes, the transformer injection method was used as in the HP4342A.
I used a very high mu tape core which worked well down to 1 KHz.
Getting the low impedance drive (less than 5 milliohms) was difficult at the higher frequencies.
I found that I had to use a large copper strip at the secondary. I believe that the tape core
essentially "disappeared" at the higher frequencies, leaving only a low ESR strip.

I also tried to use the LC circuit under test as an oscillator to generate the test signal.
A broadband amplifier with an AGC was used.? It monitored the drive voltage and kept it constant.
This was done with a 400EL voltmeter, using the DC output to provide feedback for the AGC.
The Q is then read on the RF voltmeter across the LC circuit.
The oscillator needs to have 360 degrees phase shift to work properly.
The first 180 degrees is easy.? The LC circuit will provide 90 degrees phase shift at resonance.
This means that another 90 degrees of phase shift needs to be provided externally.
Ideally it should be a broadband phase shifter and possibly require a fine tuning at high Q and higher RF frequencies.
I did a prototype of this scheme that worked in the X10 KHz range back then, for my employer.
It was fun !? But the project didn't go any further.

Jacques

Hi Mikek,

You wrote:
< the negative effect of reactance of the drive transformer. I haven't understood, how the reactance affects the measurement >

When you resonate the Q meter the coil inductance L, and the source Ls add up so that:
- The resonant frequency will decrease. (compared to not having Ls)
- This increases the total reactance XL, since we have two inductors in series.
- The? Q = XL/Rs? so the Q will increase slightly by a small amount.? More increase above 10 MHz and with high resonant C.
The effect of Rs is generally larger and overall the measured Q factor decreases .
My Excel sheet has been revised to add the corrections for both Rs and Ls, the source resistance and reactance,
for the HP4342A and the Boonton 260? Q Meters.? (For the Boonton I assumed Rs=0.02 ohms and Ls = 5 uH, but it can be changed)



Hope this helps,
Jacques

开云体育

Ls=5uH ????? 5uH is? the inductance of a? thin wire in a circle of 1m diameter, 5nH might be more appropriate.

73.

Alan G8LCO

On Tue, Jan 24, 2023 at 05:19 AM, Alan wrote:

Ls=5uH ????? 5uH is? the inductance of a? thin wire in a circle of 1m diameter, 5nH might be more appropriate.

I'm pretty sure that's a typo, in his previous post with the PowerPoint attachment, he has .05nH and 3nH examples.
Good catch though!
???????????????????????????? Thanks, Mikek
???????????????????????????????????????????????

开云体育

?Since we are fighting inductance, Does a lower impedance transformer help reduce output inductance?
If not, my point is not helpful.
Alan G8LCO, suggested a 10Ω to 1 turn transformer with a 10Ω drive impedance, in the original "making a Q meter" thread
? Alan? said the following,
"Don't think much of the 50:1 transformer at all, HP used it as a cheap way of producing a low source impedance up to mid HF. Using a 2-mix core is not great either, I went for a 3C90 core for an adapter that I mentioned on another group, 10 turn primary and 1 turn secondary transformed a 4:1 pad of 39 and 10 ohms down to 0.08 ohms source
that was flat from 3KHto >>150MHz. Inserting a second cascaded similar transformer would produce a 100:1 stepdown in voltage but a 1/10,000 reduction in impedance without great transformer issues or drive amp complexity.

Generally it is fair to say that non-transmission line transformers become "difficult" when the ratio goes beyond 10 or that many more than 20 turns are involved, stray reactance then tends to become problematic? at very wide bandwidths. However if you stick to modest ratio's and few turns on the smallest core available ( have a look at minicircuit's wideband transformers) then you can have an easier time than others. This leads to using ferrites rather than dust iron cores ( not 2-mix for example)."

This it what I think he did, then he suggest another 10 to 1 transformer to go even lower output impedance.

?????????????????????????????????????? Mikek

?

?

?

Hello Jacques, MikeK,

I have been following this thread with great interest and I have been thinking about constructing my own Q-meter.
I am also interested in lower frequencies and therefore choose a tape wound core to obtain wide bandwidth.
The first test was with 97 turns 0.50mm on VAC Vitroperm 500 T60006-L2025-W523, AL 28.4uH, 25 x 20 x 10mm.
This resulted in a low frequency limit of 84Hz in a 50 Ohm system.

To determine the achievable leakage inductance I wrapped the primary winding with a secondary wound with thin copper foil. This secondary was shorted and the secondary leakage inductance and resistance of a one turn secondary (as close to the primary as the foil) was determined by measuring the primary inductance and resistance and dividing by the primary turn squared: 0.3nH and 1.15 mOhm at 10MHz. At the HF end the first resonance is at 40MHz. This is probably due to the high number of turns. Apart from resonances of parasitic inductances and capacitances I don't think we will get beyond the frequency for which a quarter wavelength equals the length of the primary winding: one of the few things I remember from transmission line theory is the a quarter wave length of transmission line inverts the impedance: a short on one end is measured as an open on the other end. So reducing the transformer size and cascading transformers should help to increase the usable HF bandwidth

I also wanted to construct an oscillator from the resonant circuit under test by using the circuit current for feedback with the proper phase. The circuit current could be measured the a transformer similar to the drive transformer. The obvious disadvantage is a doubling of the series resistance and leakage inductance. But this can be circumvented if the level feedback is tapped after both transformers.

I should do some simulations!

Best regards,

Gerhard

Way to go, Gerhard. I'm looking at a foil secondary too, and have been working on various schemes to build an appropriate mechanical structure for it. I don't know if you did the foil wrap only for test purposes, but I recommend using it for the actual secondary too. The trick is in working out a good structure that maintains low Z to the DUT connection.

Regardless of the core size, the center post should be as fat as possible, with a very close fit to the finished ID of the winding assembly. Then, a sheet of Cu foil can be wrapped tightly around the outside. One end of the wrapping becomes the approximate midpoint of the winding, bonded electrically to the center post. The other end of the wrapping can be cut open to clear the post, and sculpted into shape to form the ground connection (which is essentially a ring with the center post passing through it). And of course, you have to sculpt in a passageway for the primary leads. During all the forming and mounting steps, the foil can be finger-pressed to conform as tightly as possible to the toroid assembly. Also, depending on the methods and parts, some insulating materials like thin Kevlar or PTFE sheet may be needed in certain spots, especially soldering zones.

The version I'm presently working on is using an unknown junk box core similar to T130-1 or -2, as far as I can tell. My estimate of A sub L is about 14 nH, which is between the -1 and -2. The main reason for this core is that I happen to have some nice silver plated brass resonator posts out of a big old junked cavity filter. They're about 11/16" OD, which looks just about right to fit the core after it's loaded with the 0.030" D (whatever AWG that is) wire I have on hand. I just have to pick the right length to cut from the base. The resonators are tapped in the center for 8-32, but I'll redo it for 10-32, the shank size of the binding posts I'm using. So, the test voltage post will thread into the transformer's center, forming the low-Z source output, and the bottom end of the secondary foil will solder to a ground plane section. The back side midpoint foil of course has to be bonded well to the post by soldering or compression, and a big washer and 10-32 screw or stud, and include features to mount on an insulator, and a sound mechanical structure. The entire outer foil is "hot" with the test voltage, with about half on the back end where it mounts, so it has to be insulated to prevent undesired ground currents. It can be further contained by an added outer shield structure if needed. Since the voltage is small, it shouldn't be too much trouble, but you never know.

I've been picturing a small version too, with say a 1/4" D post made from a threaded brass spacer or shaft stock. The structure would be the same, except that the ground end of the foil would be soldered to a brass 1/4" ID x 3/8" threaded shaft bushing during construction. The bushing would be drilled out a little over-sized to provide clearance for a Kevlar insulation wrap, and the ground plane mount would be the bushing, so it attaches like a pot to a panel.

Ed

I would like to see work towards a transformer or even a very low impedance drive resistance that is compatible with Banana connectors mounted 1" apart, hoping to keep it compatible.
In the end a PCB supply house could take a design and send boards instead of everyone having a different design.
?However that does take some agreement on what the actual project is, because I doubt that 1" spacing is going to work for small value inductors at high frequency. (100MHz)
? Jacques has a simple approach in just knowing the errors and correcting for them.
?I would just like a more accurate instrument even if it is limited to 100kHz to 30MHz. The HP4342A was designed over 50 years ago, electronic components have changed size and many high impedance, high accuracy ICs have been developed, even, processors are cheap. I think with 50 years of progress, the HP 4342A can be improved on, with most of the components mounted on one PCB.
? Here is part of a board for using banana connectors and a low resistance driver that I started to draw up when someone mentioned using 7 SMD resistors to make a low inductance load. As they pointed out 7 is not magic. The info is in the old thread "Making a Q Meter" I don't remember the resistor values, but they were tiny to get us under 0.02Ω.
Same applies with mounting a transformer, center leg is the banana Connector.
Just my 2 cents :-)
??????????????????????????????????????? Mikek

Mikek,

I don't know if you're aware of a radial form of this kind of thing that's been around for a long time, used especially for coaxial terminations. It's just two concentric circles of Cu, with a bunch of paralleled SMD (or leaded) resistors arranged radially from the inner ring to the outer. It's assembled with conventional methods, with everything on-plane. The interconnects, feature sizes, ground planes etc depend on the purpose and environment.

Ed

On Wed, Jan 25, 2023 at 06:01 PM, Ed Breya wrote:

I don't know if you're aware of a radial form of this kind of thing that's been around for a long time, used especially for coaxial terminations.

No, I am not aware, do you have an example.
????????????????????????????????????????? Mikek

Mikek said "do you have an example?"

It's not something you would find and buy - rather a component layout arrangement. If you look at microwave/RF coaxial terminations and attenuating devices, you can see that they often use discus type resistors or even cylindrical complete sub-assemblies, where there are uniform sheets of resistive material, with current flowing from a center conductor to an outer one.in radial fashion. You can approximate this structure with discrete components. A bunch of parts in a radial arrangement is an easy way to get pretty close to the ideal continuous structure. A large or infinite number of them would do nicely, except that they would have to be shaped like slices of pie, while your common SMD parts are rectangles. So, you put a bunch of them in a radial pattern, like cylinders on a radial internal combustion aircraft engine. You can only fit so many in a given layout structure.

Ed

Hi,

I did further simulations and calculations on the idea of measuring the voltage ratio of the
Q meter voltmeter voltage TO the source voltage, to eliminate the source voltage variations.
I found that this will work ONLY if the Q voltmeter shunt resistance is very high, that is
more than 100 times the reactance of the resonating capacitor.
I found that it is possible to calculate the correct Q even when the Q voltmeter shunt resistance is low.
This can easily occur with the Boonton 260? Q meter.
I have revised my Excel spreadsheet to take into account all residual parameters:
The source resistance, inductance, capacitor residual inductance and the load resistance across the
tuning capacitor.

See:

The details of the calculations and simulations:


Jacques
VE2AZX