7106 based Ammeter problem.

Thread Starter

#JohnMaya

Joined Feb 10, 2021
9
Hi all,

I have a TC7106 based 100A ammeter (with an FSD of 100mV). It all works reasonably well but shows a permanent offset of about -7A.

The actual circuit appears to be much the same as the one on the datasheet.

The circuit is powered by a 7805 regulator connected to a +12V supply. The reading is still there when Vin+ and Vin- are shorted together.

When powered by an isolated battery the problem disappears.

I understand the cause is an offset between the negative on the shunt that connects to Vin- and the negative of the supply to the 7805 regulator.
Theoretically all the negatives should be at the same potential, but the circuit is being used in a boat and quite often [in marine electricals] the different negatives can be a few mV apart and it is extremely difficult to correct this issue.

In the circuit GND, ANALOGUE COMMON, Refin- and Vin- are all connected together and in turn connected to the negative of the regulator.

The "negative" of the shunt sits about 6mv above the negative of the regulator which is introducing the offset.

The ideal solution is to have the input floating with respect to GND so that the offset doesn't interfere with the reading.

Is it just a matter of disconnecting the GND from ANA. COMMON and all the negative inputs on the 7106?
 

Thread Starter

#JohnMaya

Joined Feb 10, 2021
9
Hi there Dodgydave,

Thanks for taking the time to respond.

Pretty much - the device is the 44 pin MQFP package so the pin numbers differ.

In my case the battery is replaced by a 7805 supplying V+ and GND. V- is connected to a capacitor diode combination to provide about -3v (That's what I measured). As explained, GND is also connected to COM and Vin-.

Looking at the diagram can I just disconnect GND from COM and Vin- to make the input floating? That would be a nice easy solution.

I find the datasheet explanation of the various pins that can/should be connected to supply negative puzzling and I remember blowing several of the old ICL7107s up through incorrect grounding... :(

Your help is appreciated!
 

Thread Starter

#JohnMaya

Joined Feb 10, 2021
9
Hi AlbertHall,

Thanks for this suggestion. It's a little more complicated than I had hoped, but maybe it's the way to go.

Annoyingly there was an old ICL7107 based ammeter in the same position that didn't suffer from this problem. It was destroyed in a lightning strike and I never thought to keep it for comparison.

I did think for while that the lightning strike had affected one or more of the grounds in the boat, but I don't believe this is the case.

I'll the article you linked too and see if I understand it.

thanks for your time..... :)
 

crutschow

Joined Mar 14, 2008
27,171
In my case the battery is replaced by a 7805 supplying V+ and GND. V- is connected to a capacitor diode combination to provide about -3v (That's what I measured). As explained, GND is also connected to COM and Vin-.
What is a "capacitor diode combination"?
Post your exact circuit schematic.

The 7106 sets the input common to 3V below the V+ supply voltage

1613066205113.png

so operating it at a 5V difference may be the problem.

The 7107 can operate with a plus and minus 5V supply so that may work better for you.

1613066393739.png
 

Thread Starter

#JohnMaya

Joined Feb 10, 2021
9
Hi crutshow,

I'm using one of the cheap chinese made meters labelled "YB5135A" and as such have no circuit diagram. Studying the PCB it seems as though it is 90% the "stock" circuit from the datasheet (as posted by DodgyDave above) with changes to the potential divider to get the correct range.

The "diode/capacitor combination" was probably a misleading thing to say. Reverse engineering the PCB, V- is generated from a pair of FETs driven from the clock signal on the chip. The buffered clock signal from the transistors is capacitively coupled in the same way as the output(s) from the CD4009 shown in your helpful "figure 9-1".

The IC is therefore being powered from +5 volts on V+ and -5(ish) volts on V- as you suggested.

I'm unsure of the purpose of ANALOG COMMON as on this PCB it is connected directly to Vin- and GND. This effectively holds it at GND potential and so I don't believe it can influence the circuit - or am I wrong :(

What if I were to separate Vin- from GND and ANALOG COMMON? Wouldn't that make Vin- float at whatever potential was connected to it? In my case the "negative" end of the shunt. This is the part of the operation of the 7106 that I don't understand.

Any input (good or bad) you have is of course appreciated!
 

crutschow

Joined Mar 14, 2008
27,171
I'm unsure of the purpose of ANALOG COMMON as on this PCB it is connected directly to Vin- and GND. This effectively holds it at GND potential and so I don't believe it can influence the circuit - or am I wrong :(

What if I were to separate Vin- from GND and ANALOG COMMON? Wouldn't that make Vin- float at whatever potential was connected to it? In my case the "negative" end of the shunt. This is the part of the operation of the 7106 that I don't understand.
Yes, it's not completely clear about how to handle that.

Here's some info on the differential input:
1613085458385.png
From that I think you suggestion is correct, so you have the ANALOG COMMON connected to the power supply GND (connection between the plus and minus supply), and no ANALOG COMMON connection to the differential input.
The differential input thus is connected only to the shunt, and the common-mode rejection should eliminate the offset from any common-mode voltage between the GND and the shunt common.
 
Last edited:

Thread Starter

#JohnMaya

Joined Feb 10, 2021
9
From that I think you suggestion is correct, so you have the ANALOG COMMON connected to the power supply GND (connection between the plus and minus supply), and no ANALOG COMMON connection to the differential input.
The differential input thus is connected only to the shunt, and the common-mode rejection should eliminate the offset from any common-mode voltage between the GND and the shunt common.
OK - after a bit of board butchering I have done this - it worked!

The meter reads the correct(ish) current but shows progressive non-linearity as the current (mV from the shunt) increase.

For example, with 1mV from the shunt, adjusting Vref to show 1A (1mV = 1 amp). When the current increases to 8A (8mV) the meter shows 7.8A. 12A (12mV) shows 10.5A. Maybe this is because Vref is generated from a voltage divider, not a stabilised supply. It's not significant to be a problem, but it is annoying.

The last digit is also a bit twitchy; maybe I'll increase the value of the auto-zero cap Caz to see if that helps.

Anyway, thanks for your input! Problem mostly solved :)
 

crutschow

Joined Mar 14, 2008
27,171
For example, with 1mV from the shunt, adjusting Vref to show 1A (1mV = 1 amp). When the current increases to 8A (8mV) the meter shows 7.8A. 12A (12mV) shows 10.5A.
You said you adjusted Vref for 1mV = 1A.
You should adjust it for the full scale current (12A or whatever that is) not the minimum current.
With only one calibration adjustment you allows adjust the full-scale value.
That will give the minimum error over the full range of readings.
 

Thread Starter

#JohnMaya

Joined Feb 10, 2021
9
The meter reads the correct(ish) current but shows progressive non-linearity as the current (mV from the shunt) increase.
Forget this - I had two meters, one is the LED version and the other the LCD version. I was doing the mods on the LED version (circuit boards are pretty much the same) to make sure I didn't damage the LCD one. The same changes applied to the LCD one works perfectly!
 

Thread Starter

#JohnMaya

Joined Feb 10, 2021
9
You should adjust it for the full scale current (12A or whatever that is) not the minimum current.
You are, of course, quite correct. I was so excited that the thing was actually working almost properly that I went for the easy option.

Full scale is 200A, but it's not easy to quickly get that much current draw AND be in a position to adjust.

As I already said the LCD based one works just fine....:)

Thanks for your continued help with this.
 

crutschow

Joined Mar 14, 2008
27,171
The last digit is also a bit twitchy; maybe I'll increase the value of the auto-zero cap Caz to see if that helps.
That's likely due to noise pickup on the input.
You might be able to reduce that by adding a filter at the input (try one 10kΩ series resistors in each line with a 10μF capacitor across the input connection.)
 

Thread Starter

#JohnMaya

Joined Feb 10, 2021
9
That's likely due to noise pickup on the input.
You might be able to reduce that by adding a filter at the input (try one 10kΩ series resistors in each line with a 10μF capacitor across the input connection.)
I shall try that. Here it is In-situ

20210213_163710.jpg

Reading 3amps, with a Vin of 3.0mV - the owner will be really happy!
 
Top