Switch Mode Power Supply

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.


Paid Member
2005-01-09 2:32 am
I have a 19VDC3A Laptop Power Supply powering a JFET BOZ preamp and it sounds very nice. For the hell of it, I hooked up two 9V batteries in series and it sounded quite a bit better. Clearer and deeper is all I can say. I do not think it sounds better due to the difference of 1V.

Now the question is this. Can I make a dull sounding powersupply like this Laptop SMPS better by just adding filters? Could it be all about better filtering or is there something more to it?

I could easily build a supply with a tranny, rectifiers and CRC filter, but I have a quite a few of these laptop power supplies (eight to ten of them) and wouldn't mind using them.

I have parts to make a CRCRC arrangement with C=2200uF and R=.5 ohm on this laptop supply, and before doing it decided to ask here first.

I also have connected two of these same laptop PS to get dual rails and a center tap. Works great as a 19v - 0 - 19v at 3A supply. I have been thinking of using it to power a Zen V5, until I tried the battery thingy mentioned above. I also have enough parts to further filter the dual rails for this supply with a CRCRC arrangement also.

Am I kidding myself on thinking these supplies are usable in Audio?
the SMPS power supply probably has some sharp transients on its output. A trick used to clean these up is to use a simple linear regulator on the output of the power supply.

Better would be to put in a Jung Super Regulator --- Old Colony has the circuit boards -- I have used these with several opamps including the AD797, AD825 and LT1115.

Pass has several regulator circuits on his (their) website using the left-over HEXFETs from the matching process. One of these would also serve you well in augmenting the PSRR of the BOZ.
Jack's post-regulator suggestion is known to be a very good method, for cleaning-up SMPS outputs.

I suggest also placing an LC or CLC lowpass filter between the SMPS and the post-regulator. That should be able to knock down the high frequency switching spikes extremely well, probably dividing them by something on the order of 1000x, depending on your SMPS's switching frequency.

There is probably an output C in the SMPS, already. So adding an LC would make a CLC filter.

I have had good results using a Bourns/J.W.Miller toroidal inductor, such as the 2100HT-100V-RC (or 100H instead of 100V, if you want horizontal mount version). That one is rated at 9.9 amps, .012 Ohms, 0.86-inch max diameter, and is $2.44 ea qty 1 or $1.69 ea qty 10 at mouser.com .

You would put the inductor in series with the SMPS output and add a capacitor to ground after the inductor, probably 1000 uF or 2200 uF. But you can adjust the L and C values.

There are also higher-value inductors in the 2100-series, but with smaller max current ratings. They also have 2200 and 2300 series, with larger current ratings. e.g. the 2300 series 10 uH is rated at 20 A and has DC resistance of .005 Ohms and 1.28-inch max diameter. All three series have inductors up to 1000 uH.

Last time I checked, both mouser.com and digikey.com had all of them.

If you are dealing with relatively-small power supply currents, there are lots of other (smaller) types of inductors that should work well, too. You should be able to use a larger inductance value and a smaller capacitance value, also.
A big problem with laptop supplies is that the usually have lots of common mode noise and coupling to the mains, besides the spikey output. With a simple, single ended LC filter this common mode stuff cannot be filtered out, you need symmetrical filters -- coils in each leg of the supply. These coils must be big, but low parallel capacitance types (that is, high self resonant frequency), and the shunt caps must be low ESR/ESL to be effective in filtering the spikes -- I would use low valued (10nF...1uF) ceramic types, nonlinear Y5V or similar. Those have only like 20% rated capacitance at rated voltage, though. If the spikes are the real problem (bigger that the common-mode hash), then a two stage filter could be considered (LCLC, symmetrical of course). RCRC (not: CRCRC! Don't load the SPMS with a cap, especially not with a good one) would surely help also (with low valued, high-Q caps, not big 'lytics) a bit, but L-based filters are way better -- if you stay well below the resonant point of the coils.

The typical linear series reg (78xx etc) is quite transparent for HF spikes, and totallly non-existent for the common mode stuff. Only series L in both lines (hot and "GND") does help against this.

- Klaus
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.