DIY 30 Amp shunt/amplifier for battery charger led display

Hi Everyone,

I am building a 10 LED display to replace the current meter of a 24v battery charger. The plastic case of the meter became deformed due to high contact resistance of the push-on terminals.
I have designed and built a display using TTL because LM3914 is not available and I want to make use of my TTL chips before they die of old age.
Briefly, IC1A clocks the 74160 counter and IC1B compares the staircase waveform produced by the R/2R d to a converter with the input signal. When the ramp exceeds the input signal, the transistor is turned off so the remaining leds do not light. High brightness leds are used because the duty cycle of each led is only 10% . The 7442 bcd to decimal decoder grounds the lower end of each led one at a time depending on the address from the 74160.

The circuit has the advantage that a multi channel display can be made by adding extra comparators, leds and transistors using the same staircase waveform.

The charger is rated at 20 Amps (30Amp max) and has a voltage control circuit consisting of a unijunction transistor based PWM circuit driving a thyristor. Line frequency is 50Hz.

Suitable shunt resistors are not available so I intend to use a suitable length of copper wire.

My questions are :

1. How would I tailor the response of the current shunt amplifier to match the response of the original moving iron meter.

2. Would temperature compensation be necessary due to the non zero resistive temperature coefficient of copper.

Thanks in advance,
Timescope.

Hi,

If you want to make a current shunt it doesnt matter what any other meter had read in the past, you'll be reading Amperes.

Copper isnt such a great metal for a shunt, brass is a little easier to use because it has about four times the resistance. You can get brass strips at the hobby shop or hardware store. You can calculate the resistance of a brass strip by assuming a resistance 4 times that of copper.

The temperature will affect the shunt, but how much it bothers your reading depends on how much accuracy you really need, and you probably dont need super perfect accuracy. If you want more accuracy, you'd have to mount a sensor next to the shunt in thermal contact with the shunt metal but not electrically in contact. This way you can measure the resistance of the sensor and determine how to adjust the reading.

But another issue that is quite important is the input offset of the op amp being used. You'll want to use an op amp made for very low input offset and low drift like a chopper stabilized op amp. That will give you the best results for measuring a DC current using a shunt. You can adjust the gain to match the shunt resistance.

A brass strip 0.0167 inches thick and 0.25 inches wide and about 1.5 inches long (a flat strip) has a voltage drop of 1mv per amp (0.001 ohms resistance) at 20 degrees C. The temperature coefficient of some brass is one half of that of copper, but unfortunately for some other brass it can be as high as twice that of copper. But using a sensor for feedback it doesnt matter as much.

A piece of AWG #12 copper wire about 7.6 inches long has resistance of close to 0.001 Ohms, and as the temperature rises by 10 degrees C the resistance increases by about 4 percent, so it goes from about 0.00100 ohms to about 0.00104 ohms. That would produce an error of 4 percent if calibrated at 20 degrees C room temperature. If that's too much then a sensor would have to be used to estimate the shunt temperature. A sensor might be able to be made from a much thinner piece of copper wire wound around the shunt wire to detect the temperature of the main wire. A short calibration procedure would make it all more accurate still.

A four wire connection is normally used here. That's where the (say) 7.6 inch piece of copper wire is extended to maybe 12 inches, and the current is run through that, but the connections for the sense wires are made right on the wire itself somewhere in from the ends such that the sense wire connections are 7.6 inches apart even though the wire is much longer. The current is fed into the ends, but the sense voltage is taken from anywhere between those ends but not directly on either end. The current feed wires have to be heavier gauge (like 12 gauge or more) but the sense wires can be #22 for example.

Interesting and well designed circuit.

I've seen this done with just an inverter, but it wouldnt be as accurate.

MrAl, thanks for the very detailed explanation and the brilliant idea of using the tempco of a thinner wire. I will look into the drift/ input offset issues.

The charger transformer output is rectified but not smoothed so the output current is not pure dc but current pulses the width of which depends on the conduction angle of the thyristor which in turn depends on the average battery voltage. The moving iron meter smooths these pulses out ( except when the battery is fully charged, the pointer vibrates visibly due to the very low duty cycle). My question rephrased is : what would be a suitable frequency response or integrator time constant for the amplifier.

dr pepper, thanks for your comments.

Thanks,
Timescope.

You could if you want to save time build the circuit on breadboard make an educated guess at the size of integrator cap, then look at the o/p on a 'scope with the worst ripple i/p you'd expect then adjust the size of the cap till the ripple just goes.
Highly unorthodox but it works.

Edit; just thinking about it, why not bung full wave rectifried ac straight to the i/p of your bargraph, if you halve the led limit resistors that will compensate for brightness.
This should also increase the resolution of the display, with dc on the i/p an led is either on or off giving you as many steps to the readout as there are leds, with ac the leds will go up and down each half cycle, as the display graph gets higher the top led will stay on longer and longer giving a fade out/in effect, multiplying the resolution of the display as the brightness will vary of the highest led.
At 100hz there wont be any visible flicker at least not in main vision (possibly a little in peripheral wilth ultrabright leds).

Hi,

There's nothing too critical about the integration time constant (size of cap) as long as it is much longer than one cycle, which in this case is 120Hz. It's only critical when we want to build a very fast responding circuit that also has to smooth out the ripple. For this we can get away with slow response and thus good ripple smoothing.

I dont think we have to mimic the behavior of the original analog meter either, as long as we see good averaging. That should tell us enough about what is happening in order to make decent judgments about whether or not the circuit is charging correctly.

Thanks for guiding me to a starting point. I will calibrate the shunt using a number of 21 watt auto lamps (2 Amps each) connected to a 12v battery (my bench supply only goes to 3 Amps). I have a UT70B meter that has a temperature probe and a 30 Amp panel meter.

We do not have a proper hardware store here to purchase the brass strips but I will see what I can get at "Big Market", a place that has all sorts of odds and ends.

I'll let you know how it goes,

Thanks,
Timescope

Hi again,


Oh ok great, it's always nice to hear about how it worked out, or didnt and then do some troubleshooting.