Page 1 of 1

SSB BW Settings...

Posted: 30 Apr 2022, 15:27
by DL3MHT
Hi Guys!

I am using FW 2.00i Beta and listening on 40m (LSB of course).
To my ears it sounds like there is a mismatch between shown and actual BW settings.

4k0 Setting could be ok
3k0 on Display sounds like 1k8 - Definitely the narrowest / brightest setting.
2k4 Setting could be ok
1k8 Setting could be the real 3k0 setting

Anybody else experiencing this?
Maybe I still have some bug in the RX path somewhere, but audio is actually generally really pleasant.

Thanks,
Tobias

Re: SSB BW Settings...

Posted: 01 May 2022, 16:32
by msax
HI,
The same here. I don't believe it's a bug, but rather a limitation of the mcu's processing power. However, it still sounds better to me on 1k8 for weaker signals than ft-450 in any dsp setting.

Re: SSB BW Settings...

Posted: 01 May 2022, 19:12
by DL3MHT
Thanks for confirming, then i don't have to mess around with the HW anymore.

I still hope it's a minor SW bug that is easily fixed.
Instead if going down with the BW as displayed it actually goes up.

1.8k -> 2.4k -> 3k -> 4k -> 1.8k -> ...
But displayed as.
3k -> 2.4k -> 1.8k -> 4k -> 3k -> ...

And i totally agree:
The RX audio quality is excellent. I made a small YouTube Livestream yesterday and recorded it from the headset socket into a Motu M2 audio Interface. Especially the last 5minutes or so of that video show how amazing this thing works.

Now I just need to check my TX path. I have to get a feeling of how loud i have to speak to get decent modulation. But maybe the guys also included a digital microphone AGC 😂.

Tobias

Re: SSB BW Settings...

Posted: 01 May 2022, 20:09
by msax
Try running that audio through fft scope on the pc since it's connected already, and you'll see filtering in the action. I think it's just perceived width that maybe sounds wrong, but it actually is 3k, 2k4 and 1k8.

Re: SSB BW Settings...

Posted: 01 May 2022, 20:39
by John
I thought that was odd, as well. But if I understand things correctly, (and I may not,) these digital audio bandpass filters are no longer just 'low-pass' filters of varying bandwidths like analog filters used to be, but are actually designed to be 'band-pass' filters that cover just the range of frequencies necessary to improve intelligibility, tossing out both high and low frequencies that aren't useful, giving them a different sound than the old low-pass circuits had, confusing those of us who have grown accustomed to the bassy sound of the old analog filter circuitry.

Re: SSB BW Settings...

Posted: 02 May 2022, 07:33
by DL3MHT
Good idea with using the FFT of e.g. Audacity.
I am just moving all data and applications from that Laptop to a new machine though. Maybe i do it if the new PC is ready to use.

I think i will just post the question to that other Q&A thread, then the author's can say: You are right or you are wrong.

Not sure how sophisticated the digital bandwidth filter on an 8-Bit MCU can be John. If it would be something psychoacoustic-clever, then i would not mention a KHz number at all and just call it Audio BW setting #1,2,3,4.

-Tobias

Re: SSB BW Settings...

Posted: 02 May 2022, 08:19
by Faraaz VK4JJ
DL3MHT wrote: ↑02 May 2022, 07:33 Good idea with using the FFT of e.g. Audacity.
I am just moving all data and applications from that Laptop to a new machine though. Maybe i do it if the new PC is ready to use.

I think i will just post the question to that other Q&A thread, then the author's can say: You are right or you are wrong.

Not sure how sophisticated the digital bandwidth filter on an 8-Bit MCU can be John. If it would be something psychoacoustic-clever, then i would not mention a KHz number at all and just call it Audio BW setting #1,2,3,4.

-Tobias
Tobias,

I have deleted your post from the Q&A topic. If you think there is an error in the bandwidth filters, please provide some evidence i.e. spectrum analysis, scope outforms, so that can be presented to the developers who have very limited time.

Re: SSB BW Settings...

Posted: 02 May 2022, 08:42
by DL3MHT
My apologies Faraaz!

For me it is very obvious that 3kHz setting is not 3KHz bandwidth and as a SW developer myself it did not seem to be such a big task to check if the referenced text strings in the FW align with the actual used mode.

Anyway, as I also wanted to test FT8 with an USB Soundcard, I and am happy to take a few screenshots with the different bandwith settings when I get to it, that should definitely show if I am right or deaf :D.

-Tobias

Re: SSB BW Settings...

Posted: 02 May 2022, 15:48
by DH2LAB
DL3MHT wrote: ↑02 May 2022, 08:42

Anyway, as I also wanted to test FT8 with an USB Soundcard...
I just ran it with a USB sound card and for me the widest filter setting worked best.
But I haven't done extensive testing yet.


https://www.ebay.com/itm/172709701419

Re: SSB BW Settings...

Posted: 26 May 2022, 10:32
by DL3MHT
Follow up on my complaint that the 3KHz bandwidth setting is the "brightest".

Since (tr)uSDX is no longer open source, I had a peek at the uSDX source code as Guido probably didn't rewrite everything from scratch, it might still be close enough to get an idea what is going on inside the atmega.

In a way I have to eat my words. Marko is kind of right, that the label shown on the display is just a name given to a set of parameters for a digital filter. The name should reflect what the filter does, but of course there are infinite ways to design a digital filter and it having a bandwidth/label that says 3KHz does not imply that it has to be a flat filter.

In the uSDX source code you find the stuff in function filt_var. A comment says the filters were designed with Micromodeler DSP. You also see that the text labels are just using the variable "filt" to lookup "filt_label". So filt 1 refers to "3000" label and ends up in switch-case 1 for function filt_var, which has a comment that says it is a 4th order IIR filter 0-2900Hz. So all looks normal in the uSDX code. There is also a comment above that says a bug was fixed in the common section for all filters that introduced 6dB attenuation and hence the coefficients were adapted.

Having said that, I still recorded a QSO with the different bandwith settings and using Audacitys FFT function one can see that the area 1KHz - 2KHz is greatly amplified with the 3KHz BW setting.
Alas in order to do this PROPPERLY I would need a test transmitter with known tx audio envelope and a noise generator. Or somehow inject a noise signals to the I and Q ADC inputs. All a hell of an effort and I don't have the proper tools. So I can't give you the definite 100% correctly measured frequency responses of the four digital filter settings on firmware revision X. :(

So what does that mean? I still think the 3KHz setting sounds the brightest. Apart from the possibility of the label string being out of sync with the actual filter coefficients, there is the possibility that there is a bug in the coefficients, or that the 3KHz filter was deliberately designed in a way to give an extra presence boost. I don't know.

-Tobias