Ku interferometer experiments

[Marcus D. Leech]

This is the result of my experiment today:

http://www.sbrac.org/files/ku-interf.png

This is with a pair of small dishes (55cm), two PLL-321S2 LNBFs, and a
common 27MHz clock, being provided by a 27MHz TCXO, low-pass filtered with
a 3-pole 28MHz low-pass, and 100-ohm series terminations towards the
LNBFs.

The measurement was made with a USRP B210, operating at 8.5Msps,
computing both total-power on the individual antennae, and the cross power,
via a conjugate-multiply between the two halves. The channels are
then integrated and recorded.

The astute will observe that the cross-power channel (shown as COS:(A*B)
in the chart) appears to contain no useful fringes, and only apparent random
noise.

This tells me that the LNBFs are not usefully coherent, even in the
presence of a shared 27MHz clock. This is either due to my clock being
of too-poor
quality for them (containing too much noise of its own), or the
loop-bandwidth on the PLL synthesizers being very large.

I don't currently have a usable oscilloscope, so I can't tell if my
clock is garbage, although my frequency counter is perfectly happy with
it. I'm
a bit stumped....

[Michiel Klaassen]

Use the unsquared power outpts

Michiel

--
--
You received this message because you are subscribed to the Google
Groups "Society of Amateur Radio Astronomers" group.
To post to this group, send email to sara...@googlegroups.com
To unsubscribe from this group, send email to
sara-list-...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/sara-list?hl=en
--- You received this message because you are subscribed to the Google Groups "Society of Amateur Radio Astronomers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sara-list+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Marcus D. Leech]

On 05/17/2015 01:43 PM, Michiel Klaassen wrote:

Use the unsquared power outpts

Michiel

???

[Marko Cebokli]

What was the observed source?

You can try to debug your system with an artificial source, a CFL lightbulb is
often good enough.

Marko Cebokli

[Marcus D. Leech]

On 05/17/2015 02:50 PM, Marko Cebokli wrote:
> What was the observed source?
>
> You can try to debug your system with an artificial source, a CFL lightbulb is
> often good enough.
>
> Marko Cebokli
>

The Sun. 55cm dishes, so Sun and maybe Moon are the only realistic
targets. Baseline was about 1.5m, so below the threshold where the Sun is
being resolved at this frequency.

The same back-end setup has been used on another, C-band, interferometer
with great success, but with LNAs "up front", instead of LNBFs. So I'm
tentatively blaming the LNBFs and/or their shared clock.

[Marko Cebokli]

You can try LNBFs only (no dish) with a CFL.
Just put the LNBFs 20..30cm apart, and move a CFL about a meter in front of
them, should produce fringes. This way, you can try more things faster, no
need to wait for the Sun to transit!

Probably, there is really some problem with your clock/locking.
Judging by the the big amplitude humps you got, you should have had nice
fringes there.

Marko Cebokli

[Marcus D. Leech]

On 05/17/2015 03:13 PM, Marko Cebokli wrote:
> You can try LNBFs only (no dish) with a CFL.
> Just put the LNBFs 20..30cm apart, and move a CFL about a meter in front of
> them, should produce fringes. This way, you can try more things faster, no
> need to wait for the Sun to transit!

Yes, I'll probably end up doing that.

>
> Probably, there is really some problem with your clock/locking.
> Judging by the the big amplitude humps you got, you should have had nice
> fringes there.

I agree--given the good detector outputs, fringes should have been
obvious and with good SNR.

I have no idea what the loop bandwidths are that are used on these
LNBFs, but they can't be all that wide, or satellite signals would have
problems due
to the increased phase noise.

[Paul Oxley]

Marcus

Most LBBFs are at least 500 MHz wide. The receiver tunes to specific transponders within this band. The intelsat types are even wider.

Paul

 Marcus D. Leech <patchv...@gmail.com>
To: sara...@googlegroups.com
Sent: Sunday, May 17, 2015 3:17 PM
Subject: Re: [SARA] Ku interferometer experiments
 

--
--
You received this message because you are subscribed to the Google
Groups "Society of Amateur Radio Astronomers" group.
To post to this group, send email to sara...@googlegroups.com
To unsubscribe from this group, send email to
sara-list-...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/sara-list?hl=en
---
You received this message because you are subscribed to the Google Groups "Society of Amateur Radio Astronomers" group.

To unsubscribe from this group and stop receiving emails from it, send an email to sara-list+unsub...@googlegroups.com.

[Marcus D. Leech]

On 05/17/2015 03:38 PM, Paul Oxley wrote:

Marcus

Most LBBFs are at least 500 MHz wide. The receiver tunes to specific transponders within this band. The intelsat types are even wider.

Paul

Yes, that's true.  But I'm talking about the loop bandwidth on the synthesizer they use.  To a first order, the smaller the loop bandwidth, the
  better the phase noise.

To unsubscribe from this group and stop receiving emails from it, send an email to sara-list+...@googlegroups.com.

[Nathan Towne]

Is it possible that the two data streams are not aligned? Have you
tried cross correlating them against each other to see if there is a
relative delay that puts them into alignment?

Nathan

[Marcus D. Leech]

On 05/17/2015 09:48 PM, Nathan Towne wrote:
> Is it possible that the two data streams are not aligned? Have you
> tried cross correlating them against each other to see if there is a
> relative delay that puts them into alignment?
> Nathan
>

(A) Not unless a bug has shown up very recently in UHD and the B210
hardware. It's *designed* for coherence.
(B) If there is any, it's small, which would just change the
phase-center on the sky--no different than having cabling that isn't
quite matched in length.

[Michiel Klaassen]

From your graphs; you used the squared outputs; A*A and B*B; and averaged them; so you lose phase.

Use the raw signlas A and B, and then do A*B and do some averaging.

I am trying to set up an amateur VLBI with data coming from several amateurs, so a large global system with high resolution can be made.

Can you send me the two raw data files (A and B) so I can do some analyses on them?

Michiel

[Marcus D. Leech]

On 05/18/2015 05:06 AM, Michiel Klaassen wrote:

From your graphs; you used the squared outputs; A*A and B*B; and averaged them; so you lose phase.

Use the raw signlas A and B, and then do A*B and do some averaging.

That's what the COS:(A*B) channel is.   The COS part of a complex-conjugate multiply.

I am trying to set up an amateur VLBI with data coming from several amateurs, so a large global system with high resolution can be made.

Can you send me the two raw data files (A and B) so I can do some analyses on them?

Michiel

Unfortunately, I have  no way of doing that with my current setup.  There's no way I can send 2 x 8.5Msps data streams over my 100Mbit
  ethernet, and I can't record the basebands on the Odroid XU3 that's doing the processing. 

[Wolfgang Herrmann]

Marcus,
From our experience, not all LNB work for this kind of exercise. Some have
substantial phase jitter when an external reference clock is used. So you
may just have been unfortunate by picking a type which does not work.

Regards
Wolfgang



-----Ursprüngliche Nachricht-----
Von: sara...@googlegroups.com [mailto:sara...@googlegroups.com] Im
Auftrag von Marko Cebokli
Gesendet: Sonntag, 17. Mai 2015 21:13
An: sara...@googlegroups.com
Betreff: Re: [SARA] Ku interferometer experiments

[Marcus D. Leech]

On 05/18/2015 12:50 PM, Wolfgang Herrmann wrote:
> Marcus,
> From our experience, not all LNB work for this kind of exercise. Some have
> substantial phase jitter when an external reference clock is used. So you
> may just have been unfortunate by picking a type which does not work.
>
> Regards
> Wolfgang
>

I'm using a type that uses the RDA3560, which uses a 27MHz reference
clock. The loop-filter is inside the chip, so ANY LNBF that uses this
chip will have the same properties. You had reported that an LNBF
with a 27MHz external ref apparently worked, so I assumed it was
the RDA3560 chip, but perhaps there's another chip that uses a 27MHz
reference. Perhaps you could talk to your guys to confirm which
chip is used in the 27MHz LNBF that they've had success with?

I'm currently improving my clock distribution, with better buffering,
and some gain, so that each unit will get about +5dBm clock signal.

[Marko Cebokli]

G4JNT reports very good results (for narrowband 10GHz HAM work, not interferometry) with RDA3560m (m?) based LNBs:

 

www.g4jnt.com/PLL_LNB_Tests.pdf

www.batc.org.uk/forum/viewtopic.php?t=3122

 

 

Some more stuff about PLL LNBs by HB9AFO:

 

http://www.hb9afo.ch/articles/pll-lnb/10ghz_pll-lnb.htm

 

I've got a couple of PLL LNBs with 25MHz reference. Haven't yet played much with them, just hooked one to the spectrum analyzer and observed the output of the analyzer's calibration comb generator. Did not look as clean as G4JNT's results.

 

Marko Cebokli

[Marcus D. Leech]

On 05/18/2015 01:41 PM, Marko Cebokli wrote:

G4JNT reports very good results (for narrowband 10GHz HAM work, not interferometry) with RDA3560m (m?) based LNBs:

 

www.g4jnt.com/PLL_LNB_Tests.pdf

www.batc.org.uk/forum/viewtopic.php?t=3122

 

 

Some more stuff about PLL LNBs by HB9AFO:

 

http://www.hb9afo.ch/articles/pll-lnb/10ghz_pll-lnb.htm

 

I've got a couple of PLL LNBs with 25MHz reference. Haven't yet played much with them, just hooked one to the spectrum analyzer and observed the output of the analyzer's calibration comb generator. Did not look as clean as G4JNT's results.

 

Marko Cebokli

 

 

Yes, thanks.  I've seen all of those, which is why I went down this path....

The only difference I can see is that they remove the 0-ohm resistor, which isolates the incoming clock signal from the crystal loading capacitor
  on that side.  I chose instead to remove the crystal, and just tie in to the crystal pad that corresponds,leaving the loading cap in place.

--

[Michiel Klaassen]

As you cannot verify the correct result from the A*B function, I am wondering if your processor can handle the task.

What speed does it have and can you decrease the sample and troughput rate.

Michiel

[Marcus D. Leech]

On 05/18/2015 04:58 PM, Michiel Klaassen wrote:

As you cannot verify the correct result from the A*B function, I am wondering if your processor can handle the task.

What speed does it have and can you decrease the sample and troughput rate.

Michiel

The computer is not the problem.   If it were, I'd be getting massive numbers of overruns.  I'm not.

I've tried at lower sample rates, same results.   Further, this *SAME* setup is used in another interferometer with great success.
  The issue appears to be lack of phase-coherence in the LNBFs, not the back-end receiving system.