Stable ADC readings

[dream-]

I am trying to get stable ADC readings.

I got a PIC 16F688 hooked up to a battery with a voltage regulator and with proper bypass capacitors. Each time I poll the ADC using ADCIN I get a different reading. Even averaging 50 samples is not enough to get a stable value. I have tried both internal and external oscillators (up to 20MHz).

I am testing it using a simple voltage divider with two resistors of the same value so that I get a value around 512 (ADC in 10 bit mode).

Is there anything I can do to stabilize the readings? And what is the source of all the noise?

Typically I am getting values oscillating from 432 up to 457, that's about a 5% oscillation.

There has to be a way to make the readings more accurate.6

[Jumper]

If you use VDD as the voltage reference for your AD then it is most likely a not stable powerline. I know you said battery and voltage regulator but have you tried to measure the VDD with an ocsillocope?

If you really need to get high accuracy reading you might have to consider an external high precition voltage reference. That one you can connect to the VREF pin and then you can set up the AD to use that one.

Also if you are switching other stuff while doing the measurement it might affect the power supply to the pic as well.

I would start with making sure the power nice and clean and batteries are not the best source when you start doing new designs.

[Acetronics2]

Hi,

May we get a scheme , plus your program ???

Crystal ball is out of range ...

Alain

[sayzer]

Adding to what Jumper said, I also face similar issues.
Usually, using 7805 or 78L05 may not provide a stable 5V.
Say it is actually between 4.98V and 5.02V. (approx 5% datasheet statement)

So, if you are using 10-bit reading there is your problem.
Eventually, Jumper's suggestion of having a high precision voltage reference is a good idea.

Another issue :
Pausing 2ms after each ADCIN command may help.

[mackrackit]

And...

More CAPs.
A capacitor from the ADC pin to VSS will help smooth things out. Will not make up for a bad power supply though.

[Acetronics2]

Hi, Sayzer

I am testing it using a simple voltage divider with two resistors of the same value so that I get a value around 512 (ADC in 10 bit mode).

That means he works in ratiometric mode and supply changes do not interfere here ...
so, problem is elsewhere ...

Alain

[sayzer]

Hi, Sayzer



That means he works in ratiometric mode and supply changes do not interfere here ...
so, problem is elsewhere ...

Alain

Yes Alain, I missed that.

Then, may be he should pause ~2ms after each read.

Also, for example, reading lower voltages and reading higher voltages may behave differently.
If reading higher voltages are accurate, but lower voltages are not, then there is something else, too.




---------------------------

[mister_e]

Input impedance too high, too short of ADC aquisition time, noisy PSU, unstable PSU, no averaging routine, multiplexed function on the ADC pin, loose/dirty bread board contacts, poor wiring, poor oscillator, poor decoupling, etc... quite a few possibilities...

[dream-]

As mentioned I am decoupling, using averaging, and testing the ratio of two resistors hooked up to the same battery after the regulator.

Let me draw the schematic and I will post.

But before that, let me say that I switched from ADCIN to doing a manual ADCON0.1 = 1 reading, and now I get a stable value. So my problem is an erroneous use of the ADCIN command or a problem in the logic of my routine.

BEFORE:

DEFINE ADC_BITS 10 ' Set number of bits in result (8, 10 or 12)
DEFINE ADC_CLOCK 3 ' Set clock source (rc = 3)
DEFINE ADC_SAMPLEUS 50 ' Set sampling time in microseconds
...
WSum = 0
Pause 100

For i = 0 to 49
ADCIn PinProbe, W0
Pause 10
WSum = WSum + W0
Next i

W0 = WSum / 50

NOW:

Pauseus 50 'Stabilize
ADCON0.1 = 1  'start the ADC
WHILE ADCON0.1 : WEND
  
W0.HighByte = ADRESH
W0.LowByte = ADRESL

It works, now the reading stays fixed and it is much faster. (I even didn't have to do any averaging)

[Acetronics2]

Hi,

What about reinitialising your Wsum value to 0 @ each 50 samples batch ??? don't you run this loop in another continuous loop ... ??? " par hasard " ...

BTW " PinProbe " is the alias for which pin ???

Alain

[dream-]

Hi,

What about reinitialise your Wsum value to 0 @ each 50 samples batch ??? don't you run this loop in another continuous loop ... ???

Alain

Oh I do. I left that part out of the code I pasted, but it is indeed running in an outside continuous loop.

PinProbe VAR PORTA.2
...
myLoop:

WSum = 0
Pause 100

For i = 0 to 49
ADCIn PinProbe, W0
Pause 10
WSum = WSum + W0
Next i

W0 = WSum / 50
...
LCDout $fe, 1 'Clear Screen
Lcdout $fe, $80
Lcdout "Reading: ", DEC W0

Goto myLoop

The readings I get with the ADCIN method are in the ballpark but move around too much. With the "manual" ADC method I get a stable value.

[Acetronics2]

ADCIN also polls for the ADCON0.1 bit ...

sooo .... aheum, how to tell it ...

Alain

[mister_e]

ADCIN also polls for the ADCON0.1 bit ...

sooo .... aheum, how to tell it ...

Alain

Yeah but we already say some odds with ADCIN in the past on some specific PIC, Microchip are good to change the way register are done, and how things work for PIC to PIC model...

This said, I never use ADCIN myself, but manually write/read ADC registers... it is also a bit less code hungry... down side "may" be when you port your code on another PIC who's not working the same way, or have different register name, bit assignment. big deal

[dream-]

Well if I understood what the problem was I would not be asking here

The fact is that the ADCIN code gives me an oscillating reading, and the manual method gives me a stable value.

I don't know the reason, but I would like to know. This is a 16F688.

[mackrackit]

For what it is worth I agree with Steve.
I tried ADCIN a couple of times and did not like it. Others swear by it though. ..

Keep doing it the way you are.

[mister_e]

I would love to see the whole code with ADCIN. Sometime if you play with some ADC registers, and you place them Before or After the DEFINEs, it may end-up with unexpected results.

[Acetronics2]

Hi,

Found in the Datasheet ...

Note:

The GO/DONE bit should not be set in the
same instruction that turns on the ADC.
Refer to Section 8.2.5 “A/D Conversion

Procedure”.

a beginning of an explanation ???

Alain

[Bruce]

ADCIN doesn't work with port pin aliases. Use ADCIN 2,W0, or define PinProbe as a constant. PinProbe CON 2 would work.

Also ADCIN doesn't setup TRIS, ANSEL or ADCON1. So you end up with Fosc/2 for your A/D clock because ADCON1 defaults to all 0's. So you're way out of spec for a 4MHz osc. Fosc/8 would be a good one to use with 4MHz (see datasheet).

[Acetronics2]

Bruce ... 1K points ...

as always !!!...

Shame on me.

Alain

[mister_e]

Blonde moment here too

Disclaimer: "Blonde moment" is a deliberate sentence that belongs to Mister_e user and plah plah plah, shall not be considered as offensive plah plah plah. Complaints about that shall not be send to Admin, Mods, user but only to Mister_e user plah plah plah....

[Acetronics2]

Hi, Steve

I Think the "Blonde" term use rights already belong to KarenHornby ...

so, you'll have to buy a licence ...

Alain

[mister_e]

busted .....

[HankMcSpank]

So to summarize 'in speak' that won't scare folks who thought if/then statements were an uinsurmountable hurdle.....are we saying that using the ADCIN is not optimum...and if not could someone please post a chunk of code to do this in an alternative method?

[mackrackit]

http://www.rentron.com/PICX2.htm

[HankMcSpank]

Excellent.....many thanks, I shall use that code gratuitously.

[Ioannis]

Have you tested this fast averaging with Hysterisis by Darrel:

http://www.picbasic.co.uk/forum/cont...ith-hysteresis

Ioannis

[HankMcSpank]

Yes I have (& very good it is too) , but I'd rather have this jittery AtoD values 'cleaned up' at the source (rather than putting averaging in place to address unstability caused by avoidable factors)

[Bruce]

Just include an RC filter on the A/D input, and put the PIC to sleep just before enabling the A/D conversion. This returns very stable readings in most applications with a stable power supply or Vref.

[fratello]

Nice subject !
I use a similar code for reading the over-current of one DC motor :

check:           
 vt  =0
 vs  =0
 grup=0
 cnt =0
 pause 50					
 for cnt = 1 to 10
 	adcin 3, adval
 	vt=adval * 5000
 	vt=div32 1023
 	grup=vt+grup
 	pause 5
 next cnt
 vs=grup / 10
      if (vs > checkin) or (T1Count > 50) then 		   
  	     	gpio.0=0
           	gpio.5=0
      endif

If the averge of 10 readings it's bigger than checkin value (or time exceed 5 seconds) then gpio.0 and gpio.5 goes low.
It's one good ideea to use ADCON instead ADCIN ? How to do this and keeping average of 10 readings ?

[fratello]

It's this correct ?
In header of code I add :

    ADCON0 = %10001101      ' ADC input on GPIO.4/AN3/pin3 (page 41)
                            ' MSB 1 = Right Justify                   "1"    
                            '     0 = Vdd as voltage reference        "0"
                            '     0 = n/a                             "00"
                            '     0 = n/a
                            '     1 = select channel 03 (bit 2 of 2)  "11"
                            '     1 = select channel 03 (bit 1 of 2)
                            '     0 = conversion status (see routine) "0"
                            ' LSB 1 = A/D converter module is on      "1"

And I modify the "check" subroutine like that :

check:           
 vt  =0
 vs  =0
 grup=0
 cnt =0
 pause 50			
 for cnt = 1 to 10
      ADCON0.1 = 1
      NOT_DONE:
		if ADCON0.1 = 1 THEN NOT_DONE
            Pause 10
            adval.highbyte = ADRESH
            adval.lowbyte = ADRESL
 PAUSE 10
 	vt=adval * 5000
 	vt=div32 1023
 	grup=vt+grup
 PAUSE 10
 next cnt
 vs=grup / 10
      if (vs > checkin) or (T1Count > 50) then 		   
  	     	gpio.0=0
           	gpio.5=0
      endif

The hardware it's in picture.
Thanks !

[fratello]

I've tested the code ; it's OK.

[HankMcSpank]

I've used ADCIN pretty much from day one, but I noticed I was getting some *very* jittery readings (255,243,246,255,243,246 etc)...I couldn't get to the bottom of it (spent a good while trying the usual stuff), I'd forgot about this thread but thankfully foiund it with Google)

It's the first time I've used the manual ADC method......but I'm pleased to also report my ADC readings are now solid as a rock.

Just for the record, here's what I used on my 12lf1822 (I'm only using 8 dits ADC)...

ADCON0 = %00000100     ' select ch1 (AN1 Pin 6)
ADCON1 = %00000000     'set ADC +ve reference to VCC 
' 
main:
gosub Do_ADC
pause 100
hserout [DEC ADC_RESULT,13,10]
goto main
' 
'ADC sub routine...
Do_ADC:
PAUSEUS 50              ' Wait for A/D channel acquisition time
ADCON0.1 = 1           ' Start conversion
WHILE ADCON0.1       ' Wait for it to complete
WEND
ADC_Result = ADRESH
return

Many thanks to you all on this thread...I shall never go back to the ADCIN method!

[HankMcSpank]

I wish the edit feature was open longer - I forgot to turn the ADC on above, sohere's the correction....

ADCON0 = %00000101     ' select chan1 (AN1 Pin 6), turn ADC On Bit 0

[mazoul72]

Hello,
i made this to programm ADC

bmf_comm1 VAR PORTB.0
bmf_comm2 VAR PORTB.1
'Variables
bmfsD VAR WORD
bmfs1D VAR WORD
iled var byte
blink1 var byte 'pour clignotement 1ère led
blink2 var byte 'pour clignotement 2ème led
'configuration du port pour lire l'entrée analogique
TRISA = %00000011 ' RA0 et RA1 en entrée
TRISB = %11111100 ' RB0 et RB1 en sortie
ADCON1 = $10000010 'RA0 à RA3

test_depart:
Low bmf_comm1
Low bmf_comm2
High bmf_comm1
high bmf_comm2
pause 2000
goto mainloop
'lecture de la valeur
getad:
PAUSEUS 50 ' Wait for A/D channel acquisition time
ADCON0.2 = 1 ' Start conversion
WHILE ADCON0.2 ' Wait for it to complete
WEND
Return

' Subroutine to get pot x value
getx:
ADCON0 = $41 ' Set A/D to Fosc/8, Channel 0, On
Gosub getad
bmfsD = ADRESL + (ADRESH*256)
Return

' Subroutine to get pot y value
gety:
ADCON0 = $49 ' Set A/D to Fosc/8, Channel 1, On
Gosub getad
bmfs1D = (ADRESH*256) + ADRESL
Return
val_leds:
low bmfs_comm1
low bmfs_comm2
gosub getx
gosub gety

if (bmfsD > 1000) Then Goto val_leds
if (bmfsD > 750) Then Goto boucle_led
if (bmfsD > 500) Then Goto boucle1_led
if (bmfsD > 250) Then Goto boucle2_led

High bmf_comm1
High bmf_comm2
pause 1500
goto val_leds
boucle_led:
iled = 0
while iled < 5
if iled = 1 or 3 or 5 then high blink1:low blink2
if iled = 0 or 2 or 4 then low blink1:high blink2
iled = iled + 1
pause 1000
wend
goto val_leds
boucle1_led:
High bmf_comm1
low bmf_comm2
pause 3000
goto val_leds
boucle2_led:
low bmf_comm1
high bmf_comm2
Goto val_leds
End

The leds works but always lightning , action on button (RB0 only for the moment) don't change anything.
No blinking .
I don't know what .
thanks.

[mister_e]

Salut et Bienvenue sur le forum

I haven't check the whole thing in deep but I think you should modify the boucle_led section to

boucle_led:
        iled = 0
        while iled < 5
                if iled = 1 or 3 or 5 then 
                        high blink1
                        low blink2     
                        endif
                        
                if iled = 0 or 2 or 4 then 
                        low blink1
                        high blink2
                        endif
                iled = iled + 1
                pause 1000
                wend
        goto val_leds

Probably it will solve your problem though... let me know the PIC you're using and your PBP version.

EDIT: Your code have 2 compilation error
1) no mainloop label
2) adcon1: the way you have written it, it's value should start with a % instead of a $

[mister_e]

Looking at your code, I guess it's your first program, There's quite a few mistakes in and it is also hard to follow what you want to do.

No problem, been there...done that. Tell us what you want to do and we could build something to help you to get started.

Tell us the PIC your're using and post your schematic if you have any, or describe it.

[mazoul72]

Sorry for double post but i tried many many modifications.
my pic is 16F877A , i desire to put multifunctions buttons of steering wheel and decode for command radio car system .
i make a test with leds for ADC value input .
Oh,oh,oh,
i have just see that i don't put MCLR to any connection , a resistor of 4.7kohms to +5V ?

It is a very simple schematic for test but the entire schematic is for steering wheel command and FIS Tachometer car display modification in live .
example : see the kwp1281 protocol codes of line K (not can bus) like VCDS soft but more simple.

This is the first part ( BMF )

For mainloop , it is because i modify the labels and i forgot it (val_leds) .
For $ , it is a mistake , THANKS.
i will try now

Regards

[mazoul72]

my leds light off after 2 sec , it is ok for starting (mainloop replaced by val_leds)
But after i have only two leds lightning together , it is not good with all boutons
and i have no blinking leds

And now i have no actions on leds (off always) , i don 't undertand why , AD not stable or other....
it is the same with or without MCLR .

[mister_e]

Ah ok I see, you read a voltage then act upon. Cool, I did something like that awhile back but it was for car remote starter purpose (Door lock, alarm bypass, Windows roll-up/down) Oh well, those were the days

Yes MCLR must be tied to Vdd. 4.7K is fine.

not sure of what you want to do and how it should flash your LEDs but try this one

        bmf_comm1   VAR PORTB.0
        bmf_comm2   VAR PORTB.1
        
        LEDBlinkRate CON 200
        
        'Variables
        bmfsD       VAR WORD
        bmfs1D      VAR WORD
        iled        var byte
        
        'configuration du port pour lire l'entrée analogique
        TRISA = %00000011       ' RA0 et RA1 en entrée
        TRISB = %11111100       ' RB0 et RB1 en sortie
       
        DEFINE ADC_BITS 10      ' Set number of bits in result
        DEFINE ADC_CLOCK 3      ' Set clock source (rc = 3)
        DEFINE ADC_SAMPLEUS 50  ' Set sampling time in 
        ADCON1 = %10000010      ' RA0 à RA3

test_depart:
        PORTB = 3               ' Turn ON PORTB<1:0>
        pause LEDBlinkRate      '  
       
val_leds:
        PORTB = 0               ' Turn OFF PORTB<1:0>
        PAUSE LEDBlinkRate        
        ADCIN 0,BMFSd           ' Read CH0
        ADCIN 1,BMFS1d          ' Read CH1
        if (bmfsD > 1000) Then val_leds
        if (bmfsD > 750)  Then boucle_led
        if (bmfsD > 500)  Then boucle1_led
        if (bmfsD > 250)  Then boucle2_led
        
        goto test_depart
        
boucle_led:
        PORTB = 1
        for ILED = 0 to 5
                PORTB = PORTB ^ 3
                PAUSE LEDBlinkRate
                NEXT
        goto val_leds
        
boucle1_led:
        High bmf_comm1
        pause LEDBlinkRate
        goto val_leds
        
boucle2_led:
        high bmf_comm2
        PAUSE LEDBlinkRate
        Goto val_leds
        End

seems to work for me here, 4MHz oscillator, some different led blinking opattern depending of the value of AN0.

[mazoul72]

Hello,

it still doesn't work ....I don't understand why .
The leds are still off whenever i select any buttons .
I see 3.96 V between middle point of my buttons and external pull up resistors (on RA0 and RA1) .
When i push buttons , i have 1.44V / 0.80 V / 0.40 V .
Whith quantum 4.88xxx mV , it gives to me 295 , 164 , 82 .
For 3.96V , 811 aproximatively .

LED replace resistors out (to test)