Changes between Version 4 and Version 5 of WikiStart

Timestamp:

Mar 16, 2017, 6:54:50 PM (7 years ago)

Author:

bouyer

Comment:

--

WikiStart

@@ -11 +11 @@
 - 4 GPIOs to connect with the SoC being evaluated
 
+We came up with the schematic at [[https://www-soc.lip6.fr/trac/soc-conso/export/1/hardware/pic_conso.sch.pdf]]. Hardware and software sources are available in the source section of the wiki, or via svn at https://www-soc.lip6.fr/svn/soc-conso/.
+
 A 1.350V precision diode is used as a voltage reference for the ADC. The input voltage is feed to ADC input 0 (AN0) via a voltage diviser. To cover the input range with enough accuracy 3 different values can be selected with a jumper.
 
@@ -19 +21 @@
 Each sample reported to the host is composed of 3 values: GPIO state (integer between 0 and 15), voltage (AN0 value between 0 and 4095) and current (AN1 value between 0 and 95), in the format: ''gvvvcccX'', with ''g'', ''v'' and ''c'' being hexadecimal digits and ''X'' being the letter X used as separator. With 8 bytes per sample, at 921600 bauds we could report in theory up to 11520 samples/s. To be safe we choose to run at 5000 samples/s (which requires 10000 A/D conversions/s).
 
+== calibration and performance ==
+
+For the calibration process, I used a good linear power supply (both current and voltage adjustable) and a 4000 points digital multimeter. A switching power
+supply was not good enough, ripple noise impacted values reported by our
+circuit.
+
+For voltage input, we selectable 3 divider values: 4.92, 11 and 19.33,
+which gives ranges of 6.63V, 13.5 and 26.1V respectively.
+
+For current input, we have 3 selectable shuts: 0.05, 0.1 and 0.2 ohms.
+The shunt's volatge is then amplified by 10 by the AD623 amplifier.
+This gives ranges of 2.68A, 1.34A and 0.67A respectively.
+
+For each tension or current range, we take 8 samples between 0 and a value
+close to the max of the range. For each sample we get the pic's output for
+about 5 seconds, and we note the real value using the digital multimeter.
+
+From this process, we found that that the AD623 output can't go down to 0,
+it stays positive by anout 6mV.
+This means we can't measure currents smaller than 3mA on the 0.67A scale,
+and 12mA on the 2.68A scale.
+We also find that, assuming an ideal power supply without noise or ripple,
+we have a +/-7 unit error on the ADC's output. This is less than 0.18% of the
+full range (or +/- 4.8mA on the 2.68A range input).
+
+
+