1
\$\begingroup\$

I have a PWM load which draws its current from a regulated DC supply. The supply and load will be separated by some distance, and I'm worried that the rapid changes in current through the current carrying wire between the supply and load will create excessive EMI.

The first idea was to have this:

schematic

simulate this circuit – Schematic created using CircuitLab

But the problem is on start-up the current drawn through R2 and C1 would be too great. Ideally the peak current drawn from the power supply would never be greater than 3A.

I'm looking for a simple circuit that just uses no more active components than a couple of BJTs to reduce rapid current changes and the initial current surge, while ideally not reducing the voltage to the load by more than about 2V. Any advice would be very helpful.

Thanks

\$\endgroup\$
  • 1
    \$\begingroup\$ Looping wire through a ferrite ring should help reduce EMI by limiting the rise and fall times of any pulse current. \$\endgroup\$ – JIm Dearden Oct 7 '13 at 16:21
  • \$\begingroup\$ what problem does startup current cause -blown fuses? \$\endgroup\$ – Andy aka Oct 7 '13 at 16:27
  • \$\begingroup\$ Thank you. I will have a microprocessor with a BJT doing the PWM on the load's side, so should I worry about the inductance creating damaging voltage spikes? (And perhaps fix it with diode/zener clamps?) \$\endgroup\$ – CL22 Oct 7 '13 at 16:27
  • \$\begingroup\$ I don't know what the power supply's circuit is, so don't know if the startup current could damage it \$\endgroup\$ – CL22 Oct 7 '13 at 16:28
2
\$\begingroup\$

You're talking about two separate issues, the EMI and inrush currents. EMI filtering won't help much with inrush current.

As far as inrush, it looks like you need to slow the charging of that honking big (10mF) cap. There are two common approaches:

  • Thermistor This is basically a resistor with a negative temperature coeffcient. It starts out with high resistance like 10 or 20 Ohms, then as current flows in it, will warm up and end up with resistance of something like 0.2 to 0.5 Ohms. It is a pretty simple approach. Problem is it relies on thermal time constants, so if system is operating and is turned off and then turned back on again over a time period which is long for the electrical time constants but short for the thermal time constants (usually 15 to 60 sec or so) the system will see a full inrush.

  • Active inrush limiter. These usually use a P or N channel FET to limit current during start up. These are made by many manufacturers like Linear Tech, or TI. Here is an example from Linear Tech, the LTC4356 . There are so many choices, with different features, that it makes no sense to go into it further here.

For EMI you will need carefully designed filters that, yes as JIm Dearden implies, will involve inductors. Every subsystem within your system will need an appropriate power and interconnect filter. All the filters will need to be designed so that they don't have impedance mismatches that could cause oscillation and interactions inside the system. What that means, and how involved it is, really depends on a lot of interrelated issues that will be system specific.

\$\endgroup\$
1
\$\begingroup\$

It sounds like you have everything at the load end to help you make startup current a minor issue. But, I will say that if you don't know what the power supply is capable of delivering as a surge then how can you help yourself properly let alone anyone on EE helping you. OK I've had my nag!

Putting that to one side why don't you program the micro to "soft-start" the load over a second or so. If, for whatever reason the load needs an instant start then use the micro to feed energy into a big capacitor, again using soft-start PWM techniques then, use an IO pin to activate a MOSFET or a relay to switch the load across the charged capacitor.

\$\endgroup\$
  • \$\begingroup\$ Thank you - it was a hard choice deciding which approach to take, I would have given +2 for the soft start using the micro if I could! \$\endgroup\$ – CL22 Oct 8 '13 at 11:40

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.