# LED circuit stopped working when expanded

I built a small prototype as shown in my diagram with the Si926 n-FET, 250 ohm pull up resistor. In my prototype board I have only 4 channels connected (only two shown). Only one LED is ever turned on at a time. When I set 3 outputs off (ground) and only 1 on, the FET turns on and I am able to supply current close to my max of 100mA for specific LEDs. I can do this for any of 4 channels, one at a time with the other channels off. However, I had a pcb made with 32 channels, 32 FETS and 32 LEDs (only 16 pull ups) and 2 decoder chips. Now my circuit behaves oddly. If I turn on only one channel and others, as described, then slowly turn up the current source, the current supply drops out at 1ma. Drop out, is when the current source detects an open circuit or extremely high impedance and subsequently disables its output.

NOTE: I am not using the PWM of max6964. The brightness of the LEDS is controlled by the current source. The constant current source is part of a design restriction. It is supplied by a constant current supply that is controlled elsewhere. It is a bench top current source. One source for all LEDs.

MAX6964 Data Sheet

Si926dl Data Sheet

Is it possible that having 32 of these output circuits is somehow effecting (lowering) the FET gate voltage with the decoder chip output low (ground) and thus causing the FET to not switch on?

Do I need a different pull up resistor? I wanted to keep it small so the voltage drop would be in the range of Vgs 1 to 2.5v.

Would changing the FET to Toshiba SSM6N43FU with Vgs 0.35v to 1V do the trick? Toshiba SSM6N43FU

• The MAX6964 has provisions for driving LEDs directly, is that not an option? If you suspect that it might be too low a gate voltage, I'd strongly suggest you measure the gate voltages, set some LEDs to full brightness (so you don't have any PWM oddness that might confuse the meter) and check the FET gates. But given that the MAX6964 is basically a bunch of 50mA capable pull down FETs and some logic, I suspect that something else is going on here. – Sam Feb 24 '17 at 23:10
• I am not using the PWM of max6964. The constant current source is part of a design restriction. It is supplied by a constant current supply that is controlled elsewhere. I was afraid i may have provided too much information :-) The focus of my question: why did the circuit work with 4 FET-LED pairs but no longer works with 32. – JoeChiphead Feb 24 '17 at 23:19
• Now when you say "It is supplied by a constant current supply that is controlled elsewhere" do you mean that there's one big CC supply somewhere feeding the LEDs or that each LED has it's own external CC supply? Either way, I'd still recommend setting the MAX6964 to "ON" or whatever setting is supposed to turn on all the Si969 FETs and check the gate voltages. If you see really low gate voltages, you're first thought shouldn't be "I'll need a lower Vth FET" it should be "Why am I getting no gate drive?". Your 32 250Ohm pull-ups resistors will suck a good 400mA off the 3.3V rail, is that ok? – Sam Feb 25 '17 at 0:01
• One big CC supply feeding all the LEDs. Only one LED is ever turned on at a time. Only 16 resistors, so ~200ma. Vg is ~2.3V on the channel that is turned on (decoder output high). – JoeChiphead Feb 25 '17 at 0:20
• Ah, I see now. But still, to only get ~2.3V from a 3.3V supply is quite odd. The MAX6964 has open drain output so it shouldn't be loading the gates of the FETs when it's outputs are off, the 250ohm pull-ups should drag them straight to 3.3V. From your schematic each of the LEDs should be independent from the others (assuming only one really is on at any one time). So the low gate voltage is indicating that somewhere something is either pulling down the 3.3V rail (does it measure 3.3V?) or something's not doing what it's supposed to be doing. Is the MAX6964 sinking current (it shouldn't be)? – Sam Feb 25 '17 at 0:37

When you have 31 LEDs off, your current draw from the 3.3V rail will be:

$$I=31\frac{3.3V}{250\Omega}=409.2mA$$

Whereas with only 3 LEDs off (that's what happened when you only had 4 channels):

$$I=3\frac{3.3V}{250\Omega}=39.6mA$$

Because you are pulling up the N-channel MOSFETs with low valued resistors, having to turn off 28 additional LEDs adds 360mA to your current requirements. If your power supply can't do that, it will start pulling down the voltage, and then the MAX6964 won't be able to output 3.3V to drive the gates (because it won't be getting them at its Vcc pin in first place).

This adds up to the fact that Si969 is only marginally capable of switching at 3.3V. The RdsON at VGS=4.5V is 3 Ohm. At 3.3V you can expect it to be higher, but it may still let enough current to light up the LED. However, if the power supply gets pulled down to, let's say 2.5 V, RdsON may be too high (higher that the 10 Ohm resistor of each LED channel, possibly) or it may not be able to switch at all.

What can you do about it?

1. Redesign your circuit using P-channel MOSFETs and active HIGH outputs from MAX6964. This will reduce the consumptions from your pull-ups no almost zero (only the pull-ups from the channels turned on will draw current). This may be traumatic for you so you can do another thing instead...

2. Increase the value of your pull-up resistors. 10x, even 100x. 250 Ohm is way too low. Unless you need blazing fast switching, a 4K7, 10K or even 22K pull-up is perfectly fine.

3. As a last resource, increase the current rating of the power supply (=buy a more capable one). Not recommended!

• I added LED data to my post. Re suggestion 1: Note that boards are already made, thus this question - Assume I could find a P-channel FET with lower VGS, same package size and pin-out, would the circuit work with the LED and limiting resistor on the high side? This forum typically shows P-FET with load on low side. Suggestion 2: Won't increasing the pull up value reduce the voltage seen at the base of the FET? – JoeChiphead Feb 26 '17 at 17:00
• If the boards are already made, then forget about suggestion #1 because it implies redesigning the board as the P-channel MOSFETs must be in the high side. OTOH, regarding suggestion #2: NO, increasing the pull-up value won't reduce the voltage at the gate of the MOSFET, because a) the gate is high-impedance, and b) it's the output of the MAX6964 what actually drives the gate, the pull-up only serves as a way to force a "default" state when the MAX6964 is powered off and doesn't provide any output. Really: don't worry about weakening the pull-up, even 22k will be fine. – Enric Blanco Feb 26 '17 at 19:21
• Since I am so close to the Vgs threshold of the si926, hours I swap out the FET to the SSM6N43FU with Vgs 0.35v to 1V? – JoeChiphead Feb 28 '17 at 11:34
• I would probably be a good idea, yes. Its lower RdsON will also help reduce the thermal stress in the MOSFET, as power dissipation will be lower. – Enric Blanco Feb 28 '17 at 12:25

I agree with Enric

• Vgs should be low logic level type <1.5 not 1.5 to 2.5 for a 3.3V supply in order to achieve low RdsOn
• I know that 100mA LED's are rated for ~ 300mW and this means ESR ~ 3 Ohms and MOSFET RdsOn ought to account for lower voltage drop. Typically << load ESR or i.e. <1Ω
• 250Ω pullup is ridiculously wrong low for this gate capacitance. You don't need 250Ω*20pF= 5ns risetime (Ciss~18.5pF typ)
• improve your attention to details on Ohm's Law calculations, incremental resistance (ESR) and Voltage measurements on every pertinent pin to resolve issues faster.

Take a look at the typical operating point on the datasheet. It should be obvious this switch is insufficient for the supply or visa versa. With a better switch , you only need 3V and no current limiter, just 0~3 Ohms in series.

If you want a precision design then take accurate measurements or calculations.

• and fix your CC supply or voltage drop issue. It shouldn't cut out. You set the voltage, no load and raise the CC to full voltage.

You can verify my nominal values in the answer. There will be a tolerance range.

SSM6N43FU is <1Ω which is OK for 3Ω LED is a wiser choice.

But how are you planning to improve your power supply without specs?