Which Power Supply Design Is Best for Low Noise Applications?

AI Thread Summary
The discussion focuses on designing a power supply for a project requiring a regulated 5V and a low noise 12V output, particularly for an RFID reader. Two designs are considered: Design 1 utilizes a 15V linear supply with a low noise linear regulator for the 12V output, while Design 2 employs a simpler 12V linear supply with filtering for low noise. Design 1 is favored for its potential to better clean up noise from the switching regulator, emphasizing the importance of careful PCB layout to avoid shared impedance in grounding. The conversation also highlights the need for appropriate output capacitance to manage load transients effectively, suggesting that the regulator primarily handles medium-frequency transients while output capacitors address higher frequency demands. Overall, careful design and component selection are crucial for achieving the desired performance in low noise applications.
j777
Messages
148
Reaction score
0
Hello,

I'm trying to decide how to design the power scheme that will supply the various voltages within my project. The two big requirements are that there is a single main supply, regulated 5V @ 1A (typ. load .5A), and extremely low noise regulated 12V @ 1A (typ. load .4A). The low noise 12V is for an RFID reader that reads 134.2kHz transponders.

Below are the two designs that I'm contemplating.

Design 1: Main supply is regulated 15V linear supply;5V switching regulator (260kHz switching freq.) with ferrite beads on both input and output to reduce conducted noise;12V low noise linear regulator (LT1763) with filtering added.

Design 2: Main supply is regulated 12V linear supply;5V switching regulator (260kHz switching freq.) with ferrite beads on both input and output to reduce conducted noise;filter 12V main supply to provide extremely low noise 12V

Design 2 is the least complex but the second linear regulator in design 1 should help to further clean up the 15V main supply and attenuate any noise caused by the 5V switching regulator.

Please offer your thoughts on these two designs and which one you might choose. Any and all thoughts/recommendations are appreciated.

Thanks.
 
Engineering news on Phys.org
In my experience, you definitely need a separate regulated rail for the RF analog section, and you need to be very, very careful how you floorplan and lay out your PCB. You will want to bring the power input (15Vdc) in near the the middle of the board floorplan, and have the RF regulator feed the RF circuitry in one direction, and the other 5V regulator feed the other circuitry in the other direction. You do NOT want to have any shared impedance in the grounding of the two circuits, and you might consider even cutting the two planes apart, except for the one bridge area where the signals cross from one section to the other (and yes, filter those connections).

So I'm saying Design 1 is the better option, plus you need to really think about your floorplan carefully, and be willing to spin the PCB design if you find that the first try is still not giving you the sensitivity that you want. I've seen 10's of dB lost in sensitivity by not being very careful about this.
 
Thanks for the help berkeman.

I will split the 2 planes and only bring them together right at the power input. There aren't any signals that need to cross between the 2 planes so I don't have to worry about filtering them (I will however do filtering on the input and output of the 5V and 12V regulators). The RF analog section will only contain the 12Vdc supply for the RFID reader.

I really appreciated your input and look forward to testing this design out.
 
The 12Vdc RF supply needs to handle load transients of .5A and have only a small deviation from the nominal output voltage under these conditions. What is the best way to determine how much output capacitance is required to keep the deviation within a desired limit? Say for example 50mV...
 
j777 said:
The 12Vdc RF supply needs to handle load transients of .5A and have only a small deviation from the nominal output voltage under these conditions. What is the best way to determine how much output capacitance is required to keep the deviation within a desired limit? Say for example 50mV...

I don't think you'll be supplying that extra 0.5A transient from the regulator's output capacitor -- it will be mostly through the regulator. The smoothing of medium-frequency transients in a linear regulator situation is mostly handled by the regulator itself. Higher frequency transients (like if there is a step in the output current demand) will be smoothed by the output capacitors. The value of the capacitor is important there, but also its ESR (equivilant series resistance) value. If you know the dI/dt value of your current steps, then you can figure out the value of the output capacitor and the ESR required. Check the linear regulator's datasheet to find its transient response characteristics.
 
The only information that I could find in the datasheet is depicted in the attached image. Would you mind giving an example of how you would calculate the value of an output capacitor given dI/dt?
 

Attachments

  • trans_response.JPG
    trans_response.JPG
    43 KB · Views: 490
Those graphs are already showing triensents around 50mV, so the capacitances they list are probably fine for you to start with. Those transients are also almost full scale (off to on), versus starting at 0.5A and having a transient to 1A. An output current transient will generally be handled better by the regulator if it is already conducting a pretty good current before the set up. Be sure to use low ESR caps like the tantalums they mention in the figures.
 
It will help to keep your 12v sub regulator on the same ground plane as your module and as close to the power connections as possible.
Also, ESR is subject to parallel resistance rules.
All else being equal two 50uf caps will work better than one 100uf cap.
 
Back
Top