Are there any manufacturers of floating point ADCs? There seems to be a lot of literature on floating point ADCs, but I couldn't find one on Analog Devices or Texas Instruments. What I am trying to get is a ADC with a 10^4 precision for voltage from 1mV to 10V - needing 27 bit (Min) ADC, which TI does have - but at very low frequencies than required. And I am also trying to avoid using logarithmic amplifiers. Any suggestions and guidance are highly appreciated...

• What frequencies do you require? Commented Sep 23, 2012 at 14:30
• Are you sure you want 27 bits?? A good A/D (not even floating) advertised as 24 bits might give you 18. 10^4 is 13-14 bit, which is a lot cheaper. I would suggest a set of amplifiers, and switch between the amplifier outputs (some might saturate, but who cares). Commented Sep 23, 2012 at 14:39
• minimum of 15kHz - it is to be used as an input to a digital PID feedback against ambient noise, which dies off o acceptable limits beyond 15kHz
– YSSP
Commented Sep 23, 2012 at 14:46
• @Wouter - the precision needs to be 10^4 for both an input of 1mV as well as for an input of 10V. So the LSB should correspond to ~.1uV. Hence the 27 bits. Unless of course there is a commercial FP-ADC.
– YSSP
Commented Sep 23, 2012 at 14:52
• @YSSP I assume there is some intelligence after the A/D. My guess is that a 27 bit ADC, if feasible at all, will be VERY expensive. The problem is that you want to get by with one ADC for the whole range, if you had some form of auto-scaling you could get by with a much much cheaper ADC. So why not have a bunch (let's say 5 ) amplifiers that amplify 1, 10, 100, 1000 and 10000 times, each feeding into a ADC that is 10^5 accurate? The intelligence will choose the 'lowest' ADC that does not saturate. I might be mistaken, but my guess is that such a system will be much cheaper. Commented Sep 23, 2012 at 15:09

One approach is to use two converters in parallel, with different fixed-gain amplifiers for each. You use software to combine the two sets of samples into a single floating-point stream.

For example, you could have two 24-bit converters, one with a 10V full scale range and the other with a 1V full scale range. When the first one indicates that the signal is less than 1V, use the output of the second one. The software uses the outputs of both converters where the ranges overlap to maintain an accurate relative scaling factor between them.

The trick is to make sure the second converter (and its amplifier) recovers quickly from saturation when the signal goes outside its range. If the saturation characteristics are well-known, the software can help compensate for them.

• See my comment for essentially the same suggestion, but without the warning about saturation. Commented Sep 23, 2012 at 15:50
• Thank you guys for the alternate solution to the FP-ADC. We shall try to get it to work indeed...
– YSSP
Commented Sep 23, 2012 at 16:27

If you want something that works really well, you should try a Voltage-Controlled Amplifier (VCA) between your signal and the ADC. Depending on how quickly need to adjust between big and small signals, you can have either a P or PI controller that adjusts the VCA.

This works really well since the ADC is never saturated and the VCA operates extremely quickly (since they're designed for modulating HF signals). It may be a bit overkill for your application, but it's not too hard to implement, and only increases development time marginally.

I've done this entirely in analog-land using an op-amp to control the VCA gain based on the output signal's value, and then simply ran the output of the VCA and the gain control into my DSP.

Similar to what Jay Carlson suggested, rather than look for an extreme ADC, you could consider using an instrumentation amp as a pre-amp, with gain switching. Here's an example from an Analog Devices "Circuit Note":

The AD620 or the updated alternate AD8221 maintain bandwidth above 15 kHz for gains from 1 to 100x (The published circuit includes gains ranging from 1 to 1600).

And the ADG1611 switch they've recommended has a maximum switching time on the order of 300 ns. So if you're 4x over-sampling (60 kSa/s sample rate), the switching time is only 0.018 of the sample period. The AD620 and AD8221 don't have specs for settling time after a gain switch, but the response chart included in the Analog note implies that that with well-timed control signals there should be minimal if any switching artifacts in your sampled outputs.

• This is exactly what we did with a PSoC a while ago. The chip contains a programmable gain amplifier. We reduced the gain as the signal increased, and increased the gain as the signal decreased. This gave us a few more bits of resolution for small signals, which is just what we needed. Commented Oct 9, 2012 at 11:24