Skip navigation
Electronic Design

Bob's Mailbox

Dear Bob: Recently I had problems paralleling MOSFETs in a power current source for an electronic load that has to absorb the power generated by four strings of two series-connected BP585 PV modules: (MaxPower_worst-case) = 4(36 V 5 A), about 720 W. The MOSFETs operate in their linear regime, but almost all parts I can buy from my company suppliers are optimized to switching applications. (That's what their maker says, but that does not mean they cannot be used for linear uses. /rap) They have nice, low RDS(ON) with positive temperature coefficients. But I have read that in linear operation, the critical parameter is the negative temperature coefficient of the VGS that causes positive feedback (I found words like electro-thermal instability, hot spots, activation of parasitic BJT, nice drain-source short circuit) and eventually the destruction of one of the MOSFETs. I found the US5057719 patent and several documents from IXYS, but I would like to know your opinion about how to deal with the problem. (We have no curve tracer to match devices. Sorry.) (You could run some parts at 2 A and 10 V and just read the VGS to see how many you could match into bins. If you always have more than 5 V, then your solution of ballasting is pretty good. /rap) At the moment I decided to use old parts in old TO-3 packages, so several IRF150s with a resistor (positive temperature coefficient) connected in series at the source are my actual choice.
Roberto Giral
Pease: Sounds good to me. You could even use some Darlingtons! WIth emitter ballast resistors. Parallel four of those. Just watch out for the SOA curves, right?

Bob: We have an oil eternal candle at church. It holds about two quarts of oil. The oil vessel is short-cone-shaped on top and longer-cone-shaped on bottom. It has about a 1-in. hole at the top where the wick assembly goes where you can fill the reservoir. Filling it is a problem, because you can't see down the fill hole, the oil is very silky and makes no gurgling noises, and the cone shape causes the fluid level to rise very fast near the full point. We've had lots of overflows and spills. Currently, we use a soda straw with a finger over the end to capture the fluid depth, but you have to measure frequently, which is a pain when you are up on a ladder and the thing is swinging from a chain and you are trying to pour candle oil. I'm working on a fluid-level sensor to help. I'm using an earphone glued to an appropriately machined small metal cylinder in turn glued to a 3-in. long, 0.125-in. diameter glass tube. Driving the earphone with a 555 timer, I can tune to the resonant frequency of the glass tube (about 1388 Hz). This gives the loudest output. When the end of the tube contacts liquid, resonance is destroyed, and you get a noticeable drop in volume. It works in the lab (my basement...). I've started packaging the electronics up using a PIC chip instead of my breadboard 555 system. The new design has buttons that let me alter the tuning up and down, store a frequency, and reset to the default (supposing that barometric or temperature changes may make one want to adjust things—I haven't studied whether that is a legitimate concern). (I would not want to guess that the glass rod's resonance would change around a lot. /rap) As I am designing the PIC program, I started thinking it would be really cool if I could make this thing self-tuning. I would have to detect resonance based on some current/voltage relationship with the driven element. I have a 51-? resistor in series with the earphone to reduce the volume to a reasonable level. Measuring the RMS voltage across the resistor does not show me the resonance point. Lower frequencies tend to have higher RMS voltages, and higher frequencies lower RMS voltages (using an old Beckman digital VOM to measure with). I suspect the mechanical system overall has a very low Q factor. Nevertheless, is there a way to condition the signal across the series resistor to tell when I am at resonance?
Steve Fraser
Pease: If it was an analog system, I would ask you to show me your schematic. Why wouldn't you set up some ac signal with a coax capacitor? Or, parallel twin-line. As the coax fills with oil, the capacitance would change in an analog way (just as you hear the pitch change when you pour fluid into a jug). Maybe you should hold on to that idea, in case your resonance doesn't work so well. This could be a matter of analog sound volume or of oscillating frequency. Use a micropower op amp as an R-C oscillator.

Hello: I saw your article on GetHuman (, ED Online 14414) and thought you might be interested in a project we have been working on called Our mission is to help users skip phone trees and connect with a real human on the customer-support phone lines at many companies throughout the U.S. Users simply choose the company they wish to call, and we'll dial the company directly, navigate its phone tree, and call users back when they are in queue for an operator or customer service representative. The service is available for free, and we've gotten some great feedback so far.
Marcin Musiolik
Pease: I don't complain about all phone trees. The number for finding out when a UA flight is arriving (800-864-8331) isn't bad. Others are hopeless, like some jury-pool numbers. You can't help with those, can you?

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.