Hello all!
I have to implement a Schmitt trigger in software on a DSP. I was thinking
a while about a good implantation, but I just came to the simple solution
with has nested if conditions (See Pseudo code below). Does anyone has a
faster code??
-Thomas
for n = 1:length(x)
if flag
if x(n) > th
do something
flag = 0;
end
else
if x(n) < -th
do soming else
flag = 1;
end
end
end
Schmitt trigger in software
Started by ●November 13, 2009
Reply by ●November 13, 20092009-11-13
Table lookup like State = Array [0..1] [0..15] of 0..1 State[0] = '0000000000011111' State[1] = '0000011111111111' Y := State[Y] [X] X limited between 0..15, else bigger table Christen Fihl
Reply by ●November 13, 20092009-11-13
tom00 wrote:> Hello all! > > I have to implement a Schmitt trigger in software on a DSP. I was thinking > a while about a good implantation, but I just came to the simple solution > with has nested if conditions (See Pseudo code below). Does anyone has a > faster code?? > > > -Thomas > > > > for n = 1:length(x) > if flag > if x(n) > th > do something > flag = 0; > end > else > if x(n) < -th > do soming else > flag = 1; > end > end > end^^^^^^^^^^^^^^^^^^^^^^^^^ Fie, stupident solution. Here we go: const s16 fubar[] = { nonsense, -nonsense }; const s16 void (*do_something[])() = { something_ , else_ }; s16 *x = blablabla; s16 th = 0; while(length--) { th = fubar[u16 state = ((u16)(*x++ +th)) >> 15]; do_something[state]; } VLV
Reply by ●March 24, 20132013-03-24
On Saturday, November 14, 2009 1:26:12 AM UTC+13, Christen Fihl wrote:> Table lookup like > > State = Array [0..1] [0..15] of 0..1 > > State[0] = '0000000000011111' > State[1] = '0000011111111111' > > Y := State[Y] [X] > > X limited between 0..15, else bigger table > > Christen FihlVery nice i like it :)
Reply by ●March 25, 20132013-03-25
if state change is rare, I'd try also to switch between two separate loops, one for each threshold. C's "goto" might come in handy, if nobody near is allergic to it.
Reply by ●March 25, 20132013-03-25
>Here we go: > >const s16 fubar[] = { nonsense, -nonsense }; >const s16 void (*do_something[])() = { something_ , else_ }; > > >s16 *x = blablabla; >s16 th = 0; > >while(length--) > { > th = fubar[u16 state = ((u16)(*x++ +th)) >> 15]; > do_something[state]; > } > > >VLV >assert( (*x > (s16)0x7FFF - nonsense) || (*x < (s16)0x8000 + nonsense) )
Reply by ●March 26, 20132013-03-26
On 3/25/13 1:56 PM, dszabo wrote:>> Here we go: >> >> const s16 fubar[] = { nonsense, -nonsense }; >> const s16 void (*do_something[])() = { something_ , else_ }; >> >> >> s16 *x = blablabla; >> s16 th = 0; >> >> while(length--) >> { >> th = fubar[u16 state = ((u16)(*x++ +th))>> 15]; >> do_something[state]; >> } >> >>Vlad, does that make it a Schmitt trigger. or do you have to switch -nonsense and +nonsense in fubar[]? you want a comparator that more sticky than a static comparator. or maybe instead of switching the nonsense in fubar, you subtract the threshold from x instead of adding it. but i think you got the polarity mixed up. you make a Schmidt trigger with an op-amp using positive feedback.> > assert( (*x> (s16)0x7FFF - nonsense) > || (*x< (s16)0x8000 + nonsense) )in case there's some wrap around overflow that breaks Vlad's Schmidt trigger? -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
Reply by ●March 26, 20132013-03-26
>On 3/25/13 1:56 PM, dszabo wrote: >>> Here we go: >>> >>> const s16 fubar[] = { nonsense, -nonsense }; >>> const s16 void (*do_something[])() = { something_ , else_ }; >>> >>> >>> s16 *x = blablabla; >>> s16 th = 0; >>> >>> while(length--) >>> { >>> th = fubar[u16 state = ((u16)(*x++ +th))>> 15]; >>> do_something[state]; >>> } >>> >>> > >Vlad, does that make it a Schmitt trigger. or do you have to switch >-nonsense and +nonsense in fubar[]? you want a comparator that more >sticky than a static comparator. or maybe instead of switching the >nonsense in fubar, you subtract the threshold from x instead of adding >it. but i think you got the polarity mixed up. you make a Schmidt >trigger with an op-amp using positive feedback. > >> >> assert( (*x> (s16)0x7FFF - nonsense) >> || (*x< (s16)0x8000 + nonsense) ) > >in case there's some wrap around overflow that breaks Vlad's Schmidt >trigger? > > >-- > >r b-j rbj@audioimagination.com > >"Imagination is more important than knowledge." > > >It's tricky because the actual threshold for *x is equal to '-th' not 'th'. Basically, bit 15 is the sign of *x + th. When th == nonsense, the sign changes when *x is less than -nonsense, and vice versa. Yep, you'll wanna keep an eye on the range of *x values. It's admittedly both clever and hard to read, though. I like it.
Reply by ●March 26, 20132013-03-26
On 3/26/2013 10:58 AM, dszabo wrote:>> On 3/25/13 1:56 PM, dszabo wrote: >>>> Here we go: >>>> >>>> const s16 fubar[] = { nonsense, -nonsense }; >>>> const s16 void (*do_something[])() = { something_ , else_ }; >>>> >>>> >>>> s16 *x = blablabla; >>>> s16 th = 0; >>>> >>>> while(length--) >>>> { >>>> th = fubar[u16 state = ((u16)(*x++ +th))>> 15]; >>>> do_something[state]; >>>> } >>>> >>>> >> >> Vlad, does that make it a Schmitt trigger. or do you have to switch >> -nonsense and +nonsense in fubar[]? you want a comparator that more >> sticky than a static comparator. or maybe instead of switching the >> nonsense in fubar, you subtract the threshold from x instead of adding >> it. but i think you got the polarity mixed up. you make a Schmidt >> trigger with an op-amp using positive feedback. >> >>> >>> assert( (*x> (s16)0x7FFF - nonsense) >>> || (*x< (s16)0x8000 + nonsense) ) >> >> in case there's some wrap around overflow that breaks Vlad's Schmidt >> trigger? >> > > It's tricky because the actual threshold for *x is equal to '-th' not 'th'. > Basically, bit 15 is the sign of *x + th. When th == nonsense, the sign > changes when *x is less than -nonsense, and vice versa. > > Yep, you'll wanna keep an eye on the range of *x values. It's admittedly > both clever and hard to read, though. I like it.The point of this exersize is about not using if() in a loop. VLV
Reply by ●March 26, 20132013-03-26
On 3/26/13 12:42 PM, Vladimir Vassilevsky wrote:> > The point of this exersize is about not using if() in a loop. >well, i guessed i missed that point. the thing i was worrying about is that i think the feedback was negative feedback, and that will give you a stable amplifier with rails. i think you messed up the polarity in the loop gain, Vlad. i'll check it again, but you need to be doing positive feedback to do a Schmitt trigger. -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
