Help getting started on programmable microcontrollers?

[[email protected]]

Hello folks,

I'm working on a project where I need a circuit that reads in a
voltage, and then outputs a voltage based both on the signal, and also
the past history of the signal. Sort of like PID, except that it is
open loop, and the Vin to Vout function can be arbitrary and
programmable. Ideally, I would like the system to be self contained,
needing a computer or laptop to reprogram the algorithms if
necessary. It only needs to run up to about 1 kHz.

My knowledge is as follows:
I know analog electronics OK, and can build op amp circuits and a few
transistor circuits.
I also know Labview real time, which can pretty much do exactly what I
want, but is overkill in money and also requires a computer to be
there all of the time.

I have a big gap in this middle area involving digital stuff, so I
would appreciate suggestions on where to start. Ideally, I would like
the cost to be low and the learning time to be not too long. I know
it needs some kind of ADC and DAC.

For example, here are two scenarios that I am capable of, both of
which are suboptimal:
1) I can build an entirely analog circuit, with a bunch of knobs to
adjust various parameters, but that's too much of a PITA, and isn't
very flexible.
2) On the other end, I can throw together a cheapo DAQ card and the
smallest computer possible, and install labview real time target on
it, as shown here:

http://zone.ni.com/devzone/cda/tut/p/id/2733

This is close to what I want, but was wondering if there was anything
cheaper or simpler.

Thanks

 

[Rich Webb]

Hello folks,

I'm working on a project where I need a circuit that reads in a
voltage, and then outputs a voltage based both on the signal, and also
the past history of the signal. Sort of like PID, except that it is
open loop, and the Vin to Vout function can be arbitrary and
programmable. Ideally, I would like the system to be self contained,
needing a computer or laptop to reprogram the algorithms if
necessary. It only needs to run up to about 1 kHz.
[snip...snip...]

This is close to what I want, but was wondering if there was anything
cheaper or simpler.

Consider the one of the Arduino boards. It's an open source hardware and
software platform so the cost of entry is relatively low. On the other
hand, the boards don't have an onboard DAC, just PWM. On the other,
other hand, an onboard 3.3 V DAC may not do you much good, anyway. The
boards have UARTs, SPI, and I2C-compatible comms, so a 1 KHz update rate
to an external DAC and analog board would be quite possible.

http://www.arduino.cc/

Otherwise, jump over to http://www.sparkfun.com/commerce/categories.php
or http://microcontrollershop.com/ and look at what dev boards they have
available that strike your fancy. Several do have onboard DACs that
could be tied into analog gain stages. Most will require separate
compilers, some have bootloaders, others require dedicated programmers.

 

[Spehro Pefhany]

On Sat, 28 Mar 2009 01:59:10 -0700 (PDT), the renowned

Hello folks,

I'm working on a project where I need a circuit that reads in a
voltage, and then outputs a voltage based both on the signal, and also
the past history of the signal. Sort of like PID, except that it is
open loop, and the Vin to Vout function can be arbitrary and
programmable. Ideally, I would like the system to be self contained,
needing a computer or laptop to reprogram the algorithms if
necessary. It only needs to run up to about 1 kHz.

My knowledge is as follows:
I know analog electronics OK, and can build op amp circuits and a few
transistor circuits.
I also know Labview real time, which can pretty much do exactly what I
want, but is overkill in money and also requires a computer to be
there all of the time.

I have a big gap in this middle area involving digital stuff, so I
would appreciate suggestions on where to start. Ideally, I would like
the cost to be low and the learning time to be not too long. I know
it needs some kind of ADC and DAC.

For example, here are two scenarios that I am capable of, both of
which are suboptimal:
1) I can build an entirely analog circuit, with a bunch of knobs to
adjust various parameters, but that's too much of a PITA, and isn't
very flexible.
2) On the other end, I can throw together a cheapo DAQ card and the
smallest computer possible, and install labview real time target on
it, as shown here:

http://zone.ni.com/devzone/cda/tut/p/id/2733

This is close to what I want, but was wondering if there was anything
cheaper or simpler.

Thanks

1kHz is actually pretty fast for a small microcontroller to do much in
the way of general-purpose calculation. PID ends up being just a few
simple (addition and multiplication ONLY) operations per sample the
way it is usually implemented, even when you hang algorithmic
barnacles off of it to make it behave reasonably at start-up etc.

For example, if your arbitrary function could include transcendental
functions, you would not be able to use MOST of the microcontrollers
out there at 1kHz ( most would be able to do the calculations-- though
perhaps only at 10Hz-100Hz).

You might want to look into something like a DSP (but at board level
or higher, given your background) if you need to have the ability to
implement sophisticated algorithms at high speed. You should also
consider the ADC and DAC requirements. 10 or 12 bits is a LOT easier
than trying to use 24-bit ADCs/DACs and getting anything like that
number of effective bits.

It might not be a clear cut choice between your default digital
scenario and using a DSP.

OTOH, if you just need a 10-12 bit resolution, less accuracy, and can
live with slower and/or simpler calculations, there are one and two
chip solutions possible.




Best regards,
Spehro Pefhany

 

[linnix]

Hello folks,

I'm working on a project where I need a circuit that reads in a
voltage, and then outputs a voltage based both on the signal, and also
the past history of the signal.  Sort of like PID, except that it is
open loop, and the Vin to Vout function can be arbitrary and
programmable.

I am doing something similar to this. I need an output proportional
to the integral of the input, using a pre-generated jump table. For 8
bit accuracy, the jump table takes up only 2K of a 16K ROM. I can
still generated 6 other arbitary functions on the chip.

 Ideally, I would like the system to be self contained,
needing a computer or laptop to reprogram the algorithms if
necessary.  

I am thinking about storing the lookup table in serial EEPROM, to be
loaded at boot.

...
This is close to what I want, but was wondering if there was anything
cheaper or simpler.

I am doing it for under $2.

 

[Nobody]

This can be done with a micro controller, for example a PIC with analog
inputs. The analog output can be made by using PWM and a low pass RC
filter.

Some of the dsPIC33 chips have built-in sigma-delta DACs.

For the 8-bit chips, it's either PWM or an external R-2R ladder.

 

[MooseFET]

Hello folks,

I'm working on a project where I need a circuit that reads in a
voltage, and then outputs a voltage based both on the signal, and also
the past history of the signal.  Sort of like PID, except that it is
open loop, and the Vin to Vout function can be arbitrary and
programmable.  Ideally, I would like the system to be self contained,
needing a computer or laptop to reprogram the algorithms if
necessary.  It only needs to run up to about 1 kHz.

My knowledge is as follows:
I know analog electronics OK, and can build op amp circuits and a few
transistor circuits.
I also know Labview real time, which can pretty much do exactly what I
want, but is overkill in money and also requires a computer to be
there all of the time.

I have a big gap in this middle area involving digital stuff, so I
would appreciate suggestions on where to start.  Ideally, I would like
the cost to be low and the learning time to be not too long.  I know
it needs some kind of ADC and DAC.

For example, here are two scenarios that I am capable of, both of
which are suboptimal:
1) I can build an entirely analog circuit, with a bunch of knobs to
adjust various parameters, but that's too much of a PITA, and isn't
very flexible.
2) On the other end, I can throw together a cheapo DAQ card and the
smallest computer possible, and install labview real time target on
it, as shown here:

http://zone.ni.com/devzone/cda/tut/p/id/2733

This is close to what I want, but was wondering if there was anything
cheaper or simpler.

If you use the development board for the SiLabs F120 or the like, you
will have a multiple input ADC a couple of DACs and a fair amount of
CPU power to play with.

The ADC and DAC are just connected as special function registers so
programming it is fairly simple. The 8051 core is easy to write for.

 

[[email protected]]

Thanks for the replies, everybody.

So far, it looks like my first choice would be an Arduino board using
the ATmega328. The main reason being that a "kit" with USB input is
pretty cheap at $30. Also, the software is free and open source, so
there seem to be a lot of people using it.
However, I need to be sure that it can do what I want. Here's what I
am imagining in more detail:

Input: samples 2 or 4 channels at 500Hz or 1kHz, 10 bit resolution is
fine, but slightly higher is preferred.
Processing: Instead of it computing complicated functions, I'd upload
some kind of lookup table for Vin and Vout. Maybe use interpolation?
I would also add some other contributions to the result, like some
value based on its past history, and maybe based on the other
channels.
Take some kind of derivative if possible.
Output: to 2 or 4 channels at 500Hz, 8 bits is fine, but higher is
preferred. I saw that you can use the I2C protocol and hang DAC's off
of the board.

Anyway, I just need to know if this microprocessor can carry out these
computations at 500 Hz, and if not, I'd have to move to a more
powerful chip, like the 0851F120.

Thanks again

 

[Jasen Betts]

Thanks for the replies, everybody.

So far, it looks like my first choice would be an Arduino board using
the ATmega328. The main reason being that a "kit" with USB input is
pretty cheap at $30. Also, the software is free and open source, so
there seem to be a lot of people using it.
However, I need to be sure that it can do what I want. Here's what I
am imagining in more detail:

Input: samples 2 or 4 channels at 500Hz or 1kHz, 10 bit resolution is
fine, but slightly higher is preferred.
Processing: Instead of it computing complicated functions, I'd upload
some kind of lookup table for Vin and Vout. Maybe use interpolation?
I would also add some other contributions to the result, like some
value based on its past history, and maybe based on the other
channels.
Take some kind of derivative if possible.
Output: to 2 or 4 channels at 500Hz, 8 bits is fine, but higher is
preferred. I saw that you can use the I2C protocol and hang DAC's off
of the board.

Anyway, I just need to know if this microprocessor can carry out these
computations at 500 Hz, and if not, I'd have to move to a more
powerful chip, like the 0851F120.

Thanks again

the avrs get close to one instruction ber clock cycle and from your
description of the task that sounds like plenty.

I did 4 channel DDS at ~10khz sample rate on a smaller AVR
running at 10Mhz.

 

[Dennis]

For a first level approximation:

The Arduinos all seem to run at 16 MHz on a 5V board. The ATMega328's
A/D peripheral needs a clock that's between 50 and 200 KHz. With the
available clock prescalers, that results in an AD clock of 125 KHz.

The AD cycle time, exclusive of interrupt handling overhead, is 13 AD
clock cycles, or 104 usecs, or about 9.6 KHz. There are 1664 main clock
instruction cycles in this time that are available for computation,
general I/O, interrupt handling, and for reading, storing, and then
restarting the AD conversion process.

Reading four AD channels gives about a 2.4 KHz sample rate per channel,
again neglecting other overhead.

You'll have to run the equivalent numbers for the I2C DAC you select. I
suspect that an SPI DAC will permit achieving a higher output rate.

I did a test on the Arduino ADC where I read 100 samples into an array
and did a timestamp before and after. I got a sample rate of just over
8KHz for the set. That will give you a feel for the overhead in addition
to the the 9.6KHz conversion rate above.

 

[[email protected]]

Thanks everybody. I ended up ordering an Arduino