ADC conversion using Instrumentation Amplifier

Hi DC,

I doubt the drift of the DAC will be the limiting factor for your system.  The effect of the DAC drift will be divided by the gain of the instrumentation amplifier.  I pulled up an inexpensive DAC from the ADI website:  the AD5601, which has 5 uV/C typical drift.  If you divide this drift by your gain of 4000 it is a tiny tiny 1.3 nV/C.  The drift of your load cell and instrumentation amplifier will far exceed this.

If you want better performance, I would recommend concentrating on your load cell and instrumentation amplifier.  If you are able to get a load cell with greater sensitivity (perhaps by using a load cell with a lower max weight), this would likely give you the best improvement.

You may want to consider the following instrumentation amplifier for better performance:

AD8221:  similar to AD620 but better drift and noise

AD8228:  Outstanding gain drift.  Max gain of 100.

AD8231:  Excellent offset drift and good gain drift.  Single supply.  Max gain of 128, but comes with extra op that you can use for more gain.

AD8230: Outstanding offset drift.  10V supply.  Higher noise than other in amps on this list, so you may need a lower bandwidth filter.

AD8293G160:  Excellent offset drift and good gain drift.  Single supply.  Fixed gain of 160.  Inexpensive.

Matt

DC,

  Another trick I saw years ago was to pulse the drive to the load cell.  In other words, instead of driving it with 5V, you drive it with 20V or more.

You get a lot more output from the bridge, but now you have to worry about self heating.  So people would use a five or ten percent duty cycle.

This adds a lot of complexity:  You need the pulse circuit, and then you either need a sample and hold, which adds more parts, noise, etc.,

or you need a faster ADC to convert during the on time.  This was back in the day of 10 and 12 bit ADCs, so it worked out o.k.  With low

cost 14 and 16 bit ADCs, it's probably not worth it, but thought I would pass it along.

  A major source of error is going to be I*R drop.  You have to be very careful with the layout.  For every component, look at where the ground

currents go.  You don't want any common traces for two or more components.

Harry

Hi,

What do you think of about my approach basically.

Alternative 1

Using 2 power sources +/- 10V and +5V. Feeding Bridge with 10V and op-amps with +/-10V. Using 5 volts to power DAC. Use +/-10V input type ADC.

Alternative 2

Using 5V only. Excite bridge with 5V. Use single supply instrumentation amplifier rail to rail type ( ad8553 ). Again rail to rail type op-amps. and DAC.

As I check single supply op-amps, their spec are not good as dual supply op-amps right? Or they are more expensive compared to dual supplies. But with this design only 1 voltage source is necessary.

My concern is that I probably can not design circuit like evaluation boards. There will be always some tricky points that effects the performance of IC's. I prefer the 10V alternative because after instrumentation amplifier my loadcell signals swings 0 to 8 volts for 1000gr weight. I have more room for mistakes. But for 5V, the output from the loadcell is 0,05uV for 0,05gr. It will be harder to deisgn it. What do you think?

Also for -10V, it is hard to make low noise low drift voltage sources. I have used lm317 before, but you set voltages with resistors, typical drift for such an source is 50ppm. Other alternative is using ref02 and boost my current with high current op-amps. But it sounds complicated a lot of components to use. I could not choose which one to try.

Regards,

DC

Ofcourse no problem,

http://www.zemic.nl/Products-Single-Point-load-cells/product-l6h5.html

10kg model.

I am not considering by the way I must use it!

Regards,

Dogan

Hi DC,

Do you feel comfortable posting a link to the datasheet of the load cell you are considering using?

Matt

Hello DC,

I think you bring a good point about complexity vs. dynamic range.

In the case of single supplies: keep in mind that going from +5V to +/-10V you've only increased your dynamic range by a factor of 4 or 12dB (assuming all the error sources remain constant). This is not a huge advantage, the same design techniques for the signal chain would apply regardless of your power supply choice. Also, typical linear regulators such as LM337 (I'm assuming you meant 337 not 317 that for the negative regulator) require at least 1.5V beyond the power rail to work. Do you have bipolar supplies available for this? Would you need to add a switching supply to boost the voltage rails? Depending on what supplies you have available, you can decide how much complexity, extra power dissipation, and cost is added by going this way. Some of the inamps that Matt proposed are quite inexpensive vs. the additional regulators and components, which are not free.

In the case of dual supplies: If you're concerned about the drift of the positive or negative rails, for a regular load cell this is a DC common-mode voltage variation, which is typically rejected by the instrumentation amplifier. Even if you get 100ppm/C if you have as low as 80dB CMRR (divide CMV by 10^4 = 10,000) the variation becomes 0.01ppm/C.

Also keep in mind that +/-10V input ADCs are harder to find. The good news is that there are alternatives to this. For the sensitivity you mentioned, you are going to need at least an 18-bit ADC, unless you consider Harry's approach. If you feel like the single supply is not the best way to go, I recommend you take a look at the following circuit note (you can use the AD7690 or any other differential input ADC if you don't need the speed):

http://www.analog.com/en/circuits-from-the-lab/CN0180/vc.html

If you need to reject a vibration frequency (and in the best case is periodic) you should consider a sigma-delta ADC such as AD7791 or similar. You can also do the same with a SAR, but you may need to average on the digital domain. Depending on whether or not you have this processing power, you can decide what to do. But most of the time, you'll get better performance with a sigma-delta since it will look at the signal for the longest period of time (of course, you'll pay the price in total system throughput). Since you need to look at a few load cells, you're going to need an analog multiplexer such as ADG409. One more advantage of the SAR ADC in this case will be that it would allow you to interleave samples even if you average at the end, thus increasing your system throughput. The circuit note above should be capable of switching channels with high-slew rates so it won't be an issue.

I hope this helps,

Gustavo

Hi DC,

As I check single supply op-amps, their spec are not good as dual supply op-amps right? Or they are more expensive compared to dual supplies. But with this design only 1 voltage source is necessary.

Speaking very broadly, you can divide our in amp portfolio into two types of technologies:

- BJT based inamps:  these parts can operate on wide supplies (+/-15).  Depending on the design chosen, they may or may not work well on single supply.    BJT based in amps have excellent noise vs. supply current and bandwidth vs. supply current tradeoffs.  Examples of these parts are the AD620, AD8221, AD8226, and AD8429.

- CMOS based inamps:  These parts are typically limited to single supply operation.  These designs typically use autozero/chopping technology to cancel out offset error, which gives them excellent offset and offset drift specifications.  Examples of these parts are the AD8553, AD8293, and AD8231.

While prices can vary considerably in both categories, on average CMOS parts are typically less expensive than BJT based designs.

What do you think of about my approach basically.

Alternative 1

Using 2 power sources +/- 10V and +5V. Feeding Bridge with 10V and op-amps with +/-10V. Using 5 volts to power DAC. Use +/-10V input type ADC.

Alternative 2

Using 5V only. Excite bridge with 5V. Use single supply instrumentation amplifier rail to rail type ( ad8553 ). Again rail to rail type op-amps. and DAC.

If you can use a 5V design, you won't have to have so many power supplies and may make for a simpler design.  But you are correct that powering the bridge from a wider supply will generate a larger signal, and therefore lesson the requirements on the signal conditioning.

I don't know which errors in your system will dominate, so I can't offer you much advice on the route to choose.   For example I don't know if the errors are dominated by the mechanical aspects of the system and the load cell, or whether the signal conditioning circuitry plays a significant role.  Assuming you had a perfect mechanical system and low bandwidth, alternative 1 might be better because autozero/chopper amps can have very low offset drift, even when measuring smaller signals.  If you bandwidth is high, alternative 2 might be better because at some point you will want the better noise performance that a BJT part (along with the larger voltage out of the bridge) can provide.

If you do end up using a +/-10V power supply for the amplifiers, you can still use a 5V ADC if you wish.  We have parts like the AD8275 and AD8475 that are designed to convert large industrial signals into the range of modern single supply ADCs.

Also for -10V, it is hard to make low noise low drift voltage sources. I have used lm317 before, but you set voltages with resistors, typical drift for such an source is 50ppm. Other alternative is using ref02 and boost my current with high current op-amps. But it sounds complicated a lot of components to use. I could not choose which one to try.

Often instead of making a perfect supply, you can deal with variations in the supply by driving your ADC reference with a voltage that varies with the supply.  For example, you could use the AD8275 to sense the 20V supply across the bridge.  The AD8275 would attenuate this voltage to 4V, which you could then use as a reference for your ADC.

As an example, let's say your power supply voltage dropped by by 5%.  This means the voltage that your in amp reads from the bridge will also drop by 5%, and this error will eventually get read by the ADC.  However since the ADC's reference voltage also dropped by 5%, the digital value coming out of the ADC will remain the same.

You may also find this article or this reference helpful.

Matt

This question has been assumed as answered either offline via email or with a multi-part answer. This question has now been closed out. If you have an inquiry related to this topic please post a new question in the applicable product forum.

Thank you,
EZ Admin