Two series spiral inductors

[cnm]

For bias line choke design, I am trading between using two small spiral inductors in series vs. one large spiral inductor. Any one has any comment on the pros and cons between the two approach?

 

[vfone]

This depends by a lot of factors as frequency, layout, afferent decoupling capacitors (or u-strip stubs), etc.
Generally using two inductors in series to build the choke you get wider bandwidth, but in the same time at very high frequencies the footprint parasitics of two inductors instead of one, are greater.
Sometimes to get wideband chokes I put two ore more inductors in series (e.g. 10nH + 100nH + 1uH; the inductor closest to the source has the largest inductance and the inductance closest to the load has the smallest inductance).

 

[G0HZU]

This depends by a lot of factors as frequency, layout, afferent decoupling capacitors (or u-strip stubs), etc.
Sometimes to get wideband chokes I put two ore more inductors in series (e.g. 10nH + 100nH + 1uH; the inductor closest to the source has the largest inductance and the inductance closest to the load has the smallest inductance).

I would expect that there could be a nasty series resonance at UHF with that arrangement?

I guess it depends on the manufacturer's inductor series you use but I would think that the above arrangement would cause problems. Do you have a combination of inductors (i.e. manufacturer's part number) that give good performance?

 

[vfone]

When is about expectations always I have a doubts.
There is a famous quote of Albert Einstein, which generally is guiding my way: "The only source of knowledge is experience".

 

[G0HZU]

When is about expectations always I have a doubts.
There is a famous quote of Albert Einstein, which generally is guiding my way: "The only source of knowledge is experience".

Not sure what to make of your reply but a 1uH choke ceases to be a 1uH at high frequencies and (depending on construction) it can look like a capacitor up at UHF.
This capacitor then resonates with the series 100nH + 10nH and you don't have a choke at that resonant frequency any more, you actually have a poor quality acceptor circuit.

That's why I asked if you actually have found real 10nH + 100nH + 1uH components that actually work in a group. I've seen engineers simulate a string of chokes using s parameters but when they actually test it for real there is usually a resonance that spoils the performance at one or more frequencies.

Can you give a part number of the 1uH you used? Is it a secret part that you don't want to share?

I think the OP may have similar issues with two spiral inductors in series. There could be a resonance that spoils the wideband performance.

Note that if the bias choke is used in the output of a high power transmitter (eg >20W) you only need to have a very faint resonance (seen as a very small and narrow dip in S21) in the series chokes to transfer enough power to destroy the bias choke(s).

 

[cnm]

Not sure what to make of your reply but a 1uH choke ceases to be a 1uH at high frequencies and (depending on construction) it can look like a capacitor up at UHF.
This capacitor then resonates with the series 100nH + 10nH and you don't have a choke at that resonant frequency any more, you actually have a poor quality acceptor circuit.

That's why I asked if you actually have found real 10nH + 100nH + 1uH components that actually work in a group. I've seen engineers simulate a string of chokes using s parameters but when they actually test it for real there is usually a resonance that spoils the performance at one or more frequencies.

Can you give a part number of the 1uH you used? Is it a secret part that you don't want to share?

I think the OP may have similar issues with two spiral inductors in series. There could be a resonance that spoils the wideband performance.

Note that if the bias choke is used in the output of a high power transmitter (eg >20W) you only need to have a very faint resonance (seen as a very small and narrow dip in S21) in the series chokes to transfer enough power to destroy the bias choke(s).

I agree with vfone in that using multiple spiral inductors will create larger footprint, which means larger overall inter-winding capacitance and large shunt to ground parasitic capacitance. That means SRF drops faster for the multiple spiral case. However, if two series inductors are both very small, they both will have higher SRF compared to one large spiral. After connecting the two small ones in series, their resultant SRF is probably still higher than a single super large inductor. But this sounds a little like an extreme case. Not sure if it is a fair comparison?

GOHZU's comments are definitely excellent, which is exactly what I was looking for. Johnanson has measured S-par parameter data. I will play with the inductor combinations using their data when I get a chance, just for fun

Please comment if you have other things to add or disagree with any of the above. I am learning from you guy's comments. :grin:

 

[biff44]

It is tempting to use two different sized coils for broadband applications. That way you get two different SRFs, and one coil can be a large one for filtering out the low frequencies. I would either space them out, or layout at right angles so one's magnetic field does not couple to the other one.

The do have conical inductors that work better for large bandwidth applications. See Piconics.

 

[G0HZU]

I've never tried stacking two spiral inductors but I'd imagine the results will be poor for the reasons I gave earlier.

I'm assuming you mean microstrip spiral inductors? These are effectively a spiralled transmission line so if you put two (different sizes) in series then at some frequency point(s) one is going to be looking capacitive and one is going to look inductive so there will be an acceptor response somewhere in the frequency band. So the result will not be broadband.

Can you define the frequency band you are considering and also what bandwidth you want it to work over?

---------- Post added at 17:23 ---------- Previous post was at 16:52 ----------

It is tempting to use two different sized coils for broadband applications. That way you get two different SRFs, and one coil can be a large one for filtering out the low frequencies. I would either space them out, or layout at right angles so one's magnetic field does not couple to the other one.

The do have conical inductors that work better for large bandwidth applications. See Piconics.

Where I work we did try playing with the Coilcraft conical inductors for use as a bias tee. The s parameter files do look really encouraging but when you actually test one on a VNA then you see lots of internal resonances. These are typically only a fraction of 1dB but for our purposes this was no good. You don't see these resonances on the simulation because the s parameter data is coarsely stepped every 50MHz

I've yet to find a better bias tee choke than the old miniature axial chokes that date back to the 1980s. eg 1uH can work extremely well to beyond 3GHz. We did try adding smaller chokes in series but this generally spoils the performance below 3GHz due to small resonant suckouts etc.

 

[cnm]

I've never tried stacking two spiral inductors but I'd imagine the results will be poor for the reasons I gave earlier.

I'm assuming you mean microstrip spiral inductors? These are effectively a spiralled transmission line so if you put two (different sizes) in series then at some frequency point(s) one is going to be looking capacitive and one is going to look inductive so there will be an acceptor response somewhere in the frequency band. So the result will not be broadband.

Can you define the frequency band you are considering and also what bandwidth you want it to work over?

---------- Post added at 17:23 ---------- Previous post was at 16:52 ----------



Where I work we did try playing with the Coilcraft conical inductors for use as a bias tee. The s parameter files do look really encouraging but when you actually test one on a VNA then you see lots of internal resonances. These are typically only a fraction of 1dB but for our purposes this was no good. You don't see these resonances on the simulation because the s parameter data is coarsely stepped every 50MHz

I've yet to find a better bias tee choke than the old miniature axial chokes that date back to the 1980s. eg 1uH can work extremely well to beyond 3GHz. We did try adding smaller chokes in series but this generally spoils the performance below 3GHz due to small resonant suckouts etc.

Let's assume a common case of building a choke for DC to 10GHz amplifier application -just to make it a little challenging and more fun.

Yes, the spirals are planar spiral inductors on board or substrate.

Why do you say conical inductor spar looks encouraging but when you test it, it shows resonances? Did the vendor's spar miss (or hide) the resonances by using coarse f steps?

 

[G0HZU]

I think it depends on what you want from the bias tee. The conical chokes are going to be fine for many users as the small resonances are insignificant to most people. We had a flatness spec that was very strict over any small subset of the band so it failed where the resonances were. However (from memory) the resonances were very small wrt 1dB.

I don't think you will cover LF to 10GHz with a set of spiral inductors!

Is the LF-10GHz for a low level amplifier? If so then I guess the Mini Circuits UWB choke forms a benchmark over 50-10,000MHz.

https://www.minicircuits.com/pages/s-params/ADCH-80A_GRAPHS.pdf


However, I rarely design stuff for more than 3GHz and I have a good 1uH bias choke for this range. I also have a reasonable 3GHz VNA at home so I suppose I could start the ball rolling with a design that doesn't meet your spec but still performs well!

I'll post up a plot of my favourite 1uH choke. Note that this is 'slightly' cheating as I'll just fit the choke in shunt on a test fixture rather than plot a proper bias tee. So no series DC block caps to spoil the response.

Is it permissible to use shunt reactance in the bias tee? eg series L, shunt C, series L, shunt C etc?

Edit:
Here's a plot of the 1uH over LF to 3GHz.
As you can see it performs very well. The plot is taken on a test fixture and this has precision 20dB pads included in the calibration in order to reduce ripple due to test equipment limitations.

As you can see there are no suckouts anywhere and the choke stays nicely reactive with no measurable resistive loss anywhere

I can't measure beyond 3GHz here at home but I do know it lets go above 3GHz so it won't meet your 10GHz spec. But it is very good for work up to 3GHz.

Click on the link below to access the doc with the plot.

 

[vfone]

Yeah, I met lot of people that never tried something but they have doubts, because they are stuck in theory. In fact, unfortunately, I live with them everyday.

Different inductors in series, or conical shape choke, or logarithmic type winding choke, (all of these approaches for increasing bandwidth) are used in microwave circuit design by more than 50 years. There are tens of patents about this ideas.

 

[cnm]

I think it depends on what you want from the bias tee. The conical chokes are going to be fine for many users as the small resonances are insignificant to most people. We had a flatness spec that was very strict over any small subset of the band so it failed where the resonances were. However (from memory) the resonances were very small wrt 1dB.

I don't think you will cover LF to 10GHz with a set of spiral inductors!

Is the LF-10GHz for a low level amplifier? If so then I guess the Mini Circuits UWB choke forms a benchmark over 50-10,000MHz.

https://www.minicircuits.com/pages/s-params/ADCH-80A_GRAPHS.pdf


However, I rarely design stuff for more than 3GHz and I have a good 1uH bias choke for this range. I also have a reasonable 3GHz VNA at home so I suppose I could start the ball rolling with a design that doesn't meet your spec but still performs well!

I'll post up a plot of my favourite 1uH choke. Note that this is 'slightly' cheating as I'll just fit the choke in shunt on a test fixture rather than plot a proper bias tee. So no series DC block caps to spoil the response.

Is it permissible to use shunt reactance in the bias tee? eg series L, shunt C, series L, shunt C etc?

Edit:
Here's a plot of the 1uH over LF to 3GHz.
As you can see it performs very well. The plot is taken on a test fixture and this has precision 20dB pads included in the calibration in order to reduce ripple due to test equipment limitations.

As you can see there are no suckouts anywhere and the choke stays nicely reactive with no measurable resistive loss anywhere

I can't measure beyond 3GHz here at home but I do know it lets go above 3GHz so it won't meet your 10GHz spec. But it is very good for work up to 3GHz.

Click on the link below to access the doc with the plot.

Cool that you have your own VNA at home. It is probably needed to use shunt C for most applications as dc bypass. Very nice to see the plot of your 1uH choke.

 

[G0HZU]

Yeah, I met lot of people that never tried something but they have doubts, because they are stuck in theory. In fact, unfortunately, I live with them everyday.

Different inductors in series, or conical shape choke, or logarithmic type winding choke, (all of these approaches for increasing bandwidth) are used in microwave circuit design by more than 50 years. There are tens of patents about this ideas.

Are you having a pop at me?

Show me how you can stack two (decade different value) inductors without introducing small resonant suckouts. Sure. you can extend the bandwidth but there is a price to pay. Take a look at the plot of my 1uH bias choke. It looks GOOD. If I try and extend it by adding a 100nH in series then It will lose performance in the 20-3000MHz range due to suckouts.

You still haven't given me an example of your own 1uH choke + 100nH + 10nH.

Also, I think the problem will be much worse with two spiral inductors in series.

I met lot of people that never tried something but they have doubts

I've certainly tried stacking lumped inductors in series and that's why I offered my experience on this.
If the bias tee is used for a high RF power application then these small suckouts are a problem.

eg a 0.5dB suckout at 25W can transfer a couple of watts to the choke. Often this is enough to destroy it. It gets worse if the circuit feeds an antenna and there is a VSWR peak at the resonant frequency where the suckout is. So stacking chokes in series can give problems to some applications. So basically, I'm not saying 'don't do it' but I am saying there could be problems with suckout resonances that could ruin the performance.

In fact, unfortunately, I live with them everyday

I also meet people like you where I work but not every day

 

[cnm]

Are you having a pop at me?

Show me how you can stack two inductors without introducing small resonant suckouts. Sure. you can extend the bandwidth but there is a price to pay. Take a look at the plot of my 1uH bias choke. It looks GOOD. If I try and extend it by adding a 100nH in series then It will lose performance in the 20-3000MHz range due to suckouts.

You still haven't given me an example of your own 1uH choke + 100nH + 10nH.

"If I try and extend it by adding a 100nH in series then It will lose performance in the 20-3000MHz range due to suckouts" - why this behavior? I thought 100nH has a much larger SRF beyond 3GHz? Maybe the added shunt C from the 100nH is working on the large inductance to cause this!

---------- Post added at 18:42 ---------- Previous post was at 18:40 ----------

Yeah, I met lot of people that never tried something but they have doubts, because they are stuck in theory. In fact, unfortunately, I live with them everyday.

Different inductors in series, or conical shape choke, or logarithmic type winding choke, (all of these approaches for increasing bandwidth) are used in microwave circuit design by more than 50 years. There are tens of patents about this ideas.

The interesting part is to compare notes and see if they are consistent. We use each other's ideas as touchstone for our own ideas. That is the best part of a forum. Let's stay cool, every one.

 

[G0HZU]

"If I try and extend it by adding a 100nH in series then It will lose performance in the 20-3000MHz range due to suckouts" - why is this? I thought 100nH has a much larger SRF beyond 3GHz?

It's because at UHF the 1uH choke is no longer a 1uH choke. At UHF and higher it's reactance will be somewhere on the outside of the smith chart and at one frequency it will have capacitive reactance that will resonate with 100nH and suddenly you have a small capacitor in series with 100nH at that frequency. Usually this doesn't mean a deep suckout resonance as there is high ESR in the resonance but it can mean a 1dB dip in the response.

I'll stick a handwound 100nH choke in series with the 1uH choke and plot it over 3GHz. Give me a while to do this...

 

[cnm]

It's because at UHF the 1uH choke is no longer a 1uH choke. At UHF and higher it's reactance will be somewhere on the outside of the smith chart and at one frequency it will have capacitive reactance that will resonate with 100nH and suddenly you have a small capacitor in series with 100nH at that frequency. Usually this doesn't mean a deep suckout resonance as there is high ESR in the resonance but it can mean a 1dB dip in the response.

I'll stick a handwound 100nH choke in series with the 1uH choke and plot it over 3GHz. Give me a while to do this...

I totally agree. will be interesting to see some data on this, when you get a chance.

---------- Post added at 19:39 ---------- Previous post was at 19:24 ----------

I think it depends on what you want from the bias tee. The conical chokes are going to be fine for many users as the small resonances are insignificant to most people. We had a flatness spec that was very strict over any small subset of the band so it failed where the resonances were. However (from memory) the resonances were very small wrt 1dB.

I don't think you will cover LF to 10GHz with a set of spiral inductors!

Is the LF-10GHz for a low level amplifier? If so then I guess the Mini Circuits UWB choke forms a benchmark over 50-10,000MHz.

https://www.minicircuits.com/pages/s-params/ADCH-80A_GRAPHS.pdf


However, I rarely design stuff for more than 3GHz and I have a good 1uH bias choke for this range. I also have a reasonable 3GHz VNA at home so I suppose I could start the ball rolling with a design that doesn't meet your spec but still performs well!

I'll post up a plot of my favourite 1uH choke. Note that this is 'slightly' cheating as I'll just fit the choke in shunt on a test fixture rather than plot a proper bias tee. So no series DC block caps to spoil the response.

Is it permissible to use shunt reactance in the bias tee? eg series L, shunt C, series L, shunt C etc?

Edit:
Here's a plot of the 1uH over LF to 3GHz.
As you can see it performs very well. The plot is taken on a test fixture and this has precision 20dB pads included in the calibration in order to reduce ripple due to test equipment limitations.

As you can see there are no suckouts anywhere and the choke stays nicely reactive with no measurable resistive loss anywhere

I can't measure beyond 3GHz here at home but I do know it lets go above 3GHz so it won't meet your 10GHz spec. But it is very good for work up to 3GHz.

Click on the link below to access the doc with the plot.

GOZHU, in your plot posted, were you plotting S21? If I understand correctly, you were measuring a two-port path, with your shunt L in the middle?

 

[G0HZU]

Yes, I was plotting S21 of a short run of microstrip on Rogers 4003 PCB material with the choke in shunt. Obviously this doesn't represent a thorough test of the choke but it is OK for a quick test. At work we have proper test fixtures and multiport VNA capability but I only have a basic VNA.

Here's a doc with what I see on the VNA with the 1uH and 100nH in series. I've taken plots with both orientations of the two chokes. eg 1uH + 100nH and also 100nH + 1uH.


I get suckouts with both ways of connecting them although the suckouts are layout sensitive and move with respect to height above the PCB. You can see than in both cases the 20-3000MHz performance is degraded. Sadly the 3GHz VNA can't show if the performance improves above 3GHz.

View attachment 1uh100nh.doc

Maybe I'll try 100nH 1uH and 10uH to scale everything down in frequency.

 

[cnm]

Yes, I was plotting S21 of a short run of microstrip on Rogers 4003 PCB material with the choke in shunt. Obviously this doesn't represent a thorough test of the choke but it is OK for a quick test. At work we have proper test fixtures and multiport VNA capability but I only have a basic VNA.

Here's a doc with what I see on the VNA with the 1uH and 100nH in series. I've taken plots with both orientations of the two chokes. eg 1uH + 100nH and also 100nH + 1uH.


I get suckouts with both ways of connecting them although the suckouts are layout sensitive and move with respect to height above the PCB. You can see than in both cases the 20-3000MHz performance is degraded. Sadly the 3GHz VNA can't show if the performance improves above 3GHz.

View attachment 50501

Maybe I'll try 100nH 1uH and 10uH to scale everything down in frequency.

These are excellent data in revealing what happens to these two series inductors! I can only say "wow!". To think this further, I am wondering if it is always true that one spiral is better than two spiral in series for building bias chokes?

I suspect at higher frequency, things will be even worse. Don't you agree?

 

[G0HZU]

It's certainly obvious with the chokes I have selected but I have found that it is worth experimenting with different chokes for bias tees. I have tried lots of different chokes and there isn't one choke that is suitable for all needs.

The conical chokes offered by Coilcraft may well do what you want but I very much doubt you will get anywhere with 2 or more series spiral chokes made from microstrip.

By the way, if you make a very crude model of a 1uH inductor eg 1uH in parallel with 0.4pF and give it a low Q and then put it in series with a 100nH choke then you will see a similar plot to my measured data if you analyse on a linear simulator (i.e. the first suckout agrees). However, in reality, I find that it is worth trying lots of different chokes as the difference in behaviour up at UHF can be quite marked. Also, with SMD chokes the PCB thickness and dielectric affect the spurious responses as there can be other resonant modes introduced.

That's why I am fond of the old school miniature axial leaded chokes as they sit slightly above the PCB surface

I was kind of hoping that vfone had found a combination of 10nH + 100nH +1uH that works and that's why I asked about specific manufacturers' parts because maybe there is a combo that works reasonably well.

---------- Post added at 22:30 ---------- Previous post was at 22:20 ----------

To think this further, I am wondering if it is always true that one spiral is better than two spiral in series for building bias chokes?

By all means try two spiral inductors but I am fairly certain you will get a frequency where there will be opposite reactances and therefore a suckout. Because spiral inductors are coiled up transmission lines I think the suckout will be quite noticeable because the lumped choke will have poor ESR at resonance but I suspect that both spirals will provide a pretty decent low ESR suckout in several places (depends on bandwidth and spiral construction).

 

[biff44]

It has been a while, but as I recall the trick is to have the smaller coil near the active circuit, then have a small capacitance to ground (maybe 22 pF), and then the 2nd larger series coil. In low current circuits I have seen series resistors added too to dampen any resonances (like 47 ohms in series).

The "theory" of the conical inductor coils is that there WILL be many resonances, but they will be small since the aspect ratio of each turn of the coil is changing. If you just have one big cylindrical coil, each loop is the same as its neighbor in physical size, so the resonance caused by interloop capacitance with each coil turn causes a number of resonances at the same frequency.