[SOLVED] How can i bias Varactor Diode in my design?

[redcrystalline]

Hi
I'm Designing an phase shifter and i have to use 2 Varactor Diode
How can i bias This phase shifter without any dc block in phase shifter.
phase shifter has 2 port by the way and I'm wondering how can i bias this Varactors?

 

[D.A.(Tony)Stewart]

Shunt Anodes to signals and common cathode positive bias for small signals with relatively large bias?

 

[redcrystalline]

Shunt Anodes to signals and common cathode positive bias for small signals with relatively large bias?

Two Varactor have been grounded Using Two Vias and both have reverse bias. DC voltage is in 1-25 V range and about ac amplitude i have not any idea because its for research purpose.

 

[D.A.(Tony)Stewart]

https://www.google.ca/search?q=vari...ecVcmXEIi9yQSl7IPABg#q=varactor+phase+shifter

 

[volker@muehlhaus]

Your varactors look very large compared to the coupler size. They will not behave like capacitors if you build this.

You can try a lower permittivity substrate to have larger line/coupler dimensions. Then you also have more room for bias coupling/decoupling.

 

[redcrystalline]

Your varactors look very large compared to the coupler size. They will not behave like capacitors if you build this.

You can try a lower permittivity substrate to have larger line/coupler dimensions. Then you also have more room for bias coupling/decoupling.

Actually your Right. But permittivity is about 3.38 and lower than this make no differ. So I suggested to my costumer to lower frequency[For now Frequency is 8 GHz]
by the way i had a look on Varactor datasheet. it says varactor Works to 20GHz. but in that frequency Varactor is very large. Its kind of confusing.

 

[volker@muehlhaus]

Actually your Right. But permittivity is about 3.38 and lower than this make no differ. So I suggested to my costumer to lower frequency[For now Frequency is 8 GHz]
by the way i had a look on Varactor datasheet. it says varactor Works to 20GHz. but in that frequency Varactor is very large. Its kind of confusing.

Yes, something looks wrong here. Are you sure the varactor package dimensions in your 3D model are correct?

 

[redcrystalline]

Yes, something looks wrong here. Are you sure the varactor package dimensions in your 3D model are correct?

Yes I'm sure because Varactor dimension is about 2 mm and line width is 1.3 mm, the problem seems to be frequency and this Varacor dimension is not right for this frequency range. i have tried to find smaller Varactor or smd ones but there is not such a part!
and one question stand yet, if this varactor cant be used for 8 GHz so how can somebody use it at 20GHz [based on it datasheet]
Varactor part number: ND3000

 

[volker@muehlhaus]

and one question stand yet, if this varactor cant be used for 8 GHz so how can somebody use it at 20GHz [based on it datasheet]
Varactor part number: ND3000

It the inductance/length can be built into the electrical circuit design, it should work. The part looked really large in your 3D model, but it looks much better in the datasheet (not so very high).

From EM Modelling experience, I would say that your 3D model with the large component size behaves different from the real package: (1) In your 3D model, you have large (high/thick) metal pads that don't exist in reality and (2) your discrete port in the 3D model is only valid if the port dimensions are small. Port dimension of 2mm = lambda/10 between the pads @ er=3.38 is borderline ... might work, but not accurate.

 

[redcrystalline]

It the inductance/length can be built into the electrical circuit design, it should work. The part looked really large in your 3D model, but it looks much better in the datasheet (not so very high).

From EM Modelling experience, I would say that your 3D model with the large component size behaves different from the real package: (1) In your 3D model, you have large (high/thick) metal pads that don't exist in reality and (2) your discrete port in the 3D model is only valid if the port dimensions are small. Port dimension of 2mm = lambda/10 between the pads @ er=3.38 is borderline ... might work, but not accurate.

Thanks for your comments. So there is not any choice. I should change frequency [Being varactor in Design is Essential]

 

[volker@muehlhaus]

I don't know about your customer and design, but frequency for my projects was always fixed, not negotiable.

What I tried to say is that your EM model of the varactor looks inaccurate to me in multiple aspects - the varactor device might be ok. It will be important to model the varactor as a 2-port with proper reference planes [*] and then see what it gives with the actual 2-port data that include physical length.


[*] I don't know how to do that in CST. I am using Sonnet and Momentum where it is possible to use accurate reference planes and accurate port calibration.

 

[redcrystalline]

I don't know about your customer and design, but frequency for my projects was always fixed, not negotiable.

What I tried to say is that your EM model of the varactor looks inaccurate to me in multiple aspects - the varactor device might be ok. It will be important to model the varactor as a 2-port with proper reference planes [*] and then see what it gives with the actual 2-port data that include physical length.


[*] I don't know how to do that in CST. I am using Sonnet and Momentum where it is possible to use accurate reference planes and accurate port calibration.

Yes most of times frequency is fixed but in this case its an Academic project[i should try to convince the costumer!!!].
about CST i'm concern about its length because wavelength is about 8 mm so landa/10=.8 mm and varactor length=2mm.
so its obvious 2>0.8 (varactor length>landa/10). so varactor length can have some influences that can;t be predicted with any kind of simulation and datasheet has not any data about this situation[by the way i think that this is same problem with any other software like sonnet ].
After all your right changing frequency is worst choice for a designer.

 

[D.A.(Tony)Stewart]

Is this for steerable antennae or bit shifter??
What parameters are important?

Rx or Tx ? Levels?
Noise factor or Loss factor ?
S parms?

Many requirements must be defined before design can begin and then material selection and layout can begin.

Including printed coil design, linear or nonlinear substrate ...
If linear then Polyamide, teflon or ceramic?
Then microstrip or stripline?

Is test coupon avail. for varactor? 50R network or?

I agree with other comments.

 

[redcrystalline]

Is this for steerable antennae or bit shifter??
What parameters are important?

Rx or Tx ? Levels?
Noise factor or Loss factor ?
S parms?

Many requirements must be defined before design can begin and then material selection and layout can begin.

Including printed coil design, linear or nonlinear substrate ...
If linear then Polyamide, teflon or ceramic?
Then microstrip or stripline?

Is test coupon avail. for varactor? 50R network or?

I agree with other comments.

It's a voltage controlled phase shifter. return loss should remain under -10 db and the phase can be controlled with DC voltage.
and about other staff i don't have any idea!
there is so many parameter that you have mentioned and freaked me out!

 

[volker@muehlhaus]

about CST i'm concern about its length because wavelength is about 8 mm so landa/10=.8 mm and varactor length=2mm.

I think your wavelength calculation is wrong. Wavelength in air at 8GHz is 37.5mm. Wavelength in substrate is 37.5mm / sqrt(3.38)

so varactor length can have some influences that can;t be predicted with any kind of simulation

This is not correct. Yes, the simple 1-port modelling that your CST picture shows is only valid when the component is short, but as said, there are better ways to do this. Good models do include the effect of electrical length, for example Modelithics component libraries. And correct EM-circuit co-simulation will include these effects by simulating the devices as 2-port elements, with the proper S2P data including parasitics (incuding length effects). This is what I do often in my EM consulting work and yes, it works just fine.

 

[D.A.(Tony)Stewart]

It's a voltage controlled phase shifter. return loss should remain under -10 db and the phase can be controlled with DC voltage.
and about other staff i don't have any idea!
there is so many parameter that you have mentioned and freaked me out!

Then considering your experience learn how Varactor tuned VCXO or LC f tune works below 100MHz

then learn stripline and microstrip for microwave design using Smith Charts and RLC parameters

then use modeling to understand geometry of layout of each 2port component with different components, materials and structures.

phase shift must use a fixed frequency , while self resonant , shifts freq.

 

[redcrystalline]

I think your wavelength calculation is wrong. Wavelength in air at 8GHz is 37.5mm. Wavelength in substrate is 37.5mm / sqrt(3.38)

Your right. my mistake!

This is not correct. Yes, the simple 1-port modelling that your CST picture shows is only valid when the component is short, but as said, there are better ways to do this. Good models do include the effect of electrical length, for example Modelithics component libraries. And correct EM-circuit co-simulation will include these effects by simulating the devices as 2-port elements, with the proper S2P data including parasitics (incuding length effects). This is what I do often in my EM consulting work and yes, it works just fine.

I've done a simple example using CST to see how it treats with lumped elements.
based on this I've simulated design for different element length=S;
It shows that CST treat with lumped elements like 2 port elements[its sensitive to lumped element length].

 

[D.A.(Tony)Stewart]

This only simulates time delay vs length. Constant slope.

 

[volker@muehlhaus]

It shows that CST treat with lumped elements like 2 port elements[its sensitive to lumped element length].

No. This is popular mistake in EM modelling. For accurate results, it is not sufficient to have one port with (+) on one pad and (-) on the other pad. You then see the series path only, with shunt parasitics floating. To include the full S2P data including shunt path parasitics, it is required to place two ports, where each port is referenced to ground. Proper placement of reference planes is needed to include the device length once (no include it twice in both S2P data and EM Model ...)

A while ago Modelithics and Agilent had published an appnote on EM ports for SMD models where they also made this mistake. I demonstrated to them why this is wrong in theory and also in some practical cases, and they changed their appnoted as I advised. But of course it's your model, so feel free to use whatever you like ...

https://muehlhaus.com/support/ads-application-notes/momentum-port-global-ground-or-differential

 

[redcrystalline]

This only simulates time delay vs length. Constant slope.

Your right. So lumped element is not simple lumped element. capacitance is fixed as lumped element but when length of this capacitance changes, CST consider this changing in length as phase shifts. then CST is OK for lumped elements simulation in RF designs.

- - - Updated - - -

No. This is popular mistake in EM modelling. For accurate results, it is not sufficient to have one port with (+) on one pad and (-) on the other pad. You then see the series path only, with shunt parasitics floating. To include the full S2P data including shunt path parasitics, it is required to place two ports, where each port is referenced to ground. Proper placement of reference planes is needed to include the device length once (no include it twice in both S2P data and EM Model ...)

A while ago Modelithics and Agilent had published an appnote on EM ports for SMD models where they also made this mistake. I demonstrated to them why this is wrong in theory and also in some practical cases, and they changed their appnoted as I advised. But of course it's your model, so feel free to use whatever you like ...

https://muehlhaus.com/support/ads-application-notes/momentum-port-global-ground-or-differential

Thanks your comments are deeply useful.