Superband update

[John Williams]

Early Adopters et.al.

Due to the stellar work of Joe Price WA9CGZ, we have some data that we
can discuss. I want to thank Joe for this help. The use of Superband is
not on Joe's plans, so tearing up his current radio to test this was a
huge gift of time and talent to our community.

Early testing of the 60/40/30/20 boards shows that the output filter
causes a strange interaction with the bandpass. Thus this was removed in
all subsequent tests. We removed the output transformer which helped as
well.

First of all, the 160/80/75 combo passed all tests. Joe had to change
the T1/T2 core to a BN43-2403. Joe suggests we change all of the boards
to that core, although it is not as significant on the higher
frequencies. For the all band Megaband design, we should us the
BN43-2402. He also used a T50-2 core for a couple of the inductors so
that all of the windings would fit.

Joe also tried a dual core BN43-202 similar to the funkamateur 5W
design. Results, in my opinion were not conclusive on that change. I
will let Joe comment on that.

We should place a 3db attenuator in the R10/R11/R12 spots to reduce
input power to the amp to keep it more linear and prevent over-driving
the amp. The values to use are 2 270 ohm to ground and 1 18ohm in series.

The 60/40/30/20 design results were not as successful. On the 1.1 board,
40 and 20 meet spurious goals. 60 does not suppress the second harmonic
enough to pass the -43db goal. 30M just meets that goal with no margin.
At this point, I suggest that you do not run the 60 or 30 bands without
some sort of external filter or an amplifier that has its own set of
LPFs. I would expect the 1.2 boards to have similar results.

IMD tests were fine.

The attempt to change the filter values was a complete failure, so I now
suggest returning the values to the initial ones. Joe noted that the
filter fall-off was not very dramatic, and opined that perhaps board
layout is causing some stray capacitance and/or coupling to the other
filter that is causing the fall off to not meet the demands of removing
the 2nd harmonic. He further notes that when he grounds the two large
pins of the BNC connector directly to the back-plane ground, he sees
some positive impact.

Thus, I will proceed with V1.3 of the board.

My list of changes is as follows -

Remove output filter comprised of L9/L11 and associated caps.
Move L10 to open board space to free up congestion near the regulator
and mounting hole.
Remove ground jumper and output transformer.
Add back in the protection diode on the RX path.
Add ground plane shielding on the top and bottom of the board under the
BNC connector. Move under-board via.
Shorten vias, if possible to reduce stray capacitance around filter area.
Populate attenuator with 3db pad

Hopefully this will give is a better trailing slope on the filter curves
and resolve the 60/30 harmonic. If this does not succeed, we will not go
any further.

Duane Brantley is testing the 17/15/12/10 board so that will be fun,
given the recent conversations on the 31.75MHz spur. This testing will
need the oscillator change.

My thoughts are that the superband is a moderate success. However, I am
beginning to view this current board as a prototype for HL v1.21 only,
and do not see it as a long term product. Given the discussions on v2,
we are likely going to move the PA to the main board, and use pluggable
cards for the filters. A smaller one for superband and a larger one with
the Megaband filters. While this is not yet firm, I am leaning in this
direction. The PCI-E connector is a novel approach, but I think
maintaining compatibility to the current Superband boards is not a good
plan.

So, my suggestion to other potential builders is that if you want to
build one for V1.21 Hermes, feel free. But do not expect that we will
provide a long term plan to allow the boards to be used in future builds.

Note that the Megaband uses a more traditional LPF filter design that
should not be as critical as these bandpass filters with respect to
reproducability. There is some concern that the Superband bandpass
filters may not yield consistent results from build to build. I share
this concern.

John - AC9HY

[Steve Haynal]

Hi John,

Thank you and Joe for all the hard work you've put into this. I suspect most DIY amplifiers don't go though testing as rigorous as you did.

Another option to reduce input power to the amp would be to adjust Rset (R1) on the Hermes-Lite main board. This adjusts the maximum current used with its power amp. Perhaps this can be adjustable in next version.

Here is an idea I've been toying around with for some time now: use a Z-match or perhaps two in series as a simplified final filtering for a 5W or less PA. I often use this Z-match http://emtech.steadynet.com/ with a long wire and have been impressed by how selective it is. I can turn the RX LNA gain up higher than with my dipole before I see clipping. The design only uses a single fixed inductor, two if two would be needed to suppress harmonics. (I don't like winding toroids or having many relays.) The poly film variable capacitors could be replaced by a switched array of parallel capacitors. You would also get an antenna tuner for free. The tuning could happen automatically as you change the frequency of the radio via noise measurement or simple SWR measurement. What I wonder about is if this design would be able to suppress harmonics enough, and if the the switched array of parallel capacitors would do the job. Has anyone seen anything along these lines?

73,

Steve

KF7O

[Zl2APV]

Some excellent Z-Match info at this site ...
http://users.tpg.com.au/users/ldbutler/ZCurves.htm

73 Graeme ZL2APV

[Steve Haynal]

Hi John and Joe,

I was looking at the filters and they are like the Peaberry. Why do you think they work (or do they really work) with the Peaberry? Is the signal from the Hermes-Lite just not as clean as the one from the Peaberry? If that is the case, there are some parameters in the RTL that can be changed. One tries to balance these parameters to provide what is needed but conserve resources. It may be that the signal can be made a bit cleaner at the expense of resources.

The softrock also supports superbands but with a shared low pass filter for RX and TX. Do you think this would help? I'm not talking about for the long run, but for my build if I hand mod with wires the filter to match the softrocks.

73,

Steve

KF7O

[John Williams]

Steve,

Addresssing both your notes...

Peaberry uses a push-pull amplifier that does not have even harmonics (or greatly reduced). Thus, the second harmonic on 60 or 30 is not an issue with the filter slope.

Softrock uses a shared LPF, and requires a external filter on 40 to meet FCC specs for the 40/30/20 build. Same thing on other 3 band designs, the lower freq band needs an external filter because it does not filter out the third harmonic adequately. They only support 3 bands, I was trying to do 4 bands. We could solve this with a more complex push pull design amplifier and keep the superband filters. I like the current PA with it's very low parts count.

Your board should work just fine on 20/17/15 (we still need to check 17 second harmonic) and be almost ok on 30. When I get my stuff back from Joe, I plan to make a few more changes and see if it can meet the -43db on 30M. Still a few tricks in the bag.

But longer term, we need to think about the timeline of the v1.21 and how we plan to address v2, and whether the PCI-E is a good long term interface design. I prefer stacked, with vertical interconnects.

On the LPF size issue, I think you are seeing LPFs for much larger amps, that take larger components. I think that splitting the megaband, or shrinking it by removing the PA will meet our needs nicely. I can do the Rx in 40x40 I think. The Tx can be done in 90x40 or so. So, envision a 100x100 mainboard, with a 40x40 plugged into one side, and a 40x90 plugged into the other, with the PA in the open area (roughly 60x50) all sliding into a small 100mm wide box, with connectors sliding thru holes in the back, and status lights in the front...for those that like them. Pretty compact. If you don't want the fancy filter boards, just plug in a single band pair filter "chicklet" that does not require relays.

I am familiar with Mobo, and looked at that design closely.

John

--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Steve Haynal]

Hi John,

I was thinking that the single filter set on a RXTX ensemble gives you two bands, and since your superband board has a relay for two filter sets, we could get 4 bands using 2 RXTX ensemble low pass filter sets. The stronger even harmonics of the PA may still be an issue.

Heat may be a problem if the PA is on the same board as or even close to the oscillator. We've already had issues with heat. It would be nice to use the aluminum enclosure as the PA heatsink somehow. I have several 100x80 enclosures that I would love to use. They are only 2.6 cm tall though. How much height would you like? One could use a 3D printer to make custom front and back plates for these enclosures.

73,

Steve

KF7O