Hackrf self induced RFI

[Michiel Klaassen]

Hi All,

I did some more experiments with the Hackrf SDR.

When you just connect a 50 Ohm stub onto the input terminal, then you would not expect to receive anything.

Well that is not the case; you do get load signals from somewhere.

Ferrite clamps and Googling did not help, opening the box did reveal things.

It shows a horror layout; control tracks crossing under the input SMA connector. Also LEDs and buttons trscks running very close onto the RF input SMA track.

All these lines come directly from the high speed processor and carry the processor operating clock signals and harmonics.

You can check that also by connecting a short SMA connected coax with at the end the brading removed and the center wire cut and not bare from the isolation.

Move this 'antenna probe' over the board and watch your SDR# display. In this way you can see the amplitude of the RFI increase when hovering the culprit track.

In my opinion this layout design was made automatically and with no human brain present.

So this is how a company can ruin its own product.

So how can this be improved.

Well I removed the headers in the LED tracks and soldered .1uf smd grounded capacitors on every line. 

For the lines under passing the input track, I loaded them with grounded isolated 4X4cm copper foil.

The results were much better, but still not solved completely.

In my software I now use a zap mask; nulling the constant frequency spikes from the own induced RFI. 

Regards,

Michiel

parac.eu

[Paul Oxley]

Michiel

A resistor at room temperature will develop a signal that is equal to its temperature.

Paul

--
--
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.
To view this discussion on the web visit https://groups.google.com/d/msgid/sara-list/CAABBjRDiOtAM%3DdUDaS%3DDa%2Buje1yGfdzgf0Z_YvnLfU8sqa876g%40mail.gmail.com.

[Michiel Klaassen]

Hi Paul,

I am talking about RFI.

https://en.wikipedia.org/wiki/Electromagnetic_interference

ITU definition;

"The effect of unwanted energy due to one or a combination of emissions, radiations, or inductions upon reception in a radiocommunication system, manifested by any performance degradation, misinterpretation, or loss of information which could be extracted in the absence of such unwanted energy".

It is the enemy of all of us in this group.

To block all external signals, the antenna input was terminated by the nominal impedance, so now all displayed resulting spike signals must come from within.

I hope this helps to understand.

Regards

Michiel

parac.eu

Op zo 28 feb. 2021 om 17:10 schreef Paul Oxley <oxl...@att.net>:

To view this discussion on the web visit https://groups.google.com/d/msgid/sara-list/1076292033.1288621.1614532198713%40mail.yahoo.com.

[fasleitung3]

Hi Michiel,

we had investigated a number of SDRs for the presence of spurious lines. It seems that all SDR suffer from this to a larger and lesser extent when operated at their sensitivity limit. Our testing was done with a -150dBm/Hz white noise at the input and gain setting to maximum. We also found, that the Hackrf Clone was worse than the original. A way to avoid this is to have sufficient gain in the LNA chain so that the spurious lines become negliable compared to the background noise.

Best regards,

Wolfgang

[Marcus D. Leech]

On 03/01/2021 07:16 AM, 'fasleitung3' via Society of Amateur Radio Astronomers wrote:

Hi Michiel,

we had investigated a number of SDRs for the presence of spurious lines. It seems that all SDR suffer from this to a larger and lesser extent when operated at their sensitivity limit. Our testing was done with a -150dBm/Hz white noise at the input and gain setting to maximum. We also found, that the Hackrf Clone was worse than the original. A way to avoid this is to have sufficient gain in the LNA chain so that the spurious lines become negliable compared to the background noise.

Best regards,

Wolfgang

HackRF was designed by someone with no previous RF experience, so its performance in a number of dimensions reflects this.

The approach of adding enough LNA gain ahead of the device is what some of us refer to as "swamping".  You want the sky noise
  to "swamp" any internal receiver artifacts by a goodly margin.

The fact is that it's nearly impossible to avoid spurs in a modern receiver architecture that has digital signals in it--at least not at pricing
  levels that are tolerable.  Even very-expensive laboratory spectrum analyzers have spurs, and the control system knows where they
  are and maps them out.

To view this discussion on the web visit https://groups.google.com/d/msgid/sara-list/f1acaf879e119f0822fa89f97b8d1d9daf08d0e7.camel%40googlemail.com.

[Michiel Klaassen]

Hi All,

I think as a principle; the manufacturere should provide a list of its own spurious frequencies with signal levels when selling their SDR.

Because on those frequencies the unit does not perform as stated.

It is a bit strange; when a second unit is interfering a first one, then it does not comply to the FCC rules.

If the unit is interfering itself, then this is allowed?

@Frans; If you look at the spectrum with a radio amateur cap on it lookes like that, but if you put on your astronomy hat and integrate 50 seconds then it lookes like fig1 and fig2.

You see that the +2MHz feature is also visible just like in your screen dump.

The -4MHz feature is caused by the extra internal x-tal board option.

So, for spectrum measurements this can be a problem. For pulsar/FRB measurements perhaps less problematic as I have shown on the Element forum. 

@Wolfgag and Marcus; Yes, I used 'swamping' before, but I called it 'blasting through'. 

With a pulsar simulator setup I tried several Hackrf settings and for me the best settings for this test were 

gain_l=40dB, gain_g=22dB, a=1, powerlevel about -56dB(Sharp#). It took a lot of time to find out.

Regards,

Michiel

parac.eu

Op ma 1 mrt. 2021 om 15:18 schreef Marcus D. Leech <patchv...@gmail.com>:

To view this discussion on the web visit https://groups.google.com/d/msgid/sara-list/603D05B1.3020807%40gmail.com.

[Lamar Owen]

On 3/3/21 6:20 AM, Michiel Klaassen wrote:
> I think as a principle; the manufacturere should provide a list of its
> own spurious frequencies with signal levels when selling their SDR.
> Because on those frequencies the unit does not perform as stated.

Hi Michiel,

First, thanks for bringing this particular issue out.  We are using an
AirSpy R2 here for HI line spectroscopy, and wouldn't you know it that
unit has a spur at exactly 1.4GHz?  I haven't added gain ahead of the
unit as yet, but this is the standard way of dealing with such things,
as others have already mentioned.  The bigger problem is the non-flat
noise floor, but that can be dealt with.

Radio astronomy requires more careful design and measurements than other
radio disciplines.

These inexpensive SDR units typically originated as hobby projects
without the resources to do a full characterization.  The R&D to meet
the specifications of radio astronomy is quite expensive; the much
larger amateur radio and SDR experimenter communities aren't nearly as
concerned about some of these issues, and so to keep the cost down for
hobbyist affordability compromises are made.  So, yes, manufacturers
'should' do this, but would the customer be willing to pay for those R&D
costs to provide this information?

High-end equipment gets a whole package of test results shipped with it;
none of my AirSpy R2, AirSpy Mini, SDRPlay 1, SDRPlay 1A, and even my
original USRPs with the old DBS-RX module at much higher cost came with
such detailed test data; test instruments good enough to make such
measurements (like the R&S FSU-26 I have in our operations center here)
and the continued up-to-date calibrations are EXPENSIVE.

> It is a bit strange; when a second unit is interfering a first one,
> then it does not comply to the FCC rules.
> If the unit is interfering itself, then this is allowed?

Yep, it sounds strange, but self-interference is not regulated for the
most part.  The idea is that the manufacturer has economic incentive to
prevent this from happening to the degree necessary for the desired
customer demographic.

[Marcus D Leech]

These inexpensive SDR units typically originated as hobby projects without the resources to do a full characterization.  The R&D to meet the specifications of radio astronomy is quite expensive; the much larger amateur radio and SDR experimenter communities aren't nearly as concerned about some of these issues, and so to keep the cost down for hobbyist affordability compromises are made.  So, yes, manufacturers 'should' do this, but would the customer be willing to pay for those R&D costs to provide this

I’ll note that later USRPs have performance data published. 

https://files.ettus.com/performance_data/

[Lamar Owen]

On 3/3/21 9:45 AM, Marcus D Leech wrote:
>
>>
>> ... So, yes, manufacturers 'should' do this, but would the customer

>> be willing to pay for those R&D costs to provide this
> I’ll note that later USRPs have performance data published.
>
> https://files.ettus.com/performance_data/

> <https://files.ettus.com/performance_data/>

Side effect of Nationl Instruments' ownership perhaps.... good to know,
thanks!

[Marcus D. Leech]

On 03/03/2021 11:02 AM, Lamar Owen wrote:
> I’ll note that later USRPs have performance data published.
>>
>> https://files.ettus.com/performance_data/
>> <https://files.ettus.com/performance_data/>
>
> Side effect of Nationl Instruments' ownership perhaps.... good to
> know, thanks!
>

But I'll also note that said performance data doesn't include the spur
map, because the spur positions and magnitudes depend A LOT on
the precise operating conditions and parameters, which means that an
exhaustive "map" would likely take a good chunk of forever to
assess and be exceedingly costly to produce.

For example, on many models of USRPs, the master-clock rate is variable,
with potentially hundreds or thousands of different clock rates
available, which drives the internal DSP "goo" in the FPGA whose
spur-production potential would depend both on the master clock rate,
and things like sample-rate delivered to the host, etc. That's quite
apart from whatever spurs the synthesizers generate, and that would
depend on the reference clock--which on some models may be variable,
and on others, fixed. This completely ignores whatever spurs
and "goo" the computer you're using this with may also generate--and
THAT "goo" may depend very much on exactly what software
load you're running on the computer.

My driving point is that with modern, "mixed-domain" RF electronics,
you'll have spurs. With the constant drive to make front-ends more
sensitive, you'll "notice" more and more of these spurs. They are
impossible to eliminate, and very-expensive to even reduce
noticeably--things
like separate EM-tight enclosures for different sub-systems,
fiber-optic connections between sub-systems, etc, etc. Board layout and
"ground discipline" can *HELP* but it isn't a panacea. Your
oh-so-careful ground and signal lay-out may reduce the spurs from 15dB above
the notional Tsys for the receiver to only perhaps 5dB above the
notional Tsys--still quite noticeable, particularly if you aren't "swamping"
this stuff through outside influence to any great extent.

[Paul Oxley]

Marcus

Someone on here suggested using an amplifier ahead of the USRP. This is a good suggestion. A good LNA would establish a good signal to noise ratio and bury the spurs below the floor.

Paul

--
--
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.

To view this discussion on the web visit https://groups.google.com/d/msgid/sara-list/603FE2F4.4040602%40gmail.com.

[Marcus D. Leech]

On 03/03/2021 02:34 PM, Paul Oxley wrote:

Marcus

Someone on here suggested using an amplifier ahead of the USRP. This is a good suggestion. A good LNA would establish a good signal to noise ratio and bury the spurs below the floor.

Paul

Absolutely. I cannot imagine doing anything else.  Most SDRs have a noise figure that is "OK" for communications purposes, but not adequate
  for most radio-astronomy applications, plus you have the added loss of the feed-line ahead of the USRP.

My own habit is to put about 40dB "up front" in LNA, then feed-line, then the SDR.  This is usually enough to "bury" most spurs and
  other weirdness, but not always.  The harmonics of the 10MHz clock on the AirSpy are quite loud...

To unsubscribe from this group and stop receiving emails from it, send an email to sara-list+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sara-list/783880999.267809.1614800095015%40mail.yahoo.com.

[Lamar Owen]

On 3/3/21 2:38 PM, Marcus D. Leech wrote:
> On 03/03/2021 02:34 PM, Paul Oxley wrote:
>> Marcus
>>
>> Someone on here suggested using an amplifier ahead of the USRP. This
>> is a good suggestion. A good LNA would establish a good signal to
>> noise ratio and bury the spurs below the floor.
>>
>> Paul
> Absolutely. I cannot imagine doing anything else.  Most SDRs have a
> noise figure that is "OK" for communications purposes, but not adequate
>   for most radio-astronomy applications, plus you have the added loss
> of the feed-line ahead of the USRP.
>
> My own habit is to put about 40dB "up front" in LNA, then feed-line,
> then the SDR.  This is usually enough to "bury" most spurs and
>   other weirdness, but not always.  The harmonics of the 10MHz clock
> on the AirSpy are quite loud...

So, I get two really useful things out of this:

1.) Enough gain should be ahead of the SDR to override the majority of
the spurs.  Now,I likely don't have enough myself in our L-band 1.42GHz
receiver, since I see the AirSpy's 1.40GHz spur well above the noise,
but I haven't added anything as yet.   I do have sufficient gain for our
S-band system, with 76dB at the feed (LNA is a Kuhne LNA 222 AH with
30dB of gain feeding an isolator then a steep rolloff bandpass filter,
then a Minicircuits  amp with 22 dB of gain, then another isolator, then
a Minicircuits ZRL-2400LN+ which is specified for +24dBm output, also
feeding an isolator; the whole chain has very low passband ripple); we
then have 30dB of loss in the feedline (LMR600 on the main segment down
the antenna and to the building entrance bulkhead and lightning
arrestor, then 150' of Andrew 1/2 inch heliax) before it gets to our
operations center where I have another steep filter and isolator prior
to the SDR. I've thought about adding 20dB at the bulkhead, but haven't
done so.


2.) A need for a relatively simple and straightforward testing procedure
where spurs can be characterized so that the gain level in item 1 can be
determined.  Using a calibrated noise source at the input of the SDR,
either lab-grade HP/Agilent stuff or the Radio Astronomy Supplies
1.42GHz noise source, of which we have a few, is likely the best
method.  These noise sources typically have test data provided; use one
of these on the input and then take a spectrum with the SDR software at
different SDR gain settings; this gives a much lower noise temp than a
simple ambient temperature line termination resistor; now, a cooled
resistor can be used effectively but those can be more expensive than
calibrated noise sources.


One item I will add is that the SDR itself needs to be in a very
well-shielded enclosure (where 'well-shielded' is wavelength dependent),
with good power supply decoupling and signal cable/analysis computer
shielding.  This is where a Raspberry PI or similar small form factor
computer can really come in handy, because the whole unit can be
shielded economically and then controlled over ethernet (which can be
over fiber, providing even more RF isolation).  The SDR capture device
needs to be treated as if it were a computer (and it can be, as is the
case with the ADALM Pluto SDR or even that USRP which has the FX2
microcontroller) and shielded/isolated as such.  Marcus, I believe
you've done some of this with the AirSpy R2 and an ODRoid, iff I
remember correctly.

[Marcus D. Leech]

I use the products of https://g8fek.com/index.html

Use them both for my own purposes, and for a government contract from a
couple of years ago. Very nicely built, and about 1/10th the
cost of "the usual suspects".

>
>
> One item I will add is that the SDR itself needs to be in a very
> well-shielded enclosure (where 'well-shielded' is wavelength
> dependent), with good power supply decoupling and signal
> cable/analysis computer shielding. This is where a Raspberry PI or
> similar small form factor computer can really come in handy, because
> the whole unit can be shielded economically and then controlled over
> ethernet (which can be over fiber, providing even more RF isolation).
> The SDR capture device needs to be treated as if it were a computer
> (and it can be, as is the case with the ADALM Pluto SDR or even that
> USRP which has the FX2 microcontroller) and shielded/isolated as
> such. Marcus, I believe you've done some of this with the AirSpy R2
> and an ODRoid, iff I remember correctly.
>

Yes, I routinely package Odroids and the like with SDRs of various
flavors. I invariably package them in separate sub-enclosures, with
ferrites on things like USB cables and power supply lines.

[Paul Oxley]

Lamar & Marcus

Great progress in the right direction.

Paul

--
--
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.

To view this discussion on the web visit https://groups.google.com/d/msgid/sara-list/604BDEB9.804%40gmail.com.

[Michiel Klaassen]

Hi All,

 

The hackrf project is open source hardware and software.

I found the schematic diagrams; see att, and now I am going to try to improve the circuits myself.

One idea is to skip all the internal amplifiers and connect the antenna signal upstream as far as possible.

I could use a RTL dongle for that as an signal entrance module. Perhaps the radio part also can be skipped and coupling into the balanced ad converter is possible.

The control settings then have to be done via 2 usb channels.

If I fry one; so be it.

Michiel

parac.eu

Op vr 12 mrt. 2021 om 22:23 schreef Paul Oxley <oxl...@att.net>:

To unsubscribe from this group and stop receiving emails from it, send an email to sara-list+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sara-list/866550541.408224.1615587818885%40mail.yahoo.com.

[Paul Oxley]

Michiel

I don't understand how your new arrangement would improve the spurs. Can you explain?

Paul

To view this discussion on the web visit https://groups.google.com/d/msgid/sara-list/CAABBjRDCtM4Y4U9fFrS%3DZuyma2FYjNDNjoybF2G%3Dpj60GL9vcA%40mail.gmail.com.

[Michiel Klaassen]

Hi Paul,

Well, the spurs are generated by the circuit board itself. They come from the processor clock frequencies and the mixer frequencies etc. This is why I said in the post subject "self induced" they do not come from outside the SDR.  

So where are these signals then coupled into the circuit; I do not know. However I see that the places they can be inserted are many.

Unlike with the simple RTL SDR, where the distance between input connector and the input IC pin is less then 10mm, the distance on the hackrf board is very large. It goes from switch to amp to switch to amp etc.

The quest for the input frequencies <500MHz is even longer; they are upconverted via a mixer also. 

Along the road, the tracs can be irradiated/contaminated directly or via the supply voltage connections or via the control connections of the switch and amplfier ic's.

So, I want to skip that long electronic road as much as possible and do the amplification outside the hackrf module.

I hope this explanation helps.

regards,

Michiel

parac.eu

Op za 13 mrt. 2021 om 15:26 schreef Paul Oxley <oxl...@att.net>:

To view this discussion on the web visit https://groups.google.com/d/msgid/sara-list/568266280.569114.1615649157791%40mail.yahoo.com.

[Paul Oxley]

Michiel

The tone spurs are likely generated by the Hackrf. As explained previously the best way to reduce or eliminate the spurs is to improve the signal to noise (spurs in this case) ratio. This can be done with a LNA preamp to raise the signal level before the Hackrf. The spurs will be buried in the thermal noise from the LNA which is at a higher level. The desired signal is also raised in level. 

I did not follow your proposal to see if this is being done. A sketch or sequential description would clarify the issue. 

Paul

To view this discussion on the web visit https://groups.google.com/d/msgid/sara-list/CAABBjRAaqZA%2BQdH%2Bh-DYx0rnsLMp-1iSpHxL%2Bb%3DZPKpEU_j6Hg%40mail.gmail.com.

[Lamar Owen]

On Mar 13, 2021 5:24 AM, Michiel Klaassen <vmin...@gmail.com> wrote:

Hi All,

 

The hackrf project is open source hardware and software.

I found the schematic diagrams; see att, and now I am going to try to improve the circuits myself.

Not a bad idea.  Reducing the spur coupling into the input will improve the dynamic range and reduce the intermodulation distortion of the system, something that is not addressed by the 'add gain' method.  I look forward to seeing your results.

[Marcus D. Leech]

My experience suggests that as long as you have digital and analog systems on the same board (even with RF-decoupled ground-planes),
  you'll have spurs from the digital bits.  Better engineering can reduce that coupling a bit, but not entirely eliminate it.

In "high end" systems, the analog piece is in a separate box from the digital piece, and ideally uses fiber-optic couplings for the IF, and
  often the control and reference-clock signals as well.  That doesn't sound cheap to me :) :)

[Paul Oxley]

All

I agree with Marcus on the source of the spurs. They are inherent in the digitally derived sources.

RE: Intermodulation - Intermodulation occurs when the total power drives the amplifier into a non-linear region. In the case of radio astronomy, the desired signal is not significantly different in level from the noise floor. Thus the dynamic range is capable of handling the total power. If intermodulation becomes an issue, it could be reduced by narrowing the signal which would lower the total power.

Communications systems operate with the desired signal much much larger than the noise floor. Therefore the intermodulation is of more concern.

Paul

--
--
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.
To view this discussion on the web visit

https://groups.google.com/d/msgid/sara-list/604D27F1.8080206%40gmail.com

.

[Marcus D. Leech]

On 03/13/2021 05:00 PM, Paul Oxley wrote:

All

I agree with Marcus on the source of the spurs. They are inherent in the digitally derived sources.

RE: Intermodulation - Intermodulation occurs when the total power drives the amplifier into a non-linear region. In the case of radio astronomy, the desired signal is not significantly different in level from the noise floor. Thus the dynamic range is capable of handling the total power. If intermodulation becomes an issue, it could be reduced by narrowing the signal which would lower the total power.

Communications systems operate with the desired signal much much larger than the noise floor. Therefore the intermodulation is of more concern.

Paul

In fact, in a "reasonably designed" weak-signal RF chain, any intermod that happens is overwhelmingly likely to occur in the 1st LNA stage, since
  you really need to "filter early, filter often".  The 1st LNA stage cannot usually be filtered, since that would imply unacceptable insertion
  losses ahead of it, and thus badly degrade Tsys.

But once you have 20-30dB of low-noise gain, you can stick in an aggressive, and even somewhat-lossy, filter, and any possibility of intermod
  would tend to go away completely, unless you have strong in-band signals that are driving things into non-linearity further down. In that case,
  it's not really intermod you're worried about, simply system dynamic range.  It's really unlikely that a few low-level spurs would be by themselves
  a cause for loss of dynamic-range and linearity.  But they can "participate" in intermodulation from very-strong in-band signals--which for
  a radio astronomy application, you hopefully don't have.

[Michiel Klaassen]

After some snooping around the board, I found this place where a party is going on.

Frequency 1420 MHz. The probe is a central core coax with only the brading removed; so no galvanic conact.

probe inserted in P28 pin nr 6; counting from top left.

The pins of this P28 function as a radiating antenna.

With the R&S spectrum analyser (3Gmax) the measured frequencies are 0.2,0.4,0.6,0.8,1,1.2,1.4,1.6,1.8,2,2.2,2.4,2.6,2.8GHz.

When you select a frequency range with the Hackrf and display that with SDR#, then you will notice that the rfi spikes shift when the selected windows shifts.

That is an indication that the rfi is inserted before the radio chip.

However there are also spikes not moving while selcting another window; they stay on one spot.

That is an indication that the rfi is inserted after the radio chip.

The radio chip converts the input frequency range to 2.3 to 2.7 GHz, so inserted rfi in that range will give a static window frequency response.

Will be continued..

Regards,

Michiel

parac.eu

Op zo 14 mrt. 2021 om 20:17 schreef Marcus D. Leech <patchv...@gmail.com>:

--
--
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.

To view this discussion on the web visit https://groups.google.com/d/msgid/sara-list/604E6F57.3080706%40gmail.com.

[Lamar Owen]

On 3/13/21 5:00 PM, Paul Oxley wrote:
> I agree with Marcus on the source of the spurs. They are inherent in
> the digitally derived sources.
>

For a spur to be visible in the signal, it must be coupled into the RF
chain in one or more places.  Spurious signals have many possible
sources, one of which is intermodulation of a very strong out-of-band
signal (such as a digital clock) with other signals on the board.

As Marcus mentioned, better board design can reduce these
internally-generated and coupled signals.  In fact, all you can do is
reduce these signals; no attenuation is ever perfect, nor is any
shielding or filtering ever perfect.  You can only improve spur
rejection, never make it perfect.  If a board re-layout reduces spur
injection by 10dB, well, that is a definite improvement.

> RE: Intermodulation - Intermodulation occurs when the total power
> drives the amplifier into a non-linear region.

...

> In the case of radio astronomy, the desired signal is not
> significantly different in level from the noise floor. Thus the
> dynamic range is capable of handling the total power.

I had a long discussion on the subject of dynamic range a number of
years ago with a radio astronomer whose main focus was pulsars.  It was
his opinion that, at least in terms of pulsar radio astronomy, more is
better.  But that's an edge case.

Closer to most radio astronomy observations is mitigation of inband and
out-of-band RFI from outside the receiver system.  Dynamic range
improvements increase the RF chain's headroom in dealing with
interfering signals; no filter is perfect, and even with high-end
filters some of these new satellite downlinks and terrestrial
communications systems can infiltrate the RF chain.

For instance, suppose you're doing some continuum work in the 2.2-2.3GHz
band, doing something like studying extreme scattering events (see:
https://www.tandfonline.com/doi/abs/10.1080/10556790701610498?journalCode=gaat20#
).  Now, there are two extraordinarily bright RFI sources in the 2.3GHz
range, and they're located on the celestial equator.  These sources have
a flux that is >20dB brighter than the Sun.  These sources are the
megawatt-class EIRP downlinks of two of the satellites used by Sirius/XM
radio.  The signals are 10dB above the noise floor in the 2.2-2.3GHz
band, even after filtering, and even out of beam of our 26-meter
antennas (in-beam the signals are >60dB above the noise floor, after the
filter (a Lark 2B 2200-220-8AA)).  The Sirius/XM system is designed to
have sufficient link budget even for 0dB gain antennas; takes a lot of
EIRP to do that.

Smaller antennas will have a bigger problem with these sources. It's
easy enough to run an FFT and drop the bins occupied by those signals,
right?  Not so fast: with insufficient dynamic range intermod can be
generated by these interference sources.  I ran a characterization of
our 2.2-2.3GHz RF chain using an FSH3 spectrum analyzer, since it is
equipped with a tracking generator;  the filter rolloff is pretty steep,
down 5dB at 2.3GHz, and down 45dB at 2.4GHz , but the Sirius/XM signals
still dominate the spectrum, no matter where in the sky the antenna is
pointed with the exception of positions where the satellite is in the
first null. A tighter filter with a lower upper cutoff might help;
sufficient dynamic range in the RF chains allows FFT bin 'editing' and
effective removal, by digital post-processing, of the interference
signal, so we left this particular filter in place, as there was enough
attenuation of those signals to do what was needed as part of that
concluded project.

The same thing can happen in the 1420MHz band with GPS satellite
downlinks.  The RAS 1420 filter, for one, is very good, but GPS signals
are so strong that the signals are still present post-filter, albeit
attenuated significantly.  Anyway, due to the requirements of a recent
project we don't have any filtering in place in the RF chain of our
1.42GHz receiver; the RF is split between an RAS SpectraCyber and an
AirSpy R2; the SpectraCyber is used as a comparison standard for
calibration of the R2 within the SpectraCyber's tuning range.  The
project?  Ever tried to observe HI for 3C273? We had an undergrad
attempt just that, along with other low-redshift and low-blueshift
extragalactic sources.  While he wasn't successful at repeatably seeing
the HI content of 3C273 (at 1226.18-1226.30MHz!), he was fully
successful with repeatable observations of three of his extragalactic
sources, M81, M86, and M87.  M87's HI spectrum showed up at the expected
place, between 1414.32 and 1414.38MHz.  A tracking filter would have
been nice to have had, but it was out of the budget range for this
project, and unless a very narrow filter was used would not have helped
the 3C273 attempt.  So we removed any pre-receiver RF chain filters.

Since the GPS signals are modulated carriers, if the dynamic range of
the entire RF chain is not sufficient to prevent intermodulation of
these signals from occurring at any amplifier in the cascade, including
the SDR front-end, an intermod product could end up in the desired
passband (and actually could explain the anomalous HI result from 3C273
that was observed once but was not repeatable; one of the multitude of
GPS satellites could have been in-beam at the time of the observation). 
Our current unfiltered 1.42GHz RF chain is both sensitive enough to do
hydrogen line spectroscopy of extragalactic sources and has enough
dynamic range that a GPS signal shouldn't create intermod issues, unless
the GPS signal is in the main beam of the antenna.  The unfiltered
observations from the extragalactic HI observations required lower
pre-SDR gain to keep the GPS downlinks from creating intermod at the SDR
input amplifer.  Of special concern to the extragalactic project was the
L2 GPS signal at 1227.6MHz; the RF chain is very capable of receiving
that signal as well as L1 at 1575.42MHz, and it is quite close to the
expected redshifted HI line from 3C273.

Increased dynamic range will help mitigate these sorts of interference
issues; lower gain is better in a high RFI environment.

So, no, the noise temperatures for typical continuum sources don't
require much dynamic range at all; this is true.  HI spectroscopy
doesn't really require much dynamic range, either.  The jury is still
out on how much dynamic range is desirable for pulsar observations; I
haven't revisited the discussion I had several years back, as we've not
participated in another pulsar observation since then.  But in the
presence of interference, whether filtered or not, RF chain dynamic
range has an impact on the degree of RFI that can be successfully
mitigated in digital post-processing.

[Lamar Owen]

On 3/15/21 11:40 AM, Lamar Owen wrote:
> ...

> Smaller antennas will have a bigger problem with these sources. It's
> easy enough to run an FFT and drop the bins occupied by those signals,
> right?  Not so fast: with insufficient dynamic range intermod can be
> generated by these interference sources.  I ran a characterization of
> our 2.2-2.3GHz RF chain using an FSH3 spectrum analyzer, since it is
> equipped with a tracking generator;  the filter rolloff is pretty
> steep, down 5dB at 2.3GHz, and down 45dB at 2.4GHz , but the Sirius/XM
> signals still dominate the spectrum, no matter where in the sky the
> antenna is pointed with the exception of positions where the satellite

> is in the first null. ..
Attached is a photo of the FSU spectrum analyzer screen of the
2.2-2.3GHz environment here (the FSU can export to floppy, but I don't
have any blank 3.5 floppies handy and I've not done an export to USB
before with it....but I need to try it out, don't I?  So please excuse
the blur, as I took the photo with my cell).  On the left, 4G LTE
terrestrial sources are seen (I have two stages of filtering going on;
the Lark 2B 2.1-2.3GHz filter is in the feed, and I have a 2.2-2.4GHz 
filter down at the spectrum analyzer to bring the LTE signals between
2.1 and 2.2GHz down to a manageable level).  On the right are the two
Sirius/XM signals, and this is with the antenna at zenith, not pointed
anywhere near the two satellites.  The markers show the relative offset
from 2.3GHz and the relative amplitude in dB referenced to the noise
floor at 2300MHz. The rolloff of the filter is clearly visible beginning
immediately above 2.3GHz, and it's rolled -5dB by the time you get to
2320MHz.  But then the Sirius/XM signal rises to a point >14dB above the
reference level at 2300MHz.  At marker 4, even though the filter has
rolled almost -20dB the signal still makes it up to only -7dB referenced
to the 2300MHz level, and that's 50MHz away from the band edge.

[Lamar Owen]

On 3/15/21 11:40 AM, Lamar Owen wrote:

> On 3/13/21 5:00 PM, Paul Oxley wrote:
>> I agree with Marcus on the source of the spurs. They are inherent in
>> the digitally derived sources.
>>
>
> For a spur to be visible in the signal, it must be coupled into the RF
> chain in one or more places.  Spurious signals have many possible
> sources, one of which is intermodulation of a very strong out-of-band
> signal (such as a digital clock) with other signals on the board.

> ...

> Our current unfiltered 1.42GHz RF chain is both sensitive enough to do
> hydrogen line spectroscopy of extragalactic sources and has enough
> dynamic range that a GPS signal shouldn't create intermod issues,

> unless the GPS signal is in the main beam of the antenna.  ..
So, another data point.  Attached are two PNG images of the raw spectrum
from our 1420MHz SDR, an AirSpy R2, using the 'AstroSpy' application,
with a one minute integration time and an RF gain of 19dB.  The first
one is a baseline, with the entire RF chain from the feed down powered
off.  The second one is with the RF chain powered on (and, remember,
this is unfiltered, just the RAS feedhorn, the RAS LNA, and LMR600
feedline to a two-way wideband splitter).  Both are taken with the
antenna at zenith and the drives off.  The massive spur right at 1420MHz
is very noticeable.  But the remainder of the noise floor is relatively
clean.  Note in the second image with the RF chain on the spurs spaced
every 10MHz that reduce in amplitude away from the central spur.  And
the HI 'bump' is clearly visible, right where it should be.  The ripple
in the passband might be mitigated by adding an isolator after the LNA;
next time I have the bucket truck out and the feed in service position
I'll be prepared to add one.

[Michiel Klaassen]

Hi Lamar,

I do not have an 'astrospy' app, I know airspy and astropy; so I cannot copy you measurement.

Anyway, with the SDR# app you can correct the central spike by clicking the 'correct i/q' selector.

I am still fighting with my hackrf board; I have removed a lot of connectors; these function as transmitting antenna's to the front end.

regards

Michiel

parac.eu

Op di 16 mrt. 2021 om 16:32 schreef Lamar Owen <lo...@pari.edu>:

--
--
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.

To view this discussion on the web visit https://groups.google.com/d/msgid/sara-list/6dded6c3-cc8b-0053-27a5-eb0b67edd79c%40pari.edu.

[Lamar Owen]

On 3/16/21 1:11 PM, Michiel Klaassen wrote:
> Hi Lamar,
>
> I do not have an 'astrospy' app, I know airspy and astropy; so I
> cannot copy you measurement.
>
> Anyway, with the SDR# app you can correct the central spike by
> clicking the 'correct i/q' selector.

Hi Michiel,


The astrospy app comes with the install package for SDR#, and only works
with an AirSpy SDR.  It doesn't have as many controls as the full SDR #
app does, and correct i/q isn't one of them, although I haven't looked
at its configuration file to see if that setting is there.... I had
looked for that setting, since that's something that typically needs
doing for any I/Q SDR device.  I need to try out my AirSpy Mini and see
if it gives similar results.