Hi, We're using the 12-bit ADC for 5 sensors. Reading the spec sheet of the ADC, it says that it can sample at 8kHz and more. The model specifies an impedance of 2000kohm and 40pF. Our sensors have first-order low-pass filters. The output is not buffered by an amp. That is, the output of low-pass is fed directly to the ADC input. I calculated the time constant. when the internal switch of the ADC closes, I assume that the low-pass capacitor(10nF) is an open-circuit. The 32kOhm low-pass resistor is then in series with the 2kOhm. The time constant is then: R*C=(32k+2k)*40pF=0.00000136 sec. Setting the sampling time to 10*R*C=0.0000136 sec. If a drive the ADC with a calibrated DCO at 4 MHz, this amounts to 54.4 clock cycles. However, the signal is still very noisy at 64 cc. It takes 50 ms of sampling before the signal stabilises. If I increase it to 1024 cc, the signal stabilises faster but is still very noisy. If I put a 1uF capacitor(low bandwidth) and sampling time of 1024 cc, the signal becomes much clearer but not as clean as it should be. I believe that this is due to 2 things: 1)the low-pass capacitor exchanges charges with the ADC capacitor. The lower the capacitor, the more a loss of charge affects its voltage. However, I just can't figure out the equation that shows that. Anyone does? 2)cross-talk between sensors is amplifed by not having a buffered output. Is this true? Thank you for your answers. Yvan Newtrax Technologies www.newtraxtech.com
ADC sampling time problem
Started by ●April 3, 2005
Reply by ●April 3, 20052005-04-03
Your 32Kohm+2kohm tell me that your sensor is an ADXL accelero, which should be considered a current source.If you don't buffer it then you should allow enough time (SHT0_10 for example) for the ADC12 input caps to charge. The noise that you see while sampling is normal according to data sheet. Karl. >From: "yvancastilloux" <kayoux@kayo...> >Reply-To: msp430@msp4... >To: msp430@msp4... >Subject: [msp430] ADC sampling time problem >Date: Sun, 03 Apr 2005 22:05:04 -0000 > > > >Hi, > >We're using the 12-bit ADC for 5 sensors. > >Reading the spec sheet of the ADC, it says that it can sample at 8kHz >and more. The model specifies an impedance of 2000kohm and 40pF. > >Our sensors have first-order low-pass filters. The output is not >buffered by an amp. That is, the output of low-pass is fed directly >to the ADC input. > >I calculated the time constant. when the internal switch of the ADC >closes, I assume that the low-pass capacitor(10nF) is an open-circuit. >The 32kOhm low-pass resistor is then in series with the 2kOhm. > >The time constant is then: R*C=(32k+2k)*40pF=0.00000136 sec. >Setting the sampling time to 10*R*C=0.0000136 sec. If a drive the >ADC with a calibrated DCO at 4 MHz, this amounts to 54.4 clock cycles. > However, the signal is still very noisy at 64 cc. It takes 50 ms of >sampling before the signal stabilises. If I increase it to 1024 cc, >the signal stabilises faster but is still very noisy. If I put a 1uF >capacitor(low bandwidth) and sampling time of 1024 cc, the signal >becomes much clearer but not as clean as it should be. > >I believe that this is due to 2 things: > >1)the low-pass capacitor exchanges charges with the ADC capacitor. >The lower the capacitor, the more a loss of charge affects its >voltage. However, I just can't figure out the equation that shows >that. Anyone does? > >2)cross-talk between sensors is amplifed by not having a buffered >output. Is this true? > >Thank you for your answers. > >Yvan >Newtrax Technologies >www.newtraxtech.com > > > > > >. > > >Yahoo! Groups Links > > > > > > > _________________________________________________________________ Dont just search. Find. Check out the new MSN Search! http://search.msn.click-url.com/go/onm00200636ave/direct/01/
Reply by ●April 12, 20052005-04-12
Hi, thank you for your answer. However, our hardware consultant made simulations with the complete model. He was able to approximate the following formulas: C2 = Lowpass cacitor C1 = ADC capacitor (40 pF) R1 = Sensor output resistance (32kOhm) R2 = ADC capacitor (2 Kohm) The initial sampled value when the sensor is first sampled is equal to :C2/(C1+C2) The first order settling time is: R2 * C1 * 10. It represents the minimum time to charge C1. The second order settling time is R1 * C2 * 3 It represents the minimum time to charge C2. For C2 = 10nF : The initial sampled value will be at 99.6% of the final sampled value ~ 8 LSB with a settling time of 1ms For C2 = 100nF The initial sampled value will be at 99.96% of the final sampled value - 0.8LSB with a settling time of 10ms. These values were observed experimentally as well. The 8LSB noise value is much larger than the theorical one. Yvan --- In msp430@msp4..., "Karl Adler" <specific2@m...> wrote: > Your 32Kohm+2kohm tell me that your sensor is an ADXL accelero, which should > be considered a current source.If you don't buffer it then you should allow > enough time (SHT0_10 for example) for the ADC12 input caps to charge. The > noise that you see while sampling is normal according to data sheet. > Karl. > > >From: "yvancastilloux" <kayoux@h...> > >Reply-To: msp430@msp4... > >To: msp430@msp4... > >Subject: [msp430] ADC sampling time problem > >Date: Sun, 03 Apr 2005 22:05:04 -0000 > > > > > > > >Hi, > > > >We're using the 12-bit ADC for 5 sensors. > > > >Reading the spec sheet of the ADC, it says that it can sample at 8kHz > >and more. The model specifies an impedance of 2000kohm and 40pF. > > > >Our sensors have first-order low-pass filters. The output is not > >buffered by an amp. That is, the output of low-pass is fed directly > >to the ADC input. > > > >I calculated the time constant. when the internal switch of the ADC > >closes, I assume that the low-pass capacitor(10nF) is an open-circuit. > >The 32kOhm low-pass resistor is then in series with the 2kOhm. > > > >The time constant is then: R*C=(32k+2k)*40pF=0.00000136 sec. > >Setting the sampling time to 10*R*C=0.0000136 sec. If a drive the > >ADC with a calibrated DCO at 4 MHz, this amounts to 54.4 clock cycles. > > However, the signal is still very noisy at 64 cc. It takes 50 ms of > >sampling before the signal stabilises. If I increase it to 1024 cc, > >the signal stabilises faster but is still very noisy. If I put a 1uF > >capacitor(low bandwidth) and sampling time of 1024 cc, the signal > >becomes much clearer but not as clean as it should be. > > > >I believe that this is due to 2 things: > > > >1)the low-pass capacitor exchanges charges with the ADC capacitor. > >The lower the capacitor, the more a loss of charge affects its > >voltage. However, I just can't figure out the equation that shows > >that. Anyone does? > > > >2)cross-talk between sensors is amplifed by not having a buffered > >output. Is this true? > > > >Thank you for your answers. > > > >Yvan > >Newtrax Technologies > >www.newtraxtech.com > > > > > > > > > > > >. > > > > > >Yahoo! Groups Links > > > > > > > > > > > > > > > > _________________________________________________________________ > Don't just search. Find. Check out the new MSN Search! > http://search.msn.click-url.com/go/onm00200636ave/direct/01/