PSG tone period update

Below a test program that uses an USR.
It writes the low byte first, and it makes click when going from 255 to 256, but the other direction is no click.

It is clicking only in cases that match the theory that it is a phase reset
The other direction, the error tone is an octave down (and one can nicely hear this tone when in BASIC one adds a = sin(a) for a some more delay)
BUT with an asm update in ~10 microseconds, the error tone is inable to EXIST! it has too long wavelength

10 clear 200,&hbf00

100 restore 9200 : read a : defusr = a : gosub 9000

200 screen ,,0 : sound 8,8 
210 f = 255 : ?"use cursor up/down"
220 defusr1 = f : rem passing the parameter in usr1 0xf3c9
225 x = usr(0)
230 a = stick(0)
240 if a = 1 then f = f + 1
250 if a = 5 then f = f - 1 
260 ?f
300 goto 220


999 end

9000 a$ = ""
9020 if a$="" then read a$

9032 if left$(a$,1)="$"  then a$=mid$(a$,2) : return
9034 if left$(a$,1)="_"  then a$=mid$(a$,2) : goto 9020
9040 if left$(a$,2)="1>" then a$=mid$(a$,3) : l1 = a : goto 9020
9042 if left$(a$,2)="2>" then a$=mid$(a$,3) : l2 = a : goto 9020
9050 if left$(a$,2)="<1" then a$=mid$(a$,3) : b$ = hex$((l1-a-1)AND255) : goto 9110
9052 if left$(a$,2)="<2" then a$=mid$(a$,3) : b$ = hex$((l2-a-1)AND255) : goto 9110

9100 b$=left$(a$,2) : a$=mid$(a$,3)
9110 poke a,val("&h"+b$) : a = a + 1
9120 goto 9020

9200 data &hc000
9210 data 3e00_d3a0_3a9CF3_d3a1__3e01_d3a0_3a9DF3_d3a1__C9$

Wow this is cool I thought one will have to compare the new value with the register contents. But to write 255 to the low byte does fullfill the requirements "never write a too low value, while a too high value is harmless".

Thanks to you too, I will keep this simple solution in mind for future projects!