Visual notch

User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Visual notch

Postby ramdor » Wed Feb 08, 2023 6:49 pm

Just having a little dabble, a welcome break from the meter monotony.

Visualisation of the notch. Note: The mid point is dropped by 100dB. Visually it is a simple high/low pass filter either side of centre, and not an exact representation of the notch filter. There is an option to use it in setup.

Cheers !

visual_notch.png
visual_notch.png (139.25 KiB) Viewed 1259 times
Trucker
Posts: 304
Joined: Wed Nov 03, 2021 5:16 pm

Re: Visual notch

Postby Trucker » Wed Feb 08, 2023 7:26 pm

Very nice. My Kenwood TS-890 has that for the manual Notch filter. It is a useful, visual aid for me.
James
WD5GWY
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: Visual notch

Postby w-u-2-o » Wed Feb 08, 2023 9:25 pm

ramdor wrote:Visualisation of the notch. Note: The mid point is dropped by 100dB. Visually it is a simple high/low pass filter either side of centre, and not an exact representation of the notch filter. There is an option to use it in setup.


Richie,

Unfortunately, I have to strongly recommend this not be incorporated as a feature. It's a gross approximation and thus violates the tradition and paradigm of technical accuracy and precision in openHPSDR radios.

There is no indicator, readout, dial, gauge or display in Thetis that is not based on actual measured data of a real signal. Things are generally accurate down to fractions of a dB, and even those few measurements that are forced to be approximations (e.g. the noise floor) are still within a dB or so of precise measurement thanks to the use of some outstanding algorithms.

While I realize there will be an option to enable or disable this feature, if people turn it on the vast majority will believe it is a true representation of what the notch is doing to the passband, and that will be wrong.

I would urge you to provide a technically accurate indication of notch filter operation. To that end, the spectrum display should be fixed (it looks broken to me at the moment) and then perhaps be made available as is, or maybe a "pana-spectrum" display, or perhaps even better as a picture-in-picture inset in the spectral display area.

Respectfully,

Scott
User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Re: Visual notch

Postby ramdor » Thu Feb 09, 2023 2:39 am

@Scott

I spoke to Warren about this back in early 2022, the consensus being that it would be a lot of work to do it 'properly'. Some of what follows is info from email exchanges between Warren and myself regarding this:

[Warren]
First, for reference, you may want to refer to the RXA block diagram on p. 16 of the “WDSP Guide”:
https://github.com/TAPR/OpenHPSDR-wdsp/blob/master/WDSP%20Guide%2C%20Rev%201.20.pdf

1) The data for the panadapter/waterfall is taken BEFORE any of the processing in this block diagram occurs, it is unfiltered, unprocessed data.
2) On the RXA (receiver) block diagram, the block at the right side of the top row is very important for our discussion – it’s called the “Notched Bandpass”. This is literally a SINGLE filter I dynamically design that both cuts the bandpass out of the entire spectrum AND cuts any number of manual/fixed notches that are specified into the bandpass. The benefits of a single filter are that it requires only the compute cycles of a single filter and only the latency of a single filter to do all that. This design can also create very deep and well-controlled notches. (Alternatively, I’ve seen designs that have used a bandpass filter and then a bunch of separate IIR notch filters to create often sub-optimal notches.)
[end quote]

Warren goes on to say that it is not possible to have a flag on in GetPixels() and see the notches in the data which is used for the panadapter/waterfall display – because the notches do not exist at this point. The notches only exist when the passband is cut out of the spectrum.

There is another method to obtain/display the data, and it can be obtained as part of the 'post filter display send', and is used to display the spectrum display that you mention (broken?). We also talked about displaying the auto notches, and that would be even more complicated and require a bunch of work and may not result in something worth using or viewing.

So as you can see, it has been something that has had a bit of thought behind it. At the end of the day it is just a visual approximation, and will be noted as such on the Setup->Options tab to enable/disable it.

Also, with respect, you have to remember that as this is my version I can put anything I like in it. Anyone can get the source, rip stuff out, rebuild it and release it anywhere, even here with moderator approval.

Cheers.
User avatar
W2PA
Posts: 166
Joined: Sun Apr 09, 2017 6:34 pm
Location: LaGrangeville, NY
Contact:

Re: Visual notch

Postby W2PA » Thu Feb 09, 2023 2:07 pm

Richie,

Thanks for sharing this interesting discussion. It makes perfect sense that the entire contours of the bandpass should be constructed as a single filter.

And this all relates to a suggestion I made a while back about possibly using the auto-notch result to create a new manual (multi) notch. I can see how this might be difficult.

Anyway, as a visual aid the shading of a notch in the display, as it's done now, is completely adequate for me to see the notch. And to Scott's point, you're seeing the *setting* of the notch, not simulating the result. The modifying of the panadapter data wouldn't add anything useful visually beyond that for me. Just another opinion.
73,
Chris, W2PA
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: Visual notch

Postby w-u-2-o » Thu Feb 09, 2023 2:38 pm

ramdor wrote:@Scott

I spoke to Warren about this back in early 2022, the consensus being that it would be a lot of work to do it 'properly'.

Yes, I remember that, and I also remember it would be a ton of work to do it properly on the existing panadapter and waterfall displays.

There is another method to obtain/display the data, and it can be obtained as part of the 'post filter display send', and is used to display the spectrum display that you mention (broken?). We also talked about displaying the auto notches, and that would be even more complicated and require a bunch of work and may not result in something worth using or viewing.

Yes, exactly! As described on page 29 of the WDSP Guide 1.20. This is the spectrum passband output post passband filtering and post notch filtering. It might even be post ANF.

It's not clear how using this would be more complicated since it already exists and contains the data necessary. However, the current implementation to display it by means of the "Spectrum" entry on the spectral display pull-down menu appears to not be working correctly. If that could be made to work, and then a another split screen mode be added, call it "pana-spectrum", then you could add/modify/delete notches all day long using the existing UI mechanisms you developed on the panadapter and see the real, correct results on the spectrum display.

So as you can see, it has been something that has had a bit of thought behind it.

Never said it didn't, and didn't mean to imply otherwise.

At the end of the day it is just a visual approximation

And that's what I'm objecting to.

Also, with respect, you have to remember that as this is my version I can put anything I like in it. Anyone can get the source, rip stuff out, rebuild it and release it anywhere, even here with moderator approval.

That's being a bit unfair, I think. I've defended the developer's rights in that respect many times. Yours in particular, if you remember.

Anyhow, I don't want this to become an argument. I'm merely providing some input on why this might not be a good idea. No different than discussing any other new or modified functionality, like you and I and others have many times in the past. In this case it violates the long-held paradigm that the program delivers real answers, not made-up answers. Some portion of the user base will not be sophisticated enough to understand that this is a made-up answer. And new users who come after implementation might never understand. They will turn it on and think it real. Obviously it's completely within your rights to agree or disagree.

Cheers,

Scott

Return to “Thetis”