Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - vpd

Pages: [1]
1
TFTs / Re: NHD-5.0-800480TFT-ATXI#CTP TOUCH problem
« on: May 27, 2015, 10:30:05 AM »
I want to rise the problem again. Now I have a NHD-7.0-800480EF-ATXL-CTP that had been working inside our testbench for several months and now it's touch suddenly stopped functioning. I want to underline that we didn't change or play with its schematics for a long period.

So, the touch does not work, but I2C device is present on the bus in lunux. And I can read it using i2cget utility or try to use i2cdump.
However, once I try to read registers I see not FT5x06 register data but a copy of register address. If I try to dump all registers in one go I usually get total i2c core hangup, no registers are read at all, timeout is caught instead.

As I understand, reading register addresses instead of data could be caused bu FT5x06 MCU not running while the I2C is already set up as a slave interface. Once it shifts in the register address it sends interupt to MCU core to update I2C transmit data register. Howerver, MCU does not precess this (internal) interrupt so the shift register remains unchanged holding a previous value been shifted. That value is exactly an address so we just get it back.

I tried to play with a WAKE signal, to put it low and then high as mentioned earlier. This does not help. Touch chip just stops answering to any I2C read request even if it returned address before WAKE pulse.

Now we need to know it this is a bug in FT6x06 or it is a problem of our schematics. This is already a second case when the display touch stops functioning after device has been assemblied, tested and configured. We are discouraged by this and also poor documentation of not only FT5x06 chip itself but also a touch controller based on it.
For example, there is no explanation between WAKE signal and RST. Why the WAKE signal is marked RST on touch cable? Why the WAKE should be zero during power up? Why the INT, SDA, SCL should be zero during power up if they all are pulled up signals? Why the chip disappear from bus by manual pulsing WAKE low? What happens there?

3
TFTs / Re: Capacitive touch problems under linux (FT5x06)
« on: July 11, 2014, 05:54:45 AM »
Michael,

Ops! Sorry, that was a graceful explanation but it's not true  ;) I made a conclusion comparing two different timings but once programmers added delays on ARM linux nothing actually changed.

But! Then I proposed to pull the flat cable out of connector and push it back while the u-boot is running. And it started to work properly!
The difference between these two cases is that when it is a 'cold' start processor has undefined state on GPIO pins. Most probably they are tri-stated. And there are no pull-up resistors on INT and WAKE, so they float. But then u-boot starts and places GPIO pins in a known state. When we re-plug the cable we power up FT5306 with WAKE being high, not undefinied. And it works.

Now we patched the starter kit by adding 10K pull-up to WAKE. And it works now.

It seems that FT5306 monitors WAKE during its startup and enters some test mode if WAKE is low. Probably, you can ask FocalTech what do they know about this? And what do they think?

Please, check this issue with your displays if possible. If so, this 'feature' must be documented.

Vladimir

4
TFTs / Re: Capacitive touch problems under linux (FT5x06)
« on: July 10, 2014, 06:17:00 AM »
Michael,

It seems we found a reason of the strange behavior. The problem is how fast I2C host goes to reading stage of I2C transfer after finishing its write state. FocalTech chip in 4.3" display is slow so it is not ready with data when the ARM processor starts read part of transaction. Now I tested it with another host which can wait for 1 ms between writing register address and issuing next start condition. In this case display works.

I think you should also test this and add this information into display datasheet. It is TOO IMPORTANT to be in NHD datasheet, not in an application note. Also, you should clearly state in datasheet what chip is used for each display.

I looked through FT chip datasheet and there is no information of this issue. All the delays reported in the spec are about 4-5 mks that is about 2 bit intervals on 400KHz bus and half bit interval of 100 kHz bus. I saw using the scope that we actually have reading delay not less then 2-3 byte intervals but this doesn't work. I think that FT5306 processes its I2C interrupt too slow to update the I2C transmit register so we get the value that has been written in write stage of the transaction.

We will try to play with reading delay on ARM processor to clarify the typical value of delay that works. Now we see that 5406 and 5306 have different delays, possibly because they have different clock frequencies, 5406' being higher that one of 5306.


5
TFTs / Re: Capacitive touch problems under linux (FT5x06)
« on: July 08, 2014, 01:51:37 PM »
To say trouth, the marking on PCB is as follows:
1) 4.3" - WAKE
2) 7" - RST
However, i see /WAKE signal in both datasheets. That's why I was confused.

Ok, we will try to issue a reset to wake signal. Actually, we did it manually by wire (sporadic pulse, not equal 100ms) but this didn't help. Probably we need something else?

6
TFTs / Re: Capacitive touch problems under linux (FT5x06)
« on: July 08, 2014, 10:55:22 AM »
Michael,

Unfortunately, the WAKE signal is poorly documented. As I could understand, it is either interrupt to Touch MCU or just a reset. It is called /WAKE in datasheet but actually on the ribbon pcb of display is marked as RST so we are confused. There is a period during boot when this signal has about 2V, not 3.3 and not 0. Moreover, this voltage is strong enough - when we tried to short /WAKE to adjacent INT pin on the connector (being 3.3V and that time) we got just 2.3V, not 3.3. I suspect that all this issues could force FT5x06 to hang. So, do you know the way to get it back to initial state? Will the one ground pulse of 100ms to WAKE signal be enough?

Vladimir

7
TFTs / Re: Capacitive touch problems under linux (FT5x06)
« on: July 08, 2014, 10:13:02 AM »
Michael,

1) We use the same code to work with both displays. As you understand this code is either linux or u-boot (based on the same linux) i2c bus framework that has been testing for many years.

2) We used an oscilloscope to inspect the signals on the bus. We didn't see any artefacts on signal timings. Also, we saw the display responds with an ACK so this behavior cannot be thought as just undefined bus.

We need some advice from you what can we check to solve this issue. Probably, there are some issues with an /WAKE signal that works differently for 7" and 4.3" display. This signal is tied to ARM GPIO and goes between pulled up, pulled down and undefined state sereval times during linux boot.

Best regards, Vladimir

8
TFTs / Re: Capacitive touch problems under linux (FT5x06)
« on: July 08, 2014, 09:41:30 AM »
We have three displays with the same behavior.

We tried to write to mode register hoping that it will hear and exit test mode but that didn't help.

9
TFTs / Re: Capacitive touch problems under linux (FT5x06)
« on: July 04, 2014, 10:08:46 AM »
Michael,

We could set up 7" sensivity and now it works with 4 mm glass cover. However, for this case threshold is too close to sensor noise margin. Spontaneous cursor movement can be seen with a threshold of 3 and glass strart to work at threshold 5. So, 7" touch screen works OK.

The most urgent problem remains with 4.3" display. As we discovered it does not work as described. In fact, it is present on i2c bus but it allways reads ad last byte written to it. For example, if we write in u-boot i2c tool:
i2c read 38 80.1 2 80200000
it reads two bytes: 80 80
if we asked it to read 5 bytes it would read 5 times the same byte, 80. The byte is always equal to register number being written.

7" display reads correctly: reading 5 bytes gives us 5 consecutive registers starting from given register number.

We tried both 400kHz and 100kHz bus speed. The rresults are the same.

So we still have problem with it.


10
TFTs / Capacitive touch problems under linux (FT5x06)
« on: July 02, 2014, 04:46:12 PM »
Hi,

now we are prototyping couple of devices using TI AM3358 Starter Kit. (http://www.ti.com/tool/tmdssk3358)
Original kit goes with resistive touch but we purchased couple of displays with capacitive touch from digikey:
1) NHD-7.0-800480EF-ATXL#-CTP
2) NHD-4.3-480272EF-ATXL#-CTP

Then we tried several FT5x06 drivers found on internet. One driver was good enough to work with touch screen. Most of them failed due to firmware version mismatch. However, several problems still exist.

1) Now driver works only with 7" display. 4.3" does not. Driver recognizes both displays via I2C, handshakes, etc. but in case of 4.3" there are no interrupts from FT5x06 during touch activity. We tried 3 different 4.3" displays from one purchase and result is the same. Then we supposed that driver has set touch threasholds too high so it does not recognize events. We tried to write different values of amplifier via the linux driver but this does not help. So, 4.3" touch does not work even without any glass.

2) Display of 7" works if direct touched but does not when placed beyond 4mm glass. Similar to item 1 we could not find solution how to improve sensivity.  We see in linux kernel log that 7" and 4.3" displays report slightly different descriptors during handshake. Now I cannot remember exact phrase but something about different number of rows in matrix.

Is there any guide how to set up touch controller settings to force it to work? We actually need it to be placed beyong glass as thick as 2 mm (or even 4 mm).

Best regards.

Pages: [1]