[Elecraft] More on NoiseNuller Latency

[David Gilbert]

ChatGPT tells me that almost all of the latency is due to the frame size 
of the FFT processor.  At the application's default sample rate of 2024 
that works out to be just barely under the 43 msec I measured.  That 
sample rate is needed to do the actual processing with decent results, 
so the best workaround seems to be a two path approach ... use a much 
lower frame size when the sidetone is detected and the default frame 
size when it isn't.  Theoretically the latency could be dropped to maybe 
less than 5 msec.

And as you say, if everything was implemented inside the radio, the 
keying itself could control that with almost zero latency.

73,
Dave   AB7E



On 7/29/2026 10:30 AM, Walter Underwood via Elecraft wrote:
> The algorithm probably has inherent latency regardless of where it is implemented. This is one of the challenges of digital processing.
>
> But…if implemented internal to a radio, the sidetone could bypass the processing.
>
> wunder
> K6WRU
> Walter Underwood
> CM87wj
> http://observer.wunderwood.org/ (my blog)
>
>> On Jul 29, 2026, at 10:03 AM, Gwen Patton via Elecraft <elecraft at mailman.qth.net> wrote:
>>
>> Can it just turn off when sending CW? When you go back to listening, turn
>> it back on? It doesn't need to be on while you're manually sending CW, you
>> can't hear the noise anyway because you're transmitting.
>>
>>
>> -=-=-=-=-=-=-=-=-
>> 73,
>> Gwen, NG3P
>>
>>
>> On Wed, Jul 29, 2026 at 2:19 AM David Gilbert via Elecraft <
>> elecraft at mailman.qth.net> wrote:
>>
>>> A ham that has been experimenting with my NoiseNuller application
>>> pointed out to me where the current latency is indeed a problem, and
>>> that is when listening to the sidetone while sending CW with a key or
>>> paddle.  The sidetone audio coming out of the computer is delayed enough
>>> to throw off your keying at anything other than pretty slow speeds if
>>> you depend upon that for feedback.  It wouldn't be a problem for keying
>>> from a keyboard, but it is when manually keying.
>>>
>>> I'm working to improve that.
>>>
>>> In my opinion, though, it's just another reason why this kind of
>>> processing really belongs inside the rig itself.
>>>
>>> 73,
>>> Dave   AB7E
>>> ______________________________________________________________
>>> Elecraft mailing list
>>> Home: http://mailman.qth.net/mailman/listinfo/elecraft
>>> Help: http://mailman.qth.net/mmfaq.htm
>>> Post: mailto:Elecraft at mailman.qth.net
>>>
>>> This list hosted by: http://www.qsl.net
>>> Please help support this email list: http://www.qsl.net/donate.html
>>> Message delivered to ardrhi at gmail.com
>> ______________________________________________________________
>> Elecraft mailing list
>> Home: http://mailman.qth.net/mailman/listinfo/elecraft
>> Help: http://mailman.qth.net/mmfaq.htm
>> Post: mailto:Elecraft at mailman.qth.net
>>
>> This list hosted by: http://www.qsl.net
>> Please help support this email list: http://www.qsl.net/donate.html
>> Message delivered to wunder at wunderwood.org
> ______________________________________________________________
> Elecraft mailing list
> Home: http://mailman.qth.net/mailman/listinfo/elecraft
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:Elecraft at mailman.qth.net
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
> Message delivered to ab7echo at gmail.com