Variable reluctance magnetic pickup signal conditionner

[Spehro Pefhany]

It probably will be higher at higher speeds due to mechanical issues
(vibration in the gap between the sensor and the gear teeth so the
edges are not clean, for example).

Of the two integrated schemes I see for automotive sensors, the first,
from OnSemi, uses a fixed hysteresis of 320mV nominal, and the second
from National uses adaptive hysteresis:

http://www.onsemi.com/pub/Collateral/CS1124-D.PDF
http://cache.national.com/ds/LM/LM1815.pdf

Best regards,
Spehro Pefhany

 

[Fred Bloggs]

It probably will be higher at higher speeds due to mechanical issues
(vibration in the gap between the sensor and the gear teeth so the
edges are not clean, for example).

Of the two integrated schemes I see for automotive sensors, the first,
from OnSemi, uses a fixed hysteresis of 320mV nominal, and the second
from National uses adaptive hysteresis:

http://www.onsemi.com/pub/Collateral/CS1124-D.PDF
http://cache.national.com/ds/LM/LM1815.pdf

The NS LM1815 is designed to handle the case where high speed noise
amplitude exceeds low speed signal amplitude- what else to do. It looks
like these variable reluctance pickup circuits can get quite
sophisticated in the way of locating exact angular position of shaft,
diagnosing bad bearings, and/or bent blades on turbine flow meters- all
based on statistical analysis of the output pulse zero crossings.

 

[John Fields]

Hi John,

Try this:



8V>--------+----------------------------------------+
| |
[10K] |
SENSOR | |
+--+ | [2K]
| |----+--------+----------[10K]----------|-\ |
| | | | >--+-+--->OUT
| |----+--[100]-|-------+--[10K]--------+-|+/ |
+--+ | | | | |
[1000pF] [1000pF] [10K] +---[2M]--+
| | |
GND>-------+--------+-------+----------------------------->GND

Not knowing more about the sensor it's difficult to say what exactly
the component values should be, but with the values shown and assuming
the sensor has a resistance of 100 ohms, the output should be
quiescently low and go high when the sensor senses a metallic object
and generates an output pulse of at least 100mV. If you need greater
sensitivity decrease the value of the 100 ohm resistor or, if you want
to make the sensitivity variable, make it a rheostat. The 2M
hysteresis resistor will give you about 10mV of hysteresis at the
output transitions and should (hopefully) eliminate chattering for
slow edges out of the sensor, if you have any.

I'd place a resistor between coil and 1000pF caps on each leg. Or at
least a couple of inductors. He is running the non-shielded cable along
an engine and that can be pretty noisy. The 2M could be taylored to
whatever signal the coil generates so he has max hysteresis protection
against spikes. Maybe to 1/4th of signal swing or so.

---
Unfortunately, doing that (adding the extra resistors) would cut down
on the sensitivity of the sensor (maybe not a bad thing!) because of
the attenuation caused by the rail-splitting resistors.

I agree with the inductor part, but what I'd do would be to replace
the 10k series resistors into the comparator inputs with chokes since
the resistors aren't really doing much anyway. I also agree with the
hysteresis part, and between your suggestions and JW's twisted pair
suggestion it seems the thing should work.

 

[Jacques St-Pierre]

As I told, here some results:

Adding bypass cap on the original design did not help. In fact it push the
circuit to oscillate.
Adding a 390K to pullup the + input also turn the circuit in oscillator
mode.
So we try John circuit and it was much better, but a lot less sensible. No
noise, no oscillation appear, but we had to change the 100 ohms for a short
to get enough sensibility. The circuit work in the lab, so now it will be
tested on the motor.
I also look at the IC that you give me the link, they may be the perfect
solution, but I don't have them in stock, for now, so I will wait to see
what append with the LM339 circuit.
I will let you what result we got.

Bye (and thanks to all)
Jacques

 

[John Fields]

As I told, here some results:

Adding bypass cap on the original design did not help. In fact it push the
circuit to oscillate.
Adding a 390K to pullup the + input also turn the circuit in oscillator
mode.
So we try John circuit and it was much better, but a lot less sensible.

 

[Fred Bloggs]

As I told, here some results:


Adding a 390K to pullup the + input also turn the circuit in oscillator
mode.

Did you add that 1K between the the 390K pullup and the 10K-10K divider?
Did you twist and shield the wires from the comparator to the pickup?
The fact that the low "sensiblity" ckt did not "oscillate" means that
you did not, your signal is dirty, it is not oscillation, it is noise
amplification.

 

[Lasse Langwadt Christensen]

Hi John,

Try this:



8V>--------+----------------------------------------+
| |
[10K] |
SENSOR | |
+--+ | [2K]
| |----+--------+----------[10K]----------|-\ |
| | | | >--+-+--->OUT
| |----+--[100]-|-------+--[10K]--------+-|+/ |
+--+ | | | | |
[1000pF] [1000pF] [10K] +---[2M]--+
| | |
GND>-------+--------+-------+----------------------------->GND

Not knowing more about the sensor it's difficult to say what exactly
the component values should be, but with the values shown and assuming
the sensor has a resistance of 100 ohms, the output should be
quiescently low and go high when the sensor senses a metallic object
and generates an output pulse of at least 100mV. If you need greater
sensitivity decrease the value of the 100 ohm resistor or, if you want
to make the sensitivity variable, make it a rheostat. The 2M
hysteresis resistor will give you about 10mV of hysteresis at the
output transitions and should (hopefully) eliminate chattering for
slow edges out of the sensor, if you have any.

I'd place a resistor between coil and 1000pF caps on each leg. Or at
least a couple of inductors. He is running the non-shielded cable along
an engine and that can be pretty noisy. The 2M could be taylored to
whatever signal the coil generates so he has max hysteresis protection
against spikes. Maybe to 1/4th of signal swing or so.

---
Unfortunately, doing that (adding the extra resistors) would cut down
on the sensitivity of the sensor (maybe not a bad thing!) because of
the attenuation caused by the rail-splitting resistors.

I agree with the inductor part, but what I'd do would be to replace
the 10k series resistors into the comparator inputs with chokes since
the resistors aren't really doing much anyway. I also agree with the
hysteresis part, and between your suggestions and JW's twisted pair
suggestion it seems the thing should work.

The resistor will limit the input current at high input voltages, the output
from the sensors normally used as crank sensor can 100+ volts at higher RPMs


-Lasse