Interfacing with accelerometer

sorenriis · Sep 4, 2008

interfacing accelerometer with mux

I would like to create an interface between my accelerometer and an embedded computer.

The accelerometer is of this type:
- Measurement Specialities ACH-01: http://www.meas-spec.com/vibration/documents/datasheets/OEM_ach01.pdf

The embedded computer is this one:
- Atmel STK1000 (http://www.atmel.com/dyn/products/tools_card.asp?tool_id=3918)

Anyone who knows something about this topic?
Also i'm very interested in books, articles etc. about it.

Thanks in advance.

Added after 5 hours 17 minutes:

Another question is:
What is the easiest to interface to?

Are we talking the serial-port, LAN, USB or connecting it directly to the A/D-converter of the STK1000?

DrBob13 · Sep 4, 2008

usart ads8320

Hi sorenriis,
You need to incorporate some sort of amplifier, usually a charge amplifier is used and then digitize the AC signal that is generated for reading by your microcontroller unless the microcontroller has an integrated ADC, this link has some useful application data on your accelerometer together with a suggested interface amp.

http://www.msisensors.com.cn/manage/sensortypes/ACH_01.pdf

I am not familiar with that ATMEL microprocessor board but a quick look at its schematics suggests that it will need either a little slave microcontroller such as the Analog Devices ADuC848 (which has a built in ADC) or a bit of FPGA state machine logic to control a separate ADC and output the data as a digital stream to the USART or to the USB with a suitable interface.
Hope this helps.
Bob.

sorenriis · Sep 5, 2008

low pass filter iphone accelerometer

Thanks alot Bob.

You are right that the STK1000 does not have a integrated ADC - I was wrong on that point.

My thoughts about it right now is to interface to the SPI- or the I2C-bus. A guy told me that there aren't many decent RS232 ADC's, so I've skipped the serial-port option. But maybe thats not true?

The requirements for sampling from the accelerometere are (at least):
- samplingfrequency: 5 kHz
- resolution: 16 bit

Added after 13 minutes:

Also - actually it's not just one accelerometer I would like to interface with.
Actually I need 3 (x-,y-,z-axis), but how is that done?

Again i'm very interested in references to book, articles etc.

raptor1981 · Sep 5, 2008

Sensors (MEMS) can not be compared with other standard blocks because sensor can not give output directly as given by analog or digital circuit. Consider the caes of of accelerometer, It can be ,odelled as a variable capacitor. Its capacitance is a function of the external force applied(assuming force feedback).Hence a analog circuit has to be used where the variable capacitor is used in negative feedback. The output of complete analog block is related with the responce of sensor.
For further details refer to uc berkley website

DrBob13 · Sep 5, 2008

Hi again,
Yes I agree, either the I2C or SPI port is the best way to go to talk to your ADC device(s), I usually use SPI as it can be faster, it does however require external enable signals for each SPI slave but that is easy with a microcontroller like yours. I have not seen one with a RS232 serial interface for a while. The ADuC848 device that I mentioned earlier would not be good enough for your application as its onboard ADC it is not fast enough. You need a four channel ADC with a sampling frequency of better than 10kHz per channel. The Linear Technology LTC2487 looks like a possible candidate (with I2C interface) for your app.

As described in the previous post by raptor1981 you will need to incorporate an analog interface similar to the one in my previous post FOR EACH accelerometer. Accelerometers with the signal conditioning amps built in and even some with digital interfaces are available but these are a lot more expensive than the ones that you have identified.

Cheers.
Bob.

sorenriis · Sep 7, 2008

Thanks both of you - helpful answers.

I've been trying to simulate one accelerometer and the interface you mentioned, but somehow it's not working. For some reason the output-signal is just constant.
Could you tell me what i'm doing wrong here?

I've simulated the accelerometer as a simple AC-source (sinus formed), but maybe thats not correct?

sorenriis · Sep 7, 2008

Problem seems to be solved. The reason was that somehow OrCAD didn't understand "2M", but was ok with "2000k".
Screenshot is updated to match the latest changes.

So now the "only" thing I have to do is to make 3 identical acc-gains and then connect them to the A/D converter?
Also - is the I2C-bus fast enough for managing 3 accelerometers (5kHz/16bit)?

halls · Sep 7, 2008

sorenriis said:
Also - actually it's not just one accelerometer I would like to interface with.
Actually I need 3 (x-,y-,z-axis), but how is that done?

Do you have to use that exact model of accelerometer?

The thing is that you could use a 3 axis accelerometer, and so, you simplify much more your circuit.

sorenriis said:
The reason was that somehow OrCAD didn't understand "2M", but was ok with "2000k".

Just a little point: "2M" means "two millis" for orcad, while what you would have to have written, was "2Meg"

good luck!

sorenriis · Sep 7, 2008

halls said:
Just a little point: "2M" means "two millis" for orcad, while what you would have to have written, was "2Meg"

Hehe... Thanks for letting me know

I'm aware that I could use 3-axis accelerometers, but because the ACH-01 is cheap and suits our needs almost perfectly we have decided to use that one.
Also in the future we might want to add a 4th accelerometer to be used as reference.

Based on the shown diagram - is it still the 'Linear Technology LTC2487' and the I2C-bus that I should go for?

halls · Sep 7, 2008

From what i've read from the datasheet, i guess that sigma-delta ADC would be a good option. Actually it seems to fit very well in your application.

sorenriis · Sep 7, 2008

halls said:
...i guess that sigma-delta ADC would be a good option...

Thanks. I will stick with the LTC2487 then - unless someone else comes up with something better

Added after 24 minutes:

Three more questions:
1) Does someone know if it's possible to get the LTC2487 into OrCAD for simulation-purposes?
2) In the datasheet it states that the LTC2487 has a programmable gain. Isn't it better to use that (much more compact circuit) instead of making the gain myself?
3) As far as I can see I need to add some offset to the signal before the ADC (no negative values). How is that done in the best way?

sorenriis · Sep 8, 2008

One thing about the LTC2487.
As far as I can read from the datasheet, it has a maximum of 15 sps as output.

If I need to sample at least 5kHz for each accelerometer, then that seems too slow?

halls · Sep 8, 2008

15 Hz would actually be very low for sampling a 5 kHz signal of course, but as I understood from the datasheet, it is just a low power mode.

On normal performance i think you could reach even 30kHz sampling frequency. Just take a look at the graphs shown on the datasheet

sorenriis · Sep 8, 2008

halls said:
On normal performance i think you could reach even 30kHz sampling frequency. Just take a look at the graphs shown on the datasheet

Could you tell me what graph you are looking at?

I only found this in the datasheet:

Code:

Output Data Rate
When using its internal oscillator, the LTC2487 produces
up to 15 samples per second (sps) with a notch frequency of
60Hz. The actual output data rate depends upon the length
of the sleep and data output cycles which are controlled
by the user and can be made insignificantly short. When
operating with an external conversion clock (FO connected
to an external oscillator), the LTC2487 output data rate
can be increased. The duration of the conversion cycle is
41036/fEOSC. If fEOSC = 307.2kHz, the converter behaves
as if the internal oscillator is used.

An increase in fEOSC over the nominal 307.2kHz will translate
into a proportional increase in the maximum output
data rate (up to a maximum of 100sps).....

What does this in practice?

DrBob13 · Sep 8, 2008

Hi Sorenriis
You are correct,
The device is definitely too slow, sorry about that, when I read the datasheet again it is much too slow to capture data even in the low Audio region, this is only suitable for slow moving sensor data.

The analog devices AD7655 looks like a better choice, this has a SPI interface and is available from stock from RS components in the UK. Analog devices are usually good performing chips for af analysis.

I hope this helps.
Bob.

halls · Sep 8, 2008

Hi Soenriis,

i'm sorry for my mistake, but I'm afraid DrBob13 is right. I didn't see the "up to a maximum of 100sps" thing, and got messed with the graphs...

Analog devices is a good choice, and for my experience, I would take a look at Texas Instruments too.

Sorry again.

sorenriis · Sep 8, 2008

Thanks for the reply to both of you.

The AD7655 that I can find only has two channels?
Also it's a 48-pin - remember I might have to solder myself

I've looked around myself abit and found this:
Burr-Brown ADS8341EB
Datasheet: https://focus.ti.com/lit/ds/symlink/ads8341.pdf

Could that be working?
In Denmark I can get 2 of those for the price of one AD7655.

halls · Sep 8, 2008

Except for a little inconvenience, I think this is a good ADC too. The inconvenience I see is the serial interface. It is not a standard serial interface, like I2C or SPI, so you will have to implement your own code for handling the connection between the ADC and the micro.

I encourage you to take a look at this one: https://focus.ti.com/lit/ds/symlink/ads8320.pdf

This last one has the disadvantage of not having 4 channels, like the one you proposed, but in the other hand is a bit cheaper and consumes much less power (I don't know if power consumption is an issue for your application).

Hope it helps

sorenriis · Sep 8, 2008

Thanks. It also says that the ADS8320 is made for vibration analysis, which is exactly what i'm going to use it for.

The fact that it's only a 2-channel. How do I solve that issue?
Does this mean that I will have to make 2 seperate circuits?

halls · Sep 8, 2008

It comes to my mind a solution where you could use an analog multiplexer to multiplex 4 accelerometer outputs into one single channel to the ADC's input. For this you would use the micro to select the sampling channel.

So, your ADC would be sampling at 20 khz (for example) which would be a 50 us period. You would switch channels every period, and when finished 4 periods (4 channels) you would start again with channel four. That would be a 200 us period for every channel, this is 5kHz.

I really haven't though much about it, so I might be missing something. I'll check it up later, now I gotta go, sorry.

Welcome to EDAboard.com

Interfacing with accelerometer

Newbie level 6

Full Member level 2

Newbie level 6

Junior Member level 1

Full Member level 2

Newbie level 6

Newbie level 6

Advanced Member level 4

Newbie level 6

Advanced Member level 4

Newbie level 6

Newbie level 6

Advanced Member level 4

Newbie level 6

Full Member level 2

Advanced Member level 4

Newbie level 6

Advanced Member level 4

Newbie level 6

Advanced Member level 4

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor