Inductor Spacing/Orientation in Filter Layouts

[Steve Haynal]

Hi Group,

I am working on the filter layout for Jim's OPA2677 amp. To minimize coupling, I see the inductors are not packed as tightly as they could be and are oriented at right angles to each other. Currently, the shortest distance from any surface mount 0805 inductor to another is about 3 to 4mm. I would like to go down to 1mm. What recommendations/guidelines do people suggest regarding how close I can pack these inductors? I will orient them at right angles.

Also, as I think about the alternate TX LPF layouts, what are similar spacing and orientation recommendations/guidelines for inductors wound on T37 cores? I've seen some pretty tight packing on John's board and in commercial relay-switched antenna tuners.

73,

Steve

KF7O

  

[in3otd]

Hello Steve,
IMHO the coils spacing requirements are related to their physical size; IIRC the inductors used in the driver "prefilter" are really small and the actual coil diameter may be around 1 mm, so if you keep them 1 mm away you may end up having a couple of coil diameter of spacing, which suggests that the coupling should be not so high even when placed parallel to each other. But it will be better to do some experiments...
Same for the toroids, but here the dimensions are much bigger and there is the "hidden turn" that can cause some more coupling to be present. Keeping them at 90 degrees will reduce this coupling. Again, to get some numbers some experiments will be needed.

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Claudio,

Thanks for the feedback. There is a graph of coupling coefficient versus distance for the murata parts used here. I think I am safe to tighten up the layout, especially since only inductors at right angles will potentially be closer than 2 mm. I will expand the layout as much as allowed.

73,

Steve

KF7O

[in3otd]

Thanks for the link,Steve,
I didn't notice that the manufacturer already provides some data on the coupling vs distance for the SMD inductors.

I have done some measurement on the coupling between two T37-6 toroids, with 19 turns each, that I happened to have around to build some CWAZ filters (see also here).

The toroids were placed parallel to each other, with the centers on the same line, and the measurements were repeated for several distances between the toroids along this line. The distance, as noted in the graphs below, is between the windings, so 0 mm means that the toroids windings were in close contact.

The same set of measurement was also repeated when swapping the connections of one of the toroids, to change the sign of the coupling coefficient; see below why this gives some slightly different results.

The coupling between the toroids when driven/terminated by a 50 ohm resistance is shown in the following graph:

note that at low frequency the sign of the (magnetic) coupling coefficient makes little difference but at high frequency there may be a notch or a flat region in the coupling, depending on the orientation. This is due to the fact that at those frequencies the capacitive coupling between the toroids becomes comparable to the magnetic coupling, so the two sum or subtract depending on the sign of the magnetic coupling.

The graph above can also be converted to show the actual magnetic coupling coefficient (only the positive coupling coefficients are shown, to have a less-cluttered graph):

as you can see, the coupling coefficient are quite small; when the toroids are in close contact the coupling coefficient is around 0.015 and at 5 mm distance it drops to 0.003.

I tried to reduce the coupling even further by adding a turn around the side of the toroids (similarly to what shown here) and I saw a reduction of about 10 dB but probably the exact placement of the "compensation turn" is critical as for all things involving cancelling small effects. I didn't yet do measurement with the toroids at 90 degrees to each other.

It would be interesting to put these coupling numbers into a simulator and see which effect they have on the filters response/rejection...

[Steve Haynal]

Hi Claudio,

Thanks for these measurements. The return wire wikipedia link is very interesting. Why don't we see more amateur radio homebrew inductors wound with a return wire as shown? It seems simple enough. Also, for figure 5 with the return winding, the return winding should also add to the total inductance. Is that true?

My takeaway from your measurements is that I may be worrying too much about close inductor spacing, especially parallel toroids. My current layout has about 7-8mm between windings of parallel toroids. I would like to add the option for 7-pole filters on the ends for 1 LPF and 1 HPF. This would reduce the spacing to about 4.5mm. I could go to tight packed 7-pole filters everywhere in the same area but spacing would be 0mm between parallel toroids. The distance between end-to-end toroids, or right angle toroids is currently 0-1mm. The end-to-end and right angle toroids were packed that closely in John's working layout.

It would be interesting to know how tight parallel packing might "detune" the inductance. Would adding a capacitor to make a tuned circuit, and then measuring how the resonance of that tune circuit changes as the distance to the coupling inductor is decreased work?

73,

Steve

KF7O

[in3otd]

Hello Steve,
I think for most of the typical ham radio circuits using toroids the return wire around the circumference does not really make a difference as the coupling to nearby structures is already quite low. Then i suspect the exact wire placement is also important to achieve a significant improvement.
Regarding Figure 5 in the link, think of that as a normal toroidal winding where half of the turns are wound over the entire circumference and then the other half is wound over the first half, but going back to the starting point. So, as usual, the inductance is given by the total number of turns, "forward" plus "backward". The disadvantage is additional parasitic capacitance due to the crossing wires; note also that the winding ends where it started and there the effect of the parasitic capacitance is greater since the voltage difference between the nearby crossing wires is higher.

I did a simulation of a 7-order Chebyshev, 1 dB ripple, low-pass filter, with a 35 MHz cutoff frequency (I wasn't able to quickly find a description of the currently proposed TX filters), adding the parasitic coupling (inductive and capacitive) as for the "0 mm" case above and there are practically no changes in the passband, while there is a degradation in the stopband once the response goes below about 50 dB, see graph below:

this is assuming maximum coupling between the first and second toroid and between the second and third, with a lower coupling between the first and third.
Actually the decreased stopband rejection comes mostly from this latter coupling (makes sense, as it bypasses the center filter section).
There are several combinations possible, with different signs for the magnetic coupling: the graph above is the worst case I was able to find (but no guarantee that there are not even worse cases...)

I did some quick measurements with toroids at 90 degrees (no graphs here, yet...); the coupling, as expected, is low but only if the toroids are place symmetrically... let me try a drawing:
 __                                                 __ ┌────┐
|  |  ____                                         |  |└────┘
|  | |    |  good, low coupling                    |  |        higher coupling
|  |  ¯¯¯¯                                         |  | 
 ¯¯                                                 ¯¯
similarly, parallel toroids can have a low coupling if placed staggered
 __ 
|  |
|  | __
|  ||  |
 ¯¯ |  |
    |  |
         ¯¯
but the alignment for minimum coupling is somewhat critical in this case (and, in general, the minimum coupling is not when the center of one toroid is aligned with the edge of the other...)

A measure of the "detuning" is given by the coupling factor (see here), but for our filters here the most important effect seems to come from the coupling which bypass filtering sections and reduce the overall rejection in the stopband.

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Claudio,

For parallel toroids, should I consider some stagger as always better than none even if I can't minimize coupling?

What about for end to end? Is some stagger like this

|____|  

|        ||____|

          |        |

always better than this?

|____||____|

|        ||        |

73,

Steve

KF7O

[Steve Haynal]

Hi Claudio,

You mentioned the biggest detriment is when inductors 1 and 3 couple to diminish the stop band. But for multiple filters in close proximity, where the inactive neighbor filters are floating and not grounded because they are switched out of the circuit, is there anything you recommend?

73,

Steve

KF7O

[in3otd]

Hello Steve,

For parallel toroids, should I consider some stagger as always better than none even if I can't minimize coupling?

Yes, the worst case is when the toroid centers are co-axial, then moving the toroids laterally the coupling decreases, reaches a null as described above, then increases again (but without reaching the maximum value seen with the coaxial centers) and finally decreases again as the distance increases.
 

What about for end to end? Is some stagger like this

|____|  

|        ||____|

          |        |

always better than this?

|____||____|

|        ||        |

 uhm, here I didn't understand how the toroids are positioned...

[Steve Haynal]

Hi Claudio,

Sorry, my ASCII art didn't come through correctly and wasn't very good to begin with. I'm wondering if there is any advantage to staggering end to end inductors as seen below. Also, if these close end to end inductors are part of different filters, will the ungrounded neighbors affect the activated filter? Given John's working layout and other layouts I've seen, I would think not.

73,

Steve

KF7O

[John Williams]

Steve,

On my TX board, the filters are all grounded unless active. On the RX board they are not.

John

--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[in3otd]

Hello Steve,
the coupling is maximum when the toroids are aligned, so also for the end-to-end case staggering will lower (a little) the coupling.

I don't think grounded or undergrounded nearby inductors will have a significant effect; doing the math, if you have two equal inductors of inductance L with a coupling factor k, shorting one makes the inductance of the other one become L*(1-k^2) - so for the low k values we have between the toroids the inductance does not practically change.

73 de Claudio, IN3OTD / DK1CG

[Sid Boyce]

Hi guys,
I have no way to measure coupling and it's not a topic raised with
regards to the eb104.ru BPF and 300W PA/LPF boards.
This is how they are arranged.

I think I have seen plots for both but I can't trace them now. I'll see
if eb104.ru can send me copies.

Empirically I find them perfectly OK.

Tim (W4YN) has also used them as well as Alex with Hermes with no
noticeable difference.

By comparison the LPF unit culled from my IC-737 looks "thrown" together.
73 ... Sid.

On 06/10/16 03:34, Steve Haynal wrote:
> Hi Claudio,
>

> For parallel toroids, should I consider some stagger as always better
> than none even if I can't minimize coupling?
>
> What about for end to end? Is some stagger like this
>

> |____| Â
> | Â Â Â Â ||____|
> Â Â Â Â Â | Â Â Â Â |

>
> always better than this?
>
> |____||____|

> | Â Â Â Â || Â Â Â Â |
>
>
> 73,
>
> Steve
> KF7O
>
>
>
> Â

>
> On Wednesday, October 5, 2016 at 1:50:26 PM UTC-7, in3otd wrote:
>
> Hello Steve,
> I think for most of the typical ham radio circuits using toroids
> the return wire around the circumference does not really make a
> difference as the coupling to nearby structures is already quite
> low. Then i suspect the exact wire placement is also important to
> achieve a significant improvement.
> Regarding Figure 5 in the link, think of that as a normal toroidal
> winding where half of the turns are wound over the entire
> circumference and then the other half is wound over the first
> half, but going back to the starting point. So, as usual, the
> inductance is given by the total number of turns, "forward" plus
> "backward". The disadvantage is additional parasitic capacitance
> due to the crossing wires; note also that the winding ends where
> it started and there the effect of the parasitic capacitance is
> greater since the voltage difference between the nearby crossing
> wires is higher.
>
> I did a simulation of a 7-order Chebyshev, 1 dB ripple, low-pass
> filter, with a 35 MHz cutoff frequency (I wasn't able to quickly
> find a description of the currently proposed TX filters), adding
> the parasitic coupling (inductive and capacitive) as for the "0
> mm" case above and there are practically no changes in the
> passband, while there is a degradation in the stopband once the
> response goes below about 50 dB, see graph below:
>

> <https://lh3.googleusercontent.com/-oDKXh9WeRzs/V_VZz8M_zpI/AAAAAAAAANQ/xaYLlXym4y4olnIOuccycmMreYXuOEjvQCLcB/s1600/export.png>

>
> this is assuming maximum coupling between the first and second
> toroid and between the second and third, with a lower coupling
> between the first and third.
> Actually the decreased stopband rejection comes mostly from this
> latter coupling (makes sense, as it bypasses the center filter
> section).
> There are several combinations possible, with different signs for
> the magnetic coupling: the graph above is the worst case I was
> able to find (but no guarantee that there are not even worse cases...)
>
> I did some quick measurements with toroids at 90 degrees (no
> graphs here, yet...); the coupling, as expected, is low but only
> if the toroids are place symmetrically... let me try a drawing:

> Â __Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â
>                     __ ┌────â”
> |Â |Â ____Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â
>               | |└────┘
> |Â | |Â Â Â |Â good, low
> coupling                   | |      Â
> higher coupling
> | | ¯¯¯¯                         Â
> Â Â Â Â Â Â Â Â Â Â Â |Â |Â
>  ¯¯                          Â
>                      ¯¯

> similarly, parallel toroids can have a low coupling if placed
> staggered

> Â __Â
> |Â |
> |Â | __
> |Â ||Â |
>  ¯¯ | |
> Â Â Â |Â |
>        ¯¯

> but the alignment for minimum coupling is somewhat critical in
> this case (and, in general, the minimum coupling is not when the
> center of one toroid is aligned with the edge of the other...)
>
> A measure of the "detuning" is given by the coupling factor (see
> here

> <https://en.wikipedia.org/wiki/Inductance#Equivalent_circuit>),

> Â

>
>
>
>
>
>
>
> On Tuesday, October 4, 2016 at 1:58:39 PM UTC-7, in3otd wrote:
>
> Thanks for the link,Steve,
> I didn't notice that the manufacturer already provides
> some data on the coupling vs distance for the SMD inductors.
>
> I have done some measurement on the coupling between two
> T37-6 toroids, with 19 turns each, that I happened to have
> around to build some CWAZ filters

> <https://www.arrl.org/files/file/Technology/tis/info/pdf/9902044.pdf>
> (see also here <http://www.gqrp.com/Datasheet_W3NQN.pdf>).

>
> The toroids were placed parallel to each other, with the
> centers on the same line, and the measurements were
> repeated for several distances between the toroids along
> this line. The distance, as noted in the graphs below, is
> between the windings, so 0 mm means that the toroids
> windings were in close contact.
>
> The same set of measurement was also repeated when
> swapping the connections of one of the toroids, to change
> the sign of the coupling coefficient; see below why this
> gives some slightly different results.
>
>
> The coupling between the toroids when driven/terminated by
> a 50 ohm resistance is shown in the following graph:
>
>

> <https://lh3.googleusercontent.com/-F9kXH8xBLtM/V_QUm7BIteI/AAAAAAAAAM4/AUq5Ji1CblQ4ee9NS77Tq91ADneIrf37ACLcB/s1600/T37-6_19t_coupling_S21.png>

>
>
> note that at low frequency the sign of the (magnetic)
> coupling coefficient makes little difference but at high
> frequency there may be a notch or a flat region in the
> coupling, depending on the orientation. This is due to the
> fact that at those frequencies the capacitive coupling
> between the toroids becomes comparable to the magnetic
> coupling, so the two sum or subtract depending on the sign
> of the magnetic coupling.
>
> The graph above can also be converted to show the actual
> magnetic coupling coefficient (only the positive coupling
> coefficients are shown, to have a less-cluttered graph):
>

> <https://lh3.googleusercontent.com/-bunn50yjlQs/V_QSYYm9C6I/AAAAAAAAAMs/Eeo-tlexcAk9uPqOf2psmsYLQi-lfvh5ACLcB/s1600/T37-6_19t_coupling.png>

>
>
> as you can see, the coupling coefficient are quite small;
> when the toroids are in close contact the coupling
> coefficient is around 0.015 and at 5 mm distance it drops
> to 0.003.
>
>
> I tried to reduce the coupling even further by adding a
> turn around the side of the toroids (similarly to what
> shown here

> <https://en.wikipedia.org/wiki/Toroidal_inductors_and_transformers#Total_B_field_confinement_by_toroidal_inductors>)

> and I saw a reduction of about 10 dB but probably the
> exact placement of the "compensation turn" is critical as
> for all things involving cancelling small effects. I
> didn't yet do measurement with the toroids at 90 degrees
> to each other.
>
>
> It would be interesting to put these coupling numbers into
> a simulator and see which effect they have on the filters
> response/rejection...
>
>
> 73 de Claudio, IN3OTD / DK1CG
>

> --
> You received this message because you are subscribed to the Google
> Groups "Hermes-Lite" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to hermes-lite...@googlegroups.com

> <mailto:hermes-lite...@googlegroups.com>.

> For more options, visit https://groups.google.com/d/optout.

--
Sid Boyce ... Hamradio License G3VBV, Licensed Private Pilot
Emeritus IBM/Amdahl Mainframes and Sun/Fujitsu Servers Tech Support
Senior Staff Specialist, Cricket Coach
Microsoft Windows Free Zone - Linux used for all Computing Tasks