4

I'm struggling designing a soft-start circuit for my project. It's 12 V with up to 500 mA load. I'm looking for a ~1000 ms ramp-up time from 0 to 12 V. This is the circuit I have designed, which seems to work in the simulator, but I may have overlooked something?

The MOSFET I'm intending to use is this one, with 30 and 20 V limits on the drain and gate voltages, with a threshold of 1.9 V. Will this work?

Simulation

enter image description here

Maciej Swic
  • 363
  • 2
  • 11
  • To some extent yes, but Vgsth varies from MOSFET to MOSFET. Is your load here just a placeholder or is it a real light bulb? – winny Dec 08 '22 at 13:02
  • 1
    If the load is a lamp, the current might be significantly greater than 500 mA for a short period of time. – Andy aka Dec 08 '22 at 13:02
  • @winny By vgsth you mean the gate threshold? Its 1.9V according to datasheet and entered into the simulation. The load is an Arduino, some switches and 100-200mA of LED strip. – Maciej Swic Dec 08 '22 at 13:07
  • @Andyaka The lamp is just a placeholder, see my previous comment for the actual load. – Maciej Swic Dec 08 '22 at 13:07
  • What is the inrush current if it were switched on in zero time? – Andy aka Dec 08 '22 at 13:08
  • @Andyaka No idea, is it very important? I'm trying to avoid hotplug damage in case data leads make contact before GND. Logic level is 5v and i've fried two arduinos because of accidental hotplugging. The system and plugs are not designed for hotplugging and this soft start is meant to avoid accidents. – Maciej Swic Dec 08 '22 at 13:15
  • Yes, gate threshold. Please draw your entire circuit with both LEDs and Arduino. Can you just ramp up the PWM for the LEDs instead and make it appear to be dimmed up? – winny Dec 08 '22 at 13:18
  • @winny The problem is data leads may make contact first and it appears like they got fried not because of the LEDs but the data leads becoming the ground return path. I cant draw out the whole circuit because its pretty involved, but its available here. https://electronics.stackexchange.com/questions/645100/hotplug-protection-for-serial-devices I'm intending to incorporate this soft-start between the 8P8C connector and the 12V consumers like the 5 V LDO – Maciej Swic Dec 08 '22 at 13:20
  • 1
    Switching the low side can lead to problems, for example if you have the Arduino plugged into a USB port, depending on the exact schematic. Maybe you just need some series resistors and protection Schottky diodes on the data leads. Your schematic might typically work (I'd prefer a slightly different and more robust circuit) but it's pretty hard on that little SO8 MOSFET when half-on. Most MOSFETs are not really designed to be used in the linear mode. – Spehro Pefhany Dec 08 '22 at 13:23
  • 1
    Unless everything is isolated from the floating ground you will be creating here, you really want to do high side switching. Draw a block diagram. – winny Dec 08 '22 at 13:24
  • 1
    `see my previous comment for the actual load` No. Please update your question. – greybeard Dec 08 '22 at 13:24
  • @SpehroPefhany It will not be connected to USB, only RS485 via 2x5 V data leads – Maciej Swic Dec 08 '22 at 13:37
  • I will update the question with a full schematic in a few hours. – Maciej Swic Dec 08 '22 at 13:49
  • @MaciejSwic Yes it's vital to know it actually because the MOSFET will suffer rapid thermal runaway in a linear application like this without even appearing to get warm. Most MOSFETs are not designed for this type of application. – Andy aka Dec 08 '22 at 13:49
  • 1
    Okay, RS485. The popular 75176, for example, has maximum recommended input of -7V and absolute maximum -10V. Will your arrangement not subject it to -12v at startup? – Spehro Pefhany Dec 08 '22 at 14:04
  • @winny By the floating ground, do you mean i should use a pmosfet instead? Would the AO6401A work? Its datasheet mentions being used as a load switch. – Maciej Swic Dec 08 '22 at 14:43
  • 1
    No. Look at the SOAR for it. With 12 V you are limited to 100 mA. – winny Dec 08 '22 at 14:54
  • @winny I missed that completely. Really struggling to find an SMD P-channel mosfet that meets the specificationand is available on JLCPCB. Any ideas please? Or should i just go for a TO-220? – Maciej Swic Dec 08 '22 at 20:32
  • This one looks like it should work? It says 38 A continuous or am i missing something? Seems like lot. https://datasheet.lcsc.com/lcsc/1912112237_HUAYI-HY15P03C2_C123687.pdf – Maciej Swic Dec 08 '22 at 20:51
  • A TO-220 can probably do it, but what about just a resistor in series and bypass it with a PMOS without any linear drive on it? – winny Dec 08 '22 at 21:26
  • @winny I don't have much experience in this field so I'm not really sure what you are suggesting. I would need a lot more detail unfortunately. – Maciej Swic Dec 08 '22 at 21:28
  • Place a resistor of suitable value across drain-source of M1 in tobalt’s example below. When the output voltage is say >11 V, turn on the MOSFET. – winny Dec 08 '22 at 21:51
  • 1
    "the data leads becoming the ground return path" I have no idea what you are doing, but this statement alone is enough to tell that the problem you are trying to solve has nothing to do with soft start – Maple Dec 09 '22 at 00:06

2 Answers2

4

The conventional circuit uses a PMOSFET if placed in the high-side branch. It can be mirrored and use an NMOSFET when placed in the low-side branch as you did.

The capacitor feedback ensures a linear ramp of the output voltage.

schematic

simulate this circuit – Schematic created using CircuitLab enter image description here

tobalt
  • 18,646
  • 16
  • 73
  • The n-channel was actually a mistake because i have one for the actual LED drive, i will probably use this design and accept if i do. Thank you. – Maciej Swic Dec 09 '22 at 12:23
  • I have posted an updated question with more information, i would greatly appreciate if you could take a look. Thank you https://electronics.stackexchange.com/questions/646454/serial-circuit-protection-design – Maciej Swic Dec 15 '22 at 12:21
2

Your basic approach is fine, but your circuit needs adjustments.

First, what you are doing is called "switching the ground". That is, the power control device is in the return leg of the loop that includes the load, rather than in the source leg. Some loads in some systems get cranky when they have no ground connection. For those situations, changing to a p-channel FET in the high side of the load is the solution. A disadvantage is that spec-for-spec, p-channels FETs are more expensive. IF IF IF your load doesn't care, I would stick with the n-channel FET.

This discussion would be easier if each component ha a unique reference designator (Q1, R3, etc.)

The 3.3K resistor in series with the gate can be eliminated. In high frequency switching circuits, a gate resistor reduces parasitic oscillations caused by the gate capacitance. For your type of once-in-a-while switching, this is not necessary.

Next, the R-C timer. Your calculation is correct in that you have a time constant very nearly equal to 1 s. The problem is that at that time, the capacitor voltage is over 7.5 V, but the FET turned on down at around 2.5 volts. At that low a voltage, the R-C curve is reasonably straight; the exponential curve kicks in higher up. As a close approximation, start with a design point of (2.5 / 7.5), or 0.33 x R x C. IOW, triple your resistor value.

Speaking of values, you will get better performance out of the circuit if you go with a larger resistor and smaller capacitor. Large capacitor values and tolerances drift much more with temperature and aging. Consider something like 100K and 33 uF.

Next, now that you have eliminated one resistor, you need to add one between the gate and GND, in parallel with the timing capacitor. In your circuit, there is nothing to discharge the capacitor when the circuit is off. Depending on the quality of the part, it could stay charged enough to turn on the FET immediately for several hours. Think about how quickly you need the cap to discharge. To reduce the effect this resistor has on the timing calculation, it should be at least 10x the timing resistor. There is a way to calculate the timing resistor to compensate for this.

AnalogKid
  • 16,865
  • 1
  • 13
  • 25
  • I have posted an updated question with more information, i would greatly appreciate if you could take a look. Thank you https://electronics.stackexchange.com/questions/646454/serial-circuit-protection-design – Maciej Swic Dec 15 '22 at 12:21