Frank, Rick,
I found my ".wav" files doing similar to what AB7E does. Except that his
has MP3 raspy artifacts all over the place and mine doesn't.
Take a listen to his and mine
AB7E MP3: http://www.ab7e.com/weak_signal/0_1_2_3_4_5_6_db.mp3
N3QE WAV: Sum of morse + noise: http://n3qe.org/sum.wav
Just in case you are interested, the raw noise and raw morse WAV files
before mixing below:
N3QE WAV: My raw noise (no morse): http://n3qe.org/noise.wav
N3QE WAV: My raw morse (no noise): http://n3qe.org/morse.wav
Tim N3QE
On Wed, Aug 21, 2019 at 3:34 PM Tim Shoppa <tshoppa@gmail.com> wrote:
> Will do, Frank!
>
> Here's an interesting exercise:
> https://www.npr.org/sections/therecord/2015/06/02/411473508/how-well-can-you-hear-audio-quality
>
> I just ran through it and picked the uncompressed .WAV all 6 times :-)
>
> One giveaway to my ears is things that were clear sharp impulses in the
> originals, get muddy in compression. For example the glocks at the end of
> the Neil Young sample is the easiest way to tell with that one.
>
> Other times, it's the lack of artifacts that helped me pick the original
> .WAV. For example the Suzanne Vega samples.
>
> Interestingly the Suzanne Vega clip was used by developers of MP3!
>
> On Wed, Aug 21, 2019 at 3:15 PM <donovanf@starpower.net> wrote:
>
>> Hi Tim,
>>
>> As you're well aware, perfect is often the enemy of good enough.
>>
>> Please publish your files and let us all experience how much better they
>> are.
>>
>> Thanks
>>
>> 73
>> Frank
>> W3LPL
>>
>> ------------------------------
>> *From: *"Tim Shoppa" <tshoppa@gmail.com>
>> *To: *"Richard (Rick) Karlquist" <richard@karlquist.com>
>> *Cc: *"Frank Donovan" <donovanf@starpower.net>
>> *Sent: *Wednesday, August 21, 2019 6:30:20 PM
>> *Subject: *Re: Topband: Impressive demonstration of one dB of signal
>> strength improvement
>>
>> Rick, I agree, MP3 just muddies the sound of those examples. I regularly
>> save my radio audio as 8kHz WAV and am happy with that. But I can tell if I
>> convert it to MP3 that it is "off" and "wrong". The MP3 format's
>> psychoacoustic encoding muddies the beginning and end of each dit AND does
>> a horrible job when presented with lightning or impulse noise.
>>
>> Last time Frank and I talked about these examples, I made a completely
>> independent set for my own use that was never compressed along the way.
>> (.WAV uncompressed format). Will dig up at home.
>>
>> Tim N3QE
>>
>>
>>
>>
>> On Wed, Aug 21, 2019 at 12:54 PM Richard (Rick) Karlquist <
>> richard@karlquist.com> wrote:
>>
>>> On 8/21/2019 8:16 AM, donovanf@starpower.net wrote:
>>> > These recordings are an impressive demonstration of the benefit of
>>> > one dB of signal strength improvement in a weak signal situation.
>>> >
>>> >
>>> > Click on the links on this website:
>>> >
>>> >
>>> >
>>> > www.ab7e.com/weak_signal/mdd.html
>>> >
>>> >
>>> > 73
>>> > Frank
>>> > W3LPL
>>>
>>> It seems to me that converting the files to .mp3
>>> muddies the water. Dave comments that 32 kHz
>>> is "decent fidelity". It certainly isn't (IMHO)
>>> decent fidelity for music. I don't know about
>>> noisy CW. The .mp3 format was optimized
>>> for music or possibly high S/N voice, but not CW
>>> buried in noise. The operational principle of
>>> mp3 (or any lossy compression) is "noise gating".
>>> For all I know, the abrupt change at low S/N may
>>> be an artifact of the compression.
>>>
>>> I don't understand the comment about "all audio files
>>> were converted to .mp3 format and mixed ..." AFAIK,
>>> mixing should be done before any compression.
>>>
>>> Using "band noise" also muddies the water. It would
>>> have been better to use the proverbial Additive
>>> White Gaussian Noise (AWGN), to get reproducible results.
>>> Dave comments that "spikey" noise is another hard to
>>> quantify condition. I have noticed anecdotally that
>>> 160 meters tends to have spikey noise vs AWGN on say
>>> 20 meters. This results in clearly audible signals that
>>> can't be copied by ear on 160 meters, whereas on 20
>>> meters, I seem to be able to get solid copy on signals
>>> that are just barely detectable. This is a poorly understood
>>> phenomenon.
>>>
>>> Another issue with reproducibility is the audio bandwidth.
>>> The brain tends to be able to provide some audio filtering
>>> of its own. IOW, I believe that you will find that
>>> a 250 Hz bandwidth will give less than a 3 dB advantage
>>> vs 500 Hz for the threshold of copy. At narrow bandwidths,
>>> ringing becomes an issue. Typical brick wall filters in SDR's
>>> may or may not be optimum for copying a CW in noise
>>> (and in the absence of QRM). My Flex 6700 uses a
>>> gazillion poles "because they can" :-)
>>>
>>> Rick N6RK
>>> _________________
>>> Searchable Archives: http://www.contesting.com/_topband - Topband
>>> Reflector
>>>
>>
>>
_________________
Searchable Archives: http://www.contesting.com/_topband - Topband Reflector
|