Improving harmonic attenuation of the Ultimate3S used with relay-switched LPF kit
Summary
This investigation was prompted by the November 2016 QST magasine review of the Ultimate 3S kit. The review noted that measurements had indicated that when the relay-switched LPF kit is used (for the Ultimate3S to automatically sequence through 6 bands), the harmonic attenuation doesn't meet the FCC requirement that harmonics must be attenuated by at least 43dB, on all bands. (An individual LPF on the main Ultimate3S board is Ok). The FCC is the US regulator setting radio communication laws in the US. The aim of this investigation was to verify these findings and if true, to try to find a solution to improve the harmonic attenuation.
It is important to note that the worst harmonic we could find was -36dBc. The Ultimate3S outputs a QRP signal with a couple of hundred miliwatts. -36dBc on 200mW is a truly miniscule spurious signal. So it is not a cause for panic!
Conclusions
These tests used the Deluxe 6-band Ultimate 3S set - a boxed Ultimate 3S with relay-switched LPF kit and six LPFs, for 10, 15, 20, 30, 40 and 80m.
1) The findings of the ARRL QST reviewers were verified. VHF harmonics are the problem. The worst case harmonic occurred on the 10m band and was -36dBc.
2) A revised relay-switched PCB has been designed and manufactured. The highest frequency LPF (for example, 10m) should be plugged into slot 1 and this LPF is always in-circuit. When using lower frequencies it helps to further attenuate VHF harmonics. The ground-plane layout is also improved, and the earthing of the RF output is isolated through the final LPF. The result is a worst case harmonic of -47dBc; the average is under -60dBc and the average improvement in attenuation by this change was better than 15dB. This new PCB is being supplied in all relay-switched LPF boards shipped from the end of October 2016 onwards.
3) A retrofit solution for existing relay-switched LPF is also possible. This meets the -43dBc FCC requirement on all 6 bands tested. This improvement is obtained by:
- Use metal 25mm spacers instead of the supplied nylon spacers; or, solder a 25mm wire along each of the headers so that the ground plane of the main Ultimate3S board and the relay board is connected additionally at the 4 corners
- Connect your RF output coaxial cable to the PCB pads at the end of the relay-switched LPF PCB next to the 2 x 5-pin connector to the main U3S board.
- Solder a 56pF capacitor between RF output bus and ground at the far end of the relay PCB (opposite end to the 2 x 5-pin header)
- Order the LPFs in the slots as follows: 1: 80m, 2: 40m, 3: 30m, 4: 20m, 5: 15m, 0: 10m (main U3S PCB)
4) Alternatively soldering a 6m LPF kit permanently in place right at the RF output of the box, has excellent results!
The Root cause
A lot of different investigations were performed (see below). In all cases this is using a standard Deluxe 6-band Ultimate3S set, with one BS170 in the PA and 5V PA supply. The PA bias is adjusted for zero idle current as described on the PA bias study page. The investigations aimed to quantitatively test:
- What difference does the ordering of the LPFs in the slots make?
- What difference does it make which end of the relay board the RF output is taken from?
- PCB layout of the plug-in LPF filters and other design aspects
- PCB layout of the relay-switched LPF kit and other design aspects
- Capacitance between relay poles
- Improving grounding between the relay-switched LPF PCB and the main Ultimate3S PCB
These 25 experiments are all documented below. The LPF module PCB layout is fine. The relay-switched PCB layout leaves room for improvement but it too, is not the cause of the VHF harmonics inadequate attenuation. More grounding, different LPF ordering, etc all help but are not the main cause, either. The root cause of the problem is as follows.
The plug-in low pass filter kits use a pair of 4-pin header connectors at the input and output. At each 4-way connector, 2 pins are ground and 2 pins are RF. The impedance of a 10mm wire is 6nH. Two in parallel have an impedance of 3nH. At 30MHz, the reactance of a 3nH inductor is 0.5 ohms. This means that the ground-plane of the LPF module is raised 0.5 ohms above the ground-plane of the relay board. 0.5 ohms in a 50 ohm system means RF can leak past the filter. It won't be possible to beat 40dB of attenuation at 30MHz, and it gets worse as the frequency gets higher. This means the attenuation of the harmonics at VHF is impaired.
These 4-pin connectors are inexpensive, compact, and convenient. The use of coaxial connectors would avoid the problem but the kits would be much much more expensive, and also they'd be larger. The current size fits 5 side by side on the relay board. The relay board is the same size as the Ultimate3S board, which itself is the same size as the common 16 x 2 LCD module, so they all fit nicely in a "stack". Thousands of these LPF kits are in use around the world and a change would break compatibility of the whole modular system, it would be expensive and very inconvenient, all the prices would increase, so would the size and construction complexity.
Normally I say, don't treat the symptoms, treat the root cause. But this is one rare unusual case where we should prefer to treat the symptoms, not the root cause! The solution is a re-design of the relay-switched LPF PCB, such that the highest frequency LPF will always be inserted in slot "1" and this LPF will always be in-circuit. Any VHF spurii escaping past the lower frequency LPF in use, gets attenuated by this final LPF. Additionally the ground of the output connector is isolated now from the ground-plane of the relay board. This means that both the RF and the Ground signals must flow through the final LPF (e.g. 10m band LPF) which avoids also the problem of VHF spurii escaping past the LPF module. Finally the ground-plane layout of the relay-switched PCB was improved. There are jumpers on the board so that the old configuration is also possible - to provide as much flexibility as possible
The results are good - the worst case harmonic is reduced from the base case -36dBc to -47dBc after the change. The average harmonic is below 60dBc, which is an improvement of better than 15dB.
In this situation, treatment of the symptoms is successful, it results in greatly improved harmonic attenuation at VHF, with only a change to the PCB layout - the firmware, other components, size, cost, etc., everything remains the same!
Index of tests performed
Test 1: Base test with standard 6-band Deluxe U3S kit
Test 2: reversed LPF positions
Test 3: Coax connection moved to far end of relay board; LPF positions reversed
Test 4: Coax connection is at far end of relay board; LPF positions are back to normal
Test 5: Adding a ground wire between relay board and main board
Test 6: as test 5 but with RF out coax taken at opposite end
Test 7: An 11pF capacitor added at the "near end"
Test 8: Coax connection at the "near end", 56pF added across the "far end", conventional LPF positions
Test 9: A 56pF and 18nH inductor in series across the RF output coax
Test 10: 56pF capacitor and 18nH inductor at the "far end"
Test 11: ground wires added between PCBs at all four corners, still with 56pF
Test 12: ground wires only, all else is standard
Test 13: 56pF capacitor at far end only, no grounding between PCBs
Test 14: 56pF at "far end", one grounding wire only - a repeat of test 8
Test 15: 56pF capacitor at "far end", 4 ground wires, special v3.10aH firmware
Test 16: Removal of PCB traces on the LPF PCB
Test 17: Cut-tracks, tin-shielded 10m LPF harmonics
Test 18: LPF without header pins
Test 19: Standard LPF without header pins
Test 20: An additional 6m LPF at the output, always in circuit
Test 21: Additional 6m LPF at the output, with grounding wires between PCBs
Test 22: New PCB layout with 10m LPF always in-circuit. LPF positions; 1: 10m, 2: 15m, 3: 20m, 4: 30m, 5: 40m, 0: 80m; no grounding in corners
Test 23: New PCB layout with 10m LPF always in-circuit. Same order of LPFs as in test 22, but now grounding added
Test 24: New PCB layout with 10m LPF always in-circuit; ordering of LPFs reversed; no grounding in corners
Test 25: New PCB layout with 10m LPF always in-circuit; ordering of LPFs reversed; grounding in corners
Test 1: Base test with standard 6-band Deluxe U3S kit
The 6-band Deluxe U3S kit has bands 80, 40, 30, 20, 15 and 10m. The RG58 coax connection to the BNC connector on the back panel is soldered at the RF output on the relay board next to the 5 x 2-pin header that connects to the main U3S board. The 10m LPF is placed on the main U3S board (relay 0). The others are on the relay board, with the lowest frequency furthest from the RF in/out. So from left to right, relay 1 to 5, in this order: 80m, 40m, 30m, 20m, 15m.
Test result: harmonic suppression is not good enough for regulations in some countries (e.g. FCC in US).
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 22.2 | -61.92 | -68.7 | -53 | -35.9 | -38.32 | -54.95 | -35.9 | |
| 15m | 22.9 | -60.05 | -42.75 | -46.07 | -38.37 | -37.85 | -38.75 | -37.85 | |
| 20m | 21.67 | -49.92 | -39.02 | -47.84 | -42.99 | -47.17 | -38.97 | -38.97 | |
| 30m | 22.82 | -48.22 | -38.67 | -42.92 | -46.59 | -49.37 | -44.14 | -38.67 | |
| 40m | 21.75 | -52.85 | -43.82 | -49.4 | -43.8 | -52.3 | -47.35 | -43.8 | |
| 80m | 22.95 | -37.17 | -55 | -53.3 | -50.12 | -45.4 | -44.07 | -37.17 |
Test 2: reversed LPF positions
In this test the positions of the LPFs were reversed. 80m is on the main board (relay 0) and the bands 10m, 15m, 20m, 30m and 40m occupy relay positions 1 to 5 respectively (i.e. left to right across the relay board). Highest frequencies are therefore furthest from the coax.
Test result: The harmonic suppression on the lower bands (30, 40 and 80m) is improved, but on the higher bands (10, 15 and 20m) it is worse. The RF also has to travel back and forth along the length of the relay board, so this is a reasonable result. It just confirms the existing recommendation on the band ordering: put highest frequency bands closest to the PA output and coax (as in test 1 above).
Note also the spectrum analyser screenshot of 80m which shows a peak in the harmonics, at 135MHz. In the 40m wide view (0 to 200MHz) you can see a similar thing at around 110MHz.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 21.02 | -47.52 | -30.77 | -31.82 | -37.87 | -51.94 | -56.49 | -30.77 | |
| 15m | 20.97 | -55.94 | -38.22 | -41.97 | -34.84 | -40.89 | -55.92 | -34.84 | |
| 20m | 21.85 | -46.6 | -35.47 | -41.52 | -39.72 | -38.32 | -31.52 | -31.52 | |
| 30m | 22.62 | -50.09 | -40.37 | -45.29 | -47.19 | -50.94 | -47.64 | -40.37 | |
| 40m | 23.72 | -55.64 | -60.09 | -50.02 | -44.24 | -51.42 | -49.67 | -44.24 | |
| 80m | 23.65 | -38.77 | -69.9 | -70.67 | -68.75 | -66.52 | -60.5 | -38.77 |
Test 3: Coax connection moved to far end of relay board; LPF positions reversed
In this test, the coax connection to the BNC connector on the rear panel is soldered at the "far end" of the relay board, which is the left side of the relay board, the side furthest from the 5 x 2-pin inter-board header connector. The coax location is now next to the SMA socket footprint on the relay board. The LPF positions are still reversed, as in test 2 above.
Test result: The harmonics of the higher frequency bands are improved. This shows that the important thing is to have the highest band (10m) near to the coax connection, NOT necessarily nearest to the U3S PA output. Note again the VHF harmonics peaks on 80 and 40m wideband views.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 21.35 | -68.92 | -35.57 | -39.2 | -43.3 | -52.82 | -54.45 | -35.57 | |
| 15m | 20.97 | -68.34 | -55.44 | -44.54 | -39.67 | -44.24 | -58.27 | -39.67 | |
| 20m | 21.82 | -55.79 | -43.99 | -49.97 | -45.87 | -38.02 | -33.97 | -33.97 | |
| 30m | 22.57 | -55.52 | -45.29 | -51.12 | -53.97 | -57.97 | -41.09 | -41.09 | (11th harmonic is column 7) |
| 40m | 23.42 | -56.29 | -54.47 | -52.82 | -47.29 | -44.72 | -43.87 | -43.87 | (15th harmonic is column 7) |
| 80m | 23.22 | -38.04 | -67.64 | -69.92 | -68.79 | -66.59 | -60.89 | -38.04 |
Test 4: Coax connection is at far end of relay board; LPF positions are back to normal
Here the coax connection is kept at the "far end" of the relay board - by the SMA pads; the LPFs are put back in the conventional order: 10m on the main U3S board (relay 0), then 80, 40, 30, 20, 15m respectively left to right in relay positions 1 to 5 on the relay board.
Test result: Generally the results are worse than Test 3, ecept for low frequencies 40m and 80m; it supports the view that the highest frequency LPFs need to be closest to the coax output.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 21.35 | -60.07 | -45.67 | -50.47 | -38.82 | -42.75 | -46.15 | -38.82 | |
| 15m | 20.77 | -65.49 | -47.99 | -53.27 | -35.94 | -38.24 | -38.94 | -35.94 | |
| 20m | 21.7 | -56.4 | -44.3 | -49.9 | -47.77 | -51.17 | -35.32 | -35.32 | (8th harmonic is column 7) |
| 30m | 22.42 | -55.72 | -44.69 | -49.74 | -53.19 | -52.44 | -39.57 | -39.57 | (11th harmonic is column 7) |
| 40m | 22.25 | -55.1 | -56.4 | -57.27 | -52.55 | -60.52 | -54.82 | -52.55 | |
| 80m | 23.42 | -38.19 | -63.92 | -64.49 | -55.64 | -55.57 | -54.17 | -38.19 |
Test 5: Adding a ground wire between relay board and main board
Here a wire was added from the ground plane of the relay board to the ground plane of the main board. The extra wire was just soldered between the corner hole pads in the corners nearest the 100K trim pot on the main U3S control board. The LPF relay positions are "conventional". The coax is soldered at the SMA socket end. The test is therefore the same as Test 4 but with the added ground wire.
Test result: the harmonic level is reduced across the bands; on average by several dB.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 21.32 | -64.92 | -54.77 | -52.49 | -39.22 | -41.64 | -47.17 | -39.22 | |
| 15m | 22.52 | -62.89 | -44.74 | -50.19 | -43.29 | -39.59 | -39.72 | -39.59 | |
| 20m | 21.72 | -55.12 | -42.27 | -48.72 | -45.42 | -49.52 | -41.79 | -41.79 | |
| 30m | 20.8 | -53.4 | -41.92 | -48.1 | -49.82 | -51.97 | -41.17 | -41.17 | (10th harmonic is column 7) |
| 40m | 22.22 | -54.17 | -54.74 | -58.79 | -50.44 | -61.94 | -55.89 | -50.44 | |
| 80m | 23.55 | -38.85 | -64.3 | -65.57 | -60.45 | -56.85 | -54.05 | -38.85 |
Test 6: as test 5 but with RF out coax taken at opposite end
Here the LPF order is "reversed"; 10m is next to the coax connector in relay 1 (far left) on the relay board; then 15, 20, 30, 40m in Relay positions 2 to 5 respectively, and 80m furthest from the coax, at relay 0 on the main board. The test is the same as 5 but with this change.
Test result: These results are a little mixed. 10m has a surprisingly bad 4th harmonic at -34.3dBc. 20m is also worse. 15m and 30m are better. The mixed nature of this result leads me to no firm conclusion on this particular test.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 21.15 | -59.92 | -60.22 | -34.3 | -43.77 | -54.2 | -56.72 | -34.3 | |
| 15m | 20.97 | -63.54 | -48.79 | -52.52 | -48.67 | -44.02 | -57.69 | -44.02 | |
| 20m | 21.5 | -53.77 | -41.42 | -45.67 | -44.17 | -44.57 | -35.52 | -35.52 | (8th harmonic is column 7) |
| 30m | 22.45 | -54.57 | -43.72 | -50.37 | -51.12 | -53.92 | -43.3 | -43.3 | (11th harmonic is column 7) |
| 40m | 21.62 | -54.02 | -51.14 | -49.87 | -44.04 | -51.47 | -42.07 | -42.07 | (16th harmonic is column 7) |
| 80m | 23.47 | -38.52 | -65.47 | -67.07 | -66.49 | -65.12 | -47.44 | -38.52 | (Some higher harmonic is 7) |
Test 7: An 11pF capacitor added at the "near end"
An 11pF capacitor was added across the RF, at the "near end" which is the end by the 5 x 2-pin header. The coax connection is still at the "far end" by the SMA connector.
Test result: again a bit mixed. There is a marked improvement in the 10m result, but the other bands are mixed. So it is hard to draw conclusions from this test either.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 21.95 | -63.62 | -51.45 | -51.02 | -47.55 | -61.75 | -62.75 | -47.55 | |
| 15m | 22.65 | -61.47 | -45.22 | -65.25 | -39.35 | -35.82 | -40.17 | -35.82 | |
| 20m | 21.77 | -50.82 | -40.17 | -46.49 | -43.64 | -45.67 | -36.72 | -36.72 | |
| 30m | 22.37 | -50.97 | -40.22 | -45.82 | -48.94 | -50.22 | -38.12 | -38.12 | (10th harmonic is column 7) |
| 40m | 22.17 | -53.39 | -53.24 | -51.89 | -46.44 | -55.02 | -50.04 | -46.44 | |
| 80m | 23.57 | -39.74 | -58.62 | -58.24 | -49.42 | -49.69 | -47.67 | -39.74 |
Test 8: Coax connection at the "near end", 56pF added across the "far end", conventional LPF positions
This test has the coax connection back at the "near end" RF pads, next to the 5 x 2-pin header from the main board. The LPF positions are back in their conventional location, with the 10m LPF on the "main board", and the highest frequencies closest to the RF output coax connection; this has 80m, 40m, 30m, 20m and 15m LPFs in relay positions left to right on the board, i.e. positions 1 to 5 respectively. The test still uses an extra wire connection between the ground planes of the relay and control boards.
A 56pF capacitor was added across the RF and Ground at the "far end" i.e. next to the SMA connector. The logic (such as logic exists) is as follows. This 56pF capacitor is near the 80m LPF. The end capacitor in the 80m LPF is 470pF and so the 56pF capacitor should not affect it much. On the other hand, it will help attenuate the VHF harmonics a bit more. At the other extreme, the 10m LPF is on the main U3S control board. The length of the traces along the board from there to the 56pF capacitor is around 8 to 10cm. That has a reasonably significant inductance at 10m frequencies. That parasitic inductance and the 56pF capacitor will form a low-Q resonant circuit that will resonate somewhere above 100MHz and help to attack the troublesome harmonics up there. The LPFs between 80m and 10m progressively experience some kind of interpolated effect. So it should help everyone somewhat.
Note that the 80m LPF has shown consistently bad 2nd harmonic performance and there is probably a fault in this LPF somewhere that requires investigation. Later tests used a different 80m LPF which did not show this problem. So bear this in mind, when interpreting the 80m result.
Test result: This test result for the first time, gets all harmonics on all bands under -40dBc! It is the best result so far.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 22.47 | -65.99 | -50.17 | -59.79 | -45.97 | -44.07 | -65.89 | -44.07 | |
| 15m | 22.57 | -59.72 | -51.22 | -44.89 | -41.84 | -40.37 | -42.07 | -40.37 | |
| 20m | 20.6 | -51.15 | -41.7 | -50.77 | -44.42 | -43.4 | -44.47 | -41.7 | |
| 30m | 22.07 | -49.57 | -40.14 | -45.79 | -48.42 | -50.52 | -44.64 | -40.14 | |
| 40m | 22.25 | -54.02 | -50.07 | -52.55 | -47.15 | -55.75 | -50.27 | -47.15 | |
| 80m | 23.62 | -39.22 | -58.37 | -59.34 | -50.49 | -50.72 | -48.04 | -39.22 |
Test 9: A 56pF and 18nH inductor in series across the RF output coax
Here the LPFs are in conventional positions, the RF output coax is at the "near" end next to the 5 x 2-pin header. But the "low Q resonant circuit" has now been moved and put in parallel with the coax output. The L is made from a few turns of wire wrapped around a pencil, and measured on an LC-meter as 18nH. The combination of 18nH and 56pF will resonate at 159MHz, for example see this resonance calculator.
Test result: The 10m 6th harmonic (168MHz) doesn't like it, and to a lesser extent, neither does the 5th harmonic. Other than that it's not a bad result but doesn't improve on the previous test 8.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 22.72 | -65.99 | -54.42 | -58.92 | -41.62 | -38.04 | -66.04 | -38.04 | |
| 15m | 22.75 | -61.97 | -51.67 | -49.17 | -43.15 | -41.47 | -43.57 | -41.47 | |
| 20m | 21.25 | -50.35 | -40.07 | -44.2 | -42.82 | -50.15 | -43.72 | -40.07 | |
| 30m | 21.87 | -49.99 | -40.39 | -44.37 | -45.02 | -40.84 | -52.57 | -40.39 | |
| 40m | 22.42 | -54.12 | -50.49 | -52.97 | -46.42 | -54.97 | -47.29 | -46.42 | |
| 80m | 23.75 | -39.35 | -59.22 | -58.37 | -49.65 | -50.15 | -47.65 | -39.35 |
Test 10: 56pF capacitor and 18nH inductor at the "far end"
Here the capacitor-inductor series notch combination is moved to the "far end" next to the SMA connector pad.
Test result: not an improvement; most bands are a little worse. This makes sense because the 18nH capacitor will effectively be in series with the inductance of the PCB tracks along the length of the relay PCB. That will lower the resonance frequency as far th
e higher bands are concerned, reducing the use of this additional filter where we need it (at VHF, above 100MHz).
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 22.67 | -66.24 | -53.87 | -58.44 | -42.02 | -38.09 | -54.99 | -38.09 | |
| 15m | 22.7 | -59.8 | -57.05 | -46.65 | -45.35 | -36.5 | -39.42 | -36.5 | |
| 20m | 21.12 | -49.64 | -40.89 | -48.54 | -43.34 | -47.52 | -40.64 | -40.64 | |
| 30m | 22.22 | -49.62 | -40.14 | -45.47 | -48.72 | -43.07 | -48.94 | -40.14 | |
| 40m | 22.4 | -53.97 | -49.85 | -53.07 | -47.1 | -56.2 | -50.5 | -47.1 | |
| 80m | 23.52 | -39.04 | -58.49 | -58.47 | -49.37 | -49.94 | -47.62 | -39.04 |
Test 11: ground wires added between PCBs at all four corners, still with 56pF
Here the grounding connections between the PCBs has been improved by adding ground wire connections between the boards at all four corners. I soldered wires into the corner holes. The grounding could also be achieved by the use of metal 25mm hex spacers rather than the supplied nylon ones. The test is otherwise the same as test 8; i.e. 56pF at the "far end" (by the SMA connector) and the RF coax taken from the "near end" next to the 5 x 2-pin header.
Note also that for this test (and all subsequent tests) the 80m LPF was replaced with another available one here; so the poor 2nd harmonic attenuation problem is not apparent now in this and subsequent tests.
Test result: the harmonic attenuation is the best yet; it is better than test 8 - the clear indication is that the grounding between the PCBs helps.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 22.75 | -71.75 | -64.12 | -63.45 | -47 | -44.52 | -61.85 | -44.52 | |
| 15m | 22.82 | -69.17 | -62.59 | -62.09 | -50.27 | -46.57 | -46.69 | -46.57 | |
| 20m | 21.25 | -56.12 | -45.2 | -51.45 | -46 | -41.35 | -56.97 | -41.35 | |
| 30m | 22.27 | -53.77 | -44.14 | -49.07 | -51.74 | -53.84 | -46.39 | -44.14 | |
| 40m | 22.47 | -55.92 | -52.94 | -55.64 | -49.62 | -58.64 | -53.02 | -49.62 | |
| 80m | 23.6 | -48.5 | -64.7 | -59.6 | -50.97 | -51.5 | -49.97 | -48.5 | 80m LPF was replaced |
Test 12: ground wires only, all else is standard
Here there is no 56pF capacitor. Only the ground wires between the corner points of the PCB.
Test result: The attenuation is worse than Test 11, considerably. It is however, considerably better than the "standard" base configuration, see Test 1 above; except on the 20m band. So it seems that the 56pF capacitor really does cause a good improvement.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 22.57 | -71.59 | -47.29 | -58.94 | -46.27 | -46.37 | -62.57 | -46.27 | |
| 15m | 22.67 | -56.42 | -46.07 | -44.89 | -39.39 | -39.17 | -40.19 | -39.17 | |
| 20m | 21.1 | -46.47 | -37.35 | -43.95 | -41.3 | -42.22 | -35.62 | -35.62 | |
| 30m | 21.57 | -46.49 | -37.19 | -41.94 | -45.44 | -45.87 | -42.07 | -37.19 | |
| 40m | 22.05 | -53.57 | -45.92 | -48.82 | -43.4 | -52.5 | -47.52 | -43.4 | |
| 80m | 23.62 | -47.99 | -57.07 | -53.14 | -44.67 | -45.77 | -45.02 | -44.67 |
Test 13: 56pF capacitor at far end only, no grounding between PCBs
Here there is no grounding between PCBs, only the 56pF capacitor at the "far end" across the RF.
Test result: The results are certainly better than the base test (test 1 above) of the standard configuration, but are nowhere near as good as when the grounding wires are used.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 21.85 | -65.42 | -58.62 | -54.52 | -39.92 | -37.32 | -58.32 | -37.32 | |
| 15m | 22.72 | -67.92 | -50.14 | -55.04 | -47.77 | -42.17 | -39.79 | -39.79 | |
| 20m | 22 | -57.17 | -44.4 | -51.27 | -47.7 | -52.75 | -41.57 | -41.57 | (8th harmonic is column 7) |
| 30m | 22.55 | -54.42 | -43.17 | -49.25 | -52.12 | -54.4 | -42.05 | -42.05 | (10th harmonic is column 7) |
| 40m | 22.32 | -54.77 | -51.77 | -53.62 | -48.79 | -57.64 | -52.72 | -48.79 | |
| 80m | 23.4 | -47.82 | -64.72 | -59.25 | -50.47 | -50.77 | -49.57 | -47.82 |
Test 14: 56pF at "far end", one grounding wire only - a repeat of test 8
Test result: A not exactly the same result as test 8... the results are broadly similar but with some dB of difference, and particularly 20m is worse now. It just underlines the difficulties of this kind of RF measurements... any changes can affect the results; who knows, maybe toroid physical positions changed a little
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 22.5 | -70.97 | -57.87 | -59.62 | -46.27 | -44.7 | -69.2 | -44.7 | |
| 15m | 22.57 | -60.47 | -49.49 | -53.14 | -42.67 | -42.04 | -41.89 | -41.89 | |
| 20m | 21.05 | -50.2 | -40.3 | -46.97 | -42.5 | -37.1 | -45.07 | -37.1 | |
| 30m | 22.15 | -50.17 | -40.57 | -45.67 | -48.65 | -49.72 | -42.1 | -40.57 | (8th harmonic is column 7) |
| 40m | 22.35 | -55.25 | -49.12 | -52.4 | -46.85 | -55.72 | -50.42 | -46.85 | |
| 80m | 23.47 | -48.22 | -53.64 | -56.99 | -48.29 | -49.09 | -47.82 | -47.82 |
Test 15: 56pF capacitor at "far end", 4 ground wires, special v3.10aH firmware
This is the same as Test 11, but checking to see if a certain firmware change makes any difference. The firmware change in question, is to stop the relay control lines from being "floating". In the official v3.10a (and all firmware versions from v3.08 onwards) the relay control output signals are "tri-stated" when the relay is off, rather than setting them to +5V. This was needed, or at least thought to be needed, to get the 11-band modification to work properly. In this test that firmware change was undone, such that an "off" relay has the output at 5V rather than high-impedance (a.k.a. floating, tri-state).
Test result: the results are almost identical to Test 11, except for 20m which is worse - but something might have slightly changed regarding the 20m LPF. So the conclusion must be that this firmware change makes no difference.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 22.95 | -72.1 | -62.02 | -60.55 | -47.07 | -44.57 | -64.22 | -44.57 | |
| 15m | 22.3 | -70.8 | -65.9 | -61.07 | -49.6 | -46.85 | -48.27 | -46.85 | |
| 20m | 21.85 | -57.45 | -45.87 | -52.52 | -46.72 | -38.57 | -59.67 | -38.57 | |
| 30m | 22.72 | -54.29 | -43.59 | -49.07 | -51.89 | -53.04 | -46.49 | -43.59 | |
| 40m | 22.8 | -56 | -53 | -55.77 | -49.67 | -58.8 | -53.1 | -49.67 | |
| 80m | No 80m results recorded |
Test 16: Removal of PCB traces on the LPF PCB
This test is to check whether the layout of the PCB traces on the LPF module itself, result in VHF leaking through the filter. The LPF PCB contains positions for TWO capacitors at each of the four capacitance positions; this is to support some LF versions of the LPF in which two capacitors are paralleled to make up the correct value; it also allows the use of 2.5mm-spaced pin or 5mm-spaced pin capacitor types. There is a trace connecting the two parallel capacitor positions in each case, approximately 10mm long. There could be capacitance between this trace and the ground-plane on the other side of the PCB that might cause the leakage.
Additionally in order to check whether there is unwanted coupling between the inductors, the fourth of the tests installs a tin metal shield around the centre toroid to shield the inductors from each other.
Four tests were performed using the Spectrum Analyser with tracking generator function. The values in the table below are dB attenuation.
a) Un-modified LPF, as the base test for comparisons
b) Removal two end traces; un-grounded capacitor pin and inductors wires soldered directly to the pin
c) Removal of 2 centre traces connecting the parallel capacitor positions, by cutting track
d) Adding a grounded tin metal shield around the centre toroid
Test result: There was no meaningful difference in any case; the track capacitance and lack of shielding between inductors are therefore NOT the cause of the VHF leakage. The four photographs show the four test results; as you can see, the curves are practically identical.
| 28MHz | 56MHz | 84MHz | 112MHz | 140MHz | 168MHz | 196MHz | Test | |
| a) | -0.3 | -34.37 | -33.65 | -29.72 | -28.25 | -26.02 | -27.07 | Un-modified LPF |
| b) | -0.55 | -34.3 | -33.95 | -29.8 | -28.2 | -26 | -27.02 | Removed 2 end traces |
| c) | -0.85 | -35.02 | -34.17 | -30.07 | -28.47 | -26.27 | -26.32 | Removed 2 center traces |
| d) | -0.92 | -33.87 | -33.7 | -29.85 | -28.32 | -26.22 | -27.25 | Added shield around centre toroid |
Test 17: Cut-tracks, tin-shielded 10m LPF harmonics
This is just a test of the modified 10m LPF in the Ultimate3S, to check the harmonic levels.
Test result: as expected, the harmonic levels are not substantially different to an un-modified 10m LPF.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 23.05 | -71.92 | -64.87 | -62.27 | -45.62 | -44.47 | -61.47 | -44.47 |
Test 18: LPF without header pins
This test is to check comments by Alan G8LCO. He calculates that the header pins that are used on the LPF kit, to plug it into the U3S board or the relay board, will have an inductance of approximately 6nH. Two in parallel at each end of the PCB will be 3nH. This is about 0.5 ohms reactance at 30MHz, which will lift the ground-plane of the LPF above the ground-plane of the motherboard.... and 0.5 ohms in a 50 ohm system will be 40dB... the theory is that this is what causes the leakage of VHF around the filter.
The LPF in use is the one used in tests 16 and 17: it has the cut tracks and the shielded center toroid.
Here are two photos. On the left, this is my test fixture for testing filter performance. The board inside the die-cast aluminium box is a sawn-off piece of relay-switched LPF board, connected to two BNC connectors in the box. The behaviour closely replicates the actual relay-switched LPF board because it is an actual piece of the real PCB used as the socket for the LPF. On the right, is the set-up for this test, with two BNC connectors (as used in the Arduino shield kit and Dummy load kit) that are soldered as closely as possible directly to the LPF PCB, NOT using the pin headers.
Four tests were performed and the harmonics up to the 7th were measured.
a) standard LPF test in the usual test fixture.
b) BNC connectors directly soldered to the LPF PCB.
c) Using the pin header at the input, and direct BNC connection at the LPF output
d) Pin header input, direct BNC output, but with BNC connector touching the metal case
Test result: Connecting directly to the LPF without the pin headers makes a message difference at the 3rd harmonic and above. 30-40dB of improved attenuation! See also the spectrum analyser plots below. The left plot is over a range 0 to 200MHz and was used to measure the harmonic levels with the on-screen markers. The right plot is over the wider range 0 to 500MHz, to see what the UHF attenuation looks like. The best result is by taking in/out both directly at the pin headers. Touching the connector to the box does result in leakage.
So, pin headers are very useful because they allow people the flexibility to plug in different LPFs as they wish. But on the other hand they harm the attenuation significantly at 30MHz and up.
| 28MHz | 56MHz | 84MHz | 112MHz | 140MHz | 168MHz | 196MHz | Test | |
| a) | -0.22 | -32.97 | -33.25 | -29.1 | -27.75 | -25.75 | -26.7 | Standard LPF |
| b) | -0.52 | -35.92 | -72.65 | -69.87 | -56 | -54.12 | -50.32 | In/Out connected without header pins |
| c) | -0.4 | -36.1 | -62 | -57.47 | -50.32 | -49 | -48.85 | In through headers, out without header pins |
| d) | -0.5 | -35.57 | -46.92 | -42.55 | -40.77 | -37.8 | -38.3 | In through pins, out direct, BNC ground touching box |
Standard LPF configuration:
In/Out connected without header pins:
In through header pin, direct BNC connection at output:
In via header pin, out direct to BNC, but with BNC connector touching metal box:
Test 19: Standard LPF without header pins
This is just a repeat of test 18 but using a different, un-modified 10m LPF (remember, test 18 uses the modified LPF with the cut tracks and the shielded central toroid). The standard configuration wasn't measured this time. Here's a photo of the test arrangement in 19b) with the in/out of the LPF connected directly to the BNC connectors:
Test result: It's again an excellent result when the RF is connected directly to the LPF input/output tracks on the PCB, and NOT via the header pins. The attenuations are slightly different to the previous LPF used in test 18, which can just be just due to natural differences in the way the LPFs are constructed (e.g. component tolerances, inductor winding variations, etc). The attenuation of the higher harmonics is somewhat better than in Test 18. The conclusion is that the cut tracks, and the shield around the central toroid, are not necessary to improve the harmonic attenuation.
| 28MHz | 56MHz | 84MHz | 112MHz | 140MHz | 168MHz | 196MHz | Test | |
| a) | Standard LPF | |||||||
| b) | -1.82 | -33.3 | -67.05 | -64.67 | -59 | -62.07 | -57.05 | In/Out connected without header pins |
| c) | -1.72 | -30.97 | -76.2 | -68.37 | -59.3 | -57.6 | -59.8 | In through headers, out without header pins |
| d) | -1.72 | -30.8 | -48.9 | -44.82 | -43.22 | -40.32 | -40.97 | In through pins, out direct, BNC ground touching box |
In/out connected without header pins:
In through header pin, direct BNC connection at output:
In via header pin, out direct to BNC, but with BNC connector touching metal box:
Test 20: An additional 6m LPF at the output, always in circuit
In this test the effect of a 6m LPF is investigated; this 6m LPF output was soldered directly to the BNC connector in the rear panel of the box, the LPF input was soldered to the RG58 coax coming from the relay board. The connections were directly to the 6m LPF PCB, not via the pins - following the result from tests 18/19 above.
The Ultimate 3S was otherwise completely standard - with no wires in the corners connecting the two PCBs, and with no 56pF capacitor, or any other modification.
Test result: This produces an excellent result on 10m and 15m. The results on 20m and 30m are not much improved. This is because the worst harmonic in these cases is the 3rd harmonic, which will fall below the cut-off frequency of the 6m LPF so it won't be improved.
NOTE: in this test the vertical axis of the spectrum analyser screen has been increased from 80dB to 120dB (12 divisions instead of 8) so that the very low harmonic levels could be measured accurately.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 22.42 | -68.95 | -74.73 | -79.64 | -82.12 | -73.87 | -69.82 | -68.95 | |
| 15m | 23.36 | -68.54 | -53.58 | -78.07 | -59.99 | -59.99 | -67.31 | -53.58 | |
| 20m | 21.82 | -56.84 | -40.23 | -51.59 | -60.1 | -73.34 | -60.22 | -40.23 | |
| 30m | 22.53 | -50.95 | -38.73 | -42.1 | -46.41 | -54.44 | -61.68 | -38.73 | |
| 40m | 22.57 | -54.1 | -47.2 | -51.03 | -44.05 | -53.95 | -48.52 | -44.05 | |
| 80m | 23.28 | -48.33 | -57.51 | -53.2 | -44.39 | -46.26 | -47.35 | -44.39 |
Test 21: Additional 6m LPF at the output, with grounding wires between PCBs
This is a repeat of test 20, but with the four grounding wires added between the relay-switched filter PCB and the main U3S control PCB.
Test result: The level of harmonics is very low on 10m and 15m, and even the 3rd harmonics on 20m and 30m are low... the worst harmonic across all the bands is the 3rd harmonic on 30m, and that is at -44.28dBc which is better than the level required by any regulations.
NOTE: in this test the vertical axis of the spectrum analyser screen has been increased from 80dB to 120dB (12 divisions instead of 8) so that the very low harmonic levels could be measured accurately.
This test indicates that it would be beneficial to have the highest LPF always in circuit, and for the highest LPF to be used without the header pins. Provided the highest LPF hasn't got header pins, the lower frequency LPF can still keep their header pins, this retains the modular use of the Ultimate3S kits, with easy-to-change bands.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 22.08 | -72.7 | -82 | -81.59 | -74.61 | -69.14 | -84.06 | -69.14 | |
| 15m | 22.65 | -76.27 | -63.6 | -77.1 | -82.27 | -66.93 | -69.63 | -63.6 | |
| 20m | 21.75 | -65.81 | -47.51 | -57.75 | -67.05 | -77.85 | -77.81 | -47.51 | |
| 30m | 22.57 | -56.8 | -44.28 | -50.2 | -52.45 | -63.52 | -67.23 | -44.28 | |
| 40m | 22.12 | -55.64 | -53.39 | -56.05 | -50.13 | -58.83 | -53.17 | -50.13 | |
| 80m | 23.47 | -49.42 | -65.84 | -60.25 | -52.12 | -53.05 | -52.45 | -49.42 |
Test 22: New PCB layout with 10m LPF always in-circuit. LPF positions; 1: 10m, 2: 15m, 3: 20m, 4: 30m, 5: 40m, 0: 80m; no grounding in corners
A new relay-switched LPF PCB was designed in which the LPF in position 1 is always in-circuit. Therefore it helps attenuate any VHF harmonics which escape past the lower band LPFs when transmitting on the lower bands. The ground output of the final LPF leads to the SMA and RF output connection pads, but it is not connected to the main PCB ground-plane. Therefore the RF signal and ground both flow through the final LPF and there is no opportunity for any to leak past via the relay-board ground-plane. Additionally the layout of the relay PCB was improved, with a more continuous and extensive ground-plane on both sides, and via connections linking the two sides' ground-planes at frequent intervals.
The harmonic attenuation with this configuration comfortably meets FCC requirements (-43dBc) on all bands, without any changes in components, or any other change or modification, only the new PCB layout to put the final LPF always in-circuit.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 21.82 | -46.62 | -61.87 | -48.27 | -54.79 | -66.04 | -71.82 | -46.62 | |
| 15m | 22.75 | -70.4 | -66.05 | -69.3 | -52.22 | -50.32 | -59.27 | -50.32 | |
| 20m | 22.4 | -68.17 | -56.87 | -69.3 | -57.45 | -48.6 | -52.37 | -48.6 | 5th, 7th, 8th, 9th (others below noise) |
| 30m | 22.75 | -68.2 | -51 | -58.35 | -72.75 | -72.75 | -72.75 | -51 | |
| 40m | 23.82 | -58.84 | -72.57 | -63.47 | -51.49 | -67.24 | -72.37 | -51.49 | |
| 80m | 23.75 | -50.4 | -70.9 | -72.22 | -65.87 | -66.35 | -60.4 | -50.4 |
Test 23: New PCB layout with 10m LPF always in-circuit. Same order of LPFs as in test 22, but now grounding added
In this test grounding wires were added in the corners, between the main U3S PCB and the relay PCB. There is a big improvement on 15m but some other bands are slightly worse - overall the conclusion is that there is not in general a significant or consistent improvement in harmonic attenuation. This means that there is no need to change to using metal spacers instead of the usual nylon ones which minimise cost and weight.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 21.72 | -46.27 | -61.19 | -47.37 | -54.52 | -64.54 | -46.27 | ||
| 15m | 22.7 | -68.17 | -70.87 | -70.45 | -66.2 | -60.3 | -66.57 | -60.3 | |
| 20m | 22.35 | -58.47 | -58.47 | ||||||
| 30m | 22.72 | -66.77 | -50.29 | -59.84 | -50.29 | ||||
| 40m | 23.55 | -58.95 | -69.05 | -61.95 | -51.95 | -68.27 | -51.95 | ||
| 80m | 23.52 | -49.87 | -70.82 | -71.87 | -65.02 | -65.27 | -60.14 | -49.87 |
Test 24: New PCB layout with 10m LPF always in-circuit; ordering of LPFs reversed; no grounding in corners
Here the 10m LPF is still in position "1" which is always a requirement; but the order of the other LPFs has been reversed. Position 2: 80m, 3: 40m, 4: 30m, 5: 20m, 0: 15m. The result is again quite mixed and significantly worse on 80m. The recommendation should therefore be to have the lowest band on the main U3S board, and progressively higher frequencies going away from the feed point of the 2 x 5-pin header on the relay board.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 21.55 | -46.62 | -63.57 | -44.85 | -54.92 | -65.87 | -44.85 | ||
| 15m | 22.72 | -66.19 | -66.27 | -66.19 | |||||
| 20m | 22.52 | -68.27 | -57.24 | -61.92 | -53.32 | -51.19 | -58.22 | -51.19 | 7th, 8th, 9th, 10th |
| 30m | 22.72 | -67.39 | -51.14 | -58.19 | -51.14 | ||||
| 40m | 23.7 | -58.72 | -71.27 | -61.37 | -51.27 | -67.25 | -51.27 | ||
| 80m | 23.72 | -49.47 | -72.87 | -66.37 | -60.27 | -59.47 | -44.89 | -44.89 |
Test 25: New PCB layout with 10m LPF always in-circuit; ordering of LPFs reversed; grounding in corners
This test configuration is the same as test 24 but now there are extra ground connections between the relay-switched LPF board and the main Ultimate3S board. This is quite a reasonable performance configuration, 15m and 20m are very good, other bands seem slightly worse than test 22. Again the improvements are not really consistent across bands and so not enough to recommend this configuration and use metal spacers.
| Band | Fundamental dBm | 2nd dBc | 3rd dBc | 4th dBc | 5th dBc | 6th dBc | 7th dBc | Worst dBc | Comment |
| 10m | 21.72 | -46.17 | -62.39 | -49.22 | -54.29 | -64.42 | -46.17 | ||
| 15m | 22.72 | -67.24 | -67.24 | ||||||
| 20m | 22.52 | -58.22 | -58.22 | ||||||
| 30m | 22.72 | -66.12 | -50.29 | -59.82 | -50.29 | ||||
| 40m | 23.4 | -58.85 | -67.07 | -59.82 | -51.5 | -68.42 | -51.5 | ||
| 80m | 23.65 | -49.7 | -71.8 | -68.5 | -58.95 | -59.27 | -54.92 | -49.7 |
