Low voltage ADC

[tszaboo]

I was searching around for ADCs for a portable equipment. Low voltage ones, which would work from 1.8V system votlage. And I noticed that there aren't any ADC with 16-24 bits of resolution that would be able to measure a bridge sensor or an RTD. Like none. There are some SAR converters, but they are either way too fast or way too expensive for this application. Am I missing something? Is it not possible to make a low voltage delta-sigma ADC for some reason?

[Rerouter]

It is possible, but you end up with a much lower noise by amplifying to a higher voltage first.

[tszaboo]

It is possible, but you end up with a much lower noise by amplifying to a higher voltage first.

If I have an RTD or a bridge sensor, then I drive it with the same voltage as the system voltage. I have one voltage rail, coming from a DC-DC or LDO, and the thing is battery powered. For RTD you dont amplify anything typically, cause it is a ratiometric measurement. For a bridge, I suppose you can run it through an INA333 or similar, but the delta-sigma ADCs that I was using in the past just had a PGA in them.

[coppice]

It is possible, but you end up with a much lower noise by amplifying to a higher voltage first.

If I have an RTD or a bridge sensor, then I drive it with the same voltage as the system voltage. I have one voltage rail, coming from a DC-DC or LDO, and the thing is battery powered. For RTD you dont amplify anything typically, cause it is a ratiometric measurement. For a bridge, I suppose you can run it through an INA333 or similar, but the delta-sigma ADCs that I was using in the past just had a PGA in them.

I'm not sure what the issues are with sigma-delta supply voltage. However, if you look at some of the MCUs with high sensitivity sigma-delta converters inside, the ADC often requires a higher supply voltage than the rest of the chip. For example, the MSP430 MCUs have always run from a minimum of 1.8V. However, the ones with a sigma-delta converter inside can only operate from 1.8V if the ADC is not used. When you use the ADC the supply needs to be at least 2.5V, despite the reference voltage for the converter being only 1.2V.

[Gibson486]

It is possible, but you end up with a much lower noise by amplifying to a higher voltage first.

If I have an RTD or a bridge sensor, then I drive it with the same voltage as the system voltage. I have one voltage rail, coming from a DC-DC or LDO, and the thing is battery powered. For RTD you dont amplify anything typically, cause it is a ratiometric measurement. For a bridge, I suppose you can run it through an INA333 or similar, but the delta-sigma ADCs that I was using in the past just had a PGA in them.

off topic, by why would you not amplify a radiometric measurement? You could make calculation adjustments later.

[SiliconWizard]

There's the ADS1245 which works with a 1.8V digital supply and 2.5V analog supply min.

If your system is battery operated and you have 3V min guaranteed (such as with a LiPo battery), you can easily generate the additional 2.5V supply.

You could also consider this: https://www.maximintegrated.com/en/products/analog/data-converters/analog-to-digital-converters/MAX11359A.html
(probably even closer to what you need!)
Many are wary about Maxim's parts availability though...

[tszaboo]

There's the ADS1245 which works with a 1.8V digital supply and 2.5V analog supply min.

If your system is battery operated and you have 3V min guaranteed (such as with a LiPo battery), you can easily generate the additional 2.5V supply.

You could also consider this: https://www.maximintegrated.com/en/products/analog/data-converters/analog-to-digital-converters/MAX11359A.html
(probably even closer to what you need!)
Many are wary about Maxim's parts availability though...

The problem is really that I am designing industrial stuff that is supposed to work at -40 degrees outside. Batteries are pretty much dead, with any sort of loading the voltage drops rapidly. And to make matters worse, I am doing intrinsically safe designs for explosive atmospheres. So all the fault conditions have to be safe, and every time you make a voltage rail, you are adding days of testing to your device. And additional complexity. It is not just an LDO and some capacitors unfortunately.

But I guess there is no way around it. Probably I need higher voltage.

[amspire]

If you have to go to higher voltages, you could look at the very cheap HX711 24 bit A/D used for the weight scale strain gauge bridges. Lots of cheap modules are available on ebay. The chip runs off 2.6 to 5.5 V and can measure at either 10Hz or 80Hz. 1.5mA. Differential input. SOP-16L Package. Operation temperature range: -40 ~ +85℃

[tszaboo]

If you have to go to higher voltages, you could look at the very cheap HX711 24 bit A/D used for the weight scale strain gauge bridges. Lots of cheap modules are available on ebay. The chip runs off 2.6 to 5.5 V and can measure at either 10Hz or 80Hz. 1.5mA. Differential input. SOP-16L Package. Operation temperature range: -40 ~ +85℃

I cannot use modules. I wouldnt use chinese stuff in the first place.
There is tight supply chain control, and everything is certified.
Imagine the electronics that you design is strapped to a propane tank, and then this propane tank is put in an oil refinery. A fault can kill people.
EEVBblog #603 gives you some idea.

[bjbb]

You could also consider this: https://www.maximintegrated.com/en/products/analog/data-converters/analog-to-digital-converters/MAX11359A.html
(probably even closer to what you need!)
Many are wary about Maxim's parts availability though...

That Maxim part is obsolete.

[splin]

There are some SAR converters, but they are either way too fast or way too expensive for this application.

Since nobody has yet come up with an SD convertor that operates at 1.8V, how about revisiting SARs? The ADS7054 is a 14 bit, 1MSPS, < $2 part. Its speed means you can oversample to easily increase the resolution to 20+ bits.

Oversampling doesn't improve the linearity though - max INL is +/- 3 LSBs or 183ppm. You can improve that significantly by adding dither to the signal, at the expense of reduced signal headroom and thus increased noise. Assuming you are using a microcontroller a DAC can be used to provide the dither which can be synchronised to the ADC sampling for best results. Also, because the microcontroller knows the level of the input signal, it can offset the dither voltage appropriately to keep the signal + dither level within the ADC's input range, avoiding the loss of headroom.

The dither could easily be 25% or even more of the full scale to minimize the linearity error. Verifying the level of INL improvement is your problem though - you can't quote datasheet specs!

You said that the SAR ADCs are too fast - is that because your controller is relatively slow and can't handle the serial datarate? The ADS7054 is cheap enough to use two or more in parallel to reduce noise whilst operating the ADCs at lower clock rates.

[tszaboo]

There are some SAR converters, but they are either way too fast or way too expensive for this application.

Since nobody has yet come up with an SD convertor that operates at 1.8V, how about revisiting SARs? The ADS7054 is a 14 bit, 1MSPS, < $2 part. Its speed means you can oversample to easily increase the resolution to 20+ bits.

Oversampling doesn't improve the linearity though - max INL is +/- 3 LSBs or 183ppm. You can improve that significantly by adding dither to the signal, at the expense of reduced signal headroom and thus increased noise. Assuming you are using a microcontroller a DAC can be used to provide the dither which can be synchronised to the ADC sampling for best results. Also, because the microcontroller knows the level of the input signal, it can offset the dither voltage appropriately to keep the signal + dither level within the ADC's input range, avoiding the loss of headroom.

The dither could easily be 25% or even more of the full scale to minimize the linearity error. Verifying the level of INL improvement is your problem though - you can't quote datasheet specs!

You said that the SAR ADCs are too fast - is that because your controller is relatively slow and can't handle the serial datarate? The ADS7054 is cheap enough to use two or more in parallel to reduce noise whilst operating the ADCs at lower clock rates.

I've used fast SAR ADCs in the past, but they are typically not low power, and require driving. The one that you are suggesting, for example needs a fast opamp, driving a 16pF capacitor into less than 1 LSB settling time in less than 1 uS.
And while the oversampling sounds like a possible way of doing it, I would need to characterise the entire system, including the sampling. And beat or software developers with a stick for days, until they get the algorithm right.

That Maxim part is obsolete.

Funny stuff, I was looking at this Maxim part a few years ago for a different project. It is nice, but probably maxim will obsolete it. Guess I was right.

[splin]

I've used fast SAR ADCs in the past, but they are typically not low power, and require driving. The one that you are suggesting, for example needs a fast opamp, driving a 16pF capacitor into less than 1 LSB settling time in less than 1 uS.

The ADS7054 spec is approx 900uW at 1.8V which is better than many (but not all) delta-sigma convertors for a given data rate and resolution. You don't need fast driver opamps to measure low frequency signals - use a large capacitor to provide the sampling capacitor's charge - 16uF should charge the 16pF to within 1ppm providing the ESR is low enough.

You would need extra circuitry and power consumption to provide the dither signal if the INL isn't good enough.

And while the oversampling sounds like a possible way of doing it, I would need to characterise the entire system, including the sampling.

Yes there could potentially be a significant amount of work to characterize the linearity depending on how low it needs to be. I would however hope that you'd do some testing whatever ADC you used rather than relying on the datasheet values which are frequently wrong (or at least optimistic) or only valid for unrealistic operating conditons.

And beat or software developers with a stick for days, until they get the algorithm right.

Oversampling is just averaging the input samples - adding some numbers together and dividing the result shouldn't be too taxing for even the most deadbeat developers. Getting them to generate a dither signal with a DAC and sycnhronising the sampling/averaging to it would be a bit more effort but hardly challenging compared to the effort required to properly comprehend, configure and use many modern ADCs with comprehensive configurability including PGA gains, filter types, input buffer settings, clock settings, modulator divider ratios, calibration settings etc. Take a look at the AD7768 for an extreme case.

The ADS7054 looks particularly good in this respect with no configuration required apart from the offset calibration.

Given your 1.8V requirement, have you found many other options apart from using a dc-dc convertor to provide a higher supply voltage? Have you found any suitable ADCs?

[tszaboo]

Given your 1.8V requirement, have you found many other options apart from using a dc-dc convertor to provide a higher supply voltage? Have you found any suitable ADCs?

Right now I'm thinking about increasing the supply voltage to 2-2.3V  and use an ADS1118 (or other variant of it) or an ADS1119 (better, does more).

The ADS7054 spec is approx 900uW at 1.8V which is better than many (but not all) delta-sigma convertors for a given data rate and resolution. You don't need fast driver opamps to measure low frequency signals - use a large capacitor to provide the sampling capacitor's charge - 16uF should charge the 16pF to within 1ppm providing the ESR is low enough.

Funny thing about that. I made a quick calculation about the safety of such a system. If you connect your sensor with 5m of cable, include the inductance and capacitance of the sensor, then you cannot use a capacitor that big. This intrinsic safety is tough business. Te idea is, that if you short out the capacitor, you get a tiny little spark. The spark needs to be smaller, than it wouldn't ignite the gas in the atmosphere. And 16 uF is not safe anymore.