Looking for 16-bit A/D w/Serial Output

[ElectroLawndart]

Hi All,

I realized that somebody on RCGroups might have some experience with this.

I'm looking for an A/D with serial output for my altimiter/airspeed indicator project. I would like it to have 16-bits but I think I could get away with using 14-bits.

Any help would be greatly appreciated.

Thanks in advance,

Dart

[icedog]

16 bit A/D is expensive, you can use integraph,comparator and 16bit counter to instead.

[Gary Warner]

What's the odds. I just bought one an hour ago. Analog Device's AD7705BN (14 pin DIP package) 16 bit A/D w/4 wire serial interface. It was $3.95 at Tanner Electronics (http://www.tannerelectronics.com/). I don't know if Jim ships out phone orders. Call him as see.

Gary
--

[ElectroLawndart]

I'm on it!

Thanks Gary

Dart

[Gary Warner]

I was down at Tanners again today. I was told they will ship out-of-state with a $15 min. order. I was told you should ask for Jake McElyea about this if you are interested. 972-242-8702

Gary
--

[deh6]

Gary,

Though you are well down-the-road on the project Intersema has a part with built-in 15 bit ADC which looks interesting.

http://www.intersema.ch/site/technical/ms5534.php

A few more detals in this post:

http://www.rcgroups.com/forums/showthread.php?p=4722526&highlight=MS5534#post4722526

[deh6]

Gary,

An additional thought--Everything I've seen indicates that resolution beyond about 12 (and even 10!) bits doesn't add anything since the noise level is about 1 bit with a 10 bit ADC.

Taking samples at a high rate and running them through a digital filter will get the improvement in resolution. I think the limit of improvement has to do with flicker noise which has a 1/f distribution, meaning it increases as the frequency goes lower, which in turn, means as the filter cutoff gets lower a point of diminishing returns is reached.

And while I'm in a pessimistic mood (I'm in the middle of nasty cold)--I'm also curious about vibration being picked up by the sensor if there is a motor running. The air column in the tube connected to the sensor is going to be pulsing from the vibration and that should show up in the sensor readings. However, low pass filtering should take that out as long as one is careful about aliasing.

I've still got this on my todo list, but it is not on top.

Don

[Gary Warner]

Gary,

An additional thought--Everything I've seen indicates that resolution beyond about 12 (and even 10!) bits doesn't add anything since the noise level is about 1 bit with a 10 bit ADC...

Don

I think you need to direct you comments to ElectroLawndart. I simply answered his question.

Oh,... I can't resist :D ... If you have a signal that can't be resolved +/- .5 LSB you need to look at your technical design, not the A/D converter IC. I love it when someone says 'it must be inherent noise in the A/D Ic". I always say "Yup, the manufacturer spent more than you'll make in 50 years to produce a chip that won't do what it's supposed too". :rolleyes:

Gary
--

[ElectroLawndart]

If you have a signal that can't be resolved +/- .5 LSB you need to look at your technical design, not the A/D converter IC. I love it when someone says 'it must be inherent noise in the A/D Ic". I always say "Yup, the manufacturer spent more than you'll make in 50 years to produce a chip that won't do what it's supposed too".
--

Agreed. I'm taking the result of millions of dollars of research and development and sticking on something I scribbled on the back of a cocktail napkin.

I'm expecting lots of revisions and board etching. Currently I'm poking around the 'net looking for 'low noise' tips.

Dart

[Gary Warner]

Look at Microchip's web site and follow the Web Seminars links. There is a good AV file there that gives examples on a read circuit to show that <1bit noise is possible if the right care is taken.

Gary
--

[deh6]

Gary,

My apologies. Indeed the reply should have been directed to ElectroLawndart.

If you have a signal that can't be resolved +/- .5 LSB you need to look at your technical design, not the A/D converter IC

Quite true, so increasing the ADC resolution is not the answer though the issue of increasing resolution via oversampling with low-pass filtering rests on some linearity assumptions about the ADC.

Unfortunately, there is not much that we, with limited resources, can do about the technical design for semiconductor pressure transducers. The only things I've run across are 1) pulse the transducer with a voltage that is 4x the rating (Motorola has an app note on this) to increase the signal/noise ratio, 2) low-pass filtering. 1) takes way too much electronics, leaving 2) as viable.

BTW, I got a response from Intersema's USA rep and the MS5534 is available for $22.50 (ordering samples from the Swiss mfger had a hefty S/H charge). As yet, I don't know the S/H charges for getting them through the rep in the US, nor if there is a mininum order, but with a 15 bit built-in ADC it looks like a something to try.

Don

[Gary Warner]

Gary,

My apologies. Indeed the reply should have been directed to ElectroLawndart.



Quite true, so increasing the ADC resolution is not the answer though the issue of increasing resolution via oversampling with low-pass filtering rests on some linearity assumptions about the ADC.

Unfortunately, there is not much that we, with limited resources, can do about the technical design for semiconductor pressure transducers. The only things I've run across are 1) pulse the transducer with a voltage that is 4x the rating (Motorola has an app note on this) to increase the signal/noise ratio, 2) low-pass filtering. 1) takes way too much electronics, leaving 2) as viable.

BTW, I got a response from Intersema's USA rep and the MS5534 is available for $22.50 (ordering samples from the Swiss mfger had a hefty S/H charge). As yet, I don't know the S/H charges for getting them through the rep in the US, nor if there is a mininum order, but with a 15 bit built-in ADC it looks like a something to try.

Don

That's cool Don. No problem here.

Noise IS a problem to overcome, what ever the source. I see your point.

I like that sensor. It sure sidesteps many design issues. What does it weigh?

Gary
--

[deh6]

The spec sheet doesn't show the weight, but judging from the size (about 9mm square x3mm thick) it is miniscule.

Skimming the spec sheet it looks like the 0.1 mbar resolution will give around 1 ft of altitude resolution at sea level (if I did the conversions correctly). However, a footnote says that a "steady" 0.1 mbar reading should be an average of 8 readings. The conversion time is 35 ms, so 8 readings means 0.28 sec per averaged reading. The 0.1 mbar resolution might also be satisfactory for airspeeds around 20 mph (I came up with 6 mbar as the pressure difference at sea level for 20 mph...assuming no mistakes!).

Several items did catch my eye. One was that it takes a fair amount of code to do the computations for calibration and temperature compensation (BTW one gets 0.1 deg C resolution temp readings out of it). Another is that requires a clock signal of 30 - 35 Khz, with the design center 32,768 Hz (the standard watch crystal freq). It points out that the jitter on this signal affects the accuracy of the ADC conversion. Driving it from the micro I'm using (Freescale 9S12 series) is not a problem, but another detail to dealt with. Overall it looks like an improvement in accuracy and simplicity (especially since it eliminates a hi-resolution ADC convertor) over using the old Motorola/Freescale MPX pressure sensors.