high volt DC-DC Conversion


Joined Apr 24, 2007
You'll probably want to use a step-down (Buck) switchmode (SMPS, aka switching, aka switcher) power supply circuit, of some sort, to avoid wasting a lot of power and generating a lot of heat. Essentially, you'll be making an oscillator that runs on your 154VDC, i.e. turning it into AC, and then stepping-down (or maybe pulse-width modulating) that AC voltage before rectifying it back into a lower-voltage DC.

Look at both national.com and linear.com, rather thoroughly. An IC-based SMPS can be made with as few as five components, although yours might need a few more than that, for the high current (and high voltage). National has an on-line automatic SMPS designer, and Linear has one included in their free LTspice (SwitcherCad III) software (highly recommended).

You might need to use a lower-current-rated SMPS controller IC/chip with an external pass transistor, in order to get the high current that you need. Those above-mentioned websites' SMPS chips' datasheets, and the many SMPS application notes, will have lots of example circuits. (You could also do the same thing with all discrete components, instead of using one of their ICs. [That might even be necessary, if there are no suitable SMPS ICs for your application.])

Depending on how clean the final supply needs to be, you might also want to use a CLC "Pi"-layout post-filter, followed by a linear voltage regulator IC. You can use a lower-current-rated linear regulator with an external pass transistor to handle the higher current. Datasheets for e.g. LM338 and LM317, at national.com, should have simple example circuits for doing just that. Linear regulators can be operated at high voltages, as long as the v_in-v_out difference is less than their max voltage rating. You'll need to design the SMPS and CLC filter combo to give you something that's at least a few volts higher than you need at the final output, to give the linear regulator some headroom. Make sure that the voltage at the regulator's input will not have sagged too far, when the maximum output current is being drawn. i.e. You might need to design the SMPS to be able to deliver more current than you think you need. (Also note that using an adjustable type of linear regulator, with a bypass capacitor for the adjust pin (see datasheets), will result in the least output noise and ripple.)

All of the parts you need should be available through mouser.com and/or digikey.com . The J.W. Miller high-current toroidal inductors seem to work well, for SMPS circuits, and are very cheap. (If you need a transformer in there somewhere, hopefully you can get by with an off-the-shelf model.) Note that using a higher-frequency switchmode circuit will result in smaller and cheaper inductors for the design (but might also make wiring or PCB layout more critical).

Also note that PCB (or wiring) layout can be extremely critical, in SMPS circuits, especially in terms of output noise (and also radiated noise). Both national.com and linear.com have application notes about that, which you should try to read. The basic idea is to make the paths that carry large changing currents as short and fat as possible, and to not let any other currents share those paths. Use what is called a "star ground" layout, where ground returns are not shared. And you'll want to keep any quiet or small signal and ground conductors away from all noisy or large-current conductors and components. ALSO, the enclosed geometric areas of all loops that have large dynamic currents should be kept as small as possible, to minimize radiated fields. All other loops' enclosed areas should also be as small as possible, to minimize currents induced in them by any radiated dynamic fields from other loops, or from external sources.

One thing that they might not emphasize in example circuits is that you will probably need to place a simple series RC snubber in parallel with the SMPS's main output diode, to damp out any high-frequency ringing or outright oscillation, which might arise at the moments when switching occurs. With the use of an oscilloscope, you could just pick an R value that won't waste too much current (say 1K Ohms) and then experiment with different C values. If you have no scope, definitely download LTspice from linear.com, and design the snubber with that. If you add the parasitic resistances to your capacitors and inductors, and add about 0.3 pF in parallel with each resistor, and maybe even add inductors with series resistances to model wires or PCB traces (using a rough guess of 25 nH and 0.001 Ohms per inch), then you should get a reasonably-accurate simulation of any ringing, etc.

If simulating with Transient Analysis runs in LTspice, you might also want to adjust the ESR (Equivalent Series Resistance) of the capacitors to be valid at the frequency of interest, to get the most-accurate results. Caps' datasheets may give ESR at one frequency. If you're lucky, they'll also give ESR or tan(delta) at another frequency. They you can use tan_delta(freq) = 2 * Pi * C * freq * ESR(freq) and assume a linear ESR change and calculate ESR for any frequency.

I have some downloadable LT-spice circuits at http://www.fullnet.com/~tomg/gooteesp.htm , which include automatic calculation of capacitors' ESR based on a freq parameter, for Nichicon UHE-series electrolytic capacitors. And Kemet has an ESR calculator/simulator for ceramic and other caps on their website, although modeling those probably won't be as critical as for the electrolytics. My website (above) also includes a spice model for a simple power transformer, which automatically calculates the model parameters from simple measurements, the (possibly dangerous) procedure for which is also included in the spice model, as comment text on the schematic.

If you need to make a PCB for it, and it's a one-off or small quantity job, take a look at http://www.fullnet.com/~tomg/gooteepc.htm . Otherwise, I like http://www.4pcb.com .

Good luck!

Tom Gootee



Joined Nov 23, 2007
Unless you need the 85 and 154V to be isolated from each other, a buck converter would be the simplest, as mentioned by gootee. See http://en.wikipedia.org/wiki/Buck_converter as a starting point.

At 9A you are looking at a pretty hefty inductor. What are the min & max currents? How accurate must the 85V be & how much ripple is acceptable?

You will need to consider mode of operation, how much ripple is ok, efficiency, cooling, safety, interference.

Small buck converters are easy but at the power, voltage & current levels you are suggesting, you need to do it properly with appropriate mathematical analysis or you will have some very bad experiences -magic smoke escaping etc. If you can find a power electronics book it will go through all this stuff in detail.

What are you wanting to use this for? I have a lot of power electronics experience (and textbooks) but the generic problem does need the specific details & numbers to get more than a generic answer :)