LM2672 Fixed 5V Switching Regulator - Input Capacitor Requirements

[j777]

Hi,

I'm using an LM2672 Fixed 5V (up to 1A load) regulator in a project. The regulator has specific requirements for the input capacitor ie low ESR, RMS current rating >= 1/2 DC load current, at least 15uF if solid tantalum is used. The input voltage for the board is 12V (from an external ac adapter) and is brought in onto a power plane. The 5V regulator shares this power plane with 8 external proximity sensors, 9 solenoids, and 1 photoeye. Would two 1 ohm resisters in series with the regulator's connection to the power plane provide some isolation from the other devices so that the input capacitors are only used for the regulator? Or is this something I shouldn't even be worrying about?

See the attachment for a better idea of what I'm trying to describe.

To be honest with you I saw something similar to this done before and I thought "hmm...I wonder if that's why they did it..." and so I figured I'd ask the experts.

 

[berkeman]

Where did you see this used before? Can you post the exact circuit that used it? A 1 Ohm resistance and 15uF gives a corner frequency of around 66kHz, which is near the switching frequency of the 2672, I believe. If there were a lot of noise on the 12V input near the switching frequency or a related frequency, then maybe filtering would help the stability of the switcher...

 

[j777]

Thanks for the reply berkeman.

It's done this way on a board that I had engineered by another company. Unfortunitely I never bothered getting schematics from them so I don't have them to post. The updated attachment shows the exact circuit (the only thing I don't know for sure are the values of the 4 input capacitors).

 

[j777]

Actually, the switching frequency of the LM2672 is 260kHz.

This is really a matter of curiosity so don't waste too much of your time on it. My main concern is that the input capacitors for the LM2672 in my design (which does not include those 1 Ohm resistors) won't be adequate for the stability of the switcher because of all the external sensors and solenoids attached to the same power plane. Is this a valid concern and if so is the solution to simply add a couple more capacitors to the plane?

 

[berkeman]

I guess if a 12V solenoid fires at the same instant that the DC-DC closes its switch, then the double current demand might droop the input voltage to the switcher and cause a brief regulation problem. Since the noise sources are not continuous, they shouldn't cause a stability problem, just perhaps an output ripple problem at each solenoid closure. I'd try it out and bang on the solenoids and watch the 5V output to see if there is unaceptable feed-forward of the input glitches. I doubt it would be a problem, though.

 

[j777]

I will definitely try that out. If there is unacceptable feed-forward of the input glitches what would be the best solution? Would simply adding a couple extra capacitors attached to the 12V plane solve such a problem?

 

[berkeman]

The solution would be something that would cut down the 12V droop response to current demands by the solenoids. Either filtering the supply for the DC-DC or solenoid or both, with low value resistors as you show.

 

[j777]

As you know I am quite the novice when it comes to this stuff. Would you mind explaining how the low value resistors would cut down the 12V droop? I can take some guesses at it but I would like to understand it completely.

 

[berkeman]

Let's say you put a 1 Ohm resistor in series with the caps for the DC-DC and a 1 Ohm resistor in series with a cap for each solenoid. Then each is pulling more of its current transients from their local cap(s), and less from other caps used by other drains on the 12V supply. The key is to lower the amplitude of the voltage transient seen on the DC-DC's input cap(s) due to other transient loads on the 12V supply. It's just like a set of RC filters, as long as you can stand big enough Rs and Cs to give you a useful cutoff frequency. Make any more sense?

 

[j777]

Yes, thank you. Clear and concise as usual.

The frequency in which the solenoids are turned on and therefore the potential transient frequency under normal usage is at most 2 Hz. The power consumption of each solenoid is .4W so each uses about 34mA. So the only time I would think a noticeable current transient might exist would be if all 9 solenoids are turned on at the same time.

Your probably right...there probably wouldn't be a current transient large enough to cause problems (especially not under normal usage). I guess if I really wanted to play it safe it wouldn't hurt to put a 1 Ohm resistor in series with the DC-DC's input caps. Do you think just doing this for the DC-DC and not the solenoids will make a difference? I don't want to add a cap for each solenoid unless I absolutely have to.

 

[berkeman]

I think a more usual way to handle it would be to have a cap on the 12V supply plane, to store local charge to handle the solenoid transients, and then your resistor(s) feeding another cap that is the input cap for the DC-DC. You can double up the caps if you need them to handle the ripple current specs.

 

[j777]

I already have a cap on the 12V supply plane so I just need to add 1 or 2 resisters to feed the DC-DC's input cap.

Thank you for taking the time to explain this stuff to me. I really appreciate it.

 

[j777]

There is just one additional thing that I just can't seem to figure out.

On this other board with the circuit as depicted in the attachment the resistors are 1 Ohm SMD 0805 chip resistors (the best power rating I've seen for these is .25W). How is it possible for them to survive a 1A load on the DC-DC?

 

[berkeman]

They definitely need to be sized for the current. Sounds like a design error in the previous design. If the node is running at full load for extended periods, the resistors will run hot, and their reliability (especially at high ambient temperatures) will not be good. At my company, the discovery of a design oversight like that would result in an EPR (engineering problem report), and corrective action.