After the second occurrence, we tested each of the external circuit components. We verified them individually, including the catch and boost diodes. We determined they were within spec and working correctly. The layout was verified, and so was the PCB itself for any new or previously undiscovered problems.
Confounded by the problem, I again started looking at the datasheet of the switcher. I also looked at the application notes on Linear's Website for any source of inspiration. I came across this article. It talks about how ceramic capacitors on the input of the switcher can cause overvoltage transients due to their extremely low ESR. That could potentially push it over the tolerance limits of the device and damage it.
My design had ceramic caps on the input. Based on the design constraints, we figured that an electrolytic/tantalum capacitor was unnecessary on the input side. Besides, the datasheet commented on how the device is stable with just ceramic capacitors. In fact, it even recommended their use.
When I probed the input supply pin of the switcher and power-cycled the board a few times, sure enough, I saw huge voltage transients. With the bench top power supply supplying +24V DC, I saw transients routinely at over +38V DC and a few times even as high as +52V DC. The absolute maximum rating of the device, at its input, was +36V DC. There was the smoking gun!
On replacing the switcher for the third time, and upon replacing one of the ceramic caps at the input with an aluminum electrolytic with a large enough ESR to slow down the ramp up of the input and thereby preventing the sharp transients, I was able to get the switcher working again and was able to consistently reduce the transients to under +30V DC at a +24V DC supply.
We put the modified interface back into the complete test setup, and it worked perfectly for the rest of the scheduled testing.
For future reference, you should not overlook the extremely low ESR of ceramic capacitors. In most situations, it’s a desirable feature with no potential threat at all. Yet all input stages of switchers should have mechanisms in place to take this fact into account.
This entry was submitted by Girish Ramachandran and edited by Rob Spiegel.
Girish Ramachandran has a master's degree in electrical and computer engineering from the University of Florida and has worked in electronics design for UAV applications, robotics, and medical equipment. He currently works as lead manufacturing engineer for Prioria Robotics in Gainesville, Fla.
Tell us your experience in solving a knotty engineering problem. Send stories to Rob Spiegel for Sherlock Ohms.
I wonder why the checkout was not done with the 24 volt power input, since that was where it was going to operate. A series current limiting resistor would indeed offer protection, but it would then need a capacitor across the input to the supply to make things stable.
But the primary failure was in not testing it with the 24 volt input before attempting to put iy in service. Murphy will often show up in that area.
Ironic part of it was that I never had a Class license. My next door neighbor was an active ham, and we discussed his shack many times, but I never got the urge to paddle. Left work in the office. Was into a lot of other things, but not amateur radio.
Also used 3-400Z, 3-500Z, 8877 (3CX1500), 4CX2500, 5CX5000 & 6146B as drivers (3 in parallel, 150W out PEP). And, used an Amperex glass (tetrode) "bulb", but can't remember the #, as well as some RCA ceramic (tetrode) tubes for the higher frequencies. And, other EIMACs as well. One disappointment though ..... all were air-cooled tubes. Would have liked to design a XMTR with some of EIMAC's water-cooled tubes, just for the fun of it.
You're re-read is correct. The input V goes into the 10 ohm. The "low" side into the input pin of the 78xx. Tied to this pin are a .01uf ceramic to gnd AND a 10uf to 22uf alum electrolytic to gnd. The output side of the 78xx usually fitted w/ the 100uf @ Vout + S.F. The 1N400x is wired anode to the output, cathode to the high side of the 10 ohm. This is especially useful when the load circuit has considerable Xl. Helps drain off that inductive kickback. Of course, I agree w/ you about the incidental cost increment for these extra components, but my projects were (and, still are) pricey enough that a few extra components to ensure long operational life are not relevant.
That $75 sounds about right. I never could afford to use those hefty boys in my homebrew, used 807s, 6146s, and (for VHF) 2E26s. First Xmtr was a 6AQ5 10W CW one. Never even owned a linear; I was QRP for economic reasons! Still licensed, not active, no gear. The vast majority of the tubes you can get today are from Russia. My younger son at one time worked for the US master distributor for Sovtek, the largest Russian manufacturer.
BTW, re-reading your post, I now realize that the 10 ohm 1W you mentioned was IN LINE with the input, not to ground! BIG difference. It actually acted as a fuse, and also helped protect aginst both OV transients and reduced the power dissipation for steady-state voltages near the max. Used that myself on occasion. Sorry for the error.
I remember HARRISON LABS also. Didn't work there, but knew of their work. My first engineering position was for a radio communications company. My specialty was designing linear power amplifiers in the range of 400W (output) to 5 KW, in the fregquency range from 2.0 MHz to 400 MHz. Some were auto-tune, some not, but the projects were varied & interesting. My favorite days were those when I get update inserts for the giganticly thick SPRAGUE ELECTRONICS & EIMAC catalogs. It was always a treat to see a new tube introduced...... How can I use this one? The last time I saw an 8660 catalogued (4CX1500B) was about 5 or 6 years ago in the ALLIED catalog. They were selling them for about $1100 / ea. Of course they weren't EIMAC brand anymore. If my memory serves me correctly, I think we paid about $75/ ea. in production quantity, but then again, that was about 40+ years ago!!!
I've been using similar practices for many years for both linears and switchers, but I also combine yours with one suggested by an older post. Especially when using those new extreme-K low ESR chips (10uF and up), I ALWAYS put in a series R so I can guarantee a minimum ESR for the combination. This includes both input and output caps. It's a bit of a balancing act between noise control and stability, but having the footprint in the layout allows optimizing. If it turns out to be unneeded, you simply use a zero-ohm R! Even in the most cost-sensitive applications, the cost of a chip R is a tiny fraction of a cent in high volumes now; typically, the cost of placing the part is higher, so the part itself is basically free. There are other components that also can benefit from this (like high-power audio amp ICs). This also helps a lot when designing "hot-plug" circuits.
Most of the circuits I've done couldn't tolerate a 10 ohm resistor input to ground for multiple reasons (efficiency, "off" state current draw, etc.). In the original post, at 24VDC input, this would draw 2.4ADC! That's 57.6 W, and the resistor would fail because you let the smoke out. Did you mean 10K???
P.S. I may be an OLDER curmudgeon, going back to TUBE days. Worked for the old HP New Jersey division (formerly Harrison Labs) for a few months after I graduated, designing one of the very first digitally-controlled high power lab supplies.
I'm interested in your story as a possible posting for Sherlock Ohms. If you're willing to draft it, showing the problem and the process of finding a solution, it would be a great contribution to Sherlock. If you're interested, we would also need a short bio.
If you decide yes, you can send it to: rob.spiegel@ubm.com
#1) Rob, Personally I currently enjoy the mix of technical issues with the more mundane (appliance nightmares, etc.) However, IF there has to be a tilt, I'd prefer reading about the technical issues more. It seems that they are more "entertaining."
#2) Regarding the power supplies. For decades, ever since the homely 78xx series has been used, I've always including a few additonal components. On the input side, I always place a 10 ohm, 1 W resitor. Additionally, I put a .01uf disc as close as possible to the input terminal to ground. Close by, I use a 10uf electrolytic. On the output side, I always use at least a 50uf electrolytic. Finally, I install a 1N400x (reverse) bridging the output of the 78xx to the high side of the 10 ohm R. While this may sound supurfluous or overkill, in all my designs whether static logic circuits or rf circuits of low to high power, I've NEVER had an issue of corruption, parasitic oscillation, etc. with the p/s rail. Many years ago, there was a caution about the 78xx series, suggesting that some parts could be subject to oscillation under certain operating conditions IF there was no bypass C as close as possible to the input & ground terminal of the TO-220 device.
I like the mix of tech and mundane problems. This was a good one as a lessons learned that could help others.
I'm in layout now on a board with an LT switcher. I'm not sure where I read about ceramics on the switcher input, but I made sure to have some tant on the input as well. Reading this made me go back and double check.
These new (at least to me) 10uF and 22uF or larger, low ESR, relatively high voltage ceramics (25V) are exotic creatures (compared to what we had only 10 years ago). Couple that with low inductance PWB layouts and you can blow away (pardon the pun) a lot of assumptions about how circuits behave.
Many years ago, I added a ferrite to the input power of a switching regulator. After production started experiencing testing failures (and even worse customer field failures), we found out the ferrite and ceramic input caps formed a series tuned circuit that caused ringing that exceeded the input voltage of the regulator. The solution was to use a smaller ferrite with more capacitance.
We even now occasionally place series resistors in front of larger value ceramics to reduce the Q of the resulting circuit to stop transient ringing effects.
This case is interesting in that it illustrates how what sometimes should work doesn't. Components themselves can be very finicky, and none more so than capacitors. For example, we've seen numerous cases over the past decade and a half of what can happen when cheap capacitors are used on computer motherboards. Here, the ceramic versus electrolytic is not a quality issue per se, but it certainly relates to functionality.
From Dell / Intel® New Paradigms in Design Work Scott Hamilton, vertical market strategist for Dell Precision workstations, 5/2/2013 5
Early in my career, I worked as a draftsman and remember the days of drawing on vellum with numbered pencils and Mylar with plastic lead. This was a fun experience in the sense that I ...
I've been using workstations for more than 10 years and love finding ways to get more performance from my system. With demanding professional applications that require more power each ...
A lasting memory from my first job as an engineer in an auto assembly plant is standing on hard concrete at six in the morning, vending-machine coffee clutched in hand, listening to ...
For industrial control applications, or even a simple assembly line, that machine can go almost 24/7 without a break. But what happens when the task is a little more complex? That’s where the “smart” machine would come in. The smart machine is one that has some simple (or complex in some cases) processing capability to be able to adapt to changing conditions. Such machines are suited for a host of applications, including automotive, aerospace, defense, medical, computers and electronics, telecommunications, consumer goods, and so on. This radio show will show what’s possible with smart machines, and what tradeoffs need to be made to implement such a solution.
To save this item to your list of favorite Design News content so you can find it later in your Profile page, click the "Save It" button next to the item.
If you found this interesting or useful, please use the links to the services below to share it with other readers. You will need a free account with each service to share an item via that service.
Tweet This
9 
