Overvoltage on transformer secondary

Do you have a link to article, or is it in his trouble shooting book? then give a page number.

I'm not sure I can help.. or have seen saturation like this. But years ago I was playing with a home made flux gate magnetometer. Looking at the voltage waveform as I drove the coil in and out of saturation I realized that there are two terms to the voltage. V= d/dt(L*i) = i*dL/dt + L*di/dt Maybe it's the first term that is giving the spike?

George H.

I followed links that Steve Wilson gave:

The specific text is in third paragraph on page 28 (using .pdf page numbers).

Good catch! That took some good sleuthing.

Like you, I find it hard to believe the failure occurred as described. I think it more likely happened on the previous shutdown. If the contacts opened at the wrong time, the flux in the transformer would collapse and generate a voltage spike, just like the ignition coil in your car or the flyback transformer in an old tube tv. This could charge the filter capacitor to high voltage and destroy the regulator.

The failure would be noted on the next power on, and falsely attributed to a spike at turnon.

Increasing the size of the capacitor is only a partial solution, as indicated by the residual failure rate. A better option may be to include a zener or similar protection at the input to the regulator, as well as increasing the size of the filter cap.

Note if the supply can be disconnected from the regulator, there is a risk of destroying the regulator if the power supply is attached while it is powered on. The problem is a possible resonance in the power supply cable with the filter cap and the input cap of the regulator. This is described in one of the Linear Technology application notes. I forget what they used for a solution, but I will see if I can find the article.

He was bound to get something wrong.

At best, when saturation shorts out the core's EMF, the secondary voltage responds "delayed" -- and can thus overshoot or ring. It's usually well damped, though. This would be the RLC response of the secondary, a fairly high impedance, inconsequential to most any circuit.

Tim

-- Seven Transistor Labs, LLC Electrical Eng>

Not on the half-cycle that saturates the core. But maybe on the next opposite-sign half cycle, the one that yanks the core out of saturation.

That will happen when there is a lot of magnetizing (actually demagnetizing) current.

Thanks, maybe my Xmas gift can be some sort of e-reader.

Otherwise, I've hardly read any of that... lucky me. :^)

George H.

: If the power line switch was turned off at exactly the wrong time of : the cycle, the flux in the transformer steel core could be : strored at high level. Then, if the line power switch was : reconnected at exactly the wrong time in the cycle, the flux : in the transformer would ontinue to build up until the : transformer saturated and produced a voltage spike of 70 to 90V : on its secondary.

increase of flux. SEM is proportional to derivative of flux, so SEM is limited. In fact, high current in primary is because SEM is to small to oppose line voltage.

constant times SEM on the primary. Ohmic losses mean that SEM on primary is lower than voltage on the primary, SEM on secondary is higher than voltage on secondary. I would expect similar effect from stray inductance.

I have experienced the effects of remanent magnetism and DC offset in circuit breaker test sets, which use large step-down transformers to drive high currents (up to 100,000 amps) into circuit breakers. The load as seen by the mains power (usually 480 VAC) is mostly inductive, so the output is controlled by an SCR switch that fires close to the waveform peak, resulting in minimal DC offset.

If the output is initiated at a zero crossing, the first half-cycle may have a peak as much as twice normal, and the waveform then decays until it is symmetrical about the zero voltage level.

When an odd number of half-cycles is applied to the transformer, there is a net DC component that creates remanent magnetism in the core. If the next application of voltage is in the same phase, the core will saturate and cause a high current spike in the primary. Since under such circumstances, the core is already saturated, the output voltage on the secondary would be expected to be lower than usual, except perhaps for a very brief spike. However, if the phase is opposite to the magnetization, It might be possible for the output to see a higher-than-normal voltage, especially if the voltage is applied at or near a zero crossing to produce DC offset.

At least, that's how I understand it and what I have observed.

Paul

Instead of speculating, the correct procedure would have been to measure the pulse with a scope, something that Bob P. should have done in the first place.

Guessing at the problem was very poor troubleshooting technique, as well as increasing the size of the capacitor by some arbitrary amount. Yhat was very poor engineering, as shown by the residual failure rate.

Another problem was the cascading failure. If the cap failed, it would destroy the regulator. This could damage the following circuits which depended on the regulator output voltage. One of the tenets of good engineering is to prevent cascading failure, something that Bob P. should have emphasized.

So in this case, Bob showed the worst possible examples of bad troublehooting technique and poor engineering. Not a very good result for a book on troubleshooting technique.

Just curious, why do you need an e-reader to view a PDF file? How do you read datasheets that are usually in PDF format?

What operating system do you run that doesn't have a pdf reader available?

If it's Windows, have you tried PDF Xchange? It runs on XP and later:

If it's Linux, there should be a multitude of PDF readers available. I use the HP device manager for the HP 3050A Deskjet. This installs on Ubuntu

I never use it to print anything, but it has an excellent scanner that can output PDF files. As usual for HP, it had a very low price and was further marked down for clearance. Irresistable. It was one of the more sane purchases I have ever made.

Why do you think he never scoped the voltages? He mentions "extensive investigations", not speculation.

He didn't say the cap value was increased by "some arbitrary amount."

The residual failure rate was "near zero", whatever that means. No failure rate will ever be zero.

Different people have their own definitions of "good engineering." To me, a failed box is failed, and it doesn't matter much how many parts are failed. The best design minimizes the field falure rate, not the details of a failure. I do try to prevent part failures from starting fires or blowing traces off boards, but if a failed mosfet might take out a gate resistor, I'm not going to make a big effort to prevent that.

A lower system failure rate but more parts lost per failure sounds good to me.

Bob was a pretty good engineer. He just had bad taste in cars.

Are you kidding? This hasn't received updates for 2.5 years, 4.5 if you're on the desktop version.

Do you still use Windows 98 as well?

highly dependant on what you're designing for.

NT

I don't care. I never use it to scan the web, but it is needed to run with Virtualbox 4.3.36 for legacy software. I have tried later versions but always ran into some incompatibilty problems, so I'll stay with 10.04.

I did for a long time, then switched to XP. I like it a lot better than Win7.

Just to terrorize you a bit further, my main browser is Firefox 12. It has a phenomenal range of add-ons that are invaluable for browsing the web. I can control javascript, CSS, page color, and kill tabbed browsing which is not meeded on XP. I also control the keybindings, use custom toolbar buttons, have cert bypass, and a password editor. None of these are available in later versions. I also depend on the Nir software utilities for many other invaluable functions.

Also, I never use critical passwords in FF12. They are all in another VDI that is protected by Sticky password manager.

There are a few websites that refuse to use SHA-1, and won't work with FF12. For these, I use Palemoon 26.5.0, but I reaally don't like it. None of the Nir Software works with it, so it is almost useless.

So, you say, you have no protection. How do you prevent malware?

My response is, you can't, even with the most advanced operating system and anti-virus software available. You need to recognize how vulnerable you are, and create a system where it doesn't matter if you get infected.

For me, the system is excellent backups on a platform that is not accessible to malware. That is extremely easy with Virtualbox. The backups are through Ubuntu, not Windows. Since I use RAID0, it is not possible for Windows malware to get to the backups.

In Windows, if you get infected, the usual remedy is to pave over and rebuild. This means you have to reinstall all your programs, and configure everything back to the way it was. This also happens if you lose a hard disk drive. It can take a very long time to reinstall and reconfigure, and you will never get everything back as it was before.

With Ubuntu, if you lose the operating system, all you have to do is reinstall it, and run a simple bash file to reinstall all your programs. All your config files are in the Home directory, and they will survive the reinstall.

In Windows, all I have to do is copy the backup files. This restores the operating system and all the software and config files. It only takes a few minutes for 100GB of files, then I am back to exactly where I was before the failure.

-----------------------

st

as

** As I read the account, fitting a 1000uF electro in lieu of 10uF solved t he problem completely.

OTOH, it is very unlikely for an iron core supply tranny to have higher vol tage at the secondary than its turns ratio dictates - so another thing to l ook out for is a rare or intermittent cause of PRIMARY overvoltage.

Aside from a lightning hit, this is most only likely to happen at the momen t the AC switch on the particular unit is opened OR and much worse if a swi tch or breaker opens that controls the AC supply to MANY units. All the uni ts then have their AC supply terminals connected in parallel with no actual supply in place. Imagine some of there units are inductor ballast fluoro t ubes.

If, as is normal, the supply switch contacts arc upon opening - a burst of noise with peak voltage up to double the supply peak occurs at the secondar y of an *unloaded* transformer.

I have tried this on the bench with a 100VA e-core and found that bridging a 10uf film cap across the secondary had little effect.

IME, the most likely scenario is due to arcing at switch off, possibly aide d and abetted by other items powered from by the same supply switch or brea ker. The reg IC fails instantly due to overvoltage and of course this is no t discovered until the next time the unit is powered up.

Using a 1000 uF electro on the secondary fixes the problem.

Other fixes like primary side varistors and X1 caps may help too.

.... Phil

Yep, in high power transformer based supplies you can get core saturation. One way around it is to remember which cycle (pos, neg) was last applied. It has been done.

Cheers

Forgot to mention 'Core walking' is a somewhat milder case. Capacitive coupling can alieviate that.

Cheers

I think one company's circuit breaker test set incorporated a gradually changing phase firing on the controller to demagnetize the core. But IIRC it had to be done after the breaker tripped.

The test set is designed to apply an equal number of positive and negative half-waves, at least when applying pulses of six cycles or less to determine instantaneous trip point. But when it tripped, there could be DC offset because it might trip in 1.5 cycles, or even less than 1/2 cycle.

Paul

Why do you say it was speculation? Seems to me just as likely that Pease did scope and that was how they found the magnitude and timing of the transient. Nothing in the writing says they speculated.

piglet

Saturation could have occurred rapidly and caused large dI/dT ?

piglet