1

For a more precise measurement of a battery voltage I'd like to apply auto adjustment. The input voltage ranges from 2.5V-4.2V (single cell) to 20V-72V (module). At the moment the input voltage is divided by ~30 with a differential OpAmp and fed into a 24bit/512kSPS ADC (ADS127L01).

Since the voltage for single cells does only use a fraction of the full scale range I'd like to adjust the circuit. During my research I found a rather promising solution: http://www.electronicdesign.com/test-measurement/optimize-high-voltage-measurements-self-adjusting-attenuator

I built the schematic with LTspice and simulated the desired output voltage which looks quite good. Nevertheless, I'm not quite sure because the diodes are not simulated correctly. Can anybody explain how this should work? As soon the corresponding OpAmp saturates (eg. -5V), the voltage at the negativ input pin increases and can cause damage to the OpAmp as it exceeds the positiv voltage supply. How should this be avoided by the two diodes in parallel to the resistor?

How about accuracy? I also run a noise analysis in spice but as I said, I highly doubt the results are correct. The result would be around 10nV/sqrt(Hz) for the sample rate of ~500kSPS. Is it recommended to apply aditional differential filters, also to match with the differential inputs of the ADS127L01?

Furthermore, I have a question on extending this circuit for even higher voltages up to 1000 V. If I use R4 (eg. 10Meg) and adjust all other resistors so that the stages match with ranges of eg. 100V-300V, 300V-600V and 600V-1000V this should work as well?

schematic

simulate this circuit – Schematic created using CircuitLab enter image description here

I'm looking forward to your replies!

MFH
  • 11
  • 4
  • What's the minimum voltage you want to support? I'm getting the idea that you want to support up to 1000 V. Fine. But can you accept a minimum voltage of 2.5 V? Or do you have to measure down to an exact 0? Oh. Never mind. Your sample rates appear to be very high. What I was thinking of isn't appropriate. – jonk Apr 13 '18 at 20:48
  • Looks like a poor-man's auto-ranging. How about an LTC6910-2 programmable gain amplifier (PGA) to provide gain under control of a 3-bit input code, the LTC6910-2 provides gain in binary-weighted increments. A comparator would detect overflow. (gain is set to 1, 2, 4, 8, 16, 32 or 64) – Tony Stewart EE75 Apr 13 '18 at 20:58
  • 2.0V or 2.5V is perfectly fine. Its important to detect „no cell“ which means voltage below this threshold. On pack level the minimum voltage should not drop below 100V. The idea is to have one layout and either equip for cell level OR cell + module level OR pack level – MFH Apr 13 '18 at 21:02
  • Read [Accuracy, Trueness, Precision, Repeatability, etc](https://electronics.stackexchange.com/questions/310436/how-to-improve-resolution-and-presicion-of-sensor-reading/310500#310500) and tell me what you require for all of these over that huge range. – jonk Apr 13 '18 at 21:05
  • If you can reach 24 bit resolution is questionable. The range 2.5V-4.2V multiplied by 8 gives 20V-33.6V and not 20V-70V? Your DAC has a differential input, would measure against a reference voltage be a solution? – Kitana Apr 13 '18 at 21:20
  • Its not necessary to achieve all 24bits. The resulting sample rate can be 1kSPS so that an averaging of 500 samples is possible. Only few times it hast to be faster. The desired voltage accurracy for a single cell (4.2V) should be below 1mV, so maybe it is possible to reach 4 1/2 or 5 digits in each corresponding Range or is that far from the possible? – MFH Apr 13 '18 at 21:32
  • A 24 bit sigma-delta converter with 1/1000 or 10 bit reduces to 12bit resolution without autoranging,. – Tony Stewart EE75 Apr 13 '18 at 21:33
  • The bandwidth of the LTC6910 is to low for the ADC. Some spikes might be as short as the sample rate of the ADS127L01. – MFH Apr 13 '18 at 21:40
  • What dV/dt do you expect on a battery? for a multiplier with 11MHz GBW – Tony Stewart EE75 Apr 13 '18 at 21:43
  • @Kitana there are also modules with 16cells in series, depending on the module. This is why there are three measurement ranges 1s, 8s and 16s – MFH Apr 13 '18 at 21:43
  • @TonyStewartEEsince1975 depends on the load profile ofc. I already got more than 1mV/us = 1000V/s but for very fast spikes this can be measured with an AC coupling – MFH Apr 13 '18 at 21:49
  • unless your AC result uses coax into scope is loaded with 50 Ohms, I doubt your spikes are accurate with resonant coupling errors from dI/dt.. What reactance are your step load cables mutual coupling, battery ESR*I=V etc, rise time PW50. All I see are measurement errors, false assumptions on ADC BW or problems with decoupling. – Tony Stewart EE75 Apr 14 '18 at 00:35
  • The AC spikes are correct, this was proved with a differential probe N2818A and a DSOX6002A and follows exactly what we expect, also on simple R-RC networks with the same load. Theory (simulation) and measurement fit perfectly for this setup. The measurement circuit with the 24bit/500kSPS ADC is to extend this setup by an accurate DC voltage amplitude for the time domain of few us up to several ms or even seconds. The averaging is than adapted dynamically. What are the downsides of the upper circuit, what are your concerns? The LTC6240 was chosen for simulation as its mentioned in the Appnote. – MFH Apr 14 '18 at 08:52

0 Answers0