Characterize and design the OMT

[csheehy]

Oct 25, 2016: Michigan visit summary (CDS)

From Jeff's email with my notes interspersed:

• Chris and I figured out how to install coax lines into the OMT
• Chris tested the OMT revealing performance that is inline with what is predicted for S11 and S21
  • we realized that the reflection can be as high as 10% in power for one OMT
  • this means that with two OMTs connected together we can get standing waves that are 2* sqrt(.1) = 0.6 i the E-filed, or about 40% in power, we saw this
  • simulations revealed that this is to be expected
  • Chris measured S11 and s21 and summed these in power
  • if there were no loss these would add to 100%
  • we see evidence for a dielectric loss in each OMT at the 5% level, with a systematic error bar that is at least a few percent *Chris looked up the specs on the coax we used (Micro Coax UT-085C-AL-TP-LL), we predict ~3-5% loss from this coax
  • All OMT’s behave the same, so we can repeatably assemble the OMTs
  • we noted that the vertical spacing between X- and Y- probes was larger than it needs to be and can be reduced in future versions

Jeff ran these sims in HFSS with the OMT open and the top and with two OMTs shorted together: QR_OTM_debug.pdf

I then put the OMTs on the VNA (fully calibrated) and meaured the S parameters. This is what I get compared to Jeff's sims:

Figure 1

Figure 2

Figure 3

Figure 3 implies significant loss, but S11+S22>1 also implies a large measurement systematic. Given this measurement systematic, it's not clear to me how much improvement we'll see if using a lower loss coax. The measurement of S11+s11 with S11 and S22 calibrated separately looked much cleaner and more sensible, showing ~0.5 dB loss.. I'm not sure what to make of that.

[csheehy]

Oct 25, 2016: Drawings from Remington

These drawings are supposed to reflect the OMT in the state it was when we measured it. omt.pdf

[slosar]

So, do you call these plots a success or not? I would have expected the match to be somewhat better, but I don't really know what to compare it against.

[csheehy]

We do not consider these a success yet. See the recent emails about identifying new lower loss coax. We are thinking that the discrepancy with sims in Figure 2 might result from having air instead of PTFE as the coax dialectric in HFSS. I am going to try to model the PTFE to see what happens. The source of the ugliness of Figure 3 is currently unknown, but given how different it looks with different VNA calibration methods, we're chalking it up to a measurement systematic. Best guess is ~5% +/- 3% loss, which could be almost entirely attributable to the coax.

[csheehy]

Dec 15, 2106: Including coax in HFSS model (CDS)

In Figure 2 of this posting, I showed the measured S21 compared to the simulated S21. Including the coax dialectric improves the agreement considerably.

Figure 1 - S21 including coax dialectric

I also tested whether the coax produces the expected loss by drawing a 1 ft section of Micro Coax UT-085C-AL-TP-LL in HFSS and comparing the predicted loss to what is listed in the data sheet. I had to add a new material, LDPTFE (low density), because HFSS only has standard teflon. To do this, I tuned the conductivity until the impedance was 50 ohms. I did the same for ULDPTFE because this is used for the UT-250C-ULL coax.

Figure 2 - simulated loss in UT-085C-AL-TP-LL coax

These are the loss/ft in dB from the data sheet: 0.5 GHz .134 1.0 GHz .190 5.0 GHz .431 10.0 GHz .617

The agreement is pretty good at low frequencies. Given Figs 1 and 2, I think we can be satisfied that our HFSS modeling is accurate. Given the high loss of the 0.085" diameter coax, we have decided to use the 0.25" OD UT-250C-ULL coax going forward.

[csheehy]

Dec 15, 2016: VLA OMT + transition HFSS modeling

After visiting Michigan, I began to design a transition to couple the OMT to a feed horn. We decided to base the transition on one designed for the VLA at 1-2 GHz. (Here is the paper: 05722981.pdf). Because of the dimension mismatch at the back of the transition relative to the OMT, however, I could not get good enough throughput. We probably could have tweaked the transition design to work, but it became clear a simple scaled version of the VLA OMT + VLA transition was both more compact and higher throughput than the existing OMT. We thus decided to go with this design.

The VLA design is quite fancy. There are a series of orthogonal "shorting blocks" instead of a single backshort. These reflect only the TE01 and TE10 modes, respectively, which allows the conductor height above the backshort be identical for each polarization. The TE11 mode propagates through the shorting blocks and is absorbed by some absorber placed at the back. I drew the VLA OMT + transition in HFSS and scaled it up by a factor 1.43.

Figure 1 shows S21 for the VLA design, including the shorting blocks and absorber. Figure 2 shows the same design but eliminating the blocks and absorber and drawing in a single backshort, splitting the difference between the optimal x and y coax heights. (I also decreased the coax spacing.) The main differences are that there is slightly different coupling for x and y without the shorting blocks, and that there is a narrow null in S21 at 0.83 GHz owing to the TE11 mode propagation without the absorber.

Figure 1 - VLA OMT + transition with blocks and absorber

Figure 2 - VLA OMT + transition without blocks and absorber

Figure 3 shows S21 for modes other than TE01 and TE10 (x and y), which just proves what the paper says, that it is TE11 causing the problem at 0.83 GHz. (TE11 is trans3 as labeled in the HFSS model, and I checked the mode vector pattern, it is indeed TE11.)

Figure 3 - higher order modes in the VLA sans blocks design

The design is much easier and cheaper to manufacture without the blocks and absorber, and it is only a very narrow frequency range affected by the higher order mode propagaion. Therefore, we've decided to proceed with the modified VLA design with a single backshort.

[csheehy]

Dec 15, 2016 - modified VLA + transition design

Following the previous posting, I detail the modifications made to the VLA OMT + transition design to make it work for BMX. The solidmodel can be found here: Full OMT model, including assemblies: directory, zip Only parts (to give to the shop): directory, zip

The overall weight is 32 lb. The modifications are:

• scale all dimensions up by factor 1.43 to make work at lower frequency
• change ridge thickness from 0.5"x1.43 = 0.715" to 0.75" for manufacturing ease
• increase ridge face length to 0.26" to accommodate 0.25" OD coax
• eliminate complicated blocks and absorber in favor of a single backshort, reduce coax spacing to 0.2" (center conductor is 0.08" OD).
• tune ridge spacing to maximize S21 - found 0.375" is optimal

The OMT and transitions ridges will be manufactured as once piece. Figure 1 shows the solidworks model of the assembly, cross sectioned to just show the OMT. I have also designed a simple clamp to hold the coax in place, as we found it could pull out easily as designed. (Only one clamp is shown, though there will be one for each polarization). Figure 2 shows the full design. Figure 3 is an alternate view of the full design without the upper outer walls, for visibility. Figure 4 is a cross section of the full design.

Figure 1 - cross section showing OMT only

Figure 2 - OMT + transition

Figure 2 - OMT + transition (walls hidden)

Figure 4 - OMT + transition cross section

Next, I exported the solidworks model to HFSS to make sure there were no transcription errors. Figure 5 shows S21. Figure 6 shows the predicted loss, including UT-250C-ULL coax and the aluminum ridges (but not the aluminum side walls or backshort). It also includes a 1/4" radius fillet on the inside corners of the ridge face to approximate how it will actually be machined. This makes no difference and is not included in the drawings.

Figure 5 - S21 (dB)

Figure 6 - loss (dB)

The difference in S21 between x and y polarizations could be reduced by reducing the center conductor spacing. It is at 0.2" now. The center conductor spacing is 0.0808" so I could probably reduce this to 0.1", but that would risk a short. 0.15" would probably be safe. However, I am pretty satisfied as it is.

Lastly, following the VLA paper, I ran a sweep of the ridge spacing and far ridge coax penetration length to assess the sensitivity to manufacturing tolerances. Figure 7 shows the ridge spacing sweep. Figure 8 shows the coax penetration sweep. Each line is a different frequency.

Figure 7 - S11, ridge spacing sweep

Figure 8 - S11, coax penetration sweep

The nominal ridge spacing is 0.375". S11 stays below -15 dB within +/- 0.01" of this target, which is quite doable. (The two lines that exceed -15 dB below 0.375" are for 1.5 GHz and 0.7 GHz which are out of band and the extreme low end of the band, respectively.)

The nominal coax penetration is 1.62" as measured from the inner face of the ridge. Judging from Figure 8, there is a +/- 0.07 tolerance on this to keep S11 < -15 dB. This is small, but again, probably realizable. However, I do see this is a possible source of non-repeatability if we're not careful.