Post by thread:[PIC]:16F877 A/D converter noise

I am using the 10-bit A/D converter in the PIC16F877 to digitize 2 signals at a rate of 2000 samples per second for display on an LCD screen. I seem to always have about 10 counts of noise even when I put a large capacitor bypassing the RA0 and RA1 input pins to ground. I am using a 16 mHz clock frequency that I can't reduce. I think I am following the datasheet recommendations: tad=32Tosc low impedance drive from an op-amp wait 20uS between reading RA0 and RA1 .1 uF bypass caps near the chip power pins. LM336 - 2.5 volt reference on RA3 bypassed with 10uF software is written in CCS-C and uses the built in read_adc() function Is there some secret to getting true 10-bit accuracy. I don't even have 8-bit now. -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

{Quote hidden}> -----Original Message----- > From: Eric Strauts [SMTP:spam_OUTstrautsTakeThisOuTDSL.TELOCITY.COM] > Sent: Friday, December 08, 2000 1:31 PM > To: .....PICLISTKILLspam@spam@MITVMA.MIT.EDU > Subject: [PIC]:16F877 A/D converter noise > > I am using the 10-bit A/D converter in the PIC16F877 to digitize 2 > signals at a rate of 2000 samples per second for display on an LCD > screen. I seem to always have about 10 counts of noise even > when I put a large capacitor bypassing the RA0 and RA1 input pins > to ground. > > I am using a 16 mHz clock frequency that I can't reduce. > > I think I am following the datasheet recommendations: > > tad=32Tosc > low impedance drive from an op-amp > wait 20uS between reading RA0 and RA1 > .1 uF bypass caps near the chip power pins. > LM336 - 2.5 volt reference on RA3 bypassed with 10uF > software is written in CCS-C and uses the built in read_adc() > function > > Is there some secret to getting true 10-bit accuracy. I don't even > have 8-bit now. > Just implemented some ADC code on a project and tested it today. I am using Vcc and 0v as the references and I'm getting a rock solid reading, only getting jitter on the LSB when it approaches it's switching point. This is using a 20 MHz crystal, but this is on a PCB we nice power and ground planes and I'm only using one channel at a low sampling rate (20 samples/sec). I would check the supply to your PIC, both Vcc and ground and make sure they are as quiet as possible. The ground pin of your reference should be connected as closely as possible to the ground on the PIC. Regards Mike -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

{Quote hidden}> I am using the 10-bit A/D converter in the PIC16F877 to digitize 2 > signals at a rate of 2000 samples per second for display on an LCD > screen. I seem to always have about 10 counts of noise even > when I put a large capacitor bypassing the RA0 and RA1 input pins > to ground. > > I am using a 16 mHz clock frequency that I can't reduce. > > I think I am following the datasheet recommendations: > > tad=32Tosc > low impedance drive from an op-amp > wait 20uS between reading RA0 and RA1 > .1 uF bypass caps near the chip power pins. > LM336 - 2.5 volt reference on RA3 bypassed with 10uF > software is written in CCS-C and uses the built in read_adc() > function > > Is there some secret to getting true 10-bit accuracy. I don't even > have 8-bit now. > Ahh... noise reduction. It is indeed and truly a Black Art (tm). Bypassing is good, you're on the right track there. How does your PCB layout look? What kind of noise do you have on your rails? I presume you used at least one ground plane, if not a four layer board. Multilayer boards generally reduce noise by 20dB(1). On way, programmatically, that you can improve your readings is to put the processor into sleep while acquiring. I see that you need samples every 0.5 mS. It would appear that putting the processor to sleep is not an option. I seem to recall reading that the MPU takes a few mS to start up in HS mode. Regards, -d (1) Kimmel and Gurke "EMI bullets". -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

On Fri, 8 Dec 2000 07:30:41 -0600, you wrote: >I am using the 10-bit A/D converter in the PIC16F877 to digitize 2 >signals at a rate of 2000 samples per second for display on an LCD >screen. I seem to always have about 10 counts of noise even >when I put a large capacitor bypassing the RA0 and RA1 input pins >to ground. > >I am using a 16 mHz clock frequency that I can't reduce. > >I think I am following the datasheet recommendations: > >tad=32Tosc >low impedance drive from an op-amp >wait 20uS between reading RA0 and RA1 >.1 uF bypass caps near the chip power pins. >LM336 - 2.5 volt reference on RA3 bypassed with 10uF >software is written in CCS-C and uses the built in read_adc() >function > >Is there some secret to getting true 10-bit accuracy. I don't even >have 8-bit now. Try eliminating possible noise sources one at a time to find out the root cause. A few things to try : Run at a slower clock Sleep whilst doing the conversion Put a series resistor in the op-amp output - I've known some opamps to glitch when an a/d input mux changes. Add some extra delay between setting the input channel and starting the conversion. Prove it's not your amplifier by feeding a known stable signal (e.g. 1.5v battery) into the ADC pin. Prove it's not reference noise, e.g. by using a 3v battery. Make sure your oscillator grounding is correct (osc caps direct to ground pin with nothing else on the same track branch. I've not used 10 bit mode, but Microchip's 8-bit ADCs have always been very good when used properly - you can often get the output to sit on one single value. -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads