PCF8883 Proximity Sensor occasionally locks up

[arizagnoev]

I notice that a similar question has been asked elsewhere in the forum. I have looked at the response from NXP but I am still having a problem and would appreciate some assistance.

Referring to the attached schematic PCF8883_typical_application.jpg

I am using the following component values:

Rf = 4K7 (I read in an application note that it should be under 5K)

Cf = 10pF

Rc = 75K   (this gives me 1.5V across Ccpc which is half the voltage of Vcc(integd) - I don't have a high impedance meter. I capture the voltage on an oscilloscope. The voltage does drop off rapidly but I am measuring the initial peak which for me is 1.5V. This is half Vcc(intregd) . My Vdd is 3V which is derived from a Linear regulator - no switched mode regulator)

I have the same problem irrespective of whether I use the filter Rf and Cf or not.

My sensing plate is rectangular with rounded corners. It measures approximately 30mm by 15mm. It is a thin steel plate covered with a thin sheet of Mylar. The user touches the Mylar. The sensor is connected to the resistor Rf via a free hanging wire which is about 15mm long. I prefer not to post photos as it is confidential. 

I am using it in push button mode. 

Most of the time it works well, but occasionally (maybe once every few minutes) it locks up in the off state even though I am still touching the sensing plate. To get it to work again I need to release and retouch the sensor plate a few times.

I have also seen it misbehaving in other ways but the main failure mode is the one I described above. 

[david_diaz]

Hello Ari,

Thank you very much for writing.

I assume that the issue is that when the input is touched for a long period of time, the device slowly recalibrates to a higher base capacitance level and if the user slightly releases the touch, but not completely letting go to trigger a release event the output will not be disabled.

The behavior explained is unfortunately a direct result of the auto-calibration feature within our device.  A possible workaround to eliminate the issue is to measure the CPC voltage buffered with a voltage follower.  On the PCF8883 board (OM11055) we use the LMC7101 to buffer the CPC voltage.

By buffering the CPC voltage, the voltage can be read by an A/D input of an MCU and the analog value is being used as a complement to the digital output of the capacitive sensor.  The idea is to register the CPC voltage value for instance x seconds after the cap sensor has detected hands and use this value later on as a reference.

Please let me know if this information helps.

Regards,

David

[arizagnoev]

Hi David,

A quick update.

I have a test board, much like the evaluation board you referred to.

I couldn't get a LMC7101 quickly, so I am using a TLC272 which has an input

resistance of 1 terra Ohm.

I can see that when the sensor is touched the voltage goes up by about 10mV

and slowly carries on going up for a short time. If I release my touch the

PCF Output logic goes low and the CPC voltage drops back to where it was

before touching. I notice something odd. When I release the sensor the

Output pin goes low and only 50ms after that do I see a drop on the CPC

voltage. Obviously the internal logic is looking at something else besides

the CPC voltage to determine that the sensor has been released.

I haven't used it on my device - that comes next.

The problem with the TLC272(that I used) and the LMC7101(that you

suggested) is the current consumption. I was thinking of looking for an Op

Amp with shutdown.

However I have found the Max44625. It has a current consumption of 4uA(even

lower in shutdown). I am not an expert - I think the low speed of 200kHz is

not an issue as ours is a slow application. The only potential issue could

be the input resistance of 1GOhm. This is less than the 5GOhm specified in

the NXP app note AN10832.

Do you think this could be issue - if so I will look for something else.

Please let me know if what I saw with the voltage on CPC is not what you

expect.

Regards

Ari

Ari Zagnoev

Combimouse Pty Ltd

401/8 Queens Avenue, Hawthorn, Melbourne, Victoria, 3122, Australia

Mobile - 0414 379 436

Skype - AriZagnoev

Email - azagnoev@combimouse.com

Website - www.combimouse.com

Facebook - http://www.facebook.com/Combimouse

Twitter - @Combimouse on Twitter: https://twitter.com/combimouse

[arizagnoev]

Hi David,

Sorry for all the emails. I should collect my thoughts first.

I have found a low power op amp with shutdown.

I will experiment with a LM358 to see if it solves the problem.

I initially thought there would be a problem because when the IC "latches

up" it takes a few retouches before it works again. I initially thought I

wouldn't be handle this scenario, but I now think I can. When I detect that

it has "gone crazy" I can sample CPC more regularly knowing that I can't

trust the output from the IC.

Ari

Ari Zagnoev

Combimouse Pty Ltd

401/8 Queens Avenue, Hawthorn, Melbourne, Victoria, 3122, Australia

Mobile - 0414 379 436

Skype - AriZagnoev

Email - azagnoev@combimouse.com

Website - www.combimouse.com

Facebook - http://www.facebook.com/Combimouse

Twitter - @Combimouse on Twitter: https://twitter.com/combimouse

[arizagnoev]

Sorry yet another response. I had a look at the LMC7101 spec and the supply

current of 0.5mA is way too high. It is some time since I looked at

followers. Do you think there is a better solution? Is there another op amp

you could recommend that has a current low current consumpton of say under

50uA?

Ari Zagnoev

Combimouse Pty Ltd

401/8 Queens Avenue, Hawthorn, Melbourne, Victoria, 3122, Australia

Mobile - 0414 379 436

Skype - AriZagnoev

Email - azagnoev@combimouse.com

Website - www.combimouse.com

Facebook - http://www.facebook.com/Combimouse

Twitter - @Combimouse on Twitter: https://twitter.com/combimouse

[arizagnoev]

Sorry my response wasn't quite correct. When the sensor indicates that it

has been released I could double check that it actually has been released

by comparing the AD reading with that registered previously and then

continue monitoring it until it actually has been released. Fortunately my

micro is not sleeping durinng this time and I am able to do this.

But as I said in my previous email I would like to know what happens to the

CPC voltage while it auto calibrates and incorrectly indicates that the

sensor has been released even though it wasn't released.

Ari Zagnoev

Combimouse Pty Ltd

401/8 Queens Avenue, Hawthorn, Melbourne, Victoria, 3122, Australia

Mobile - 0414 379 436

Skype - AriZagnoev

Email - azagnoev@combimouse.com

Website - www.combimouse.com

Facebook - http://www.facebook.com/Combimouse

Twitter - @Combimouse on Twitter: https://twitter.com/combimouse

[arizagnoev]

Hello David,

Thank you for your response. I had tried various things without success. I

had put it aside and was going to get back to this problem.

I must say I was going to look for an alternative IC at DIgikey, but your

device has such great specs especially the current consumption.

I plan on getting back to it soon so your email was timely. I will try what

you suggested.

What is considered a long time? Although I suspect my application might be

a long time because I could be gripping the sensor for over minute or even

longer.

I think your suggestion could work. As you say I could register the Voltage

at CPC soon after a touch is detected. I could then regularly monitor the

CPC voltage to check whether it has been released. What would I expect to

see at CPC over the course of a "long time" as it auto-calibrates? Maybe I

am lazy here and need to read the spec :smileyhappy: How will I know that the user

has removed their hand.

Thank you

Ari Zagnoev

Ari Zagnoev

Combimouse Pty Ltd

401/8 Queens Avenue, Hawthorn, Melbourne, Victoria, 3122, Australia

Mobile - 0414 379 436

Skype - AriZagnoev

Email - azagnoev@combimouse.com

Website - www.combimouse.com

Facebook - http://www.facebook.com/Combimouse

Twitter - @Combimouse on Twitter: https://twitter.com/combimouse