Neutral Hydrogen Detection - sigh.....

[Peter Masters]

Hi All,

Sorry need some help here as I think I've consumed all the stuff I can find on this, and I'm not getting anywhere.  It's definitely time to call in the experts as I'm definitely not one of them!

I've been trying to detect neutral hydrogen - we've all thought of this one - build a horn antenna, map the galaxy, detect the spin rate etc.  I followed the design from SETI - the "Horn of Plenty" http://www.setileague.org/articles/ham/horn.pdf.  I used conductive Kingspan insulation, with conductive tape.  There's no separate waveguide (read: oil can, etc) - the cone just tapers.  I've checked the connections with a multimeter - all panels connected and low resistance, and the probe is ~1.8inches long as the SETI article suggests (which is interesting, as that isn't 1/4 wavelength...).    Attached to that, I have a Nooelec SAWbird H1, which is an LNA + 1420Mhz filter (@ ~1db NF, ~20dB gain).  That is connected to a Nooelec LaNA LNA that I was using on my previous project (NOAA weather satellites anyone?).  That then goes to the Nooelec v4 RTL based SDR.  The LaNA is powered using Bias-tee, but there is then a DC blocker, and the SAWbird is powered via usb.  I've been using a raspberry pi, and I've setup Virgo https://github.com/0xCoto/Virgo (as written by Coto on this forum) to do the capture, with 2 min captures, trying both high (30) and low(5) gain in the virgo setup.  In essence, nothing is working, and I'm not really sure how to debug.  My first few tries showed spiky results, which looked like RFI, so I've move the raspberry pi away using USB extension leads, and wrapped all the electronics (Pi, LaNA, Sawbird LNA) in tin foil which helped.  I took calibration "images" by trying a capture at 1422mhz (i.e. +2mhz), and separately pointing at the floor.  I'm using Stellarium to ensure I must be looking at a galaxy arm.

Few questions:

Firstly, of course anything wrong in my setup?  I wondered if the probe length would be an issue.  I used the LaNA as I had it, but I wondered if that isn't a good gain-stage amp?  I've got a 1420 bandpass filter on order, as the standard approach seems to be LNA, Line Amp--coax----Line Amp, Filter, SDR.  I've got a satellite line amp too but the LaNA looked to have better specs in terms of noise/gain so I used that for now.

Why would the "Average power vs Time" in my attached graph reduce over time - I would assume that would be constant, but most images show that sloping down?

I'm into astrophotography, and I'm a IT professional with a lot of linux time under my belt, so that side I understand, but radio is definitely a new one on me!

Anyway, head scratching!  Send help!

Edited April 23, 2021 by Peter Masters
1db didn't say it was for the noise figure!

[Ed astro]

Hi Peter,

"Nothing works and nobody knows why" is a common problem when doing amateur radio adtronomy ; ) I must admit this hobby has a very steep learning curve. When I started in 2017/18 it took me nearly a year to get good results with hydrogen line observations. 

Looking at the pictures of your setup the first thing that comes to my mind is checking the conductivity of the insulation material. If it is not conductive, the horn antenna would not work. In that case you could try lining the inside of the horn with aluminium foil.

The sawbird HI LNA should have enough gain on its own as long as you do not have a very long and lossy coax cable between the LNA and the SDR receiver. There is already a 1420 MHz bandpass filter in the sawbird HI, so I think an extra filter is not necessary. Furthermore, losses between the antenna and the LNA have a lot of impact on the sensitivity of your system. It is therefore a good practice to keep the cable that connects the antenna to the LNA as short as possible. 

Best regards,

 

Eduard

[Peter Masters]

Hi Ed,

Thanks for the reply! - yes this is definitely a learning curve.  The number of scary scientific papers with detailed FFT algorithms that come up when you just type "radio astronomy introduction" in google is probably not a good sign!  It seems quite a small community of experimenters as I keep stumbling across the same people via different routes!

I have tried just the LNA (Sawbird), which I powered with Bias-Tee to simplify things, but still no joy.  The Antenna is definitely conductive - I've used a multi-meter to test connectivity between the 4 sides, the end and the negative connector on the cable, and that is all good, and v.low resistance.  The positive side of the probe is definitely not shorted to that, and connects fine between probe and cable.  The cable itself is ~20cm long now between the probe and the LNA.

The frustrating thing is that I'm seeing no variation in the graphs at all unless I change amplification or take elements out the circuit (i.e. no hint of hydrogen bumps).  I did go through an exercise of removing everything and then sequentially adding things back to see what the graphs did in terms of noise/signal.

In Virgo, I'm using these settings:  I was using a high rf_gain of 30 but thought maybe I was overcooking it, so I'm trying lower numbers (on the basis there's a lot of amplification already).  As you can see, I'm trying 2min captures at the moment.

obs = {
    'dev_args': '',
    'rf_gain': 5,
    'if_gain': 10,
    'bb_gain': 10,
    'frequency': 1420000000,
    'bandwidth': 2000000,
    'channels': 2048,
    't_sample': 1,
    'duration': 120
}

How "tolerant" will the probe length be?  I couldn't find any info really on that.  4.6cm long as per the SETI Horn of Plenty guide, but 1/4 wavelength would be 5.25cm.  I noticed that when I was trying to capture NOAA satellites that I'd forgotten to account for the length of cable outside the coax before it joined the actual probe (which of course is still part of the antenna).  In this setup, the excess cable is probably 1.5 cm, so maybe I'm close to the 1/4 wavelength anyway.  In any case, I don't know how much out that could be before it would make the probe ineffective.

Thanks!

[Carl Reade]

Hi, if you have the software running, when you turn on the amp the noise level should jump up. Will give you an indication things are connected.

The last thing you want to use in calibration is the ground. The noise from the ground is far higher than cold sky hence would wipe out any readings.

Top tip we have a convenient large orange noisy ball in the sky 😁 your entire noise level should rise when pointed at it. If your system can't see the sun it's unlikely it will see the H line.

Go for the sun first. Keep at it you will get there.

[saac]

3 hours ago, Peter Masters said:

 

How "tolerant" will the probe length be?  I couldn't find any info really on that.  4.6cm long as per the SETI Horn of Plenty guide, but 1/4 wavelength would be 5.25cm.  I noticed that when I was trying to capture NOAA satellites that I'd forgotten to account for the length of cable outside the coax before it joined the actual probe (which of course is still part of the antenna).  In this setup, the excess cable is probably 1.5 cm, so maybe I'm close to the 1/4 wavelength anyway.  In any case, I don't know how much out that could be before it would make the probe ineffective.

Thanks!

Hi Peter, Im following your thread with real interest as I am on my own journey on this. The plan for my antenna design was provided by the DSPIRA team and they use a 52.5 mm probe (as you said 1/4 wavelength).  I wonder, subject to it being an easy thing to do, would it be worthwhile swapping your probe out for a 52.5 mm length just to take that out of the equation.   I am very far from being knowledgeable on any of this  - my learning curve inclines behind vertical !  At the moment I'm trying to get my software running  - started off using again the DSPIRA recommended spectrometer on gnu radio platform running on ubuntu (which I know nothing about).  I've hit a problem as their setup uses the (expensive) Airspy R2 SDR receiver while I have a Nooelec NESDR dr Smarttee XR connected to a SAWbird.  I think I somehow need to swap out the driver for the SDR so that the software recognises it but I have no idea how to proceed.   That is today's task    Anyway good luck , I think a lot of this as others have hinted is going to be a battle of persistence , trial and error. 

Oh,  where did you source your aluminised thermal board if you don't mind me asking - I couldn't get my hands on any so I went down the road of lining foam board with Al foil. 

Jim 

Edited April 24, 2021 by saac

[Ed astro]

Hi,

The probe length should be a bit less than 1/4 wavelength. In my 1420 MHz feed it is about 4.5 cm. However, I do not think the probe is the problem in your setup. If the probe is a few mm too long or too short you will notice that the antenna is less sensitive but you will probably still be able to detect HI. Besides my 1420 MHz feed I have also made a 1665 MHz feed (for OH line observations), its probe is of course a bit too short for 1420 MHz but when I use it at 1420 MHz I can still detect HI with it just fine. 

[Peter Masters]

Thanks Ed,

Hmm ok, I think I might have found a "smoking gun".  One of the connecting wires was shorted out in the plug, which would certainly explain things.  I checked everything at the aerial but hadn't checked that.  I've adjusted things and I'm going to try this with just the LNA later and see what happens.  I've confirmed all the connections this time, but at least the problem so far was at a level I can understand   

Let's see what we get later!

Peter

[Peter Masters]

2 hours ago, saac said:

Hi Peter, Im following your thread with real interest as I am on my own journey on this. The plan for my antenna design was provided by the DSPIRA team and they use a 52.5 mm probe (as you said 1/4 wavelength).  I wonder, subject to it being an easy thing to do, would it be worthwhile swapping your probe out for a 52.5 mm length just to take that out of the equation.   I am very far from being knowledgeable on any of this  - my learning curve inclines behind vertical !  At the moment I'm trying to get my software running  - started off using again the DSPIRA recommended spectrometer on gnu radio platform running on ubuntu (which I know nothing about).  I've hit a problem as their setup uses the (expensive) Airspy R2 SDR receiver while I have a Nooelec NESDR dr Smarttee XR connected to a SAWbird.  I think I somehow need to swap out the driver for the SDR so that the software recognises it but I have no idea how to proceed.   That is today's task    Anyway good luck , I think a lot of this as others have hinted is going to be a battle of persistence , trial and error. 

Oh,  where did you source your aluminised thermal board if you don't mind me asking - I couldn't get my hands on any so I went down the road of lining foam board with Al foil. 

Jim 

Hi Jim,

Thanks - keep at it, sounds like you have a similar SDR to me, I have the "Nooelec NESDR SMArt v4 SDR".   As I wrote my original question, it was starting to sound like a Nooelec advert - lol, but I did like their LNA as it had the filter in it (and people had had success with it).  In terms of probe, I started out with a length of brass bar.  I rebuilt the connector though and decided to make something where I could easily change the probe (and I read that a thinner probe means a less broad reception which sounded useful).  I've now gone with an electrical "choccy block" approach, with the probe wire bent vertically out, and the other connecting through to a conductive patch on the antenna.  As you'll read above, my cable had a short, so this turned out to be useful.  It also makes it easier to swap the probe out (I'll get back ache leaning down these crazy horns!).

I've put some links here for similar projects, as I've found a few things each time that help steer my efforts:-

https://docs.google.com/document/d/1_7ZOe1Et_8QTk07bgbTd7LLNqDAtgAjmCS50JM9JRbQ/edit

https://www.rtl-sdr.com/cheap-and-easy-hydrogen-line-radio-astronomy-with-a-rtl-sdr-wifi-parabolic-grid-dish-lna-and-sdrsharp/

https://physicsopenlab.org/2020/07/26/sdr-based-receiver-for-the-21-cm-neutral-hydrogen-line/

Hoping I'll get a good "first light" tonight - the galaxy arm comes up about 5pm ish I think.  I might also see if I can grab some Sun noise as Carl suggests (thank you Carl!) to prove things out.

In terms of the insulation board, I like in the UK, so I found it in Wickes (https://www.wickes.co.uk/search?text=kingspan).  I bought a full board for ~£35 of 25mm - I measured in advance where I could split it easily to get it in the car without chopping a needed triangle in half.  The conducting tape came from ebay (https://www.ebay.co.uk/itm/154140454092), and has worked really well.  You get loads of it - which given I put the horn together wrong the first time was useful (you have to cut it at that point, as peeling off the tape will destroy the foil on the board) - hence the ghetto look in the photos - oops!  I'm pretty sure by now my family think I'm mad

Peter

[Peter Masters]

Ok, @Carl Reade Thank you for the steer, and I got some interesting results.  I just pulled up a live view of the SDR (web based, called OpenWebRX - useful on the PI when it's not plugged into a display directly!).

So I tried pointing at the sun and away, with no discernible results.  However, I then simplified my setup to just the SDR and the horn antenna (i.e. no LNA).  I saw the output change!  I then unplugged it, put the LNA back on - nothing.  Add the antenna, nothing.  I'm thinking my LNA is bust  (this is the SAWbird+H1 thing).

I have another LNA (the wideband LaNA one), so I screwed that on to the SDR first (change in signal occurred).  I then plugged in a whippy FM aerial - loads of noise.  I then unplugged the FM aerial (no real noise), then the horn aerial (low level of noise).  To my mind, this means my horn antenna is doing something, and the LaNA is amplifying correctly, but when I plug in the other SAWbird LNA, it don't work.  I tried the SAWbird with the FM aerial too - nothing at all.

Just to be clear, all this testing was done whilst watching 1420mhz, so although the sawbird is filtered, I should have seen something (the waterfall was showing 1419 - 1421 mhz).

I'm not sure how else to test the SAWbird LNA - the power is definitely working (Bias-tee provided by the SDR), and I'm told their pretty much designed to work together, so I don't think it's insufficient power.

If that LNA is bad, then I'm stuck for now as I don't have another filter for 1420, but I'll try anyway.  I don't want to jump into sending it back until I've at least done some more thinking, but this might be more ammunition to think about.

Peter

[Peter Masters]

Well, I may have seen something.  Ignore the LNA thing, the following graphs were taken with the SDR->SAWbird LNA->Antenna, and they're definitely doing something.  This is a sequence of pictures taken at 2min intervals.  The waterfall is a bit weird, and somewhen during my series of 30 pics (including these 4) the wind blew the antenna over - murphy's law would suggest it was right in the middle of the useful bit of course, but I didn't look at the time  .  Anyway, this data is promising.  The calibrations are inaccurate as I haven't recaptured them but this is definitely progress.  Thank you all for your help thus far, I'm going to try and get a more complete sequence and see what that gets me.

Peter

[Carl Reade]

Hi yes it can be difficult to know if a H line dedicated LNA is working but you should see a rise in the noise level when switching on. A wideband LNA is much easier to test as you can just test it on any signal in SDRsharp to see if it's working. If your H line Freq is clear of spikes you should get away without a filter.

I found it easier to use SDRsharp with the IF Averager plugin. It can be calibrated flat. As per one of your links. The hump can be easily seen when using it. Then go on to the more complex software. 

[Peter Masters]

Aaand back to tearing my hair out :).

So apart from that one blip that I put above (which whilst "looking" like a hydrogen bump was a very short lived spike in 2 frames, so I'm not convinced), I've not seen anything come out.  I've tried increasing/decreasing the various gain controls to see if that affects things.  I've tried SDRsharp, but again, nothing. I've confirmed that plugging in the LNA causes signal, and that then plugging in the antenna causes additional change, which suggests that there is an end-end connection.  I've also re-confirmed all the electrical connectivity on the antenna side:  The 4 sides of the antenna are connected to each other (and to the end), they in turn connect to the shield/outside on the SMA plug.  The probe connects to the pin on the SMA plug and there's no short between pin and shield.  I've even taken a small whip antenna that came with my SDR and placed that inside the horn near the probe, and used the SDR with that instead, with pretty much the same output.  I'm assuming there's no downside in the fact this insulation is foil on both sides, and both inner and outer are connected (which would kind of give me a horn inside another horn)

My latest theory is that Hampshire is in fact a faraday cage :).

Any thoughts?  I'm really not sure where to go with this right now.  Is there any way I can make this really really simple to prove things?  The SDR definitely works in that I can pick up FM (and previously had no problem with a NOAA satellite pickup with a home made dipole).  In essence, all I'm doing here is using that same setup with a different aerial and a weaker signal.

Sigh!

[Peter Masters]

Just to say that I'm also getting the same view on the waterfall with both LNAs (the SAWBird, which is filtered to 1.4Ghz and the LaNA which isn't).  I can see changes from plugging in the antenna so everything looks sensible to me, but not gettin' nada

[Peter Masters]

Just to prove what I'm saying - I've tuned my SDR to Classic FM (100.3Mhz) whilst plugged into the horn antenna (SDR->LaNA->Horn) and I can pick it up fine (this has got to be the most dumb way to pick up Classic FM I can think of).  I then unplugged the horn, and then plugged it back in again, and you can see the effect on the waterfall.

[saac]

Peter I had a complete failure over the weekend with mine making effectively zero progress due to software issues. I'm not really into coding of such an my initial approach was to use gnu spectrometer based software advised by DSPIRA. That however involves installing and running ubuntu which is where the problems started.   Long story short I eventually went back to using SDR Sharp and eventually managed to install the If Average Plugin. So my next stage is to capture the background (instrument) signal so that it can be subtracted.  I think the If Average function also allows me to take effectively a long exposure and stack the signal to bring out the weak hydrogen line hopefully !!!!   I have yet to confirm that my system can pick up a signal - waiting for good line of sight of the sun and favourable weather to test that out .  One thing is for sure - this is not easy  

Jim 

[Peter Masters]

Hi Jim,

Yes, it definitely seems to demand a pint of blood!  As a computer programmer, I'm fine with the software, but radio is a black art.  I'm mostly running this on a raspberry pi as it's more portable for putting this outside (though therefore running linux), but I do have a laptop that I use for astrophotography which I have tried with SDR sharp.  I'm not really trying to get a background signal yet - I guess this is equivalent to a dark or flat frame in astrophotography (effectively working on the (Signal+Noise) - (Noise) = (Just Signal) principle).  In theory, whilst a calibrated picture would look nicer, I just want to see a difference between frames where I have a galaxy arm in the frame vs not.  At the moment, whilst bumpy, I don't see any difference at all between any picture.  As a lot of people seem to show at least a +10dB jump when Hydrogen shows up, I figure it should be easily visible.  In my graph views (which come from Virgo), I'm really therefore only looking at the top left graph, and expecting to see a new feature popping up in some frames - I tend to take something like 120 x 2min pictures, and then skip through the frames looking for a growing/shrinking lump somewhere around the point where the red-dash line is (+/- for doppler in theory, not that I see anything anyway!).

I think fundamentally my setup works in that I don't have connectivity issues, and the math says that the antenna should be reasonable in gain, and directional.  All I can think is that either I have a local interference source to cut out, or that somehow I'm not getting gain in the way I should be.

Good luck with your setup, let me know how you get on, and I'll certainly help if I can if you want to try ubuntu - that's actually something I do understand!

Cheers, Peter

[Carl Reade]

Hi all yes I would recommend using the sdr Averager plugin to start with and take a calibration about an hour before the galaxy comes into view of the scope. Bear in mind the H line will usually only be 2db above the noise very weak. I would also ensure the dB increments on the left are as low as you can set, I think the most it goes down to is 3 dB. If aiming for the sun there will be no humps etc the entire noise level will rise.

Software settings can be half the battle.

[Ed astro]

Hi,

I agree with Carl Reade, with my 3 metre dish the hydrogen line is at most around 2dB above the noise floor. I also use IF average and it works fine for me.

I would like to add that the sun may not be a particularly strong noise source given the relatively small aperture of your horn antenna. It may be easier to use the thermal noise from the ground to get an idea of the sensitivity of your setup. If you point the antenna to the ground you should see an increase in noise of more than 3 dB. If it is less than 2 dB seeing HI will be really hard. 

[Peter Masters]

Still nope

[Ed astro]

Hi Peter,

I see your vertical scale goes from -34 to -174 dB, so you probably would not see a 1 or 2 dB H line signal.

You can use the "gain" and "level" sliders to zoom in on the vertical scale. I usually set "gain" to maximum and adjust the "level" until I have a good view of the spectrum.

[Peter Masters]

Ah ok thanks - I have to admit I wasn't quite sure what I was doing with the sliders.  I shall try again and see where that gets me!

[Peter Masters]

Ok I think I understand a bit more about what is going on with the plugin now but I've failed to calibrate it for anything more than 1000 on the dynamic averaging.  The link I had earlier (https://www.rtl-sdr.com/cheap-and-easy-hydrogen-line-radio-astronomy-with-a-rtl-sdr-wifi-parabolic-grid-dish-lna-and-sdrsharp/) suggested it would take 6-7 mins but it's still not said calibrated.  Even so, I would have expected to see some variation building in the wave though?

[Ed astro]

Hi Peter,

I would have expected to see atleast something at this point...

The '"wavy" response is also a bit strange, but it may just be the passband respons of the SDR you are using.

To get an idea of what you should be able to see I quickly put together something similar to your setup. I used a small conical horn antenna (which whas originally intended as a feed antenna for my dish but was rarely used), a nooelec sawbird HI LNA and an RTL SDR. I did not do bandpass correction. The hydrogen line shows up as a small bump in the spectrum that normally would not be there. Note that the aperture of my horn antenna is probably quite a bit smaller than yours, so you should expect to see a somewhat stronger HI signal than me.

 

 

[Carl Reade]

5 hours ago, Ed astro said:

Hi Peter,

I would have expected to see atleast something at this point...

The '"wavy" response is also a bit strange, but it may just be the passband respons of the SDR you are using.

To get an idea of what you should be able to see I quickly put together something similar to your setup. I used a small conical horn antenna (which whas originally intended as a feed antenna for my dish but was rarely used), a nooelec sawbird HI LNA and an RTL SDR. I did not do bandpass correction. The hydrogen line shows up as a small bump in the spectrum that normally would not be there. Note that the aperture of my horn antenna is probably quite a bit smaller than yours, so you should expect to see a somewhat stronger HI signal than me.

 

 

Excellent post. That's how small it is. Software is half the battle. I am also not sure what the humps are Peter on your last post, that is an issue in the receive end.

[saac]

Im finding that the drivers for my NESDR Smartee XTR SDR usb dongle are quite sensitive - Had to reinstall both SDR Sharp and the drivers several times before it settled again. Anyway,  I had a trial run last night with the IF Average plugin and tried to take a background calibration.  As Peter has said the article at (https://www.rtl-sdr.com/cheap-and-easy-hydrogen-line-radio-astronomy-with-a-rtl-sdr-wifi-parabolic-grid-dish-lna-and-sdrsharp/)  suggested the calibration would take up to 7 minutes but mine reported calibration really quickly - almost immediately - so not really sure it worked properly.    It was partly cloudy last night so I couldn't get a good fix on the Milky Way just pointed it randomly where I though Denab may be !  Anyway nothing showed  - probably not pointing in right direction anyway.   I also think I was too optimistic, I'm learning that the reading will take a few minutes to integrate before anything shows.   Was hoping to do something today but been raining non stop - need to modify my horn with a drain tap  

Getting there very slowly !

Jim