Totally Analogue or DSP mixture?

[dmccric]

Morning all,

I have been given a task to design a sensor that will amplify very low voltage signals from muscle electrode sensors.

The device is battery operated and needs to be a very small PCB footprint.

Specs:

Input
10uV to 100uV (20Hz to 450Hz)

Output
0V to 4.7V

My question is should I use 100% hardware or should I incorporate a little hardware with a dsPIC or similar?

If I use a dsPIC then I can use dsp techniques to amplify and filter signals. I can have a single device for 50Hz & 60Hz. Since we are using a microprocessor I can use an I2C bi-directional bus to transfer the results with no loss of data on the output. I can also use a coms. protocol to program the gain, etc. This method will take longer to implement the software to get it right. Some PICs have op-amps on board giving a gain of ~64 max which can help reduce pcb space as well but probably not as good as devices tailored for the situation.

However if I use all hardware I will require a bigger PCB area. The Analogue output will be susceptible to noise after all filtering, etc. And I will need 2 different designs to reject the noise for 50Hz & 60Hz. There are good instrumentation op amps to use (AD8227, etc) for signal amplification which also have good CMRR, etc.

I usually write Software and have done some complex PIC projects but I need to make a decision soon on which method I will use.

Any advice from more experienced people?

Thanks

 

[KlausST]

Hi,

I recommend a tailored analog amplifier with filter. In either case.

Then you may decide to use it as is or to process it digitally.

* If analog: then consider to transmit it differentially (signal pair) on the sensor cable. For best signal quality.

***
A bad analog signal won´t become better by processing it digitally. For sure you may filter away unwanted signal (frequencies) with digital filters...but..
Wrong anti alias filter (creating alias frequencies), wrong signal amplitude (low resolution introducing noise), or wrong sampling rate (or ..jitter) can´t be cancelled out with digital filters.

Klaus

 

[dmccric]

Adding an Anti-Aliasing filter on the output of the pre-amplifier I suppose negates having a processor doing filtering, amplifying, etc.

I never thought of a differential output though. Good idea thanks.

 

[KlausST]

Hi,

Adding an Anti-Aliasing filter on the output of the pre-amplifier I suppose negates having a processor doing filtering, amplifying, etc.

As long as you don´t use any "undersampling technique" I always recommend to use an anti aliasing filter in front of an ADC. Only WITH the filter the output is predictable.

Klaus

 

[schmitt trigger]

Another consideration you have not put on the table is cost.
The question that sales/marketing always asks is: which option provides equivalent performance level at lower cost?
Euros, Dollars, GBPs; if this is a commercial product, cost will eventually play a role.

From the technical point of view, however, I would go the all digital route with a good analog front end.

BTW, the powerline (hum) frequencies that you must filter are not 50 and 60 Hz, but rather 100 and 120 Hz.