|
Bill, Thanks very much for the on-line design guide! Please note that I have not accomplished the impossible (i.e. designing a complete SMPS with no formal training). I am using a pair of IRF 2110's to drive the mosfets. The circuit includes two very low value current sense resistors followed by op-amps and comparators for current limiting (not yet connected - I keep an eye on the current with my oscilloscope). All the power supply has to do is charge up the bank of capacitors. Thus I have no problems with stability. Also, I am not concerned with ripple current/voltage. As expected, noise is a problem. Mike  Hi Bill, Now, I am a celebrity !!! Mike Bill, Thanks for your latest e-mails. You have started me thinking. Since our last correspondence, I may have discovered a way to change the specifications that would reduce the current draw and thus make it more favorable for a CUK design. However, I have travelled a long distance away from my plastic laminated guide and I won't be back home for over a week - and I am just itching to do some calculations to see if my hunch is correct. I have spent several hours trying to remember how I calculated the values of the capacitors and inductors based on the information in the design guide. I figured out D, D' and M per your example on the old portion of the web site. But after that I am stuck. Note: some variables are missing in the example. Please remember, I have NO formal training in electronics, especially SMPS. However, I have managed to build a full bridge switcher that, at the moment, puts out over 100 Amps. (At full power, it is designed for 300 Amps.) Thanks for your help. Mike [Reads from bottom up.] Mike: Unfortunately, a large bank of capacitors does not really have a particularly low high frequency impedance. This is explained on my website under the heading of bypassing. Too many design run aground on the idea that lots of output capacitance will absorb switching ripple without producing noise. Even your converter itself will suffer from switching noise and spikes reflected from the ESR and ESL of large arrays of capacitors. Their parasitics, wiring and trace inductance, contact resistance, and so on will prevent them from presenting a particularly low impedance at fs, the switching frequency, and beyond. A further, more fundamental limitation is imposed by the laws of Physics. Given a capacitor C1 charged to voltage V1, and a second uncharged capacitor C2: Both conservation of charge, and conservation of energy must be satisfied. If you work through the math, you'll find that much energy will (since it must) be lost in the charging process, no matter how it is done! [Compare E = 1/2 CVsquared and Q = CV for the two caps taken together both before and after the transfer!] This applies to charging with a switcher, just a surely as it does to charging through a resistor, or a linear regulator. In a shunt damped switcher, the duty cycle will be slewing as charging occurs, and the damping resistor will get hot! The same must be true of current mode programmed converters, though I don't know off hand where the heat would show up! (Probably the switch would be pulled out of saturation and burn up.) This is an outstanding problem in any switcher design where either line or load voltage changes significantly. [note to self: what happens in PFC Boost converters? one hates to think....] One of the first switching converters I encountered was Bob (now Dr.) Erickson's Coupled Inductor Cuk Converter switching amplifier presented at PowerCon 5. Dr. Cuk calculated an additional series loss resistance we introduced to linearize the D/D' transfer characteristic of the amplifier, with an eye toward reducing harmonic distortion. In retrospect, I suppose it helped the amplifier operate properly by providing a safe place for such losses to occur, since a switching amp is always slewing! Infinity, a manufacturer of high end audio amps, has been trying forever to perfect a Buck (or other) -based switching amp, but experiences chronic reliability problems. I suppose they improve efficiency (which linearizes a Buck,) only to overstress the rest of the topology, which is thereby forced to dissipate the energy difference! As for bridge converters, I avoid them. They are too complicated, and a high side FET is tricky to switch. High side IC drive chips are available, but suffer from high junction counts. I.e., they use a lot of transistors and diodes to perform the function. This reduces your reliability a lot. Unfortunately, it turns out that integrating components onto a chip does not reduce their failure rate much! The recent wave of multi-feature drive and control ICs are convenient for the engineer, and offer a wide array of "sexy" functions to the customer, but plunge the lifetime of the product. Believe it or not, much of Great Britain's phone system 9run by the British Post Office, I believe) was built here in Chicago! It is electromechanical, rather than electronic. It was designed to last 50 years, and is, one supposes, still going strong! In the mid 80s, GTE was building microprocessor based troubleshooting equipment to identify worn springs and relay contacts! Our MTBF for similar electronic telephony gear was only 30 years, which we were having trouble meeting.... Best regards, Bill ----- Original Message ----- From: MMiller Sent: Wednesday, October 02, 2002 1:31 PM To: billbehen@msn.com Subject: 500 Amps Bill, All I am trying to do with this application is transfer a charge from one bank of capacitors to another across an air gap. I really don't need to have a smooth output. So, it seems to me that I would be better off using a half or full bridge converter for this application. Any thoughts? Regards, Mike Miller Mike, Good that you happened on Col. McLyman's work! It was he who first figured out that winding resistance was crucial in power magnetics design. Until his Kg method was published, people had used older transformer design techniques that relied heavily on minimizing core loss. These, including Hannah curves and "area product" calculations had been derived for other applications (RF, analog telephony, computer core memories, and so on.) But for power magnetics, one must use low perm, low loss cores, and copper loss is really the dominant problem. I would be tempted to try 100KHz, myself. Today's FETs seem just as happy there as at 20KHz, since they switch in 100s of nanoseconds. At 20KHz, one gets stuck in the electrolytic trap, as you noticed! I am still a big fan of polypropylene film caps for the energy transfer cap. Though they are not cheap, they are simple in construction = high reliability = long life. I would be less worried about expense with them. Remember you are handling 2.5 KiloWatts, somewhat over 3 HP! That should make it worth it. TRW pioneered the technology many moons ago, then backed off. They couldn't get the polypropylene film thin enough to produce anything less than a 100 Volt part. That made them physically large, even though they are light weight. At the time, (~1980) switchers ran at 20KHz, and sufficient capacity was hard to get with film caps. Now, things are different, and their ability to handle high currents makes them attractive! Switcher grade electrolytics need heavy metallization to handle high current, which makes them weigh a lot! Further, the lower values needed at higher frequencies have big ESR, while film caps have next to none! Regards, Bill ----- Original Message ----- From: MMiller Sent: Tuesday, October 01, 2002 3:12 PM To: billbehen@msn.com Subject: ETC Bill, Thanks very much for your response. Erie Technological Products is now part of Murata. www.murata.com Yes, 500 Amps IS a lot ! I have been devouring the information on the Magnetics, Inc web site. The information on the web site is duplicated in Col. McLyman's software. Although the software would be convenient, I should be able to do the analysis with a calculator. In my last e-mail I was really hoping that you could suggest what type of capacitor I should be considering. At 100 kHz, I calculated a need for ~400 uF cuk capacitor. Film capacitors are rather pricey for this amount of capacitance. On the other hand the ESR of capacitors such as OSCon will create large voltage drops due to the higher ESR. At lower Fs (say 20 kHz) I will have to use a core material that can handle high flux density so that I can use a standard size core. Of course, I will have to use a much larger capacitor. What type of capacitor should I consider for > 1000 uF ETC? Note: This is the first time in about 7 years of self study in electronics that I have had someone to answer my questions...what a relief this is! So, thanks again. Regards, Michael Miller Mike: Some more thoughts regarding your questions: The 1:300 ratio is a good number. The determination of that value took me about 20 years, believe it or not! The reason for that particular number is therefore rather elaborate, and I have no simple formula. It ensures that the magnetizing inductance is not so low that excess magnetizing ripple current flows, adding current stress to the switch. Further, too large an Lm means that the transformer will be prone to saturation under transient, unless it is physically very large. The argument is similar to that which dictates that a high perm core is not very useful as a power inductor core, and must be gapped to handle DC. You will not get good results if you stray more than 2X or (1/2)X from this value. As for the rest of the design, just design the transformer as you ordinarily would! Then introduce your gap. It sounds like this approach would be ideal for your application, since you can use about any type of core! Regards, Bill ----- Original Message ----- From: MMiller Sent: Monday, September 30, 2002 4:31 PM To: billbehen@msn.com Subject: Redcap capacitor and other questions. Bill, What is a Redcap? Is this a ceramic capacitor? Does it have special characteristics? I have a second application for the boostbuck converter topology that involves a lot of current (500 amps) at low voltage (5.0 volts) that must be isolated with a 1:1 gapped transformer. (Actually, the transformer will be designed to be separable with the primary winding on one half and the secondary on the other). Should I be considering a high frequency or low frequency design? Can you give me a starting point for cuk capacitor and transformer core selection for this unusual application? Is the 1:300 gap to core length ratio that you recommend necessary (can it be lower)? What should the inductance of the transformer be in relation to the input and output inductors. Thanks for your help and best regards, Mike Miller Background | Tips | Engineering vs. Science | | Isolation of the Cuk Converter | Historical Perspective | | Welcome | 50K! | | Return Home | Safety First | Lesson in Ethics | The Consequences of Bad Design | Real World Applications | Deals | e-mail me! | |
||