Oznog
Active Member
My project to read amp-hrs out of my camper van's deep cycle battery is finally working!
I used a 200amp, 50mV shunt from a surplus catalog with a LTC2411 24-bit ADC with Vdd connected to the +12V and gnd floated to 5V below +12V, gotten from a -5V reg. So the ADC can be hooked straight onto the shunt. There's a dual optoisolator carrying signals to the PIC which is of course ground referenced to the frame. The LTC is put into internal clk, continuous conversion mode so SCK and SDO are just fed unidirectionally though the opto.
The design was taken from Linear's Design Note 341. I think they made a mistake in their design by only using a single channel opto. The duals are available in the same damn 8 pin dip for another dime, but they pulled a wacky scheme with an external XOR gate and some roundabout code to work it. I don't see where that was called for at all.
After some rework, I realized since the SCK and SDO are high 90% of the time, it would be wise to have them pulldown the opto's LED, thus drawing power only when low. Fortunately, the opto doesn't internally make the two LEDs common anode or cathode. The opto's output is open collector too, so it now remains high and drawing no power most of the time as well.
With 8x oversamling averages, looks like I've got about +/- 15mA of noise in the reading. The offset error over a large number of samples looks like maybe 2mA! I was skeptical that I'd be able to make a circuit accurate to 500nV, but there it is!
That is so awesome- overkill, but awesome. Here's to overkill!
I've got the output going to a T6963C 240x64 graphics LCD, I already had C libraries I made for drawing graphics, graphical fonts, & doing text, so it's fairly easy to make graphs and such showing usage history, estimated time remaining, etc. It's all gravy.
I used a 200amp, 50mV shunt from a surplus catalog with a LTC2411 24-bit ADC with Vdd connected to the +12V and gnd floated to 5V below +12V, gotten from a -5V reg. So the ADC can be hooked straight onto the shunt. There's a dual optoisolator carrying signals to the PIC which is of course ground referenced to the frame. The LTC is put into internal clk, continuous conversion mode so SCK and SDO are just fed unidirectionally though the opto.
The design was taken from Linear's Design Note 341. I think they made a mistake in their design by only using a single channel opto. The duals are available in the same damn 8 pin dip for another dime, but they pulled a wacky scheme with an external XOR gate and some roundabout code to work it. I don't see where that was called for at all.
After some rework, I realized since the SCK and SDO are high 90% of the time, it would be wise to have them pulldown the opto's LED, thus drawing power only when low. Fortunately, the opto doesn't internally make the two LEDs common anode or cathode. The opto's output is open collector too, so it now remains high and drawing no power most of the time as well.
With 8x oversamling averages, looks like I've got about +/- 15mA of noise in the reading. The offset error over a large number of samples looks like maybe 2mA! I was skeptical that I'd be able to make a circuit accurate to 500nV, but there it is!
That is so awesome- overkill, but awesome. Here's to overkill!
I've got the output going to a T6963C 240x64 graphics LCD, I already had C libraries I made for drawing graphics, graphical fonts, & doing text, so it's fairly easy to make graphs and such showing usage history, estimated time remaining, etc. It's all gravy.
