dspcan said:
then, gootee, do you think the circuit will be helpful for reducing the ripple of the output when using the switcher instead of the linear regulator? (I never use the switcher, I don't know)
Hi dspcan,
That type of "linear regulator with opamp ripple-cancellation" circuit topology should be able to help, if placed after an SMPS output. But I suppose it might need to be optimized for dealing with the fast spikes from some SMPS types. Maybe you can try simulating that, with LTspice, which also has a menu option (File-->Switch Selector Guide) that will automagically design an SMPS for you. And then you can modify the SMPS circuit and run more simulations, including the cancellation circuit. LTspice also includes many faster opamp models, e.g. LT1363/LT1364 with slew-rate of 1000v/us.
You could also download the OPA541 (E version) power chipamp's spice model, from Texas Instruments, at
http://www.ti.com , and try using it as a dynamic load in your simulations. I usually do that, with the opa541 pushing large-amplitude square waves into a low-value resistive load, to "torture test" a power supply design.
And, if you add realistic wire or pcb trace impedances to the supply lines and ground returns, etc, you can see for yourself why a "star ground" topology is essential. (You can download my LTspice files that include star-ground-testing setups, at
**broken link removed** , and copy and paste what you need, into your own LTspice schematics. That will make it easy to share and un-share ground-return-current paths, to experiment.) [By the way, always remember to include at least the parasitic series resistance, in capacitor and inductor models. In LTspice, you can just right-click on any L or C component, in the schematic, to set the parasitics' values.]
But, the first thing I would probably try, for a noisy SMPS output, is an LC or CLC lowpass filter. i.e. Place a high-current inductor (probably 10uH to 100uH) in series with the SMPS output and connect a large-ish electrolytic capacitor (maybe try 1000uF to 4700uF, as a start) to ground, after the inductor. You can also connect a similar capacitor to ground _before_ the inductor, to have a CLC lowpass topology (and that capacitor might already be there, in the SMPS). For my 60 kHz boost-mode SMPS, based on linear.com's LT1270A switcher chip, 2200uF/10uH/2200uF worked well-enough.
For the inductor, you could use, for example, one of the Bourns/J.W. Miller high-current toroidal models (2100, 2200, or 2300 series, e.g. Mouser.com # 2301-V-RC, 10 uH, 20 Amps, $2.78 qty 1). And, in my case, the 2200 uF 50V Nichicon UHE-series low-ESR caps worked well, with a 10uH inductor. e.g. Mouser.com # 647-UHE1H222MHD6, $1.81 qty 1.
Of course, if you are DESIGNING the SMPS, too, then you should also make sure that you have used snubber networks, if needed. In my simple boost-mode SMPS, I definitely needed a series RC snubber across the main diode after the switch, to get rid of some serious high-frequency ringing. Ltspice showed the ringing, very well, and was also used to help optimize the R and C values for the snubber (R and C can be calculated. But, usually, you can just pick an R value that won't waste too much power and then change the C value to get the best waveform.)
For "the best of both worlds", you can also add a linear regulator circuit, after the CLC filter. I have used an LD1084 (an LT1084 "knockoff") 5-Amp adjustable linear regulator, with a small electrolytic cap bypassing the adjust pin to ground and a 150 uF electro from output to ground (and also with all of the other components recommended in the datasheet), and, in simulations at least, got the switching noise + 120 Hz ripple down to about 15 uV p-p (Reality is always a little different. But it is very quiet.).
Sorry to have blathered-on, for so long, about all of that.
- Tom Gootee
**broken link removed**