-
Hello,
I'm implementing Si2305DS P-fets in my project – using in LED matrix on a high side to drive 45 LED at a time, driven by port manipulation with MCU ~200Hz, not PWM). I've seen people using pull up resistors on a signal line (I added them as well), but is there a need to use them If I drive fets with MCU pin which will be set as HIGH/LOW and may float only when MCU powers on? I assume 10K resistors (I've seen people use 1K) will be OK due to small package and low charge (total gate charge 10-15 nC, input capacitance 1245pF)? I tested them with my hobbyist oscilloscope and 1 LED – with 10K output rising is clean and sharp, but falling is a bit "rounded", I'm guessing with 45 LEDs falling will be way "sharper" (sorry for not providing falling timings)?
Fets datasheet:
https://www.vishay.com/docs/70833/70833.pdf
Thanks in advance. -
I'm implementing Si2305DS P-fets in my project – using in LED matrix on a high side to drive 45 LED at a time, driven by port manipulation with MCU ~200Hz, not PWM). I've seen people using pull up resistors on a signal line (I added them as well), but is there a need to use them If I drive fets with MCU pin which will be set as HIGH/LOW and may float only when MCU powers on? I assume 10K resistors (I've seen people use 1K) will be OK due to small package and low charge (total gate charge 10-15 nC, input capacitance 1245pF)? I tested them with my hobbyist oscilloscope and 1 LED – with 10K output rising is clean and sharp, but falling is a bit "rounded", I'm guessing with 45 LEDs falling will be way "sharper" (sorry for not providing falling timings)?.
Hi 3dgeo,
I assume that the port driver (MCU) is operating from 5V rather than 3V3, and that thedrainsource [edit #1] of the PMOSFET is connected to the 5V supply line.
The effective gate capacitor will be much higher than you think because of the high value of the PMOSFET's parasitic capacitor from drain to gate (CDG).
Because your PMOSFET has such a low minimum gate threshold (VGS[th]) of 450mV, it would be wise to have a pull-up of 1k8.
Also, put a 22R directly on the gate terminal of the PMOSFET, using as short a lead as possible. Then connect the pull up to the 22R.
And connect a 470nF minimum ceramic X7R capacitor (physically large, not surface mount) from the PMOSFET source terminal to the 0V line as close as possible to the PMOSFET.
You would expect the PMOSFET drain voltage to rise fast and fall in a mess due to the LED (+ resistor?) load. If you would like to speed the fall, connect a 220R (or even 100R) from the drain of the PMOSFET to 0V. -
I assume that the port driver (MCU) is operating from 5V rather than 3V3, and that the drain of the PMOSFET is connected to the 5V supply line.
Thank You for your replay!
DRAIN connected to 5V??? Shouldn't DRAIN be connected to load and SOURCE to 5V??? I'm sooo confused now....
Funny as it is, today in my schematic and PCB design I flipped all sources and drains cos I was connected them wrong myself, drain was on 5V and source on load, but examples and tests shows that source should be connected to 5V and drain to load. Am I going nuts?
Yes, power line is 5V.
Can you give a bit more information on how you got these numbers and why these components have to be used? I want to learn, not only implement and forget it
Sadly, I can't add not SMD components, I only have max 4mm above PCB.
P.S. low side of the LEDs will be PWM with TLC59116, it has constant current limit, so no need for external resistors, don't know if that makes any different... -
DRAIN connected to 5V??? Shouldn't DRAIN be connected to load and SOURCE to 5V??? I'm sooo confused now....
Can you give a bit more information on how you got these numbers and why these components have to be used? I want to learn, not only implement and forget it
First the PMOSFET gate resistor. That is a gate stopper which discourages the PMOSFET from oscillating of its own accord, typically from 50kHz to 20MHz. The gate stopper plays no part in the fundamental operation of the circuit.
The gate of a MOSFET only requires voltage to control the drain current, so it follows that to turn the PMOSFET on and off fast, you need to have a fast voltage change between its gate and source. But the PMOSFET has a large real and virtual capacitance between its gate and source, so to generate a fast signal at the PMOSFET gate you need to charge the gate capacitance up and down fast, and that requires current, the more current the faster.
But there is a further complication with the gate of the PMOSFET. It only turns off at 4V2 so the MCU driving the PMOSFET gate cannot provide much current to discharge the gate capacitor and turn the NMOSFET off. This is where the pull up resistor comes into play: it discharges the gate capacitor to 5V and turns the NMOSFET off relatively fast.
When it comes to turning the PMOSFET on, the MCU has plenty of voltage swing, probably down to 0.5V. The PMOSFET, thus turns on at 4V2, but the MCU carries on down to 500mV which provides a large drive to charge the PMOSFET gate capacitance and turn the PMOSFET on fast.
Moving on to the drain of the PMOSFET. As stated, the PMOSFET has a nice big turn on gate drive so there will be bucket loads of current pouring out of its drain. That current will turn LEDs on and charge up capacitances as fast as hell.
But, when the PMOSFET is turned off (even fast) everything on its drain is just left hanging in space with no discharge path to 0V: the drain pull-down resistor provides that discharge path.
How were the particular values chosen: by precedent and compromise. The gate stopper was by precedent and compromise- I know that a resistor of around 10R to 1K is OK for gate stopping, and would normally have chosen 100R, but 100R would have slowed the PMOSFET gate drive too much, so I went for 22R as a compromise.
The pull-up resistor was also chosen by compromise. On one hand I wanted it to have a low value for fast PMOSFET turn-off, but on the other hand the minimum value was limited by the current sink capacity of the MCU.
The pull down resistor was also chosen by a compromise between speed (low value) and limiting wasted current (high value).
In general, for speed you need to charge and discharge capacitance fast, and to do that you need to source and sink high currents. To illustrate this, there are MOSFET gate driver chips which can swing 15V and source and sink 9A (see link below).
And finally, design in any field, is very much about optimizing compromises.Sadly, I can't add not SMD components, I only have max 4mm above PCB.
OK, use SM, but 1uF if you can.P.S. low side of the LEDs will be PWM with TLC59116, it has constant current limit, so no need for external resistors, don't know if that makes any different...
No- it will be fine, but be sure to fit the decoupling capacitor.
http://www.ti.com/lit/ds/symlink/ucc27321.pdf -
Thank you, it's a bit more clear now.
Well, this will probably make no different but I'll switch to Si2305CDS
To be fair, I didn't expect that I have to add so many components
I already made place for pullups, I may even squeez in 22R, but I don't have grounding in that part of a circuit 
You said "fast", I don't think 200Hz is fast, in worst case scenario what can happen if I skip 22R and caps? I'll do some more testing tomorrow to find out its oscillation at 200Hz. I'm also using bunch of 10K resistors, so it would be really nice to use those as well.... I'm not pissing away all your effort here, I just wonder, what if? 
Did I understood circuit correctly:
-
I'll switch to Si2305CDS
That would make the turn-off a touch worse.I already made place for pullups, I may even squeez in 22R, but I don't have grounding in that part of a circuit.
You will probably get away with no gate stoppers, but the decoupling capacitor is most important.I'll do some more testing tomorrow.
That's the spirit. I only hoped to illustrate the principles involved. Don't worry about pissing on my design.Did I understood circuit correctly:
Almost:
-
That would make the turn-off a touch worse.
Worse? It has lower capacitance... OK. I'll stick with old one.You will probably get away with no gate stoppers, but the decoupling capacitor is most important.
OK, I can try to squeeze in capacitors. There are 2 groups – 3 and 4 mosfers next to each other (7 fets total), only 1 mosfet out of those 7 will be on at one time, can I get away with putting only 1 cap next to those 3 and another to those 4? I have cheap CAP kit from ALI, I hope these will do, or I need special kind of caps?
I try to squeeze in 22R resistors as well, but I feel it's just waste of effort and unnecessary increase of BOM
I have to point out that with my hobbyist toy oscilloscope I didn't noticed ANY oscillation, tho as I said will experiment more tomorrow.
P.S. I know there are mosfet drivers, but I was hoping to get away without them due to low frequency, small fets and low (below 1A) switching current. I want to keep BOM as low as possible because I'm planing to make at least 100 of my boards.
Thanks again for helping to a potato like myself
-
Worse? It has lower capacitance... OK. I'll stick with old one.
It also has a potentially lower gate threshold voltageOK, I can try to squeeze in capacitors. There are 2 groups – 3 and 4 mosfers next to each other (7 fets total), only 1 mosfet out of those 7 will be on at one time, can I get away with putting only 1 cap next to those 3 and another to those 4?
Yes- good plan.I have cheap CAP kit from ALI, I hope these will do, or I need special kind of caps?
Those capacitors are ceramic but the dielectric is not shown on the Aliexpress website, so best to fit as larger value as practical: 22uf looks good.I try to squeeze in 22R resistors as well, but I feel it's just waste of effort and unnecessary increase of BOM
You would probably get away with no gate stoppers but not advisable if you are making many LED multiplexers. You may not see the oscillations on a low bandwidth scope, or any scope for that matter, because of intrusion by the scope probe.
I have to point out that with my hobbyist toy oscilloscope I didn't noticed ANY oscillation, tho as I said will experiment more tomorrow.P.S. I know there are MOSFET drivers, but I was hoping to get away without them due to low frequency, small fets and low (below 1A) switching current. I want to keep BOM as low as possible because I'm planing to make at least 100 of my boards.
I was just wondering why you are switching the LEDs on and off at such a high frequency (200kHz). The rule with switching is to use as lower frequency as possible. If the LEDs are for viewing, each LED only needs to be refreshed 60 times a second to avoid visual flicker.Thanks again for helping to a potato like myself
No sweat- anyway potatoes are a staple food of the world 
By the way, if you are making a load of these LED multiplexers, you could probably simplify the circuit by choosing a different PMOSFET or even a jellybean PNP BJT. If you would like to investigate this let me know and I will do a search. And if you could post the whole circuit for the LED multiplexer, that would be a great help. One fundamental question is, what is the maximum current that any switch will be required to source. Can you also state what the MCU output line driving the switch is. -
I was just wondering why you are switching the LEDs on and off at such a high frequency (200kHz). The rule with switching is to use as lower frequency as possible. If the LEDs are for viewing, each LED only needs to be refreshed 60 times a second to avoid visual flicker.
200 kHz? I'm positive that I haven't used "K" in there – 200 plain old HERTZ, no KILOS
I tried to point out that my refresh target is very low By the way, if you are making a load of these LED multiplexers, you could probably simplify the circuit by choosing a different PMOSFET or even a jellybean PNP BJT. If you would like to investigate this let me know and I will do a search. And if you could post the whole circuit for the LED multiplexer, that would be a great help. One fundamental question is, what is the maximum current that any switch will be required to source. Can you also state what the MCU output line driving the switch is.
Oh yes, I would really appreciate your help. Basically my circuit is a copy from
but instead of 3 * TCL5940 I'm using 3 * TCL59116 (I2C interface), and only 15 outputs per chip with 7 P-mosfets. Originally I've designed this to run on a USB with 10mA per LED (still bright enough), so 10 mA * 45 LEDs ON at one time = 450mA. MCU is ATMEGA32U4. But I also implemented a feature – if desired free TLC output will cut TLCs IREF pins resistance to ground in half and increase LEDs current to about 20mA (total ~900mA).
This project is a macro keyboard and I probably should point out that P-mosfets outputs also are used to multiplex keybard matrix.
Here is my PCB design, is 16 mm gab between fets is OK or I should put them closer to use only 1 cap?
-
Araaaaaah! That is the second time a missing/added k has got me into trouble.I was just wondering why you are switching the LEDs on and off at such a high frequency (200kHz). The rule with switching is to use as lower frequency as possible. If the LEDs are for viewing, each LED only needs to be refreshed 60 times a second to avoid visual flicker.
200 kHz? I'm positive that I haven't used "K" in there – 200 plain old HERTZ, no KILOS I tried to point out that my refresh target is very low
-
I would really appreciate your help. Basically my circuit is a copy from
Thanks for info. I will have a look at the design.
but instead of 3 * TCL5940 I'm using 3 * TCL59116 (I2C interface), and only 15 outputs per chip with 7 P-mosfets. Originally I've designed this to run on a USB with 10mA per LED (still bright enough), so 10 mA * 45 LEDs ON at one time = 450mA. MCU is ATMEGA32U4. But I also implemented a feature – if desired free TLC output will cut TLCs IREF pins resistance to ground in half and increase LEDs current to about 20mA (total ~900mA).
This project is a macro keyboard and I probably should point out that P-mosfets outputs also are used to multiplex keyboard matrix. -
I made simplifies schematic, hope this will help.
ICs (TLC59116) sinks the current (PWM) and sets constant current according to resistance from REXT pin to ground.
-
I found that there already is 22R in the BOM, so with some effort and a lot of cursing I managed to squeeze all in
China new year soon, so I have to hurry up with PCB prototyping or I will have to wait 3 weeks until china sobers out
Feedback?
-
+ 3dgeo
I have not found any radical changes to the design of the PMOS switch, except that it would be better to reduce the gate pull up resistor to 1k, so you can go ahead with the PCB procurement.
I have been looking for a PMOSFET with a higher gate threshold voltage, which would improve the turn-off performance and have found a couple of PMOSFETs, but have not completed the search yet. Will advise. -
+ 3dgeo
This is the PMOSFET that I would recommend: http://www.vishay.com/docs/67401/sq3469ev.pdf
It is a TSOP pack. Hope it will fit.
Attached below is the latest schematic for the PMOSFET switch:
-
Thank you!
Well, I have 1K resistors in a BOM, but wasted current from all 7 resistors will be 35mA, thats way above what I'm OK with. Next bigger value I'm using is 10K – 3.5mA is OK, I'm sticking with 10K. Sorry If my choice is stupid, but as you said – "design in any field, is very much about optimizing compromises". I'm already limiting LEDs to a 10mA to fit below 0.5A, can't afford to waste 35mA
Can you give me parameters of a fets I need to look for? There are hundreds of Pfets at LCSC, I will try to find most suited myself in a SOT-23.
P.S. I'm in a hurry to order only a prototype board, I'm not expecting it to be final design and I'm only ordering 5 boards, tho as I said I have to order fast (in 24h) to receive them before China new year or I'm waiting at least 3 weeks – that I can't do... So no pressure