Problem with high side switching p- channel MOSFET for a multiplexed LED Display

[Electr0nicus]

For a current project I've designed a multiplexed LED Display Board. The prototype you can see on the picture below. 


Everything works just fine. But there is a thing that bothers me and that's the switch off of the highside P-Channel Mosfets. They switch on sharp an clean, but switch off really weird. Although the Gate voltage is the same as the voltage at the source of the MOSFET, it doesn't switch off as expected. The voltage on the Drain of the MOSFET pretty much declines like a capacitor beeing discharged. 

Here is the scope- image (note that the waveform is -2.12V offset and display is blank)



The gate waveform is really clean (note that the waveform is -2.12V offset and display is blank):



Now if I switch on all segments on the display (display test) and then measure the drain of the MOSFET the waveform looks even ugglier.



Short description of my hardware Setup:

• The gates of the MOSFETs are switched on/off via a 74HC595 Shift register
• I use 7- and 16- segment LED Displays with a common anode.
• the cathodes of those Displays are connected to the outputs of the SPT16CP05 constant current shift registers.
• everything is powered with +5V
• MOSFETs used are IRML2246

Has anyone a idea what causes that weird behavior of the MOSFETs maybe the STP16CP05 devices play some role?

[KJDS]

Can you share the schematic of the LED driver?

[Electr0nicus]

Can you share the schematic of the LED driver?

Sure I can (click on image to enlarge):


Note that the RCK pin of the 74HC595 shift registers is connected to GND in the schematic, which is of course wrong. I have connected it to SCK on the prototype board, and was too lazy to update the schematic.

The LED constant current drivers (STP16CP05) and the shift registers to select the active digit (74HC595) are all daisy chained onto one SPI bus. The display then is driven as follows:

• First all outputs of the STP16CP05s and the 74HC595s are switched off. During that time the outputs of the 74HC595 Chips are Hi-Z. To ensure a stable gate signal for the highside P-Channel mosfets a 10k pullup resistor to +5V on the gate of each MOSFET is used.
• Then the display Data is shifted trough
• Outputs are enabled

[komet]

As you mention everything actually works fine, I think you're measuring bullshit caused by everything going high-impedance. What happens if you connect a resistor (e.g. 100k) between the anode and ground?

[Electr0nicus]

You have nailed it!

That was exactly the case! I added a 10k Resistor directly from the drain to GND, and the waveform looks as expected. Nevertheless I have some EMI issues. Pretty much because of the high impulse currents in a multiplexed LED display design. But I think with a decoupling cap near every MOSFET, I should get that problem under control. Because when you look on the oscilloscope image now, you can see spikes everytime a new display- digit is switched on and the other one switched off.



But anyway thanks a lot komet  and greetings to my neighboring county! Sometimes it only needs someone unprejudiced to solve a problem 

[free_electron]

your power rail cant handle the high pulse currents. local capacitor per display driver (both the top and bottom drivers)
i'd throw in a bunch of ceramic 10uf caps. they have very low esr and will feed you all the current you need ( 100nf won;t cope it as those current spikes are large in amuxed display... they'd be empty on no time)

[Electr0nicus]

your power rail cant handle the high pulse currents. local capacitor per display driver (both the top and bottom drivers)
i'd throw in a bunch of ceramic 10uf caps. they have very low esr and will feed you all the current you need ( 100nf won;t cope it as those current spikes are large in amuxed display... they'd be empty on no time)

Do you mean the STP16CP05s with "bottom drivers"? If so, they don't need extra decoupling caps. Because they have only have a digital supply. So I only have to put a 10µF ceramics close to the sources of the highside switching P-channel MOSFETs.

[Niklas]

Where have you connected the probe's grounding clip? High frequency signal content and high current can cause false readings if you have enough inductance and resistance between the local GND and the clip's GND. The local GND is the one at the circuit you measure on.