Electronics > Projects, Designs, and Technical Stuff
Low Voltage Detection & Power Cut IC
Ian.M:
In JDW's other thread I suggested:
--- Quote from: Ian.M on August 27, 2019, 09:49:39 am ---You need the capability to power-up/power-down the sensor module cleanly to prevent it corrupting its FLASH, so using a linear LDO with an enable pin to go from a nominal 5V down to 3.3V at the sensor head wouldn't be too much of an imposition. At an active current of 75mA, that would only be 130mW of extra dissipation. A small board connected to the sensor's power and I/O header could also have the voltage supervisor IC and and a buffer for the frame touch signal, as an off-board floating when inactive 3.3V logic level signal is asking for trouble in an automotive environment.
--- End quote ---
I'm not suggesting a buck regulator at the sensor, I'm suggesting a linear regulator. IMHO its advisable to have the sensor's Vcc rail monitoring and undervoltage cutoff at the sensor head, not in the main unit, as connectors fret and contact surfaces oxidise with age, which will reduce the Vcc available to the sensor and increase its impedance, so its likely that it will become unreliable with age if one tries to implement the undervoltage cutout purely in the main unit.
There's no need to isolate most of the signal wires when the sensor is powered down. With its Vcc at 0V, the two outputs from the sensor will effectively be at logic '0'. However one probably should isolate the sensor's serial RX pin, as that line probably idles at logic '1', and allowing the input protection to clamp it could result in out of spec voltages on the sensor Vcc rail and a return of the memory corruption problem. A simple N-MOSFET, source to sensor RX, gate to +5V and drain to the cable from the main unit will do nicely, or use a single gate logic buffer that can tolerate logic '1' in while powered down without loading the input signal.
Why 5V to the sensor head?, well its a low enough voltage to minimise dissipation in the local LDO, may already be needed in the base unit, and is low enough to be maintainable during cranking. If the base unit doesn't already have a 5V rail, getting 75mA @5v from a 14V power input with a linear regulator would be possible, as that's only 675mW dissipation, and that 75mA is only needed intermittently when the sensor is active. Of cource one could also use a buck regulator. There shouldn't be any need for extra power input protection as that should already be in place for the main unit's own power supply - just T into the feed after the existing protection circuit.
JDW:
--- Quote from: Ian.M on August 31, 2019, 10:02:45 am ---...one probably should isolate the sensor's serial RX pin, as that line probably idles at logic '1', and allowing the input protection to clamp it could result in out of spec voltages on the sensor Vcc rail and a return of the memory corruption problem. A simple N-MOSFET, source to sensor RX, gate to +5V and drain to the cable from the main unit will do nicely, or use a single gate logic buffer that can tolerate logic '1' in while powered down without loading the input signal.
--- End quote ---
Thank you for your reply, and yes the Rx pin does idle at 1 so it would need to be switched. But could you explain why you suggest I use a MOSFET (e.g., 2N7000G) instead of an NPN BJT for the signal switching? The required current flow would be in the microampere level, and a sufficiently high base resistor on an NPN (more than 10k) would keep current consumption on part with a MOSFET.
--- Quote from: tautech on August 31, 2019, 08:48:54 am ---And you couldn't imagine to use the IC in reply #6 with a MOSFET and/or a quad logic switch like HEF4066B ? :-/O
--- End quote ---
Thank you for proposing the HEF4066B, but I see it is an "analog" switch as opposed to a dedicated "logic" switch (even though logic can be used with analog switches, I am aware) and the datasheet says Ron(rail)=340-ohm max & Ron(peak)=2.5k-ohm max (@Vdd=5v & Isw=200A). That Ron seems a bit high to run 9600bps logic level data through it.
JDW:
--- Quote from: GeorgeOfTheJungle on September 02, 2019, 07:32:37 am ---To be sure that the problem is power glitches I would first solder a small lipo to VIN and GND of that sensor and see if it really stops misbehaving then.
--- End quote ---
"lipo"?
JDW:
Here is a simplified schematic so everyone can better see what I am attempting. (The power supply is complete and not simplified. Note it accepts either 12V or 24V vehicle power -- so about 29V when a truck's engine is running.)
This is my current bench-test setup, not the final product. The final product must move the fingerprint sensor 1.5m away using the 5 wires shown.
Note that while the UART section of the Fingerprint sensor can accommodate 5.0V just fine, the Holtek MCU side is restricted to 3.3v only. And it's power wire goes straight into that MCU as shown, with no protection at all. The same for "Out" -- it's an MCU output pin that has no protection at all.
I added a 1k pulldown on RA2 because I was sometimes getting some oddball readings without it.
I added a weak 22k pulldown on RB5 because if that wire gets disconnected from the sensor, I can't leave RB5 floating.
RB7 is kept at HI by the PIC, so no pull-up or pull-down is needed.
There is other capacitance on the 3.3v rail not shown in the schematic. 0.1uF ceramics and a couple 47uF aluminum electrolytics.
The manufacturer said that I need to power off the fingerprint sensor for Vin lower than 2.7V, since there is no RESET that I can access. It has been suggested that change my switcher power supply to 5V and then add an LDO (5.0V to 3.3V drop-down) on the PIC side. But if I drive my PIC with that LDO, I cannot switch that LDO off to power-down ONLY the fingerprint sensor. And if I drive my PIC at 5.0V and add the LDO only for the fingerprint sensor, then I would need voltage level conversion because the signals coming from the sensor would be 3.3V. Another consideration with that is that driving my PIC at 5.0V instead of 3.3V consumes more current (not a lot, but a consideration nonetheless). I would then need to add a RESET IC that would kill the 3.3V LDO (killing power to the fingerprint sensor) when voltage drops below say 2.8V, then enabling the LDO again when the voltage rises. But cutting voltage via LDO alone is insufficient, and I would need to cut the RB5 line and the RA2 line. ESD protection would be needed as well.
So if you have any new thoughts to share in light of this schematic, I am happy to hear them.
Ian.M:
--- Quote from: JDW on September 02, 2019, 03:06:30 am ---Thank you for your reply, and yes the Rx pin does idle at 1 so it would need to be switched. But could you explain why you suggest I use a MOSFET (e.g., 2N7000G) instead of an NPN BJT for the signal switching? The required current flow would be in the microampere level, and a sufficiently high base resistor on an NPN (more than 10k) would keep current consumption on part with a MOSFET.
--- End quote ---
Nobody gives a flying F-word about a few mA current consumption in an automotive application. The self-discharge of a Lead Acid battery is of the order of 4% per month, which for a 90AH battery, the smallest you are likely to find in a four wheel vehicle, is equivalent to a standing load current of 5mA. The problems start when you have *far* too many systems that draw tens of mA in standby with the ignition off, then you get a vehicle that you cant leave on airport long term parking for a few weeks, and expect to it to start when you return, without a solar panel to keep the battery topped off. When the vehicle is running, tens of mA is still negligible.
A BJT could be used, but will give *SLOW* rising edges, as all that's available (unless you add an extra pullup) to drive high is the base current *(1 + reverse HFE), and the reverse HFE) of most transistors is pretty lousy. It will also compromise your logic levels, lifting Logic '0' slightly by Vce_on and dropping logic '1' by Vbe, affecting noise immunity A small MOSFET with its gate pulled above the logic '1' level has low on resistance in both directions, and as long as the driving signal source can handle its channel to gate capacitance as a load, will have negligible effect on logic levels and nois immunity. It also takes less PCN area as it doesn't need a resistor.
N.B. a 2N7000 is *NOT* suitable unless you run your supply to the sensor at >7V as its max. Vgs threshold voltage is 3V, so it wont reliably stay on with 5V on the gate and a 3.3V logic '1' signal on the channel, which only gives 1.7V Vgs. You need one with a max threshold voltage <1.5V to guarantee a full logic '1' level at the sensor RX pin. <1.8V threshold will work with a possible slight reduction of the logic '1' level but still within spec.
You could also do the level conversion with 74LVC1T45 or 74LVC2T45 level translating buffers, (one of each or three of the 1T45) as they support partial power-down applications by going hi-Z if both sides aren't powered. That would let you run 5V logic levels over the sensor cable, for better noise immunity and let you run the PIC at 5V. The extra current consumption will be negligible for an automotive application (see above) and if it was important, by judicious use of either reducing the clock speed or sleeping when inactive, in most applications you can save more than going from 3.3V to 5V at a fixed clock speed costs you. The 74LVC1T45 is available in packages as small as the 1mm x 1mm DFN1010-6, so doesn't need much board area at the sensor.
Why do you think the sensor needs significant ESD protection on its cable (as long as that cable is screened)? If your usage case includes dumb users in nylon shell-suits fiddling with the connectors, and the 2KV HBM model ESD rating of the 74LVC1T45 isn't good enough, you *might* consider adding some, but remember the PIC is just as vulnerable so *IF* you need it, you need it *BOTH* ends of the cable.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version