开云体育

ctrl + shift + ? for shortcuts
© 2025 开云体育
  1. QRPLabs
  2. Topics

Torturing QMX+ CESSB...

 

Hans,
?
Very nice.
?
One of the first digital songs I ever heard was a copy of Thunderstruck downloaded over a 300 baud modem.? I believe it was sampled at 8000Hz and 8-bit mu-law encoded.? The quality was quite similar to this as I remember it.
?
There is definitely some "crackle" on your audio.? They might have called it "multiplicative noise" back in the day since it comes and goes with the content (i.e., is amplitude dependent).
?
It is hard to say why since there is a lot going on.
  • Original MP3 encode
  • MP3 decode on your PC
  • Digital processing on QMX
  • Modulation to 40m
  • Transfer over the coax/attenuators to the second QMX
  • Demodulation to baseband
  • Digital processing on QMX
  • Encoding into MP3 on your PC
  • MP3 decode on my PC
  • My audio path to my headphones
And you can easily see some pops the Audacity spectrogram.
Here is the part where the singer says "leave it be" (starting at 1403703 samples):
You can see the sharp narrow vertical bands in the two /iy/ vowels (e sounds in "leave" and "be").
The pops occur between the drum beats during steady vocalization.
In the waveform view it appears as the nice /iy/ vowel being replaced by a near sinusoid.
(second pop shown):
And the length of the apparent replacement is around 90 samples (around 1msec).
It makes me wonder what buffer sizes you use for your audio pipeline.? :-)
?
Probably it is not worth going after the crackle or the pops since the noise in an HF QSO will likely mask such subtle artifacts.
But I thought I'd mention it since you have the entire processing pipeline in your head and it might trigger an "aha".
?
Great work!
?
73, Mike KK7ER
?
 

But how will the pops affect digital modes without FEC?
 

Hi Mike

Don't worry about pops right now. The current developments have been a collecrion of experiments, not a well-integrared whole firmware.?

Clock domain discontinuities are a common problem in audio systems. Here we have:

1. Who knows what, on the source audio (a YouTube video of AC DC's 1979 Highway to Hell track).?

2. 48ksps streamed from PC to QMX+ which will be something close to 48,000 Hz but not exactly

3. 48ksps ADC (Receive I and Q) running on QMX+ which is also not exact. The USB audio processing (in and out) is in here.?

4. 12ksps transmit DSP in QMX+ which is actually running out of sync with the 48ksps ADC, which itself leads to two very small pops per second.?

5. Then RF to the QMX+ antenna port, 63dB attenuators, to the RF port of my QMX 80-20.?

6. The QMX also has its own different 48ksps again which again is different to the 48ksps clock domain of the PC USB port.?

All these differences are a problem and need to be dealt with somehow. Either by phase locking clock domains or by putting up with occasional discontinuities (missing a sample or interpolating one). If done carefully these discontinuities are hard to detect and not a big deal.?

Don't worry, in the final prepared SSB firmware I'll take care of any sync issues, between different clock domains.?

73 Hans G0UPL


On Fri, Oct 18, 2024, 18:39 Mike KK7ER via <groupio=[email protected]> wrote:
Hans,
?
Very nice.
?
One of the first digital songs I ever heard was a copy of Thunderstruck downloaded over a 300 baud modem.? I believe it was sampled at 8000Hz and 8-bit mu-law encoded.? The quality was quite similar to this as I remember it.
?
There is definitely some "crackle" on your audio.? They might have called it "multiplicative noise" back in the day since it comes and goes with the content (i.e., is amplitude dependent).
?
It is hard to say why since there is a lot going on.
  • Original MP3 encode
  • MP3 decode on your PC
  • Digital processing on QMX
  • Modulation to 40m
  • Transfer over the coax/attenuators to the second QMX
  • Demodulation to baseband
  • Digital processing on QMX
  • Encoding into MP3 on your PC
  • MP3 decode on my PC
  • My audio path to my headphones
And you can easily see some pops the Audacity spectrogram.
Here is the part where the singer says "leave it be" (starting at 1403703 samples):
You can see the sharp narrow vertical bands in the two /iy/ vowels (e sounds in "leave" and "be").
The pops occur between the drum beats during steady vocalization.
In the waveform view it appears as the nice /iy/ vowel being replaced by a near sinusoid.
(second pop shown):
And the length of the apparent replacement is around 90 samples (around 1msec).
It makes me wonder what buffer sizes you use for your audio pipeline.? :-)
?
Probably it is not worth going after the crackle or the pops since the noise in an HF QSO will likely mask such subtle artifacts.
But I thought I'd mention it since you have the entire processing pipeline in your head and it might trigger an "aha".
?
Great work!
?
73, Mike KK7ER
?
 

everyone says the low frequencies should be somewhat suppressed because they add little to SSB DX intelligibility.?

It depends lot on the voice in question TBH? ? A deep male voice is badly affected by low cutting too soon, Less intelligible and sounds 'very thin' locally.? ??
Best to make any roll off points adjustable - ideally with a local /DX switch?
 

I will add a LOT of configurable options in the SSB and CESSB implementation so that people can experiment with what works best for them; default values will be what I have found most suitable in my testing, but it will allow experiment.?

73 Hans G0UPL
 

>The difference between your video >and their video is monetization.
--
Good answer Shirley.
I recall that our blues band could only play originals in curtain clubs because they were not registered and dues paid with the various Performing Rights Societies.
?
regards,
Bryan, N0LUF
 

Speech is harmonically rich. Such filtering removes the fundamental for females and the fundamental and possibly first few harmonics for lower pitched males. The difference is audible but it is not as bad as one might think.
?
73, Mike KK7ER?
?
 

On Sun, Oct 20, 2024 at 04:20 AM, Hans Summers wrote:
Don't worry about pops right now
I'm not worried. No doubt you have improvements still to be completed.?
?
But I expected you to blame the crackling on sync issues. The pops are much longer duration. It's not obvious to me how 48KHz clock sync explains those. But then I don't know what you know about your processing pipeline.?
?
73, Mike KK7ER?
?
 

CESSB firmware gets released, 20m gets taken over with the
sounds of AC/DC, or maybe Rush's "Spirit of Radio" ;-)
?
?
Nice reference to Radio Caroline in the video too.
 

The voiced speech of a typical adult male will have a fundamental frequency from 85 to 180 Hz, and that of a typical adult female from 165 to 255 Hz.? If the fundamental is suppressed for preventing wasting of power, enough of the harmonic series will be present for the brain "fills in" the missing fundamental to create the impression of hearing the fundamental tone.? For speakerphones, a different audio application altogether, it was found that keeping the fundamental was important for having a more realistic voice experience rather than a "communication quality" experience.?

-Steve K1RF?


------ Original Message ------
From "Mike KK7ER" <groupio@...>
Date 10/20/2024 11:04:45 AM
Subject Re: [QRPLabs] Torturing QMX+ CESSB...

Speech is harmonically rich. Such filtering removes the fundamental for females and the fundamental and possibly first few harmonics for lower pitched males. The difference is audible but it is not as bad as one might think.
?
73, Mike KK7ER?
?
 

Hans,
?
Sounds pretty good.? My old ears couldn't detect any pops on the audio.? I'd be interested to know how hot Q507, the series modulator transistor, got at 1W PEP output and will it manage thermally at 5W PEP?
--
Peter Lee
G3SPL
 

Oh, and I meant to say, you could use some "royalty free" music available on the internet for a Youtube video demonstration to avoid copyright strikes!
--
Peter Lee
G3SPL

21 - 32 of 32