SHIFTIN SHIFTOUT timing problem
Does anyone know how to lengthen (slow down) the LOW period of the SHIFT clock?
The statement
extends the HIGH period of the shift clock to about 10 uS but the LOW period remains fixed at about 2 uS. I need to retain something close to a 50:50 duty cycle but at a lower speed with about 10 uS HIGH and 10 uS LOW.Code:DEFINE SHIFT_PAUSEUS 2
Cheers
Brian
Last edited by BrianT; - 9th April 2010 at 08:47. Reason: typo
|
I've dealt with this two ways, manually modify the shiftin/out library adding nop's in assembler to create a delay, or a better option is to bit bang the routines manually. This way you have complete control of the timing. Examples here
http://www.melabs.com/resources/samples/pbc/shift.bas
|
bit banging for less than 8 bits
Thanks for that Steve.
I had considered the bit bang approach and had found the BS2 routines on the Melabs site. I am using the Intersema pressure transducer which needs the SHIFTOUT value\bits option to clock out some 2 and 3 bit values as well as regular 8 bit bytes. I guess I'll just have to bite the bullet and dive into the assembler. The lazy option is not going to cut it.
Cheers
Brian
|
Hey Brian,
No need to use assembler depending upon yout timing requirements. You can insert pauses in the routines to meet your requirements. This also can makes it easy to read and write different numbers of bits.
|
bit banging the Intersema sensor
I could not make the PBP ShiftIn or ShiftOut routines work reliably with the latest Intersema MS5541C although the same code had worked a treat for several years. Intersema must have made some internal changes, one at least to reduce power consumption when MClk stops, or maybe meLabs has tweaked some code.
Intersema show DIn must be stable before SClk goes high. ShiftOut in PBP shows this as Mode 0. In the direction from Intersema to PIC, ShiftIn mode 2 should get the sense and timing right according to the PBP and Intersema waveform sketches. The problem seems to be that neither Intersema nor PBP specifies the setup time between data change and clock strobe. On an oscilloscope the clock follows data by a few hundred of nanoseconds which ought to be enough but I get many bad reads with the potted ShiftIn and ShiftOut routines.
I hand coded the 7 commands from PIC to Intersema and the single reply code from Intersema to PIC. The code is none too elegant but it now works 100% every time. Previously I was getting about 75% mis-reads of pressure data and some DA5541C chips would not work at all.
The minimum pauseus is 4 uS so there is now plenty of time for the data to stabilise before the clock strobe.Code:OutShift: for fa = 1 to clockbits 'OutShift adds an extra bit at the end din = (1 & iword) 'select lowest bit & pauseus 1 'will actually be 4 uS or so sclk = 1 : pauseus 1 : sclk = 0 iword = iword >> 1 'get next LSB, clock out zeros after 16 bits next fa sclk = 1 : pauseus 1 : sclk = 0 'extra clock per DA5541B_00513 page 13 return ResetIntersema: ' resets ALL pressure sensors ' shiftout din, sclk, 0, [85, 85, 0\5] ' Sense of Din is IN to 5541 output din : output sclk clockbits = 20 : iword = %0101010101010101 '16 bit data word gosub outshift return ConvertDelay: ' specific to each sensor if tank = 0 then While dout0 = 1 wend endif if tank = 1 then While dout1 = 1 wend endif if tank = 2 then While dout2 = 1 wend endif return FetchWord: ' read specific channel reply input dout0 : input dout1 : input dout2 : output sclk : iword = 0 if tank = 0 then for fa = 15 to 0 step -1 'need 17 clock bits so add one at end sclk = 1 : pauseus 1 'wait for Intersema to present next bit iword.0[fa] = dout0 'read the bit on selected channel sclk = 0 'drop clock next fa 'do it 16 times endif if tank = 1 then for fa = 15 to 0 step -1 'need 17 clock bits so add one at end sclk = 1 : pauseus 1 'wait for Intersema to present next bit iword.0[fa] = dout1 'read the bit sclk = 0 'drop clock next fa 'do it 16 times endif if tank = 2 then for fa = 15 to 0 step -1 'need 17 clock bits so add one at end sclk = 1 : pauseus 1 'wait for Intersema to present next bit iword.0[fa] = dout2 'read the bit sclk = 0 'drop clock next fa 'do it 16 times endif sclk = 1 : pauseus 1 : sclk = 0 ' 17th clock bit ii = iword.byte0 : ij = iword.byte1 return ReadFactoryCal: ' This unpacks the factory calibration coefficients from the just ' read W1 ~ W4. ' These bitmaps are unpacked into the 6 working coefficients ' C1 to C6 which are then stored in EEROM for later use. 'W1 gosub resetintersema ' shiftout din, sclk, 0, [87, 1\5] ' Send W1 pattern to all sensors clockbits = 12 : iword = %000101010111 'request W1 pattern gosub outshift gosub fetchword ' recall selected Tank reply W1.byte0 = ii W1.byte1 = ij if (w1 = 0) or (w1 = 65535) then debug "W1 error", 13, 10 goto readfactorycal endif 'W2 gosub resetintersema ' shiftout din, sclk, 0, [215, 0\5] ' Send W2 pattern clockbits = 12 : iword = %000011010111 gosub outshift gosub fetchword W2.byte0 = ii W2.byte1 = ij if (w2 = 0) or (w2 = 65535) then debug "W2 error", 13, 10 goto readfactorycal endif 'W3 gosub resetintersema ' shiftout din, sclk, 0, [55, 1\5] ' Send W3 pattern clockbits = 12 : iword = %000100110111 gosub outshift gosub fetchword W3.byte0 = ii W3.byte1 = ij if (w3 = 0) or (w3 = 65535) then debug "W3 error", 13, 10 goto readfactorycal endif 'W4 gosub resetintersema ' shiftout din, sclk, 0, [183, 0\5] ' Send W4 pattern clockbits = 12 : iword = %000010110111 gosub outshift gosub fetchword W4.byte0 = ii W4.byte1 = ij if (w4 = 0) or (w4 = 65535) then debug "W4 error", 13, 10 goto readfactorycal endif CalcCoefficients: ' this serves all three sensors. 'C1 C1 = W1 >> 3 'unpack coefficient read (108 + tank*20), z.byte0 'store read (109 + tank*20), z.byte1 if z<>c1 then write (108 + tank*20), c1.byte0 'store write (109 + tank*20), c1.byte1 endif 'C2 C2 = ((W1 & %0000000000000111) << 10) + (W2 >> 6) read (110 + tank*20), z.byte0 read (111 + tank*20), z.byte1 if z<>c2 then write (110 + tank*20), c2.byte0 write (111 + tank*20), c2.byte1 endif 'C3 C3 = W3 >> 6 read (112 + tank*20), z.byte0 read (113 + tank*20), z.byte1 if z<>c3 then write (112 + tank*20), c3.byte0 write (113 + tank*20), c3.byte1 endif 'C4 C4 = W4 >> 7 read (114 + tank*20), z.byte0 read (115 + tank*20), z.byte1 if z<>c4 then write (114 + tank*20), c4.byte0 write (115 + tank*20), c4.byte1 endif 'C5 C5 = ((W2 & %0000000000111111) << 6) + (W3 & %0000000000111111) read (116 + tank*20), z.byte0 read (116 + tank*20), z.byte0 if z<>c5 then write (117 + tank*20), c5.byte1 write (117 + tank*20), c5.byte1 endif 'C6 C6 = W4 & %0000000001111111 read (118 + tank*20), z if z<>c6 then write (118 + tank*20), c6 endif Show5541Coefficients: 'only used during diagnostics ' high txd : pause 1 ' debug 13, 10, "Tank #", #tank, ", W1 = ", #w1, ", W2 = ",_ ' #w2, ", W3 = ", #w3, ", W4 = ", #w4, 13, 10 ' debug "Derived coeffs C1 = ", #C1, ", C2 = ", #C2, ", C3 = ",_ ' #C3, ", C4 = ", #c4, ", C5 = ", #C5, ", C6 = ", #C6, 13, 10 return ReadTank: ' D1 is Pressure - D2 is Temperature. Same code for all 3. 'MUST know Tank# before calling this routine. 'D1 gosub resetintersema ' shiftout din, sclk, 0, [47, 0\5] ' Select D1 pattern clockbits = 11 : iword = %00000101111 gosub outshift gosub convertdelay gosub fetchword d1.byte0 = ii d1.byte1 = ij if (d1 = 0) or (d1 = 65535) then debug "D1 error", 13, 10 goto readtank endif 'D2 gosub resetintersema 'shiftout din, sclk, 0, [79, 0\5] ' Select D2 pattern clockbits = 11 : iword = %00001001111 gosub outshift gosub convertdelay gosub fetchword D2.byte0 = ii D2.byte1 = ij if (d2 = 0) or (d2 = 65535) then debug "Tank ", #tank, ", D2 error", 13, 10 goto readtank endif CalcTempPress: read (108 + tank*20), c1.byte0 read (109 + tank*20), c1.byte1 read (110 + tank*20), c2.byte0 read (111 + tank*20), c2.byte1 read (112 + tank*20), c3.byte0 read (113 + tank*20), c3.byte1 read (114 + tank*20), c4.byte0 read (115 + tank*20), c4.byte1 read (116 + tank*20), c5.byte0 read (117 + tank*20), c5.byte1 read (118 + tank*20), c6 ut1 = 8*c5 + 10000 dt = d2 - ut1 if dt < 0 then dt2 = dt - (dt*dt)/(128*64) else dt2 = dt - (dt*dt)/(128*16) endif ' tanktemp = 200 + (dt*(c6+100))/2048 'simpler form, less accurate?? tanktemp = 200 + dt2*(c6 + 100)/2048 offset = c2 + ((c4 - 250)*dt)/4096 + 10000 senstvty = c1/2 + ((c3 + 200)*dt)/8192 + 3000 tankpress = (senstvty*(d1 - offset))/4096 + 1000 BoundsCheck: if (tankpress > 8000) or (tankpress < 700) then debug 13, 10, "Bounds error. Tank = ",#tank, ", TankPress = ", _ #tankpress," mB", 13, 10 goto readtank 'retry until value is within bounds endif if tank = 0 then t0pressure = tankpress t0celsius = tanktemp ' debug "T0 D1, D2 = ", dec d1,", ", dec d2, ", T0 Pressure = ", dec t0pressure, ", Temp ", dec t0celsius, 13, 10 endif if tank = 1 then t1pressure = tankpress t1celsius = tanktemp ' debug "T1 D1, D2 = ", dec d1,", ", dec d2, ", T1 Pressure = ", dec t1pressure, ", Temp ", dec t1celsius, 13, 10 endif if tank = 2 then t2pressure = tankpress t2celsius = tanktemp ' debug "T2 D1, D2 = ", dec d1,", ", dec d2, ", T2 Pressure = ", dec t2pressure, ", Temp ", dec t2celsius, 13, 10 endif return
The Intersema pressure sensors are spectacular little temperature and pressure sensors and up until some recent changes have served me very well. With this new code, the conversion takes about 1 mS longer and the code is slightly bigger but I am now in control of the read & write process.
I hope someone finds this useful.
Cheers
BrianT
Last edited by BrianT; - 14th April 2010 at 01:52.
|
Hi Brian,
Glad you have it working. Can you tell a little more about what your doing with the pressure/temperature sensors? I do a lot of data acquisition work involving those types of measurements.
Thanks
Members who have read this thread : 0You do not have permission to view the list of names.
Posting Permissions
Bookmarks