EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: OM222O on January 20, 2019, 10:14:33 am
-
Hello
I need to be able to read a really small signal as a differential pair using an delta sigma ADC.
The adc itself allows for 2 differential channels but I want to eliminate the noise as much as possible using a filter before feeding it into the input of the amplifier (Something similar to this circuit which I found on a T.I datasheet):
(https://i.imgur.com/WLuA7pD.jpg)
my idea was to use a difference amplifier followed by a second order low pass filter to block any AC noise but this requires very low offset amplifiers such as the MAX4238. The signal itself is in the range of 10s of uV (measured across high current shunt using 2 wires of about 30cm length) so signal to noise ratio is pretty awful on the input side. I haven't been able to find a difference amplifier with low offset voltages so far and I know using external resistors is janky at best. I was considering having the footprint of the shunt resistor etched on the PCB so instead of using wires, I can directly touch it to the PCB to get a reading but that will be my last backup option. If anyone has a different approach to solving the issue or knows a difference amplifier with about 5uV max offset, please let me know, thanks.
-
I know using external resistors is janky at best.
Yes, but you may look for precision resistor networks. They can have very good tolerance and tempco on resistance ratio.
-
The ultra low offset/noise/tempco differential (output) opams are almost none existent. I've been struggling with it in a different thread on an AFE for 24++bit ADCs.
While reading posts on TI's engineering forum I've seen the TI's OPA388 has been recommended as the best front end opamp for the ADS1263 high-end. It is not a differential output opamp, however.
You may consider single ended ADC input, ie. with ANI- grounded, and you would need a small negative supply for the opamp as well.
Btw, there are a few ADCs which can work bipolar (ie AVcc=2.5V, AVss=-2.5V) and they have got a chopper PGA as well. Usually you do not need any AFE for it then, as your shunt is very low impedance.
Resistors: the LT5400-x series (quad resistors array on a chip) is low abs tempco (8ppm/C) with 0.2ppm/C matching tempco.
-
The ultra low offset/noise/tempco differential (output) opams are almost none existent. I've been struggling with it in a different thread on an AFE for 24++bit ADCs.
While reading posts on TI's engineering forum I've seen the TI's OPA388 has been recommended as the best front end opamp for the ADS1263 high-end. It is not a differential output opamp, however.
You may consider single ended ADC input, ie. with ANI- grounded, and you would need a small negative supply for the opamp as well.
Btw, there are a few ADCs which can work bipolar (ie AVcc=2.5V, AVss=-2.5V) and they have got a chopper PGA as well. Usually you do not need any AFE for it then, as your shunt is very low impedance.
Resistors: the LT5400-x series (quad resistors array on a chip) is low abs tempco (8ppm/C) with 0.2ppm/C matching tempco.
The LT5400 seems a really solid option! I don't really need split supply ADC but thanks for the recommendation. the small negative supply already exists in the design but thanks for mentioning it. OPA2388 would be great for the task as well without being too expensive (2 of the MAX4238s will cost more but probably perform marginally better, I need to do more testing on that). You saved my day, thanks!
-
LT5400 is not cheap, btw. :)
You may use the LT5400 for the R1-R4 with the OPAx388 (with some small -Vss) and the R5 and R6 are not critical, but they should be kept rather low res. The second opamp for the Vcm (Vref/2), the R divider at its input is not critical as the Vcm errors cancel out (diff ADC inputs).
PS: the Vcm opamp could be any standard RR low voltage one.
-
LTC6102. Offset voltage max 10µV, drift max 0.035µV /°C :
https://www.analog.com/media/en/technical-documentation/data-sheets/6102fe.pdf (https://www.analog.com/media/en/technical-documentation/data-sheets/6102fe.pdf)
-
There are chopper stabilized instrumentation amplifiers like the LTC2053 and AD8230 which could be used. The LTC1043 or LTC6943 switched capacitor building block could also be used with a separate operational amplifier.
Do not get hung up on providing the ADS1115 with a true differential signal. If provided from the current shunt, the extra active signal path will increase noise by 3dB and if provided by the output of the amplifier, will provide little or no benefit.
Offset can be calibrated out. Offset voltage drift is the problem but even worse with a current shunt will be thermocouple effects which will require careful attention. Note that offset voltage drift in a chopper stabilized amplifier is actually produced by thermocouple effects associated with the amplifier itself.
-
Long time back I got a bag of ADS1110. Its max resolution is 8uV (with typical 8uVpp noise) only, and all other params are pretty low-end (ie 60Hz suppression only), but tried High-Side Shunt Sensing, with no opapm (inputs tied directly to the shunt) and it worked (an experiment) :)
Below a single opapm Low-Side Shunt TI recommends for this 16bitter even the OP requires much better noise performance, afaik..
-
There are chopper stabilized instrumentation amplifiers like the LTC2053 and AD8230 which could be used.
Also MAX4208.
-
There are chopper stabilized instrumentation amplifiers like the LTC2053 and AD8230 which could be used.
Also MAX4208.
I have lost touch with Maxim products over the years because as a company they have always been so difficult to deal with.
-
Hey there,
Have you tried using a difference amplifier with integrated resistors like the INA190 (TI)? The bandwidth on these types of parts isn't great (50-100kHz), however, the ADS1115 isn't sampling at a very high rate either (<1ksps). These INA devices are made specifically for the type of circuit you drew. Some of them even have integrated data-converters with an I2C interface.
Curious, what's the end application?
Edit: INA233 features an internal 16-bit ADC. Also if you need a small negative voltage power supply (for those down to ground applications), the LM7705 is a little known device that generates a -0.23V rail from 3.3V or 5V input. It does have a bit of ripple on the output (single-digit mV), but the op-amp PSRR usually filters that out nicely.
Edit #2: I suppose that I should have consolidated all of my research before I started typing away. Regarding the OPA388, that's a chopper based amplifier and it does have ESD diodes across the inputs, so you can't effectively pull the inputs apart within forward biasing a diode. In most applications that's a don't care but I figured I'd let you know just the same. TI's newer precision amplifiers are starting to adopt eTrim, which is essentially EEPROM programmed at manufacturing. The EEPROM is used to set various trim elements inside the amplifier. Traditionally, laser trim has been the go to; however it does suffer from two draw backs: it's expensive, and you have to trim at the wafer stage, which means that you can't factor in package parasitics. eTrim devices are trimmed at final-test, which means that package parasitics are lumped into the measure/trim function. INA192 is an example of one such device and features near chopper level performance, without the chopper noise or excessive cost of a laser trim device. About the best device I found at TI is the OPA189. 3uV offset. Both the INA192 and INA189 also feature comparator inputs, which means there are no back-to-back diodes across the inputs. If you're muxing various gains stages together, this can be advantageous.
If you choose to run with a discrete op-amp, you'll need to match those resistors if you hope to have any CMRR. You can buy precision resistor arrays, but they're expensive. You could also add trim potentiometers in the feedback loop (less expensive, but someone has to tune each circuit). This circles us back to my original suggestion of the INA190 (resistors baked into the device). You likely know this, but resistors on silicon are very tightly matched, but have horrible accuracy. If the application is ratiometric, then all is good. Otherwise not so.
These days, about the only reason I see folks go with discrete op-amps for this application is if they need bandwidth. Lots of it. Trading precision for bandwidth is often a delicate proposition.
Good luck with the project!!
-
Do you have some ballpark figures for the range of Shunt values and currents you want to use?
As said before, the Shunt tends t be a pretty low impedance source and it seems that an extra opamp justs adds / amplifies noise.
Sometimes I think they add those opamps in AN's just to sell more components.
If your input signal is in the "10's of uV" then using a chip that is designed for thermocouples seems a better choice, for example the ADS1220, which also has a programmable PGA upto 128x.
-
If your input signal is in the "10's of uV" then using a chip that is designed for thermocouples seems a better choice, for example the ADS1220, which also has a programmable PGA upto 128x.
At high data rates and at low PGA gains the input noise increases significantly on delta-sigma converters so it can be advantageous to use an external operational amplifier depending on the application. Or an operational amplifier might be used to remove common mode noise if that is an issue.