HI data interpretation help

[JERRY TAYLOR]

I am trying to interpret the Hydrogen line data I have been collecting rather than just looking at a pretty picture.  I am trying to learn how to interpret my data and have had to resort to AI for help (I know, it can be a crapshoot for accuracy).  Anyway, I have been learning about Gaussian fits and residuals and with the help of AI have written a  python script that evaluates a data set.  I have attached the script and the data set.  I know I am asking a lot but if someone could please evaluate the script ( you will need to alter the data file path) to see if I am on the right track I would appreciate it.  This data was collected with the Pettit designed patch yagi at a declination of 40° from my location in Traverse City, MI.

Thanks!

[Mike Otte]

I have heard mention of Gaussian fitting before in the SARA group but never was too excited because of my limited exposure to gaussians in staistics. Does it look like a bell curve?  But your example has opened new possibilities with fitting multiple gaussians to a HI spectrum.  i am impressed and was reading more last night.  It shows the arms of the milkyway much more distinctly than counting the peaks and not even seeing some of the peaks.  The spectrum is certainly a mixture.

I have come across this in my previous work at a chemical separation factory.  We would do "profiles" to see where the various chemicals were coming out using liquid chromatography. 

Thankyou for the  "aha" moment. I will study this more but I am not sure how to help you. 

One thing is the use of  V_LSR. I don't see the correction. i only see the measured velocity.

On Wed, Apr 8, 2026 at 11:02 AM JERRY TAYLOR <gfaj...@gmail.com> wrote:

I am trying to interpret the Hydrogen line data I have been collecting rather than just looking at a pretty picture.  I am trying to learn how to interpret my data and have had to resort to AI for help (I know, it can be a crapshoot for accuracy).  Anyway, I have been learning about Gaussian fits and residuals and with the help of AI have written a  python script that evaluates a data set.  I have attached the script and the data set.  I know I am asking a lot but if someone could please evaluate the script ( you will need to alter the data file path) to see if I am on the right track I would appreciate it.  This data was collected with the Pettit designed patch yagi at a declination of 40° from my location in Traverse City, MI.

Thanks!

--
--
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 visit https://groups.google.com/d/msgid/sara-list/d6c33038-5228-42e9-a247-0a6e1c384ae7n%40googlegroups.com.

--

Mike Otte W9YS

[JERRY TAYLOR]

Mike,

Thanks for the reply. Yes, based on my limited knowledge a Gaussian fit is a mathematical model for fitting data within one or more bell shaped curves. Hence, if done correctly, you can pull out more detailed information regarding HI cloud characteristics 

The X axis is labeled as VLSR so I assume the software did the calculation correctly.  As has been discussed in SARA meetings, using AI generated scripts can possibly result in bad results unless you understand the physics/math behind the process.  Like you, I was just counting peaks in my spectrums but wanted to learn more so I dug into the Gaussian fitting of the data.  The problem is I knew almost nothing about the process and I don't have anybody locally here that could tutor me so I had to rely on Gemini and Grok for information.  As far as I can determine, the script and results are reasonable but I was hoping someone with more experience could confirm it.  As you research the process keep me posted on what you learn.  Maybe we can learn together.

Jerry

To view this discussion visit https://groups.google.com/d/msgid/sara-list/CA%2BG65tTp2MwEjE%3DiG4jzA_Es-Z%2Biz2nHRn%3DkM7ZTDw%3DVfZMH_w%40mail.gmail.com.

[Robert Hamers]

Gaussian fitting is widely used as a first-pass way of identifying how many components might be buried in a complex peak.  I use it a lot in my professional work.  A Gaussian peak is the same as "normal distribution".  The velocity distribution of atoms/molecule in a gas is given by a a closely-related shape of the Maxwell-Boltzmann velocity distribution. So one might expect that adding on a Doppler shift and fitting with a Gaussian is a reasonable starting point.  One danger with fitting to Gaussian peaks is that the fit will always get better as you include more Gaussians, simply because you have more adjustable parameters. So  including more peaks will always give a better fit, even if there is no real justification for doing so.  One more valid way of fitting is to also look at a statistical parameter known as the reduced-chi-squared,  which is a measure of the difference between the fit and your data, normalized by the standard deviation of your data. That's essentially like a signal-to-noise ratio of your fit.  If the reduced chi-squared is <1 then there is no longer any statistical justifacation for including more peaks.   Because the H data we're looking at has other features (like sharp drop-offs in intensity and velocity at the edge of the spiral arms), a Gaussian isn't necessarily valid in many cases.  So a good practice is to start by fitting to a small number of peaks, and increase the number of peaks as necessary until the reduced chi-square is just less than one. Then stop.  That will give you the minimum number of Gaussian peaks that can be justified as being significant in a statistical sense. 

[JERRY TAYLOR]

Thanks Robert.  As I was testing different Gaussian fits I wondered when I should stop. 

I appreciate the advice!

Jerry

To view this discussion visit https://groups.google.com/d/msgid/sara-list/b9d4eb4f-319a-4234-8685-c0f9c3d84e47n%40googlegroups.com.

[Mike Otte]

Jerry,

Wow! I just ran the Python script on one of my data sets.

That really cleans it up! And sees things I don't.

runfile /home/mmotte/Documents/GuassianFit/GuassianFitHIdata.py --wdir
Data loaded successfully: 400 channels
Velocity coverage: -144.9 to 249.8 km/s

Estimated baseline level: 0.00106

=== 4-Gaussian Fit Results with Physical Interpretation ===

1. Main Local Gas:
   Amplitude : 0.4000
   Center    : 25.00 km/s
   FWHM      : 18.84 km/s
   Interpretation: Narrow → cold, coherent cloud with low turbulence | Local Milky Way disk gas

2. Secondary HV Feature:
   Amplitude : 0.0500
   Center    : -120.00 km/s
   FWHM      : 14.13 km/s
   Interpretation: Narrow → cold, coherent cloud with low turbulence | High-Velocity / Intermediate-Velocity Cloud (approaching)

3. Broad Wing Component:
   Amplitude : 0.0800
   Center    : -60.73 km/s
   FWHM      : 58.88 km/s
   Interpretation: Broad → turbulent gas, blended components, or dynamical activity | Intermediate velocity gas

4. Additional Component:
   Amplitude : 0.0200
   Center    : 85.51 km/s
   FWHM      : 23.55 km/s
   Interpretation: Narrow → cold, coherent cloud with low turbulence | Receding gas (possibly outer arm or faint HV feature)

RMS residual (all)      : 0.07258
RMS residual (wings)    : 0.00000
Improvement vs 3-Gauss  : ~-178.7% lower RMS

I am impressed!

Had to adjust the IFAvg output file from space delimited to Comma delimited.

BR

Mike

To view this discussion visit https://groups.google.com/d/msgid/sara-list/CAMMXHa7e9Au6femWzmsAUGCLTS4rmggxvzApYPwRd02Q7OSiHg%40mail.gmail.com.

--

Mike Otte W9YS

[Mike Otte]

Still question the V_LSR calc?

--

Mike Otte W9YS

[fasleitung3]

Mike,

in order to calculate the LSR correction you would need to have the time and date of the observation and the direction where you were pointing.

Best regards,

Wolfgang

[Alex P]

HawkRAO VLSR Calculator

Calculates the topocentric radial velocity of an observer in a given direction 

(equatorial coordinates) and observer's latitude and longitude, and UTC time.

http://f4klo.ampr.org/vlsrKLO.php

Use of the Calculated Radial Velocity: The radial velocity calculated here is the observer's topocentric velocity towards the specified direction referenced to the LSR (local standard of rest).  However, in most cases the spectral line source (e.g. W3) has its own velocity w.r.t. the LSR, and so the observed velocity will be different to that calculated above.  As the observed velocity will be different for each observatory and varies over the day and year, in order to compare results between observatories and over time, a common VLSR reference is assigned to the source signal.  Use the calculators below to find a VLSR, or an expected observation frequency.

- Alex

============================================

[Mike Otte]

Jerry,

using the calculator, it asks for things that the python script has no way of knowing.  I think the V_LSR on the x axis is just velocity.

Now do we add this or subtract it?

--
--
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 visit https://groups.google.com/d/msgid/sara-list/e24e6075-3e7a-4ec9-8114-c07c6cd6a280n%40googlegroups.com.

--

Mike Otte W9YS

[b alex pettit jr]

    +- 32 km/sec  per 'day'

 The calc shows the variation = you need to Subtract this from your data

Alex

================================

On Saturday, April 11, 2026 at 07:39:39 AM EDT, Mike Otte <mike....@gmail.com> wrote:

Jerry,

using the calculator, it asks for things that the python script has no way of knowing.  I think the V_LSR on the x axis is just velocity.

Now do we add this or subtract it?

On Sat, Apr 11, 2026 at 4:19 AM 'Alex P' via Society of Amateur Radio Astronomers <sara...@googlegroups.com> wrote:

HawkRAO VLSR Calculator

Calculates the topocentric radial velocity of an observer in a given direction 

(equatorial coordinates) and observer's latitude and longitude, and UTC time.

http://f4klo.ampr.org/vlsrKLO.php

.

[b alex pettit jr]

I ran this calc to correct 24 hr drift scans by solving in 1 hr steps over the same time period,

solving a 5th order polynomial fit, and applying the correction to the data set.  quite effective

Alex

========================================================================

[JERRY TAYLOR]

Mike,

I have some reservations about that code until someone with more experience can look at it.  Your results look suspicious to me.  I think the RFI and tilted spectrum in your raw data may confuse the script.

Try this. Collect a few hours of data using Sdr# with the If average plugin.  Then process it with Alex Pettit’s Hline software.  This will flatten the spectrum and remove RFI.  Then run the python script for Gaussian fits on the csv file Hline created.  This is the workflow I use and seems to work.  Let me know how that works and if you have any questions.

Jerry

--
--
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 visit https://groups.google.com/d/msgid/sara-list/117482755.589525.1775909767373%40mail.yahoo.com.

[JERRY TAYLOR]

Mike,

I will see if I can run the data you sent through my process and see what happens.

Jerry

[b alex pettit jr]

Mike, That is very low amplitude data . What antenna/software are you using ?

HL3D

Corrected

RFI filtered

files attached

[andrew....@googlemail.com]

Below is an example of H-Line data collection in ezCol from my 86cm 4x4 Ptarmigan Triffid Dipole Array at LRO (www.astronomy.me.uk) for comparison. This is not a particularly high performing antenna – built for battlefield communications rather than satellite or sky work. Alex P. will also point out that ezCol gives relative results without specified hot/cold sky calibration. This is “calibrated” using 1423MHz offset frequency refence samples every other sample. The peak is also at approx.. max. signal intensity during 24-hour period (the Milky Way crosses the beam twice in every 24 hours during drift scans, one crossing has substantially lower signal intensity than the other – this is the higher one).

 

Another interesting aspect of H-Line monitoring is baseline shape – on this Ptarmigan Array with RTL-SDR Dongle (SDR) and SAWBird H1 LNA, I get a reasonable flat baseline as shown below – however, on my 150cm dish with RTL-SDR it is substantially curved in an upward direction at sides similar to your plots below – and if I use my Ettus B210 clone board at 10MHz bandwidth then the baseline is all over the place! Certainly, gives a good reason for using RTL-SDR as initial device for people new to this aspect of the hobby and move onto more specialist/expensive SDRs later (general rule of thumb – clearly some people are more advanced when they start! I give the advice for sake of anyone else on this mailing list who is considering starting in H-Line work – just like astrophotography start small learn your skills before buying that 14” SCT on that mountainous red hunk of steel that dares to call itself a mount!! People who do not do astrophotography will wonder what on Earth I am talking about…..)

 

Worth saying ezRA does decent job of automatically flattening the baseline – but is still limited to 1^st order polynomial equation (flattening a straight line) – most software does not provide automatic second-order polynomial flattening as standard (flattening a curve). Jason (Burnfield – member here) uses Excel spreadsheets of his own making and within those he has provided option to use 2^nd order polynomials, but I don’t know of any others. As an aside, does anyone know of any those will do this?

 

Andy

 

 

--

--
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 visit https://groups.google.com/d/msgid/sara-list/816082695.1997648.1775931698071%40mail.yahoo.com.

[Mike Otte]

As you can read on the Old data picture, this was Aug 2021.  I have to check the log books to see what amps/LNAs I was running but it was on a 3M dish. EZRA was just started to be developed and i was testing some.  but this data was  IFavg.  I don't remember if the time conversion was GMT or RA.  

The point is to test!   Jerry's  AI generated script and see if it makes sense   

It does it all: 

It quiets the noise

It  checks the extends.

It levels the field.

It converts to velocity  (I don't think it subtracts LSR)

It Gaussian Fits

It reports the data for the fits

My test was a random old data selection that looked like it had a couple peaks.

Remember that this must be spectrum data.

I used Spyder 6 to run it.

I only had to add commas to the data to make it work,   CSV.

We're on the edge of science! exciting!

I went to a wedding today, so I didn't have time to look any more.

GL

Mike

To view this discussion visit https://groups.google.com/d/msgid/sara-list/002201dcc9ea%24802a0540%24807e0fc0%24%40googlemail.com.

--

Mike Otte W9YS

[Andrew Thornett]

Forgive me, Mike. I didn't mean to imply any criticism of your work in my comments, only to give some suggestions to newcomers (not your good self who does not need them).

With regards to yourself, the only bit of my post that is relevant is the plot which is meant simply to provide a comparitor plot - and not necessarily a particular good comparitor at that! Alex was commenting on low strength signal in your plots so I was posting a plot from separate system to demonstrate strength on that system.

Again, my apologies if my comments appeared to come across as critical - they were not meant to do so.

Congratulations to the newly wed couple!

Andy

Sent from Outlook for Android

---

From: sara...@googlegroups.com <sara...@googlegroups.com> on behalf of Mike Otte <mike....@gmail.com>
Sent: Saturday, April 11, 2026 10:26:59 PM
To: sara...@googlegroups.com <sara...@googlegroups.com>

To view this discussion visit https://groups.google.com/d/msgid/sara-list/CA%2BG65tRnB6zAvpxu0tBDGwUs6fbkjMGedLkx4WyMG3-OFbgkOA%40mail.gmail.com.

[Mike Otte]

I apologize if I came across too snippy. I am sorry.  

I just thought you wandered off of the topic of the thread.

BR

no problem

Mike

 

To view this discussion visit https://groups.google.com/d/msgid/sara-list/FRWP194MB275255FC3C081531FCEC7A2DFF262%40FRWP194MB2752.EURP194.PROD.OUTLOOK.COM.

--

Mike Otte W9YS

[Eduard Mol]

Hi all, bit late to the discussion but maybe some points to add here that might be helpful:

For LSR correction, I often use this script developed by Tammo- Jan from CAMRAS. 

https://gitlab.camras.nl/dijkema/HPIB/blob/185d241ad9bd7507ed90c9fa91fe0a63009d3eee/vlsr.py

For fitting n-th polynomials through the baseline I use the numpy.polyfit() function in my python processing scripts. (I never do more than a 3rd order fit, anything more than 2nd or 3rd order has a higher risk of introducing false positives…) 

This would be fairly straightforward to implement in EZRA. 

Op zo 12 apr 2026 om 03:36 schreef Mike Otte <mike....@gmail.com>

To view this discussion visit https://groups.google.com/d/msgid/sara-list/CA%2BG65tRxZJfHh1JVx8FcU7fY8RGRcD8_cFaFTp2tTHtnHF_Gww%40mail.gmail.com.

[Mike Otte]

Thanks Eduard for the vlsr script.

********************

I got way too excited about the  AI script.

I have been line by line analyzing the script and it doesn't do what I initially thought it did

sorry

Mike

To view this discussion visit https://groups.google.com/d/msgid/sara-list/CAFcm1Q4mM5q%3DGtBm7Ly-mNrPNpRHSdGMRRJSxY-P1W393%3DO4kQ%40mail.gmail.com.

--

Mike Otte W9YS

[Marcus D. Leech]

On 2026-04-13 11:12, Eduard Mol wrote:

Hi all, bit late to the discussion but maybe some points to add here that might be helpful:

For LSR correction, I often use this script developed by Tammo- Jan from CAMRAS. 

https://gitlab.camras.nl/dijkema/HPIB/blob/185d241ad9bd7507ed90c9fa91fe0a63009d3eee/vlsr.py

For fitting n-th polynomials through the baseline I use the numpy.polyfit() function in my python processing scripts. (I never do more than a 3rd order fit, anything more than 2nd or 3rd order has a higher risk of introducing false positives…) 

This would be fairly straightforward to implement in EZRA. 

My own experience with using polynomial fits to estimate the baseline has been "rather mixed" with satisfactory results in some case, and not others.

Our current long-term project is taking daily measurements of the absorption line in M8, and I wanted to isolate the absorption spike.   I ended up using
  a high-order median filter to estimate "everything that isn't the spike", and subtracting that out, then inverting, to give me a nice clear, positive-going,
  spike.

My suspicion is that there will inevitably be some customization required both along the "individual instrument" axis, and the "what are you observing?" axis.

I think that likely the "cleanest" baseline estimator is one that keeps all the instrumentation parameters the same, and switches the front-end to a nice clean,
 flat, noise-source.

The problem with "oblivious" baseline estimation is that it has imperfect information to work with.

To view this discussion visit https://groups.google.com/d/msgid/sara-list/CAFcm1Q4mM5q%3DGtBm7Ly-mNrPNpRHSdGMRRJSxY-P1W393%3DO4kQ%40mail.gmail.com.

[Eduard Mol]

@Marcus Just to clarify, I usually do a different baseline correction procedure, recording a set of “off-target” spectra and dividing those from the “on-target” set. Often there is still some residual slope or curve left after this. That’s what I then remove with curve fitting. (This residual curve is usually also of much lower amplitude and more simple than the original bandpass response curve.)

For galactic hydrogen this does not work very well because there is not really an area without hydrogen to do the “off-target” measurement. I have found that with my setup and my airspy SDR fitting a third order curve through the background of the “raw” spectra produces acceptable results. However, this is not always the case especially with a different setup, SDR or target. Your mileage may vary…

Op wo 15 apr 2026 om 16:46 schreef Marcus D. Leech <patchv...@gmail.com>

To view this discussion visit https://groups.google.com/d/msgid/sara-list/a37d88ee-d2b9-4211-a679-8c987251299f%40gmail.com.

[Marcus D. Leech]

On 2026-04-15 12:07, Eduard Mol wrote:

@Marcus Just to clarify, I usually do a different baseline correction procedure, recording a set of “off-target” spectra and dividing those from the “on-target” set. Often there is still some residual slope or curve left after this. That’s what I then remove with curve fitting. (This residual curve is usually also of much lower amplitude and more simple than the original bandpass response curve.)

For galactic hydrogen this does not work very well because there is not really an area without hydrogen to do the “off-target” measurement. I have found that with my setup and my airspy SDR fitting a third order curve through the background of the “raw” spectra produces acceptable results. However, this is not always the case especially with a different setup, SDR or target. Your mileage may vary…

This is pretty consistent with my experience, also.

To view this discussion visit https://groups.google.com/d/msgid/sara-list/CAFcm1Q4i0s1AxyCcC3U9Tz9%3DVmm3dg3-TF6TG-nwebgq6QxBSQ%40mail.gmail.com.

[b alex pettit jr]

To remove the "system" curve, IFaverage does a good job of that.

You can either cover the feed with an absorber or  put a 50 Ohm resistor across the LNA input prior to acq

and store a "background correction" which is applied to all data.

The plots displayed & recorded look like this ( not perfect, but .. )

Dec + 40 dg  

RA:1100

RA:1300

RA:2030

Alex

Eduard :

[Eduard Mol]

Hi Alex, 

I never use the “background” option in IFaverage, because once it’s applied it cannot be undone. By recording my “on-target” and “of-target” spectra separately and doing the bandpass correction afterwards, I can always go back if  the background correction gives bad results or if the “off-target” set is contaminated with RFI. 

The trick with an absorber or a 50ohm load will work for galactic HI, but you have to keep in mind that you will be adding ~300K of noise (significantly more than the ~100 - 150K system temperature of a typical amateur HI setup). For strong galactic hydrogen signals where the SNR is typically >100, you will barely notice the difference. But for much weaker signals with lower SNR like HVCs, extragalactic HI or masers, this extra noise will be an issue.

Op wo 15 apr 2026 om 19:24 schreef 'b alex pettit jr' via Society of Amateur Radio Astronomers <sara...@googlegroups.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 visit https://groups.google.com/d/msgid/sara-list/9953686.550302.1776273838945%40mail.yahoo.com.

[b alex pettit jr]

Hi Eduard,

Yes, I was referring  to Galactic H1.. 

Actually, you don't "add" 300K.. I made an excel process which I believe 'simulates' what IFavg accomplishes.

You do capture a ~300K reference, but then normalize it that so the center 75%  averages 1.000  = Unity Gain.

That result is the Divisor to correct the System Gain variation over the bandwidth.

I made a writeup Somewhere but not as yet located it.

Here is the Excel  .. not well 'annotated.

It is a real signal from  50ohm > SawBird Hi > ASmini ..

( I then imposed attens between the SBird & SDR for the steps )

Since all files are thereafter similarly corrected, it should Not influence your data ... even at very low levels.

Alex

[Marcus D. Leech]

On 2026-04-15 16:02, 'b alex pettit jr' via Society of Amateur Radio Astronomers wrote:

Hi Eduard,

Yes, I was referring  to Galactic H1.. 

Actually, you don't "add" 300K.. I made an excel process which I believe 'simulates' what IFavg accomplishes.

You do capture a ~300K reference, but then normalize it that so the center 75%  averages 1.000  = Unity Gain.

That result is the Divisor to correct the System Gain variation over the bandwidth.

I made a writeup Somewhere but not as yet located it.

Here is the Excel  .. not well 'annotated.

It is a real signal from  50ohm > SawBird Hi > ASmini ..

( I then imposed attens between the SBird & SDR for the steps )

Since all files are thereafter similarly corrected, it should Not influence your data ... even at very low levels.

Alex

Yes, this is pretty much what I do.  I use the reference data (from an absorber) to help define the "shape" of the baseline, and then scale it appropriately to the data
  at hand.

On Wednesday, April 15, 2026 at 03:45:52 PM EDT, Eduard Mol <eddiem...@gmail.com> wrote:

Hi Alex, 

I never use the “background” option in IFaverage, because once it’s applied it cannot be undone. By recording my “on-target” and “of-target” spectra separately and doing the bandpass correction afterwards, I can always go back if  the background correction gives bad results or if the “off-target” set is contaminated with RFI. 

The trick with an absorber or a 50ohm load will work for galactic HI, but you have to keep in mind that you will be adding ~300K of noise (significantly more than the ~100 - 150K system temperature of a typical amateur HI setup). For strong galactic hydrogen signals where the SNR is typically >100, you will barely notice the difference. But for much weaker signals with lower SNR like HVCs, extragalactic HI or masers, this extra noise will be an issue.

--
--
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 visit https://groups.google.com/d/msgid/sara-list/321016270.2987530.1776283322318%40mail.yahoo.com.

[b alex pettit jr]

Vulcan Mind Meld  :)

=================

[b alex pettit jr]

Hi Eduard,

I've never made this measurement = Interesting !

This is a set of data processed via SDR# > IFavg ( background corrected )  HL3D > Rinearn

PRT 1.2m Dish System  

FFT blocksize 512

300 sec averages

12 averages ( 1 Hr ) plotted

UNFILTERED

Dec + 40 deg   RA  1130-1230 hrs

H1 Profile = 1.6K above cold sky

System Signal ~ 0.08 dB above cold sky

Regards,

Alex P

[Stephen Arbogast]

Alex,

By  IF  Averaging  do you  think  the author  may  have incorrectly  labeled  his original  software?  I have never  understood why it is called  IF  Averaging. Maybe  he  was using  a  down  converter?   I  don't know...  

I am  using  the  very  small  Discovery Dish  with  ezRA suite   with  many options  to  post process my raw data.  I am getting  good results  but  not as clear  and  strong  results  as people  with  bigger dishes   but  hanging in there.

Kind of  sad  that  we are   a  week  away  from  Discovery  Drive  funding campaign  and only   50 %   of funding  has been raised.

But  hoping  for the  best.

[b alex pettit jr]

Stephen ,

What is the ezRA suite post process option to convert the HLine spectra scaling of

Relative RMS Power  and/or  Fraction of Y Auto Scale  to   dB relative to Cold Sky   ?

Alex

=============================

On Wednesday, April 15, 2026 at 09:11:52 PM EDT, 'Stephen Arbogast' via Society of Amateur Radio Astronomers <sara...@googlegroups.com> wrote:

Alex,

  

I am  using  the  very  small  Discovery Dish  with  ezRA suite   with  many options  to  post process my raw data.  I am getting  good results .....

[b alex pettit jr]

I guess, more importantly, is the data stored in a format which can be converted to & plotted as such ?

===========================================================