Voltage Spikes when plugging in DCDC supply

[dreaquil]

Hi,
I'm using a 48V 40W medical grade switch mode supply from meanwell in one of my projects.  When I plug in the dc jack to my circuit the voltage spikes to anywhere between 80-100V (I added a 20mH inductor on the input to reduce this but it didn't stop them).  You can see it attached. Any ideas why or what I can do to stop it (I'm looking at TVS diodes ATM). The plot below was captured using a x10 probe that's why it's on the single digits.

[langwadt]

Hi,
I'm using a 48V 40W medical grade switch mode supply from meanwell in one of my projects.  When I plug in the dc jack to my circuit the voltage spikes to anywhere between 80-100V (I added a 20mH inductor on the input to reduce this but it didn't stop them).  You can see it attached. Any ideas why or what I can do to stop it (I'm looking at TVS diodes ATM). The plot below was captured using a x10 probe that's why it's on the single digits.

I'm guessing the inductance of the cable combined with peak current charging capacitors on the input of your box
the inductor probably makes it worse, you need slow the ramp down or clamp it

[Stuart Coyle]

The DC-DC converter can't have perfect step response. You need to slow the charging up of your circuit a bit and let the converter have time to settle gracefully. Using a TVS or something to clamp the spike is not going to stop all that ringing on the step response.

[dreaquil]

So did I make it worse by adding the inductor? I figured it would slow the step response? I'm rather confused.

[rx8pilot]

So did I make it worse by adding the inductor? I figured it would slow the step response? I'm rather confused.

Probably so. Suppressing the spike can be a challenge since it would require some delicate analysis of the SMPS, all the parasitics, and the load. There are a number of ways you could reduce the ring, but I cannot think of anything easier than a TVS of the appropriate size.

Without seeing the whole system, it's hard to say exactly what you need. Also, how you measure it can make a HUGE difference in what the scope shows and what is actually being fed into the circuit. Show us where and how you measuring this ringing. Use a low-inductance GND wire on the probe with a small .1uf cap across the input to your circuit.

[dreaquil]

Wow, I thought this would be a much easier fix but its seems to be a pretty detailed issue. I figured a medical grade supply meant A++ design. I don't currently have the circuit with me but I'll take a few photos of the setup as soon as I can. On the input directly before the Regulator I have a 10uF cap and a 0.1uF cap, a pptc resettable fuse, and a stp130A schottky for reverse voltage protection. This is the supply I'm using:

http://eu.mouser.com/ProductDetail/Mean-Well/GSM40A48-P1J/?qs=sGAEpiMZZMs2%252bVrH5rwl1qk8J1tlmiMtT42FG%2fn6p5c%3d


EDIT: It might be good to point out that I'm drawing a MAX of 100mA from the supply so about 4.8W

[PeteH]

Ceramic capacitors? Try adding series resistance with the bulk capacitor or use a hot swap part designed to soft start during hot plug events.

Hot plugging a live supply into a capacitive bus shuttfles a lot of current with only the line inductance to limit the current. Series inductance slows the current but also results in a huge undamped ring due to the currents involved. 48V hot plugging should really be handled by either a small passive softstsrt solution that can double as reverse polarity or a dedicated controller for 48V applications....

Edit... Series resistance on the order of 300m to 5 ohms....

Or a series line resistor since you're not doing much current. Like 2 to 10 ohms (small efficiency hit... But its insignificant ... Normally the bulk dampening capacitor is 3 to 5 times the higher frequency bypass ... And the series R can be calculated for optimal damping)

[PeteH]

In the past I've always controlled the inrush to a CCA.... If you expose the front end of the board to the will of a product that can have an unknown rise time it's hard to cover your bases. Especially if you don't control or don't have visibility into the output stage of the supply.

I've always done the minimum in terms of EFT and ESD (smaller capacitance and damping with a TVS and RevPOL) followed by a soft start stage with current limiting. (OVP, OCP, UVP). Assuming there's bulk capacitance or stability requirements (input imedance requirements) that needs to be isolated from excessive inrush.

[mikerj]

So did I make it worse by adding the inductor? I figured it would slow the step response? I'm rather confused.

Too much inductance on the input to switchers can actually make them unstable. The inductor is trying to maintain a constant current by raising the the input voltage as the current drops and vice versa, whereas a switcher is doing the opposite, i.e. reducing it's current consumption as the supply voltage rises and vice versa.

You need a soft start circuit to fix this.  IC's to drive external MOSFETS are available, or you can roll your own with care.

[dreaquil]

I'll try the small resistance on the input in order to make this prototype usable. I'll definitely consider a soft start circuit for next time.

@mikerj any IC comes to mind in particular?

@PeteH for the OVP do you use some sort of Crowbar circuit or is there a more elegant way to do it without blowing a cap?

[calzap]

Not sure how to interpret the original post.  Are you inputting DC to the Meanwell PSU and using it as a DC/DC converter?  If this is the case, where is the DC supply for the Meanwell coming from?

Or are you supplying mains AC to the Meanwell and  using its output to provide power to a separate DC to DC converter?  In this case, I'm curious as to  why an AC to final DC PSU couldn't be used.   There can be lots of good reasons.  What's the design of the DC/DC converter and its power rating and output voltage?  Does the DC/DC converter have isolated outputs, and you do see ringing on the output of the DC/DC converter?

Or are you supplying mains AC to the Meanwell (but are calling it a DC/DC converter anyway) and using its output for your project without an intervening DC/DC converter?

Was surprised to see in the specs that the Meanwell has an inrush current of 65A.

Mike in California

[PeteH]

OVP depends entirely on the characteristics of the supply.
For a regulated PS feeding this board... I would think a TVS and some kind of fusing elemnt would be enough as its not likely to overvoltage unless it's toast... Where the TVS should fail short or clamp and pop the fuse, saving the downstream stuff if selected well.

Can also use an IC that does softstart and basic OVP.
Like a LTC4367, for example

[rx8pilot]

You could try something simple like this - at least to get by.

Use an N-FET with a slow transit of the linear region to control the current during startup. Essentially the same thing as the Linear part LTC4397 (which I have used successfully), just much dumber and without all the protections.

Green trace is the gate voltage.
Blue is the load current. The load is R3 set at 10ohms as a wild guess.



The voltage divider would be set for the appropriate gate voltage, the cap C1 sets the rise time of the gate which will determine the rise time of the current. It's simple and dirty.

LT Spice attached

[Kalvin]

You could try something simple like this - at least to get by.

Essentially this is a simple shunt regulator which could use power-NPN with a 48V zener connected to base as well. Very simple solution and easy the get working.

[dreaquil]

My Mistake, I meant to write AC/DC converter. I still can't really wrap my head whats causing it  Is it the sudden introduction of a load causing the supply to try to load regulate and overshooting?

[mikerj]

@mikerj any IC comes to mind in particular?

The ones I've used are design for low voltage applications (i.e. 3.3v or 5v), but there are many devices suitable for high voltage operation.  "Hot Swap Controller" is the magic phrase to search on, here's a list of Linear Technologies devices.

[NiHaoMike]

What about a series inductor with a diode in parallel to limit the overshoot?

[dreaquil]

thanks for AN and other sources @AcHmed99

[dreaquil]

Just wanted to confirm with you guys that it was in fact the need for an electrolytic. Here's the input response with the cap!

Thanks for the help everyone!

[rx8pilot]

Wouldn't it just be cheaper particularly with LT stuff to just add damping in the form of a 5 cent 100uf or so cap. I have a 120uf polymer cap and a 4.7uH inductor at my input high Q but I put in a 470uf AL cap and just put in 0.16 oHm R in series, works like a charm everytime. If you have a low quality cap you can skip the series resistor all I had available was a Rubycon low esr so I had to add some R. I have the exact same setup as the OP thats why I posted it.

Nice, simple solution. Well done.