not equal phase feeding of patch array elements, what would be S11?

[Terminator3]

I am reading paper about beam steering. I tried to simulate two patch array with parallel feeding. If patches have equal phase, S11 looks good. But if i add feeding line length for one of the patches, then there is very strange result in s11, it becomes worse, less than 10dB magnitude. Although current distribution seems good, just not inphase. Am i doing something wrong or unequal phase patch array cant be analyzed using S11 value? Also i read other paper with series fed patch array with fixed angle steering, their simulation result showed nice result. I tried series fed two patch array too, and again S11 becomes worse if i make longer feeding line between patches. But current distribution looks nice to me. Where i am wrong?

 

[volker@muehlhaus]

Obviously, there is coupling between the antennas and they influence each other's fields.

 

[Terminator3]

I calculated array factor for patch antenna for fixed steering angle array, with distance between patches (dx). By summing up each element along (x) axis with additional phase shift (phs). As F2(t,p,1,0,0) +F2(t,p,2,0,1) +F2(t,p,3,0,2) + etc.

F2(theta,phi,x,y,n)=Exp(-i*2*Pi*((x)*dx)*Sin(theta)*Cos(phi)) *Exp(-i*2*Pi*(y*dy)*Sin(theta)*Sin(phi)) *Exp(i**phs*Pi/180);

And surprisingly obtained resulting angle exactly as in papers i learned from. Also i used Sonnet lite to simulate small part of antenna to see something interesting. As true patch sizes was unknown, i started from single patch, and tuned it height to obtain center frequency. When i connected patches in series array, S11 center frequency moved up by 500MHz. Unsure how to interpret it, maybe it is field influence you mentioned. If it is true, then my S11 result is so different because of this. Also i removed autognd port and moved feeding line directly to box wall, and it seems to calculate faster, and results are similar. But now i have an error message:

Sonnet Warning EG2550:
Potential problem with circuit.
Bottom box-wall reference plane may extend beyond a discontinuity.

Here is small example of three elements series fed array:
I come to conclusion that series fed array beam angle is very sensitive to feeding frequency. While corporate feed gives in-phase feed for every patch even if center frequency is shifted. It is strange to me why most modern sensors with VCO ability use series fed (not as on my image, a little different), as they have modulation around 200MHz. For wide band sensors corporate seems more appropriate, for single frequency CW sensor series fed is good to make good power dividing for reduce side beams.

 

[volker@muehlhaus]

It is strange to me why most modern sensors with VCO ability use series fed (not as on my image, a little different), as they have modulation around 200MHz.

I get your point, but what is the relative bandwidth? An example would be radars in the 24GHz band that use 250MHz sweep bandwidth. That is only 1% relative bandwidth, so the feedline between the patch elements is almost constant phase across the band.

 

[Terminator3]

I tried simulation with different substrates, each time scaling design of series fed 3-patch array shown above for different Er and thikness. While doing this i found out where problem comes from. For substrate with same dielectric constant, results vary very much for different loss tangent.

For example, substrate Arlon AD430. Er=4.3 at X-band. If substrate have low loss tangent around 0.003 then center frequency is at right place, S11 magnitude is good at center frequency.
If loss tangent is high (> 0.01), then center frequency moves and S11 looks weird, actually it is several small dips around -10dB.

1.0mm; Er=4.3; loss tangent=0.003 - nice result, as expected
1.0mm ; Er=4.3; loss tangent=0.01 - center frequency moved, something to do with loss tangent.

 

[volker@muehlhaus]

Your results don't make sense to me. Increasing the loss tangent should increase the antenna bandwidth, but not move the center frequency. Are you sure you changed only the loss tangent?

 

[Terminator3]

Here is my file, i just changed loss tangent from 0.003 to 0.03. But frequency is pretty high - 10GHz.
If i change it back to 0.003 then result would be better.
Also if i delete upper patch and feed line i get very nice result for single patch even with tangent loss of 0.03.

 

[Terminator3]

Dear Volker. It seems i was wrong about center frequency. But i want to reproduce another problem and show you graphs and files. Btw i scaled beam steering antenna to some Taconic TLX with 0.5mm thickness and Er of 2.55, and result is very good. Now the problem is only with Er of 4.3.. I made inset fed around 0.4mm, and single patch shows S11 of -50dB. After adding next patch in series i got only -10dB. Maybe free space distance is too small, so patches interact as you mentioned.

 

[Terminator3]

Here is results i can't understand.
Series fed of two patches, spacing is near 1.0 freespace wavelenth to ensure smaller influence (but usually it is around 0.7 in most arrays i have seen). If i put quarterwave stub in second series feeding line, there is nice dip in S11 around -50dB. But when i remove this stub, results look weird. Influence of second patch is really so big or something wrong with my setup? Files attached.

 

[Terminator3]

As volker@muehlhaus said,

Increasing the loss tangent should increase the antenna bandwidth, but not move the center frequency.

it is true. I was erroneously interpreted some graph, as with increased bandwidth there also was some other dips in S11, while main dip remained.

The most strange thing for me now is rapid change in S11 graph when i add second patch antenna. It can be seen on the images in my previous post. I tried to put patches very far away, so i am doubt any significant mutual coupling exists. Previously i thought that after designing single patch i can put them in an array, and only feeding network provide additional losses, while phasing for some angle will add gain. Can anyone explain why adding additional patch gives such big difference?

 

[volker@muehlhaus]

I tried to put patches very far away, so i am doubt any significant mutual coupling exists.

I checked that by a 2-port simulation, and the field coupling between patches is around -25dB, so this should NOT cause any trouble.

From what I see in simulation, the issue results from connecting the patches by the line. Somehow that line isn't designed as you wanted it.