0
\$\begingroup\$

I have drawn a schematic for a simple LED-driver.

It is nothing special, but how can I try circuit out?

Some parts are only available in SMD.

So my idea is that some one looks it over before I send it to the manufacturer.

With the MOSFETs in SMD, how big must be the thermal pad of each MOSFET be if it must handle 5 amperes? The final MOSFET is an IRLZ44ZSTRLPBF.

enter image description here

\$\endgroup\$
6
  • 1
    \$\begingroup\$ Did you do research which of the pins of the Atmega328P have hardware PWM support? Not all pins have native hardware support \$\endgroup\$ – Ferrybig Mar 23 at 19:04
  • \$\begingroup\$ Thanks for you comment. I have check it . I have used this pinout. components101.com/asset/sites/default/files/inline-images/… \$\endgroup\$ – fred mustermann Mar 23 at 19:05
  • \$\begingroup\$ Are MAX485 and ADM202 both mounted or one of them only? If both, it won't work. Well, the RS485 interface would not work anyway as bus has no ground reference. All I can say is double check everything, too much things found at first glance. \$\endgroup\$ – Justme Mar 23 at 19:28
  • \$\begingroup\$ thanks for you comment RS232 is for debug and RS485 is normal bus. RS485 needs no Ground ! it is differential. \$\endgroup\$ – fred mustermann Mar 23 at 19:35
  • \$\begingroup\$ @fredmustermann No, you are mistaken - don't worry, it is a common misconception that RS485 does not need a ground. It has been discussed even here many times. RS485 does need ground reference, especially since you have a completely isolated floating supply for it. \$\endgroup\$ – Justme Mar 23 at 19:52
2
\$\begingroup\$

With the MOSFETs in SMD, how big must be the thermal pad of each MOSFET be if it must handle 5 amperes? The final MOSFET is an IRLZ44ZSTRLPBF.

There are too many unknowns to solve this mathematically. But to get you started:

irlz44zpbf datasheet bottom of page 1: "62 RθJA Junction-to-Ambient (PCB Mount) = 40°C/W". This means that for the minimum pad shown at the end, the die temperature will rise 40°C for each watt of power dissipated.

When the MOSFETs are off, no power is dissipated.

When on, they act like (at best) a 13.5mΩ resistor. 5A * 13.5mΩ = 67.5mV. 67.5mV * 5A = 0.3375W. So if continuously on, each will waste 0.34W as heat. Since RθJA = 40°C/W, 0.34W * 40°C/W = 13.5°C rise in die temperature. Totally fine.

The trouble happens during the time the MOSFETs are switching on and off. If that could be impossibly fast, then it would not matter. But nothing can switch instantly, so there is a period of each transition where the MOSFET is in it's linear region (acting like a resistor, but with a value greater than 0.0135Ω.) In this period, the MOSFET drops lots of power as heat. The faster the PWM frequency, the more transitions there will be per second, so the more this will matter.

When high-speed, fast switching is required, often a MOSFET gate driver IC is used. The ATMega can probably source and sink 20mA from it's pins, which will be the dominating factor in how quickly the MOSFETs can switch. A gate driver IC however, is often designed for an amp or more of drive current, which greatly speeds up switching.

If you're not concerned about wasting power as heat due to slow switching, it may be possible to use the existing schematic with large copper pads on the MOSFETs. My gut instinct is that it will get hot though. If you want to conserve power and lower heat, gate driver ICs are the way to go.

\$\endgroup\$
3
  • \$\begingroup\$ Thanks apart from the interface problem the rest should work ? the 5A are absolutely max current in reality it is around 1A. \$\endgroup\$ – fred mustermann Mar 23 at 20:15
  • \$\begingroup\$ Always design to "max", never "typical" \$\endgroup\$ – Kyle B Mar 23 at 21:32
  • \$\begingroup\$ It should work. RS485 is differential signaling, but note that there are limits to how "differential" it really can be. This stems from the common-mode range of the receiving op-amps. If just that aspect doesn't work well, investigate that as a separate matter (most likely ground-related.) \$\endgroup\$ – rdtsc Mar 24 at 15:53
1
\$\begingroup\$

A minimum of 5cm^2 / W of heavy copper or 2x this otherwise with standard copper heatsink is required. this depends also on your maximum internal ambient and max desired reliability with junction temperatures and self heating of nearby e-caps.

Datasheets may give better formulae and recommended copper thickness and area, but this is my rule of thumb for Alumclad PCB or 2oz copper on both sides with thermal vias.

Thumbnail calculations at 4.5V indicate 400mW max could be cut in half if you had a 10V charge pump to bias the gates instead and thus reduce the temp rise to nearby parts.

Another solution would use +10V on V+ISO of U5-2 to pullup the gates with 100k and use open drain or collector for PWM. Then allow 1k leakage from ISO gnd to 0V for dc bias to work and still have reasonable isolation from serial port CMRR , if done properly balanced you wouldn’t need the isolation, perhaps. But that’s another question.

\$\endgroup\$
8
  • \$\begingroup\$ thans for you answer. And can rest work ? \$\endgroup\$ – fred mustermann Mar 23 at 19:39
  • \$\begingroup\$ thanks but VCC_ISO is only for RS485 and RS232. \$\endgroup\$ – fred mustermann Mar 23 at 19:51
  • \$\begingroup\$ IDK, there may be more problems alerted by others. STP on RS485 shudn’t need ISO . Watch out for SMPS CM noise. An RF shunt cap to earth ground or a BALun might work. Depends on MHz-m signal product and length of cable near EMI \$\endgroup\$ – Tony Stewart EE75 Mar 23 at 19:51
  • \$\begingroup\$ Thanks but what is STP. FP1 and FB2 are reducing the noise .In the datasheet is a typical aplication circuit wich I am using. \$\endgroup\$ – fred mustermann Mar 23 at 19:57
  • \$\begingroup\$ If you don’t know what STP cables are, you may have a CMRR problem \$\endgroup\$ – Tony Stewart EE75 Mar 23 at 20:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.