In short: I´m working on a battery driven handheld device with capacitive-touch screen. The capacitive-touch works perfect when the PCB connected via USB, but it loses sensitivity when driven by battery after a couple of seconds.

The setup: 10cm dia 4-layer PCB with micorcontroller, radio-module, PCB antenna, and FPC connectors to an 3.5" TFT display, on which FPC the CTP driver IC FT5346 from FocalTech is.

The PCB is mounted in a ABS housing via screws (GND connected). The display is glued on the top of that housing and connected to the PCB via FPC cables.

The phenomen:

While developing, the device is connected via a USB-cable or debugger with the PCB. When I check the capacitive touch functionality on the display, it works perfectly. Every time I touch the display with my fingertip, the capacitive touch sensor triggers an interrupt on INT.

When I unplug all cables from the device and power it only via battery, the touch functionality gets worse over time. The first 5-10s the touch reacts like before, instantly on touch inputs via fingertip, but when I wait at first like 10s and try a touch with my fingertip, there is no reaction, even if I try it repeadly. When I use my whole fingerprint, then it reacts again and directly after that the fingertip works until I wait the 10s again.

My conclusion: As long as I connect the device to an earthed connection (USB, debugger) the capacitive touch works like a charm. When I disconnect the external GND/EARTH, the battery-minus is now the system GND and some "system capacity" is running low or high on time, so my fingertip only produces touches for a short period of time.

My observations so far:

  • I tried the same on another PCB with same results.
  • I observed the INT signal via oscilloscope and saw that there is really no interupt signal from the FT5346
  • I tried to couple the display case directly via cable with the PCB GND; same results
  • I tried to increase/decrease the touch threshold in CTP (THGROUP, THPEAK, THCAL) but no (recognizable) change in sensitivity at all

Do you have further tips that Ican try? In the datasheets is something about "auto-calibration" but no information about that anywhere.

The "user experience" is bad if he realizes he needs to put the whole finger on screen sometimes and sometimes the fingertip is enough.

UPDATE_01: I've placed the Device in the middle of a metalplate (stencil 400mm x 280mm) without connection (device-foot of ABS), the problem still persists. Then I connected the PCB-GND with a cable to this metalplate, and the touch works great again. There must be a way to improve the GND on my device to improve this behaviour without metalplate!?

UPDATE_02: I´ve connected the device via usb to the laptop, which is also isolated (run by battery) and the device touch works great. I think the size of the couple-capacity of the device is to small in ratio to the capacitive touch interface. But how do they accomplish this on smart watches or smartphones? Is there a trick in increasing the the GND size on small devices?

  • 1
    \$\begingroup\$ Does the problem go away if you touch the case with one hand while using the other hand to touch the screen? \$\endgroup\$
    – Annie
    Oct 11, 2021 at 21:44
  • \$\begingroup\$ @Annie No there is no change in sensitivity if I grab the housing with the other hand. \$\endgroup\$
    – Dr_Smu
    Oct 13, 2021 at 6:04

3 Answers 3


All capacitative touch screen sensors have a reduction in sensitivity when the device is not connected to ground.

When isolated from ground the device will have a capacitance to ground that is dependent upon the size of the device - for something the size of a cell-phone that capacitance is about 4pF and effectively appears in series with the touch capacitance. Your body will have a few hundred pF to ground and your finger to screen capacitance will be of the order of 1pF although if it is doing mutual capacitance detection (almost certain) the differential capacitance that needs to be detected will be of the order of 100fF.

This controller operates by scanning through the transmit channels and sensing the signals on the receive channels. It uses a mode of operation called mutual capacitance sensing where the coupling from row to column is maximum when there is no finger on the panel. Placing a finger on the panel diverts the signal to ground and can then be sensed as a touch. When the system ground is removed the signal change that results from touching the panel is reduced as there is a small capacitor in series with your finger, that reduces the touch sensitivity.

In order tolerate the large parasitic capacitances the signals will be averaged for a few seconds to get a baseline of the amount of coupling.

When the panel is touched the signal will depart from that baseline and be reported as a touch. That is probably part of the symptoms you are seeing where it changes over time.

Are there any configuration changes possible for the touch controller, frequency of operation, integration time etc? Is there any interference that is being received - when I worked on cell-phone touch sensing the biggest problem was when the phone was isolated from ground (as you are doing) or with AC adapters that generate interference in the same frequency band as used for the touch sensing.

  • \$\begingroup\$ I'm looking on the register set of the most common IC from focaltech FT5x06, which seems almost identical link. There is are some registers to change the threshold values which i tried, but no recognized change. No option for changing the operation frequency... Is there a way of improve the GND on my hardware site? \$\endgroup\$
    – Dr_Smu
    Oct 13, 2021 at 6:35
  • \$\begingroup\$ the ca·pac·i·tive performance reduction without PE gnd is due to CM noise shunted along that path which gets converted to DM noise from unbalanced CM coupling when not connected. So reducing the CM noise after identifying it's source will help with the correct "balun" \$\endgroup\$ Oct 13, 2021 at 12:58
  • \$\begingroup\$ @TonyStewartEE75 And where will be the correct place for this balun to test around? The CTP driver is placed on the FPC to the display, so no chance to play around on that circuit. I´ve placed ferrite beads on power lines between each domain circuit (power uC; power rf; power LCD; power CTP), will see if it helps, new Prototypes in production. Could it be also good practice to connect the GND of CTP via net-tie to the common PCB GND to dismiss potential interferences/disturbance? \$\endgroup\$
    – Dr_Smu
    Oct 14, 2021 at 19:27

There must be a stable ground reference to the touch layers and a stable reference capacitor (only film) to compare the change in C voltage (ac) wrt. gnd. to be a stable sensitive sensor.

Without details, it is hard to point which ground link is missing when cables are disconnected or is there is another issue like excessive CM noise from RF(?)

  • \$\begingroup\$ The CTP don't use an external reference capacitor as seen on the application circuit link . What makes me wonder is the change in sensitivity over time. If i tip the display direct after start it works well but if i wait a couple of seconds, the displays shows worse sensitivity over time. \$\endgroup\$
    – Dr_Smu
    Oct 13, 2021 at 6:44
  • \$\begingroup\$ @Dr_Smu There is an obvious missing ground reference connection in your test results which cannot be seen in the words without more details. This still could be excessive CM noise \$\endgroup\$ Oct 13, 2021 at 12:38
  • \$\begingroup\$ Try to map the supply and return paths with external cables that make it work. If the unearthed supply works only when an RF cap connects to PE earth , then it is a CM noise issue that may be improved . \$\endgroup\$ Oct 13, 2021 at 12:43
  • \$\begingroup\$ I´ve done some tests so far. Run device only by battery and couple GND to a also isolated metal plate and a battery powered laptop, which both ends up in better results. Also i powered the device directly after the switcher via battery to get a smooth Vsys, still same problems. Where or how can i detect the CM in my system? Osci is available. \$\endgroup\$
    – Dr_Smu
    Oct 13, 2021 at 14:50
  • \$\begingroup\$ If PE is a stable ground fixes the issue, measure the noise between 0V & PE, then Vdd & PE . The noise that is the same is CM noise. But this doesn't tell me where it comes from. That needs your debug measurement skills with radiated / induced or conducted noise. Protection may be from bridging star-gnds to source or shielding etc.' \$\endgroup\$ Oct 14, 2021 at 22:11

I've got the problem solved.
The CTP driver IC FT5346 from FocalTech, goes into monitor mode after aprox. 18s (less power consumption and simpler algorythm). After that, the sensibility of the touch interface is really bad (need a larger capacity change to react).
When i press with the whole thumb, it detects the first time and change from monitor mode back to the normal acitve mode. At that point, it detects like it should.
My solution (workaround) is now, to deactivate the automatic switchover in monitor mode, so the display will stay in active mode and reacts on each little touch on my display.
The problem is only if the system is battery-powered. So an improvement of the isolated PCB GND (relation between GND-area size and touch-area size)could increase the sensiblity in capacity-change detection.
My workaround will consume a little bit more power over time, but thats my trade-off for a quick solution.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.