Hermes-Lite 2beta3 RX IMD

[in3otd]

Hello,
I've done the usual IMD measurements on the H-Lv2b3; I'm a bit disappointed with the results since they are worse than the v2b2... after having pushed Steve to change the FPGA pins used for the RX data I was hoping to see at least a little improvement.

Anyway, here is a comparison of the IM2 and IM3 at 19 dB of RxPGA gain for the various H-L versions

for the H-Lv2b3 I've measured enabling and disabling the full duplex in Quisk but then I realized this does not actually change the actual HW mode.

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Claudio,

Thanks as always for the measurements. Do you consider this difference to be large enough that component variation can't account for everything? I am having a hard time understanding how much we should expect variations in the FPGA and AD9866 to affect these measurements.

If not component variations, what about discrete part selection by Elecrow? Perhaps some of the decoupling capacitors in this area are not of the same quality as in your beta2 build. Is this something you can check?

I would be very surprised if the changes in routing created this problem. The routing for the RX lines looks very similar to beta2 but with a better choice of FPGA pins in terms of capacitance.

The change of pins supports true half duplex. I will have to enable that in the firmware to see if that improves RX. I'd like to have software HDX/FDX switch really reset the AD9866 and change modes. We may be able to do this fast enough to allow FDX during TX for pure signal, etc., but half duplex during RX for best RX.

73,

Steve

KF7O

[in3otd]

Hello Steve,
I do not really know how much components variations can affect the IMD; my current hypothesis is that the IMD is due partly to crosstalk internal to the AD9866 and partly to some contributions from "something else" coming the PCB traces. I don't know which contribution is more important or even if different batches of AD9866 behave the same. I would tend to rule out decoupling issues, I did a few experiments with decoupling on the H-Lv2b2 and didn't see a difference.
We already saw that full-duplex made the IMD performance quite worse on the H-Lv1, so having the possibility to dynamically switch between full and half duplex would surely be the best solution.

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Claudio,

Does adding a little noise as one would experience in the real world have any effect?

I think all of the beta3 testers have a HL1 or beta2. Would it be possible to convert your scripts to use a HL as the signal source, changing amplitude via volume and TX gain in Quisk? This would allow us to measure all 5 units. I think you provided some scripts in the past and I'll have to go and take another look.

73,

Steve

KF7O

[in3otd]

Hello Steve.
I'd expect adding some noise will help, as seen in the "incidental dithering" experiments done some time ago, which only confirmed what was seen before by others. So, again, the IMD may not really be an issue in practice but, well, I wanted to see some improvements, hi.

I thin it should be possible to use another H-L as source, this was what I had in mind also for the final testing if/when the H-L will go to the production stage. Need to check how one can control multiple radios with Quisk, I've never tried that before.

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Claudio,

There is a half-duplex beta3 firmware here. Github is also up to date. It doesn't allow switching between half and full duplex yet. It would be interesting to see if the IMD numbers improve with half-duplex.

73,

Steve

KF7O

On Tuesday, August 15, 2017 at 9:32:10 AM UTC-7, in3otd wrote:

[Steve Haynal]

Hi Claudio,

Please remind me again if the HL1 test in the IMD graphs with the 20160220 TxDAC firmware was full or half duplex, and with CVA9, CV or SDK.

73,

Steve

KF7O

[in3otd]

Hello Steve,

here are the IM measurements with the latest HD FW, with also the curves from the first (full-duplex) FW for comparison:

quite strange, IM2 seems worse, IM3 improved, instead.
Anyway, I have to do more experiments, I have the impression the results are a bit different when the board is in the enclosure than when placed on the bench, without an enclosure.

The measurements on the H-Lv1 in the post above were in half-duplex, with a CV A9.

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Claudio,

Thanks for the measurements. Maybe a shield over the AD9866 will help here. There is no commercial aluminum shield speced, but there are ground holes about on the four corners near the AD9866 which could be used to build some sort of shield.

I'd love to be able to run these tests using two HL2s here.

73,

Steve

KF7O

[John Laur]

I had always been puzzled why most of the open hardware SDR receivers (including all of the HPSDR lineup) were never designed with cans over the sensitive RF areas -- maybe because this is usually custom (but there are standard parts too) 

Some experiments with the hermes/angelia/orion boards with copper foil tape shielding by some very dedicated individuals has yielded measurable improvements, but I know on the AD9866 heat is also a concern. If you did want to experiment with an RF can, it might also be possible to have it double as a heat spreader. It's pretty late stage to consider adding something that dictates layout so much, though...

73, John K5IT

--
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+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Steve Haynal]

Hi John,

The KiwiSDR has a shell. I was thinking if it is helpful to shield the AD9866, then we can make a small custom 2 layer PCB board that lines up with the ground through holes around the AD9866. The board would be just a solid ground plane. We could expose a pad where the AD9866 is and with a little thermal grease also use it for thermal dissipation. Relatively thicker copper wire could be used to secure and connect the PCB to ground. I'm not sure how important the sides of the can are, but don't think they matter too much at the frequencies of concern, maybe provide a bit shorter path to ground. This may actually be less expensive than a commercial can.

73,

Steve

KF7O

[Dani EA4GPZ]

El 14/09/17 a las 08:50, Steve Haynal escribió:

> Hi John,
>
> The KiwiSDR has a shell. I was thinking if it is helpful to shield the
> AD9866, then we can make a small custom 2 layer PCB board that lines up
> with the ground through holes around the AD9866. The board would be just
> a solid ground plane. We could expose a pad where the AD9866 is and with
> a little thermal grease also use it for thermal dissipation. Relatively
> thicker copper wire could be used to secure and connect the PCB to
> ground. I'm not sure how important the sides of the can are, but don't
> think they matter too much at the frequencies of concern, maybe provide
> a bit shorter path to ground. This may actually be less expensive than a
> commercial can.

Hi Steve,

Your suggestion reminded me of the QRP Labs OCXO:

http://www.qrp-labs.com/ocxokit

It is a low cost OCXO where the oven is built by soldering together
pieces of PCB. The same construction method could serve to build a low
cost RF can.

73,

Dani.

[Steve Haynal]

Hi Dani,

Thanks for the link. It is a good reference.

73,

Steve

KF7O