Current Flow Detection

[gmr1981]

Hi All,

I'm working on a power supply than than can deliver 5A@30V, to a load which has the potential to be a long distance away (about 100m).

I want to add a basic detection circuit to the power supply that can turn on an indicator when any current at all flows to the load, to determine whether the remote end is connected. I want this to turn on when about 1mA flows and be active all the way until the full 5A is flowing.

Can anyone suggest some cost-effective circuits? I am thinking of using the Hall effect sensor like the TMCS1107A4U, but this seems a bit expensive for what I'm trying to do, I don't need to measure the current, I just want to detect when it's flowing. Current sense resistors and appropriate amplifiers also seem a bit expensive (relatively).

The power supply consists of a COTS 240V AC to 48V DC module, feeding a buck converter that drops the 48V to 30V. I'm using an LM76005RNPR for the buck converter stage.

Thanks,
Greg

[SCSKITS]

What is the minimum current that the load will require?
Use below that value as the detect threshold.
Sensing 1mA and surviving 5A (5000 to 1) could get complicated.
If you can afford a Schottky diode drop in line, maybe detect across that.
With a few mA you should still see greater than 0.3V.
Maybe put the diode in the return so the measurement can be ground referenced.

[PGPG]

If you can afford a Schottky diode drop in line

If Schottky drop is to much you can think of few wire turns around reed switch but it will not work since 1mA

[Manul]

If you are designing that power supply and it's a switcher, then perhaps observe it's frequency / pulse skipping to detect loading. Completely unloaded switcher will skip a lot compared to even lightly loaded.

[Zero999]

How thick is the cable? Is it feasible to add another conductor to sense the voltage at the load end and trigger an indicator when the voltage drop across the cable is above a certain threshold?

[ejeffrey]

Can anyone suggest some cost-effective circuits? I am thinking of using the Hall effect sensor like the TMCS1107A4U, but this seems a bit expensive for what I'm trying to do, I don't need to measure the current, I just want to detect when it's flowing. Current sense resistors and appropriate amplifiers also seem a bit expensive (relatively).

A sense resistor and amplifier should be pretty cheap, can you either show what circuit you looked at, or just give an indication of what you consider "expensive"?

Pushing 5A over 100m is a relatively large amount of current for such a low voltage.  So IR drops in the wires will be pretty high unless you use extremely thick wires.  That might be a problem, but if it's considered OK, it also means you don't have to go crazy to limit the voltage drop of your sense resistor.


For instance 200 meters of 16 AWG wire has a resistance of around 0.25 ohm which will give a 1.25 V drop at max load.

A 0.1 ohm / 5 W sense resistor plus a suitable current sense amplifier would work, and give you large enough dynamic range.  You can just go to TIs website and search "Analog current sense amplifiers" for one with suitable offset voltage and common mode range.   But since you just need on/off, I would second the recommendation to use a shottky diode instead of a resistor.  That will give you a logarithmic output voltage, so easier to sense at the 1 mA level, but saturating at 5A.

[PCB.Wiz]

Can anyone suggest some cost-effective circuits? I am thinking of using the Hall effect sensor like the TMCS1107A4U, but this seems a bit expensive for what I'm trying to do, I don't need to measure the current, I just want to detect when it's flowing. Current sense resistors and appropriate amplifiers also seem a bit expensive (relatively).

1mA-5A is a significant range, but you can get low offset opamps and low offset current sense amps.
The cheaper current sense amps run up to 26V common mode, so you would need to externally level shift those, or you can level shift a 1.8-5V opamp.
Offsets below 10uV are easy to find, if you map 1mA to say 4x that or 40uV, that's 40milli Ohms of sense resistance.
At 5A you then drop 200mV for 1w loss.

Opamp examples : lcsc have AD8628ARTZ(UMW) for 22c/50, you need to level shift that.

The offset voltage is usually spec'd at Vcc/2, and it can get worse near the rails. Not all suppliers will define that.
If you want lowest possible offset, you would float that opamp above and below the rails, and level shift with a PMOS or PNP or optocoupler

[Uunoctium]

//More OT
State of technology since years in broadcast for interconnection of cameras & BS between OB-van or studio is (was, now optical fiber):
Triax cable with detection of:
open/short, cable length, type of camera with automatic adaption of power supply (DC, AC), automatic cable equalizing, ...
Detecting and sensing at end of cable and transmit an protocol rf modulated to BS over the same cable. Bidirectional with lot of carriers in differing frequency slots.

[gmr1981]

Thanks to all for the suggestions.

I haven't decided on the answer yet, but I am leaning towards the current sense resistor and amplifier solution.

[ejeffrey]

The offset voltage is usually spec'd at Vcc/2, and it can get worse near the rails. Not all suppliers will define that.
If you want lowest possible offset, you would float that opamp above and below the rails, and level shift with a PMOS or PNP or optocoupler

Another option might be to use low side sensing.  The  you wouldn't need a high voltage amplifier or worry about level translators.  Depends on the application whether that works or not.

[MarkF]

I use a UF5402 diode and a LM393 comparator to detect locomotives on my model railroad.
They work well for my railroad power blocks.  The longest being approx. 2 meters long track.
I set the comparator threshold at 0.5V with a small pot with the diode on the low side.
A schottky diode would have a lower drop.  However, that is a pretty low threshold to meet.
Maybe the voltage could be boosted to compensate for the diode loss if it matters.

With a sense resistor, the voltage threshold to be check varies with the load current.
The diode check eliminates that variable uncertainty and variable supply voltage.
Granted you can use a small value sense resistor but then you're more susceptible to noise over the length of the cable.

The LM393 comparator has as open collector output which can easily drive a LED and current limit resistor.  It's supply voltage also goes to 36V.  With a diode capable of 5A and the LM393, you would only need two components.

[mtwieg]

The offset voltage is usually spec'd at Vcc/2, and it can get worse near the rails. Not all suppliers will define that.
If you want lowest possible offset, you would float that opamp above and below the rails, and level shift with a PMOS or PNP or optocoupler

Another option might be to use low side sensing.  The  you wouldn't need a high voltage amplifier or worry about level translators.  Depends on the application whether that works or not.

Low side sensing will likely be fine. After all, if there really is 100m of cable between the PSU and the load, then there will certainly be substantial resistance on both. Whether the sense resistor adds a little more resistance on one side or the other shouldn't matter.

However, this assumes that there are no other return paths for the supply current. If the supply current can return to the PSU via any other path (earth ground for example), then I would not sense current on the low side.

[robert.rozee]

I'm working on a power supply than than can deliver 5A@30V, to a load which has the potential to be a long distance away (about 100m).

I want to add a basic detection circuit [...] to determine whether the remote end is connected. I want this to turn on when about 1mA flows and be active all the way until the full 5A is flowing. [...] The power supply consists of a COTS 240V AC to 48V DC module, feeding a buck converter that drops the 48V to 30V.

sense the current flow between the COTS module (48v) and the input to the buck converter, taking into account the IDLE current that the buck converter consumes - this idle current should be fairly well defined.

use 2x silicon diodes (in series) paralleled with a 'sense' resistor. pick the resistor value such that it will have a 0.6v drop across it when 1mA + I_IDLE is flowing. above 2x this current, the 2x silicon diodes will be conducting and protect the sense resistor - these diodes need to be capable of passing your maximum 5A current draw (actually a bit less as you are bucking 48v down to 30v).

attach a PNP transistor:
- emitter to the high side (48v),
- base to the low side (46.8v) via a 1k resistor,
- collector to 0v via an LED and 4k7 resistor (giving approx 10mA LED current).

once you have this working, only then consider optimizing by going for 1x schottky diode and some fancy current sensing amplifier. personally, i'd be reluctant to go that far as you are unlikely to gain much better than a 2:1 power saving.


cheers,
rob   :-)

[snoop33]

You will likely find the read unreliable unless you use a higher value sense resistor, which at 5A then becomes wasteful. Ignoring the wire loss as others have mentioned if the loss is a concern at 5A I would think about using two different sense resistors, and mosfets to transfer current from/to them as the values change.

[SiliconWizard]

Of course the issue with a sense resistor here and the wide current range is that - say you use a 100 mOhm sense resistor, that's only 100 µV @ 1 mA, but 500 mV @ 5 A  (2.5 W power loss).
100 µV can be detected using an appropriate opamp or dedicated current sense amplifier, but lower than this becomes challenging to be reliable, especially if the solution must be low cost.

A dual approach would be more efficient but it's going to become significantly more complex and will cost more than a Hall effect sensor.

Note that the TMCS1107 has at best (version A4) a 400 mV/A ratio, so to detect 1 mA, that'd be a threshold of 400 µV. But the TMCS1107A4 has an typ. offset error of +/-12.5 mA. I don't think that would work.

[Zero999]

Presumably this doesn't have to be very accurate.

Use a dual Schottky diode. The diodes should be fairly closely matched and certainly will be well thermally coupled. One diode passes a reference current, equal to the trip value. I've used a generic current source on the simulation, but it could be a resistor or a current source IC.

[snoop33]

That will never work. At 5A, assuming a 0.4V loss that's 2W of heat - they'll quickly be unbalanced. It's also not going to be trimmed enough to detect 1mA reliably. This is why current sense chips use laser trimmed resistors.

Hall-effect sensors will never sense 1mA reliably either. At that current they're more likely to detect a change in the weather! They're good for isolation and AC when accuracy isn't required at low current.

It's got to be a sense resistor, OP just has to choose if being wasteful at 5A is doable - one resistor will then meet his price point. If not and the budget can't go up then he's going to have to beg for some more coin.

[Zero999]

That will never work. At 5A, assuming a 0.4V loss that's 2W of heat - they'll quickly be unbalanced. It's also not going to be trimmed enough to detect 1mA reliably. This is why current sense chips use laser trimmed resistors.

I assumed the 1mA was not critical. It could be 600µA or 1.4mA.

Does it matter if it becomes unbalanced at 5A? It will certainly read on at that point. The only issue will be if the load is removed, the threshold will change, unless sufficient time has elapsed to allow the diodes to reach the same temperature, but that might not matter, if response time is unimportant.

It's got to be a sense resistor, OP just has to choose if being wasteful at 5A is doable - one resistor will then meet his price point. If not and the budget can't go up then he's going to have to beg for some more coin.

To avoid excessive losses, that requires a comparator with a low offset/drift.

[ejeffrey]

That will never work. At 5A, assuming a 0.4V loss that's 2W of heat - they'll quickly be unbalanced. It's also not going to be trimmed enough to detect 1mA reliably. This is why current sense chips use laser trimmed resistors.

The OP claims to be just looking for load presence / absense detection.  A shottky diode sensor is the cheapest and easiest mechanism to do that.  It doesn't require an opamp or comparator with any special precision.  The accuracy won't be anything special but it will easily tell the difference between 1 mA load and open circuit (say, less than 100 uA of leakage current)

If the I*Vf losses of a shottky diode are too much they will need to do something more complicated but it's likely less than the losses in the cable.

[ejeffrey]

Stability could  be significantly improved by adding a 10 ohm resistor in parallel with the diode.  At 1 mA/10 mV no significant current will flow through a shottky diode unless it's extremely hot (over 85C), so you will have a relatively linear and stable shunt, but by 100 mV the diode will be conducting significant current and keeping the voltage low.  It's the range switching approach but with a diode as a passive switch.

[snoop33]

Sorry I didn't read it properly. If it's just any current then I agree.

Personally I'd just use a large enough resistor to detect a minimal voltage for 1mA, and adc it against a low ref. For example ATtiny1616 has a 0.55v ref. 10ohm resistor generates 0.01v which is enough.

Stick a NMOS in parallel with the resistor. 0.2A over a NMOS @ ~40mOhm will read 0.008v, ATtiny1616 will pick that up.

And toggle the NMOS based on the read making 0.2A the switch point.

[Zero999]

Sorry I didn't read it properly. If it's just any current then I agree.

Personally I'd just use a large enough resistor to detect a minimal voltage for 1mA, and adc it against a low ref. For example ATtiny1616 has a 0.55v ref. 10ohm resistor generates 0.01v which is enough.

Stick a NMOS in parallel with the resistor. 0.2A over a NMOS @ ~40mOhm will read 0.008v, ATtiny1616 will pick that up.

And toggle the NMOS based on the read making 0.2A the switch point.

That will work, if a large amount of hysteresis is acceptable i.e. on >1mA, off <100mA, but it won't work if you want it to turn off closer to 1mA, because once the MOSFET is on, the voltage across it is going to be tiny.

[PCB.Wiz]

Stick a NMOS in parallel with the resistor….

And toggle the NMOS based on the read making 0.2A the switch point.

Interesting idea, it’s more parts than a low offset opamp, but MOSFETS are low cost these days.
A MCU may be a bit slow for a dynamic load, but you could combine a MOSFET and a jellybean opamp, to give a non linear transfer.
Below some Vth/gain threshold, it looks like a resistor, above that it behaves more like constant voltage, until it finally looks like rds(on) of the fet.

[Zero999]

Stick a NMOS in parallel with the resistor….

And toggle the NMOS based on the read making 0.2A the switch point.

Interesting idea, it’s more parts than a low offset opamp, but MOSFETS are low cost these days.
A MCU may be a bit slow for a dynamic load, but you could combine a MOSFET and a jellybean opamp, to give a non linear transfer.
Below some Vth/gain threshold, it looks like a resistor, above that it behaves more like constant voltage, until it finally looks like rds(on) of the fet.

Use the MOSFET as a shunt regulator to limit the voltage drop across the sense resistor, like the Schottky diode idea, but lower voltage drop. Note this is simply a DC simulation. It will need frequency compensation to avoid oscillation.

[PCB.Wiz]

I was curious so here is the Resistor+FET with a wrapper opamp.
Linear to ~ 4mA and then follows the Vgs-Id curve.
This FET has a temperature null at ~ 1.2A and you can see the Rds temperature variation.
A 12mOhm FET show appx 29mOhms at 5A with this gate-drive scheme.