Analogue OverSampling

Download the files here: (Right click and Save As)

The attached .zip has only one file, DT_Analog.pbp. Extract it to your PBP folder (the one with PBPW.exe)

You've probably noticed that when using the Analog to Digital converter, the numbers are NEVER stable. At best, the result will always be bouncing back and forth between two numbers.\\
While it's extremely annoying, there is a reason for it, and the number of times it bounces between those two numbers is actually indicating the values of more bits of resolution that are not included in the original 10-bit result. This effect can be exploited to increase the effective number of bits in the result.
By taking a number of consecutive samples and averaging the results with a lower divisor, you can get up to 16-bit accuracy from that poor little 10-bit converter.

An explanation of the Oversampling technique can be found on Wikipedia.

Samples required:
For each additional bit of resolution, the number of samples required is multiplied times 4 (22(n-10)). The samples are accumulated in a 32-bit variable, then divided by a number that is doubled for each additional bit (2(n-10)). [This is all handled by the module.]

Maximum results:
Since the highest value provided by the converter is 1023, it can't bounce between 1023-1024.
This limits the maximum values to slightly lower than expected, which is (1023 << (bits-10))
For 16-bits, the maximum is (1023 << 6) or 65472. This value is returned in the ADmax variable after the conversion for easy voltage calculations.

A/D Bits Samples required MAX result Conv. Time1
10 1 1023 24.2us
11 4 2046 96.8us
12 16 4092 387us
13 64 8184 1.54ms
14 256 16368 6.2ms
15 1024 32736 24.7ms
16 4096 65472 99.1ms

1 Conversion time based on 5us acquisition with 1.6us Tad.

Taking 1/10th of a second to convert a 16-bit value seems like a colossal waste of time. But it's nice to know you can do it if you really need to.
To me, the real winner is 12-bit, at under 400us conversion time, with a 4 fold increase in resolution, on almost any PIC, ... I doubt I'll be using 10-bit much anymore. But even 14-bit's not too bad @6ms, although the extra time for 14-bit probably isn't worth it for most analog projects.

Conversion time:
The time it takes to complete a conversion depends on a couple of factors.

  1. The Acquisition time specified in the DEFINE ADC_SAMPLEUS statement. If there is no DEFINE, PBP will default to 50us, which is for the absolute worst case scenario with high impedance. It's very likely that you will be able to use a much smaller value. With an impedance of 1K or less, you can usually get away with 3-5us. Also, the minimum ADC_SAMPLEUS is limited to the minimum PAUSEUS delay, which at 4Mhz OSC is 24us, @20Mhz it's 3us. Use as low a value as you can possibly get away with because this time happens for every sample. With 256 or more samples being taken, it adds up pretty quick.
  2. The ADC clock source should be set as close as possible to 1.6us Tad. And it takes 12 Tad to complete a 10-bit conversion. This one's hard to explain, so I'll just list the recommended values. Use the highlighted values in your DEFINE ADC_CLOCK statement, based on your oscillator frequency.
OSC DEFINE ADC_CLOCK Clk source Tad (μs) 12 Tad (μs)
4 1 Fosc/8 2 24
8 5 Fosc/16 2 24
10 5 Fosc/16 1.6 19.2
12 2 Fosc/32 2.6 31.2
16 2 Fosc/32 2 24
20 2 Fosc/32 1.6 19.2
24 6 Fosc/64 2.6 31.2
32 6 Fosc/64 2 24
40 6 Fosc/64 1.6 19.2
48 3 FRC 4 48

Noise in the system: As previously mentioned, it's the number of times the value bounces back and forth between two numbers that allows it to gain more bits of resolution. If (due to poor design) your A/D values are swinging wildly even at normal 10-bit, this system will also help calm them down (some) since it will find an average of the samples taken. Usually it's the mid-point of those swings that you want to find anyway. Naturally, you should minimize ALL noise as much as possible to begin with.

Using the Module: You must first configure your PIC to have the desired Analog inputs and use the correct Oscillator source. This module will not do it for you. You must also set the ADC DEFINE's for the fastest available 10-bit results, depending on your oscillator frequency. Then INCLUDE the file somewhere near the top of your program.

Set the ADchan variable to the AN channel you wish to read. (ADchan = 1) Set the ADbits variable to the desired resolution. (ADbits = 12) Then GOSUB GetADC.

The resulting value will be returned in the ADvalue word sized variable. And the maximum A/D value for the selected resolution is returned in the ADmax variable.

Here's a simple example that displays 12-bit A/D values on an LCD.
Of course, by changing one number it will display 16-bit values too.

;--- Set your __configs here as needed ---
INCLUDE "DT_Analog.pbp"     ; DT's 16-bit Analog Module

DEFINE ADC_BITS 10          ; Set-up ADC for fast 10-bit results

;----[Change these to match your hardware]----------------------------------
DEFINE LCD_DBIT    0        ; Starting Data Bit
DEFINE LCD_RSREG   PORTB    ; Register Select Port
DEFINE LCD_RSBIT   4        ; Register Select Bit
DEFINE LCD_EREG    PORTB    ; Enable Port
DEFINE LCD_EBIT    5        ; Enable Bit
DEFINE LCD_BITS    4        ; Data Bus Size
DEFINE LCD_LINES   2        ; Number of Lines on LCD
DEFINE LCD_COMMANDUS  2000  ; Command Delay time in uS 
DEFINE LCD_DATAUS  50       ; Data Delay time in uS

ADCON1 = %10001101          ; right justify, AN0 & AN1 analog
PAUSE 250 : LCDOUT $FE,1    ; Initialize LCD
PAUSE 100 : LCDOUT $FE,1    ;  not all LCDs need both lines

ADbits = 12                 ; set to 12-bit resolution
  FOR ADchan = 0 to 1       ; Do both AN0 and AN1
    GOSUB GetADC            ; Get A/D value
    IF ADchan = 0 THEN
      LCDOUT $FE, $80      ; LCD line 1
      LCDOUT $FE, $C0      ; LCD line 2
    LCDOUT "CH-",DEC1 ADchan," = ",DEC ADvalue,"   "
   NEXT ADchan
GOTO Main:

Here's another example that shows all the resolutions from 1-16 bit from a single A/D channel via the USART with HyperTerminal or other ANSI terminal program.

;--- Set your __configs here as needed ---
INCLUDE "DT_Analog.pbp"           ; DT's 16-bit Analog Module


DEFINE HSER_SPBRG 10              ; 115200 @ 20 Mhz
DEFINE HSER_RCSTA 90h             ; Hser receive status init 
DEFINE HSER_TXSTA 24h             ; Hser transmit status init 
DEFINE HSER_CLROERR 1             ; Hser clear overflow automatically 

Volts   VAR WORD                  ; Volts calculated from the A/D reading

ADCON1 = %10001101                ; right justify, AN0 & AN1 analog
HSEROUT [27,"[2J"]                ; Clear screen
  HSEROUT [27,"[H"]               ; home cursor
  HSEROUT [27,"[34m",27,"[1m"]    ; bold blue text
  HSEROUT ["ADbits   DEC    HEX       BIN16         ADmax   Volts"]
  HSEROUT [27,"[37m",13,10]       ; black text

  ADchan = 1                      ; Select channel AN1
  FOR ADbits = 1 to 16            ; cycle thru all bit resolutions
    GOSUB GetADC                  ; get the A/D value
    GOSUB ShowAD                  ; show the A/D results
  NEXT ADbits                     ; do next resolution

  IF ADbits < 10 THEN HSEROUT [" "]
  HSEROUT [DEC ADbits,"-bit = ",DEC ADvalue,"    "]
  HSEROUT [27,"[",DEC ADbits+1,";16H",HEX4 ADvalue,"  ",BIN16 ADvalue, _
          "   ",DEC ADmax]
  IF ADbits < 16 THEN             ; calculate the Voltage with 4 decimals
    Volts = 50000 * ADvalue
    Volts = DIV32 ADmax
  ELSE                            ; with 16-bit, ADmax is too big for DIV32 
    ADmax = ADmax >> 1            ;   cut it in half
    Volts = 25000 * ADvalue       ;   and scale the multiplier accordingly
    Volts = DIV32 ADmax           ; PBPL & Longs, doesn't have that problem
  HSEROUT [27,"[",DEC ADbits+1,";49H", _
           DEC Volts/10000,".",DEC4 Volts//10000 ,13,10]

NOTE: This will NOT work as is on 12-bit A/D converter.

Page last modified on March 03, 2018, at 03:20 PM