On 2019-09-04, Alister  wrote:
> On Wed, 04 Sep 2019 12:27:53 +0100, Folderol wrote:
>
>> On Wed, 04 Sep 2019 10:34:21 GMT Alister 
>> wrote:
>>
>>>On Tue, 03 Sep 2019 13:06:58 -0400, Dennis Lee Bieber wrote:
>>>
>>>> On Tue, 3 Sep 2019 07:39:14 -0400, "J.B. Wood"
>>>>  declaimed the following:
>>>>
>>>>>Hello, all.  Just thought I'd revisit this topic since it's been a
>>>>>while.  Has anyone ever constructed a Pi-based standalone 1-2 channel
>>>>>MIDI sequencer?  By standalone I mean it has an integral display and
>>>>>required pushbuttons/rotary knob for device control.
>>>>>
>>>>  What do you intend to show on the display?
>>>>
>>>>  What type of control do you envision? (If you visualize a one/two
>>>> octave keyboard, you then have to specify how many notes in parallel
>>>> -- one note at a time can use a simple row/column keyboard scan -- but
>>>> when you go beyond two notes such a matrix can generate ghost key
>>>> presses; avoiding ghosts requires a linear matrix, which means LOTS of
>>>> GPIOs -- a 1x24 would need 24 GPIOs to detect key presses, while a 4x4
>>>> is only using 8 GPIOs for 16 keys)
>>>
>>>Not Necessarily so. To eliminate ghosting you simply need a diode on
>>>each switch (how do you think a quality computer keyboard manages N Key
>>>rollover?).
>>>16 io pins gets you 8*8 = 64 keys so a full piano (81 keys if i remember
>>>correctly) still only requires 9 by 9.
>>>
>>>
>> You need more than that to allow for various switches/selectors etc.
>> Also you need to scan the matrix as fast as possible, preferably in the
>> tens of microseconds range. However, just to make it 'interesting' you
>> should include some form of debounce mechanism.
>> Been there, done that, T shirt worn out!
>
> figures quoted were a minimum obviously the matrix can be made as large
> as possible.
> microseconds for a full scan is massively overkill a 10mS scan time would
> equal 100 scans a second more than enough for human interaction.
>
> My original job was with a manufacturer of custom industrial keyboards so
> I am fully aware of the need to debounce a 6mS delay when a change was
> detected was considered more than adequate.
>
> another option would be to only recognise a change after 2 consecutive
> scans which would still be more than acceptable performance with the
> above 100hz scan frequency.

Isn't MIDI time resolution 1/128 of a second?  That's pretty
close to 100Hz aka 10msec.

--
Robert Riches
spamtrap42@jacob21819.net
(Yes, that is one of my email addresses.)

--- SoupGate-Win32 v1.05