Discussion in 'Automotive Electronics' started by Litch, Jun 10, 2015.

1. ### Litch Thread Starter Member

Jan 25, 2013
86
7
Building a power monitor for my car.

The following is a "probe" for the power monitor (which will talk to this probe via I2C). I'll build a few of these, one to monitor the feed into the audio amps (0 - 600A) another to measure the batteries (-50 to 600A) and one more for the alternator feed (0A - 400A). From this I'll be able to graph and analyse the power-performance of my (significant) car audio install, the charge rate of the batteries and other interesting stuff like RPM to Alternator output performance. The below (as a start/template for the rest) will be for the amp power feed (0 - 600A)

Theory:
• Tolerance: Assume all resistors are 1%, all caps are 10% and everything else is of standard/popular values.
• [POWERIN]/[POWEROUT] is the zero-gauge wire to the audio amps running through a current transducer (LEM HTFS series).
• [OUT] is 50% of Vcc at 0A, 100% Vcc at +600A and 0% Vcc at -600A
• [REF] is 50% of Vcc (reference for 0A)
• [OUT] is fed into ACD's CH2, then grounded via a 5K resistor. (V=IR, I=V/R, I=5/5K, I=1mA drain, P=VI, P=5mW drain)
• [+12v IN] is a direct feed from the car's +12v (which will be more like 12.6V when engine off, and 14.8v when engine is running)
• This feeds a low noise DC-DC switching reg (that, in turn, powers the ADC and the current transducer via [+5V1A])
• Power draw is likely < 50mA for the circuit, it's overkill but meh - room for expansion.
• The [+12v IN] also goes to ACD CH1 via a 33% voltage divider (33% of 15V = 5V = ADC's max input voltage) and smoothed a little more via a 100nF cap.
• The ADC is capable of 200kS/s
• Ignore the +5v feed into the I2C connector - superfluous, won't be used.
Questions:
• With 12bit accuracy - do I need better smoothing of the analogue inputs?
• Is that 100nF on CH1 sufficient or should I be looking at an RLF filter?
• But at what point of smoothing do I lose the degree of accuracy I'm looking for calculating power draw of the amps - it would be nice to sample all this at 10kS/s and graph it out.
I appreciate the input as I don't have much room to play with all this (or breadboard the circuit with such small SSOP packages) so I usually just order a few pcb prototypes which isn't cheap (isn't that expensive either) but the point is, I'd like to get it peer reviewed before I go and build it.

-KB

2. ### MrAl Distinguished Member

Jun 17, 2014
2,554
515
Hello there,

I can offer a few comments here...

First, the degree of smoothing you need or should use depends on a couple of things like how fast you need to sample (input bandwidth limit) and whether or not you intend to oversample.

If you intend to oversample then you need to allow some noise into the system through the input channel. This will help get the extra bits. You dont want to filter too much if at all, or else you wont be able to get any extra bits because oversampling depends highly on the presence of some random noise on the input. Without any noise in the input you wont be able to get the results you want with oversampling.

If you dont intend to oversample then if you also have to sample fast then you need to consider the settling time to say 99 percent of actual input. The filtering can not be too severe or else the response will take too long to settle in order to get an accurate reading. This can be calculated from knowing the maximum input change rate.

If you dont intend to oversample then if you also dont have to sample fast (limited bandwidth which is typical of many DC systems) then you can use more filtering to average out the input. This will free up clock cycles for other things because you wont have to digitally filter as much if at all.

In most DC measuring systems it is better to have a smooth signal to work with, unless you intend to oversample and then all of the smoothing is done digitally with an averaging like algorithm.

In some very noisy systems you may have to add passive filtering whether you want to or not. Some tests will be in order once the system is up and running.

Last edited: Jun 17, 2015
3. ### nsaspook AAC Fanatic!

Aug 27, 2009
3,011
2,357
MrAl is right. For this application you need to have a filter of some sort so it's usually best to design the ADC stage software to oversample/digital filter at a rate at least 8X the expected ripple current frequency from the alternator to integrate the readings to a stable value so you can actually use most of the 12bit result.

Last edited: Jun 17, 2015
4. ### cmartinez AAC Fanatic!

Jan 17, 2007
3,698
2,779
I think it would be much simpler to use ampsense sensors coupled with an AD536AJQ chip to get linear output. And then to conect that to an ADC. That arrangement would give you a true RMS reading.

Apr 2, 2009
8,789
771

6. ### nsaspook AAC Fanatic!

Aug 27, 2009
3,011
2,357
I don't know, with a single supply-rail the AD536 input needs to be AC coupled.

7. ### cmartinez AAC Fanatic!

Jan 17, 2007
3,698
2,779
Check Page 12, figure 17 of the the datasheet. It shows how to connect the AD536A using a single supply. From +5V up to +36V.
The AMP25 sensor can operate with a single +5V to +10V supply too.

8. ### nsaspook AAC Fanatic!

Aug 27, 2009
3,011
2,357
Sure, they will both run single supply (I've used the AMP sensors many times with offset calibrations for the 1/2 Vcc zero current point) but my reading of the AD536A single supply circuit with a input capacitor requirement on pin 1 that the frequency response to 10hz not DC.

9. ### cmartinez AAC Fanatic!

Jan 17, 2007
3,698
2,779
Ahhh... I think I see what you mean now... the input capacitor makes it impossible to read DC... but the AMP25 sensor delivers only positive voltage, even though it vareies between 0 and 5V, with a 2.5V offset. That would mean that the input capacitor wouldn't be needed anymore, would it?

10. ### nsaspook AAC Fanatic!

Aug 27, 2009
3,011
2,357
I don't know but usually when input DC isolation is a requirement it's not simply that it has to be at a certain level, there are feedbacks.