Selecting SMPS Based On Bandwidth Noise Limits

[jmeyer84]

Hello-

I need help selecting a SMPS to power a single known component. The power profile expected is attached.

Requirements:
-10 VDC to 30 VDC input
-5 VDC @ 2A output
-Vripple less than 40mVpp
-Voltage droop for "Function 1" in the attachment cannot be greater than 0.2 Volts
-Noise should be less than the limits in the following profile:
---100mVpp from 0 to 50 kHz bandwidth
---5mVpp at 1 MHz measured in 50 kHz bandwidth
---10mVpp at 1 MHz measured in 1 MHz bandwidth
---5mVpp above 5MHz measure in 1 MHz bandwidth

I'm trying to avoid an LDO-style power supply due to heat, efficiency, and heatsink size concerns. I apologize for my ignorance, but I am really lost in the noise bandwidth area of the requirements.

Are SMPS's even capable of meeting the requirements listed above?

Is droop a concern for SMPS's based on the power profile I have attached? Do higher or lower frequencies help with droop?

Should I be looking for a high/low frequency SMPS? PWM? ZCS/ZVS?

Any assistance would be much appreciated! Thank you.

 

[jmeyer84]

I've been doing some hunting, and it looks like National Semiconductor's LMZ14203 would be appropriate for the application, but I'm still uncertain about the noise levels in the bandwidth ranges....

http://www.national.com/pf/LM/LMZ14203.html

 

[itsmoked]

You should be looking for an OffTheShelf solution. There are dozens of companies that offer these. Shop your specs with them. Fix your list because I doubt you really mean a supply voltage of "-10V to +30V" which is what you've written.

Look for a "10W DC to DC". "Switcher" - in the industry - generally refers to an AC to DC supply.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[jmeyer84]

Yes. Hyphens were a bad choice for my list. Should be:

10 VDC to 30 VDC input
5 VDC @ 2A output

Just to clear things up: I'm not looking for a efficient power supply for a surface mount application similar to the National link provided above. I'm very concerned about the noise related issues.

 

[zappedagain]

I'm a bit rusty on these as it has been a few years since I've looked into designing my own. Here is what looks significant to me:

Your noise specs are leaning toward a low frequency SMPS, as there is much higher margin at the low frequencies.

What are the dI/dt rates on your current waveform? Those look pretty fast, so I expect that response may push you to a higher frequency SMPS.

V = L * dI/dt gives you the inductance that you need to meet your input voltage requirements. The size of that inductor drives your operating frequency. i = C * dV/dt tells you how much capacitance you'll need to maintain your droop. I believe you'll need to run at a frequency much below 1/(2*PI*sqrt(L*C)) to maintain stability.

There are some good references available on SMPS design; they aren't trivial. If all this is confusing (and this is only the tip of the iceberg) then you'll definitely want to be looking for an off-the-shelf solution as Keith mentioned.

Rusty John

 

[jmeyer84]

Thanks for the responses. I've done a little homework of the weekend and determined that a DC-DC converter won't get me where I want with the requirements listed above. The plan is to follow up the SMPS with a LDO. It won't be as efficient as I was hoping for, but I will be able to get a wide input range with a low noise profile.

 

[zappedagain]

Ahh, the old SMPS followed by a LDO trick! That is typically a good solution for most low noise applications.

John D

 

[Warpspeed]

Since all switching power supplies switch, the output always need some type of ripple and noise output filtering. That is inescapable.

How well you decide to do this, or how low you wish to go with output noise, is just a case of building a suitable output filter (with an appropriate physical layout) to do the job. And it does not need to all be lumped in one stage either.

Voltage droop should not be a concern, as the voltage gain of the feedback at zero frequency can be made enormous. Even remote load sensing can be used if desired to eliminate any voltage drop in the output wiring.