IF alignment with damping of an adjacent coil

[Argus25]

Originally Posted by regenfreak

When the coupling k = 1, f2 goes to infinity. It means it has one resonance frequency. The inequality signs are typos?:

When k goes to 1 the frequency goes to 1/(root (1+1)) or about 70%.

So to give you an example, imagine two coils tightly coupled (say bifilar wound on the same core and k close to 1) each one tuned to the same frequency by the same value tuning capacitor. If you leave one capacitor disconnected, you will have some resonant frequency corresponding to the inductance of one winding and one tuning capacitor. When you connect the other capacitor across the other winding, the frequency will drop to 70% of what it was.

You get the same result of course if you simply connected the capacitor across the other tuning capacitor and just had the one winding. So its easy to see it is correct because the frequency is proportional to 1/(root C) so if you double the C the frequency change is again 1/(root 2).

So when two identical windings are very tightly coupled they behave from the alternating current perspective as one winding, even though they may be isolated from the DC perspective.This is one of the many reasons why bifilar wound IF transformers are a good choice for the video IF transformers for vintage TV sets and I used these in my version of the Argus TV set.

[regenfreak]

So to give you an example, imagine two coils tightly coupled (say bifilar wound on the same core and k close to 1) each one tuned to the same frequency by the same value tuning capacitor. If you leave one capacitor disconnected, you will have some resonant frequency corresponding to the inductance of one winding and one tuning capacitor. When you connect the other capacitor across the other winding, the frequency will drop to 70% of what it was.

You get the same result of course if you simply connected the capacitor across the other tuning capacitor and just had the one winding. So its easy to see it is correct because the frequency is proportional to 1/(root C) so if you double the C the frequency change is again 1/(root 2).

I pondered over this last night as well. My probelm is that allowing myself to remove one of the winding, or one of capacitor does not mean it is strictly equivalent when k =1. I would think of it as two electrically isolated biflar windings (magnetic field in phase) with non-negligible distributed capacitance between windings. The equivalent RC tanks are both indirect capacitive and inductive coupled.


I omitted the case (d), when both windings are identical with the same resonant frequency. This corresponds to the most typical IFs in tube radios.Below i attached the all three cases (note typo for f1, f should be 1). To get a better "feel" of the equations in case (d), I would input typical critical coupling coefficient k = 0.01, L = 1.2mH and then work out f1 and f2 assuming the typical If frequency fa=fb=455KHz. I am on holiday away from home and dont have my calculator with me.

PS: Can I increase the Q of a coil if I use biflar windings of lower strand count Litz wires to double the strand counts? I think the answer is not so obvious.

[regenfreak]

I have a suspician that there are no ' symbol for L1 in the three equations in case (d). These are probably typos.

Since M=k * sqrt(L1L2)=k*L1 (L1=L2)

Futher up in the text:
L1'=L1+M
L2'=L2+M

If I were to subsitute IF=455.1KHz, L=1.223mH, C= 100pF, critical coupling coefficient K =0.01 into the equations in case (d), M = 0.01223mH, the resulting values for the f1 and f2 do not make sense.

I would expect f1 and f2 are very close. If i were correct, there should be no dash' for the L in the equations. Using online LC resonance calculator:

f1=452.84kHz, f2=457.393KHz for IF = 455.1KHz

The probelm was that the book was typed by clerk using ball type writers back in the 1940s. I suppose, those typos were missed during proof reading.

[regenfreak]

case a should match case d,
using my crappy phone calculator, i get
approximately similar answers as above

[Radio Wrangler]

If you wind your two coils together as a bifilar pair, you get maximum coupling and you no longer get two poles, you just get one.

Litz wire can be helpful in raising inductor Q in two situations; firstly where skin effect reduces the conductivity of simple round wire, secondly where there is strong fringing flux (like near the gap of a pot core) and this flux induces transverse eddy currents in the cross section of your wire.

David

[Argus25]

Originally Posted by regenfreak

I pondered over this last night as well. My probelm is that allowing myself to remove one of the winding, or one of capacitor does not mean it is strictly equivalent when k =1.

Well, I did mention the important information was difficult to find.

Although you found reference to the basic equations, a vital piece of information was missing in your textbook reference, which I posted in mine:

https://www.vintage-radio.net/forum/...=161439&page=4

on post #76.

It explains why the higher frequency peak or resonance is present as the K value drops below 1.

After all ,you are stuck with some fixed capacitors ( in the example we have been talking about at least) so where does the higher frequency peak come from as K gets lower, say if you part the windings, and ultimately, as the K goes to 0 and the F2 frequency peak heads to infinity ?

You will notice I added the equations (1-k)Lp = Lip and (1-k)Ls = Lis, meaning say in the first case, that the leakage inductance of the primary Lip equals (1-k)Lp. (Lp is the primary inductance). Same applies to the secondary.

So it is easy to see that the leakage inductances go to zero when k=1, otherwise the leakage inductances are usually smaller values than the main inductance and this is where the high frequency resonances come from, in conjunction with the tuning capacitors.

These resonances of course are the nemesis for switch-mode PSU's and cause the ringing around switching edges when you switch voltages across a transformer. This is why often, switchmode psu's have a bifilar wound primary to help prevent this issue by having the k as high as possible on each half of the primary (and the leakage inductance as low as possible)

[regenfreak]

Argh thats interesting! That's leakage inductance equation is new to me...I have been trying to understand the leakage inducatance equation on the wiki page...more teaser for my newbie brain!

I know leakage inductance is good for limiting currents for neon sign transformers and welding transformers that are designed to be in short circuit operation.
I am at an aitport and gotta catch a flight flying back to the UK

[Argus25]

One interesting thing about magnetically coupled circuit is the way a primary neutralisizes inductance in a secondary , depending how heavily it is loaded (how low the source resistance is).

Again the equations not too easy to find but done very well by Terman. (see attached).

For example, if you short one winding out on a transformer, it neutralizes all the inductance of the other winding, at least that part it shares the main flux with, it cannot do anything to the leakage flux and this is why you are then just left with the leakage inductance which appears as an inductance in series with the other winding. That of course will resonate with the self or added capacitance.

The Terman equation can be used to work out the frequency response of an inter-stage audio coupling transformer and then it becomes clear that it is dependent on the plate resistance of the valve driving the transformer.

Of course if you switch DC across the primary, as you might in a switch-mode psu, apart from the usual rise in current that occurs and magnetic energy storage, you have what amounts to a shorted out primary (from the AC perspective) and a step voltage on the secondary, energizing the leakage inductance and causing high frequency oscillations.

I have attached a diagram as an example of two mosfets switching the primary of a transformer. Even with no obvious secondary winding, the other half of the primary acts as a secondary and you get the ringing because there is never perfect coupling between the the two halves of the primary and some leakage inductance remains. Better transformers for this application have bifilar wound primaries.

[Radio Wrangler]

Bifilar is a big help, but in many SPMS transformers you can't meet mains isolation safety regs with bifilar windings, so you're stuck with separate windings, probably with 'mummy wrap' insulation on one of them, and you wind up having to quench the leakage inductance ring with snubbers to keep your transistors in their happy zones.

Some SMPS like flyback converters, you can view the output rectifier/smoother as a clamp. The switch transistor sees the clamp voltage (=output voltage) times the transformer ratio (pri/sec) plus the unclamped leakage inductance ring. Replacing a rectifier with one that's too slow can be disastrous.

All sorts of creative winding schemes get used to try to reduce leakage inductance by improving coupling while still meeting insulation thickness and creepage distances.

In RF transformers that don't need safety isolation, twisted bifilar wire is great stuff. You can also match the wire gauges and the twists/inch to the operating impedance, so capacitive coupling takes over where the core dies off at higher frequencies. These are called 'transmission line transformers'. Ruthroff, "Some transmission line transformers" is the seminal reference. You aren't tied to BI -filar, I've done balanced transformers at RF with hexafilar winding.

These techniques give far, far more coupling than you want when you're trying to make a coupled resonator pair as a 2-pole bandpass filter where you want both poles acting to shape your passband for best selectivity. Tight coupling and forcing the second pole off towards infinity leaves you only one pole worth of selectivity, giving poorer shape factor.

David

[G0HZU_JMR]

Originally Posted by Radio Wrangler

Some of us have had to design high-order systems, and part of the job is designing an efficient and effective adjustment process. 7th Chebyshevs are mild compared to some. But with a methodical approach, they are do-able.

At work the biggest IF BPF I designed was a 13 resonator VHF jobbie. To make matters worse it was part of a multi channel DF system so there were 6 (supported maximum of 8) of these filters to align and they all had to be aligned for amplitude and phase so they tracked the same across a bandwidth of several MHz.

Every capacitor was measured and graded to a tight tolerance and put into component bags by hand to try and get the smoothest phase response across all of the IF sections. Luckily we only had to make one of these DF systems.

I'd recommend anyone to try out the Dishal method for aligning BPFs. Even if you only ever try it on a simulator it is good experience. If done correctly each resonator only has to be adjusted once. I think it's reasonable to suggest that it can't be beaten by a determined operator who tries to adjust the filter conventionally with a swept display (eg using a scalar or vector network analyser). Even if you gave them 2 hours to adjust a multi resonator filter they would never match the Dishal method. It only takes a couple of minutes at most with the Dishal method.

[Radio Wrangler]

Originally Posted by G0HZU_JMR

Even if you gave them 2 hours to adjust a multi resonator filter they would never match the Dishal method. It only takes a couple of minutes at most with the Dishal method.

In teaching someone to align filters, you LET them choose to try twiddling everything at once with a network analyser display. It's everyone's natural instinct. Let 'em flounder for an hour or two, and then show them Dishal's one stage at a time method and you'll have their full attention

With Chebyshev ripply responses, and even simple 2nd order coupled resonators with an overall M-shaped response, the temptation is to think that each bump in the response comes from a separate resonator. And it ain't like that!

David

[julie_m]

In my mind's eye. I'm seeing an arrangement with a pendulum of slightly different length, suspended from each end of a rigid rod which itself is suspended so it's free to move, and each pendulum can transfer some KE to the other with each swing. So instead of there being just two packets of energy to consider -- the KE in the speed and the PE in the height -- there is also some energy transferred through the suspension to the other pendulum.

[regenfreak]

You will notice I added the equations (1-k)Lp = Lip and (1-k)Ls = Lis, meaning say in the first case, that the leakage inductance of the primary Lip equals (1-k)Lp. (Lp is the primary inductance). Same applies to the secondary.

I can only understand complex problems if I break things down. Denote L1s as the leakage inductance,

L1s=L1-k*L1=L1-sq(M/L2)

k and L1s are only defined under the condition that the primary is unloaded, open-circuit, e.g. across an LRC meter during lab measurement.

Looking at the equation like staring at a Picasso abstract painting, I have tried to get a "feel" for the equation by doing measurement of a 240-250V HT valve transformer:

L1= 6.3H
L1s=206mH
k=0.98351
M=7.1546H

For an IF transformer, M and k are much smaller.

Again the equations not too easy to find but done very well by Terman. (see attached).

For example, if you short one winding out on a transformer, it neutralizes all the inductance of the other winding, at least that part it shares the main flux with, it cannot do anything to the leakage flux and this is why you are then just left with the leakage inductance which appears as an inductance in series with the other winding. That of course will resonate with the self or added capacitance.

If you short out the secondary, the voltage across secondary is zero and it reflects back to the primary, so the primary voltage becomes zero. The LRC meter reads the leakage inductance.

The equivalent negative inductance term by Terman can be further simplified by assuming the primary Dc resistance is small Rp=0, you get the same thing as above:

Leq=the equivalent negative inductance = -sq(k)*L1=-sq(M/L2)=-kL1

My newbie's limited understanding is that Tehman's reactance term for the negative inductance is basically the leakage reactance due to the back EMF (Lenz's law).

Tehman's model in the screen shot is not applicable to parallel inductive coupled LC tank resonance. In Henney's equivalent model the mutual inductance is added to L1:

L1'=L1+M

It looks like that the model always have to be idealised by finding the equivalent, direct inductive coupled (physically connected) LC resonance circuit.
Ok this is heavy stuff. I am no electrical engineer and I was the one as an undergraduate student skipped most of the 2nd year avionics lectures at university because the lecturer was as boring as hell

PS: k = sqrt [1-(Ls1/L1)] and M =sqrt[L2(L1-Ls1)]

[Argus25]

Originally Posted by julie_m

So instead of there being just two packets of energy to consider -- the KE in the speed and the PE in the height -- there is also some energy transferred through the suspension to the other pendulum.

Inductance of Primary, Secondary and Mutual inductance analogy perhaps.

It does seem as though resonant circuits, especially tuned to the same frequency are "desperate" to exchange energy with each other, even two tuned circuits spaced far apart on a desk.

In the early days of TRF radio, the big nemesis was what happened when two identical tuned circuits were placed in the grid & plate of the same triode. The Miller capacitance allowed energy exchange between the to resonant circuits and of course it oscillated. So Neutralization was born, largely to disappear later with the invention of the screen grid, or transistors with such low collector-base feedback capacitance that neutralization was no longer required for stability. Early transistors like OC45's have 10 times the feedback capacitance of a AF127, so you will always see neutralization in IF stages that use the OC45, but rarely for the AF127.

Of course if your coils are not properly shielded and there is mutual inductance the energy will exchange between the coils that way, then we call it an oscillator, at least if it was a deliberate attempt for the coils to have M.

[Sideband]

Just in passing, some years ago I restored a Philips 470A that used a low IF of 110khz. At the time I didn't have full alignment instructions (or maybe I did but didn't read them fully.....). Anyway I peaked up the IF transformers using a signal generator and the radio worked well......however there was a tendency towards IF oscillation and in fact from cold it would just go into oscillation at the high end of MW. I put the set aside to investigate later.

I then obtained the full Philips manual and that was when I discovered about damping the IF during alignment. I went through the alignment sequence again using the recommended damping network and at the end of the exercise (which took much longer than my original approach) resulted in a perfectly working radio with no tendency towards IF instability and much better quality sound (since the bandwidth was now wider). The radio was built in 1938 so the method of damping IF's has been around since the early days of superhets.

[Radio Wrangler]

I have an Eddystone EA12 receiver with a final IF of 100kHz.

It has variable IF bandwidth, achieved with a mechanical linkage from a front panel control, moving the coils closer together or further apart in the IFTs. This doesn't shift the IF centre, it just shifts the separation of each of the two poles of each IFT from the 100kHz centre.

Variable coupling, variable bandwidth.

For CW, a cam and a microswitch adds a crystal.

David

[regenfreak]

however there was a tendency towards IF oscillation and in fact from cold it would just go into oscillation at the high end of MW.

I read somewhere that a squeal may occur at one or two stations on the band when the harmonics of the oscillator beat with the other stations. IF of 100kHz has poorer image rejection.

It has variable IF bandwidth, achieved with a mechanical linkage from a front panel control, moving the coils closer together or further apart in the IFTs. This doesn't shift the IF centre, it just shifts the separation of each of the two poles of each IFT from the 100kHz centre.

Variable coupling, variable bandwidth.

Earlier this year I ordered some order-to-made IF cans from Reinhofer Electronics in Germany. They are probably the only few companies in the world still manufacture AM and composite AM/FM valve IF cans. The axes of the slugs are parallel but lateral mounted rather than longitudinal mounted along the axis of the shield. Their IF cans coupling can be varied easily by loosening the screw-mounted coils and changing their spacing without affecting the k and M.

In normal IF cans, there are two maximum peaks depending on the positions of ferrite core positions whether they are screwed towards outside or inside. This can mess up the coupling coefficient k and M, changing the shape of the band pass curve.

[Alistair D]

Is this the Reinhofer Electronics you were referring to?

https://www.reinhoefer.de/

Their website is a bit vague about the range of 'Spool' products that they manufacture.

Since IF cans can be easily damaged due to stuck cores I am sure members would be interested to know that replacement transformers could be obtained.

Al

[regenfreak]

yes, they are family run business but their core business is more industrial electronics.They cater for very narrow section of vintage radio and audio amp hobbyists. I will post the link when i get home. They are all hand wounded individually. They also make custom made valve transformers.Their IF cans page is hiddened somewhere. Since most european If cans dont suffer silver mica migration disease like the American, i think they get very few order so their turnover time will be a few weeks. Postal costs are high.

I have looked at an Ecko If transformer that i stripped, it was designed the same way as the new German If cans.

[regenfreak]

https://www.roehrentechnik.de/html/zf-bandfilter.html

You would need to order the adjusrment tool as the slug takes traingular tool. I have both chassis and PCB miniature versions.

I remember vaguely their minimum order is 40 euros