What Could Be Causing My Transformer to Fail During Exciting Current Test?

[kudjung]

Hi Folks,

We've been use one of the linear transformer in our product for a while and suddenly in one lot the transformer has failed the exciting current test limit at higher votage(132V).
This is a 120AC/20VAC@750mA linear transformer. Supplier is trying to figure out what went wrong but still find no clue.

-What should I investigate?
-Is there any measurement I can do to check whether the problem come from the core or the winding?
-Can this related to a core saturation? Can we look for problem with core saturation with oscilloscope?

Sorry if there seem to be too many newbies question. I was been pretty much in digital domain for so long.

Thanks,
Kudjung.

 

[berkeman]

Hi Folks,

We've been use one of the linear transformer in our product for a while and suddenly in one lot the transformer has failed the exciting current test limit at higher votage(132V).
This is a 120AC/20VAC@750mA linear transformer. Supplier is trying to figure out what went wrong but still find no clue.

-What should I investigate?
-Is there any measurement I can do to check whether the problem come from the core or the winding?
-Can this related to a core saturation? Can we look for problem with core saturation with oscilloscope?

Sorry if there seem to be too many newbies question. I was been pretty much in digital domain for so long.

Thanks,
Kudjung.

Welcome to the PF. Sounds like a problem with the power transformer in your product. Could you please post more details about the test setup, and what it means to "fail the exciting current test"? That is not standard transformer terminology, in my experience.

Sounds like the root cause could be poor lamination insulation, or some winding issue. More details would help us a lot.

 

[kudjung]

Thanks a lot Berkeman for quick reply.

Here are some more details.

This is actually a linear power supply similar to an off the shelf DC adaptor. It just that we buy a bare transformer and build the DC power supply on our own to include in our product.
So, in this unit, it only contain transformer and a bridge rectifier.

Output spec. is 20V@750 mA with 120VAC/60Hz.

On one of the test we mesure the ac input current going into the power supply unit with no load connect(excitation current). We also measure this at various AC input voltage from 110 - 132 Volts.

Normally,we got a reading for 132 Volts input at around 100 mA and it never exceed 120mA at worst. On a recent lot. of the transformer, we have some of the transformer with give a reading at around 150mA and some of even at 170 mA.

We're just wondering what could be wrong with this transformer?
Our QA also question whether will this affect UL compliance?

We're procuring this part from China and their engineer have given me no clue.

Anything that you would suggest for me to look at.

Really appreciate your help.

Thanks,
Kudjung

 

[berkeman]

Thanks a lot Berkeman for quick reply.

Here are some more details.

This is actually a linear power supply similar to an off the shelf DC adaptor. It just that we buy a bare transformer and build the DC power supply on our own to include in our product.
So, in this unit, it only contain transformer and a bridge rectifier.

Output spec. is 20V@750 mA with 120VAC/60Hz.

On one of the test we mesure the ac input current going into the power supply unit with no load connect(excitation current). We also measure this at various AC input voltage from 110 - 132 Volts.

Normally,we got a reading for 132 Volts input at around 100 mA and it never exceed 120mA at worst. On a recent lot. of the transformer, we have some of the transformer with give a reading at around 150mA and some of even at 170 mA.

We're just wondering what could be wrong with this transformer?
Our QA also question whether will this affect UL compliance?

We're procuring this part from China and their engineer have given me no clue.

Anything that you would suggest for me to look at.

Really appreciate your help.

Thanks,
Kudjung

Is the DC resistance (DCR) still within spec? Do you check the magnetizing inductance and the leakage inductance as part of your tests? With a power transformer like that, you can even check it with an inexpensive $100 LCR meter.

Are you able to see the laminations? Do they look like they are isolated as well on the failing units, compared to passing units?

 

[Bob S]

You will get a high no-load current at high input (excitation) voltage if the transformer core is being driven into saturation. This could be caused by either the manufacturer leaving a few laminations out, or not putting enough turns on the primary.
Bob S

 

[Jim1138]

Could also be caused by the wrong steel being used for laminations. The laminations must be separated with insulating varnish to prevent eddy current within the laminations which would increase input current. The varnish might be lacking to some degree and increasing eddy currents within the laminations. Also, there is usually a gap in the laminations which helps prevent saturation. The gap might not be to spec.

 

[The Electrician]

Find an earlier transformer that had low exciting current and keep it on hand to compare to the newer ones with high exciting current.

What is the DC resistance of the primary (120VAC) winding of the two transformers?

Can you post a picture of the transformer so we can see the construction method?

Put a 1 ohm resistor in series with the primary and post an oscilloscope capture of the voltage across that resistor when 120 VAC and 132 VAC is applied to the series combination of the 1 ohm resistor and the primary winding. This will show the exciting current waveform.

Show the waveform for one of the earlier transformers with low exciting current, and for one of the newer ones with the high exciting current.

Be sure to use a 1-to-1 isolation transformer to drive the combination. Otherwise, you will be connecting the oscilloscope to the mains directly, and that would be unsafe; besides posing a shock hazard, you could do serious damage to the oscilloscope.

 

[uart]

Hi kudjung. Have you recently upgraded your current measurement equipment? The magnetising current will be substantially more "peaky" than a normal sine-wave, especially at elevated supply voltages. So if for example you had upgraded to something that measures true RMS (as compared with cheaper instruments that only report adjusted absolute average) then that alone could account for a 50% increase in reported current. Even if the actual current was unchanged!

 

[kudjung]

Thanks so much for every of the answer. They're all very helpful and educational. Seem like I'll have a lot of stuffs to look at. I had asked them to shipped me some of the good and bad transformer(from China factory). I should be getting them in a few days. I'll do some more investigation and update the result, or maybe posting some more questions with some of the data that I have.

Thanks so much.
Kudjung

 

[Bob S]

One more thing to measure is whether the amps at normal voltage are real (in phase) or imaginary (90 degrees) relative to the input voltage. Loss of inductance and lamination losses (eddy current) losses produce different effects. So measure both watts and volt-amps with no load.

[added] In the United states, there is a very useful plug-in unit called "Kill-a-Watt" available at some retailers (incl. Radio Shack) for ~$30.00 which can measure both watts and volt-amps (and maybe power factor) down to 1 or 2 watts. This would be ideal to make your measurements.
Bob S

 

[kudjung]

Folks,

I just got some time to take some measurements, photo and upload it here.
Roughly, there seem to be significant different in coreloss value between the good and bad transformer measured using the RLC meter.

From the external structure, I can't really tell what the different between the two. And what was causing the other one to has higer core loss.

Any thought.

Thanks,
Kudjung

 

[kudjung]

Didn't see attachment in post. Try one more time

 

[Bob S]

Both your current waveforms are ugly. You are driving the iron into saturation.
[added]
The equation for inductance in iron transformers (without air gaps) is (from memory)

L = μ_r μ₀ N² A/ l

where μ_r = relative permeability, μ₀ = 4 pi x 10^-7 Henrys per meter, N = # turns, A = area of iron core, l = length of iron core.

See my thumbnail post #2 of iron permeability in thread

https://www.physicsforums.com/showthread.php?t=352947

You are probably exceeding 300 amp-turns per meter in core.

Most likely solutions

1) Increase N in primary (and all secondaries too)

2) Increase iron core area. Are any laminations missing?

Added #2-
Are there conducting welds across the laminations on two sides? They look pretty messy. Why do you have welds?

How are laminations stacked? Alternating "E"s and "I"s? Do you have an air gap in the iron circuit that can modify the inductance? Air gaps in magnetic circuit will change inductance, reactive current, and saturation..

On closer inspection, your "E"s and "I"s are not alternating. Is there a gap between them (under the welds), and how much? Use precision verneer caliper on both transformers.
Bob S

 

[The Electrician]

There are some references on the web you should examine:

http://en.wikipedia.org/wiki/Silicon_steel

Here you can download a material properties spreadsheet:

http://www.protolam.com/page5.html

Here is an online catalog with mostly powdered iron and ferrite specs, but there is some information about laminations:

http://www.tscinternational.com/tscwebcat2007.html

There is a page with pertinent formulas near the bottom.
-----------------------------------------------------

Good quality transformers are made with silicon steel laminations that come in different grades, with designations such as M6, M19, M27, and so forth. What they mean is this:

The number after the letter "M" is 10 times the maximum power loss per pound, with sine wave excitation at 60 Hz, and a peak flux density of 15,000 gauss.

That is, M6 laminations will have a maximum core loss of .6 watts/pound; M19 laminations will have a maximum core loss of 1.9 watts/pound, etc. (all at 15,000 gauss and 60 Hz sine wave excitation).

There do exist grades M2, M3 and M4, but they are expensive, and generally not used except in special circumstances. Typically, a good quality transformer will use M6 lams, which is probably what your transformer uses.
-------------------------------------------------

I have encountered the same problem you're having in the past. I suspect your problem is this:

M6 laminations are specified to have a maximum loss of .6 watts/pound, but as typically shipped, they will have substantially less loss than that. I measured the loss in numerous shipments of M6 material, and typically the loss was 70% of the maximum.

We had been using a certain transformer built with M6 laminations, and when the product was first designed, a person without intimate knowledge of magnetics had measured the no-load excitation current and added 10% to that value and set that number as the go/no-go limit for the transformer.

Then one day, we got a shipment that was right at the limit for core loss, and the excitation current was over that previously set limit. Everybody got all excited, believing that we had received a bad lot of laminations.

You can't determine a go/no-go limit that way.

What you have to do is get the dimensions of the core and calculate the weight of iron. Then, knowing the number of turns on the primary, the cross sectional area of the center leg, and the applied voltage (sine wave), you can calculate the peak flux density. If it's near 15,000 gauss, then you should expect a maximum core loss of .6 watts/pound (if the core material is M6), and you just multiply the core weight (in pounds) by .6 to get the expected maximum core loss.

If the peak flux density is not 15,000 gauss, then you'll have to find a curve in the manufacturer's data showing core loss per pound for other flux densities.

The weight of standard core sizes, with a correction for stacking factor, is typically given in the manufacturer's data for a square stack. I see this data in:

http://www.tscinternational.com/tscwebcat2007.html

A square stack is a stack of laminations such that the height of the stack is the same as the width of the center leg. For other stack heights, use proportions.

Finally, in the spread sheet you posted, I see that you have used an LCR meter to make some measurements, and you have listed core loss as a resistance value. That is not core loss, nor is core loss the product of the primary exciting current and the exciting voltage. Core loss is measured in watts, and must be measured with rated voltage applied to the primary (the LCR meter is only applying about 1 volt or less). To get an accurate measurement of core loss you need a special wattmeter such as:
http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=230399762953&ssPageName=STRK:MEWAX:IT

or perhaps a more modern electronic wattmeter. Your transformer is rather small, and the core loss is probably under a watt.

I see that you have an Agilent scope, and I suspect it has a math function. You can get a value for the core loss to perhaps 10% accuracy using the scope.

Measure the primary current on channel 1 of the scope, and the applied primary voltage (set to 120 VAC) on channel 2. Use the math function, set to multiply channel 1 times channel 2, the use the measure function to display the mean (or average) of the math function. This will be the total no-load loss. You must subtract the copper loss to get the core loss. The copper loss is equal to the square of the excitation current times the DC resistance of the primary winding, where the DC resistance was measured at ambient temperature. Make the scope measurements with the transformer also at ambient temperature; don't allow the transformer to sit on the bench, powered up for hours, before making measurements.

To check the calibration of the method, temporarily replace the primary of the transformer with a 2 watt, 15,000 ohm resistor. The scope should indicate a power of 120*120/15000 = .96 watts.

It is very likely that in the past you have been receiving transformers whose core material was substantially better than the rated maximum loss, and now you have received some whose core loss is right at the maximum allowed. If this is the case, you have no complaint against the transformer manufacturer.

You shouldn't really care what the loss is at 132 VAC applied, but you certainly could determine what the loss should be at the flux density that occurs with 132 VAC applied.

You should also understand by now that the excitation current isn't really what you should care about; it's the actual loss, measured by a wattmeter method, that matters. And, especially, the excitation current at an elevated primary voltage may not be a significant indicator of core material quality. The elevated primary voltage pushes the core further into saturation, which increases the excitation current peaks. This increases the no-load copper loss proportionately more than it does the core loss.

You should get a suitable wattmeter to use for incoming inspection. The Kill-a-Watt meter mentioned in another post is not suitable. You could either use a scope to measure loss, or for best accuracy (at lowest cost; about US$650 for the electromechanical Yokogawa, several thousand for an electronic one), get a Yokogawa electromechanical meter such as the one I linked to on eBay. The one on eBay is not quite sensitive enough for your application, but Yokogawa does make one that would be. The lowest current range of the one on eBay is 1 amp, and you really should have one with a .2 amp range.

It's quite typical for small transformers to run into saturation somewhat, and that's ok. I have a high quality 12VA Stancor transformer, and the second image below shows the applied line voltage, and the excitation current. You can see that the core is driven into saturation.

The third image shows the result of using the scope to measure transformer loss. The red trace is the math function; it's the product of the applied current and voltage. The measured value is .419 watts. The measured DC resistance of the primary is 50 ohms, and the scope says the exciting current is .0282 amps, for a no-load copper loss of .04 watts. The no-load core loss is then .419 - .04 = .379 watts.

The first image shows formulas. It came from:
http://www.tscinternational.com/tscwebcat2007.html

So, before you get all worried, find out what core material the manufacturer of the transformer uses, and do the calculations. I would expect the transformer manufacturer to have an engineer on staff who knows all this, and they should be able to give you a maximum core loss spec for that particular transformer.

 

[The Electrician]

It might be good to get a feeling for just how much additional loss is being produced by the welds.

Ask the transformer manufacturer to send you two transformers, both built with laminations from the same lot; one welded and one without welds.

Then you can measure total no-load loss of each and get an idea what the effect of the welds is.

Before making the measurements on the non-welded unit, gently pound the lamination stack with a rubber mallet to make sure the butt joints in the laminations are tight.

 

[Bob S]

Before making the measurements on the non-welded unit, gently pound the lamination stack with a rubber mallet to make sure the butt joints in the laminations are tight.

I attach the equations (in MKS) for a gapped inductor from National.com. A 10% decrease of the primary inductance is important. An air gap length of the order of 0.01% of the magnetic path length in the iron can decrease the inductance by 10%. See thumbnail.
Bob S

 

[The Electrician]

I have a 30 VA Radio Shack transformer that's very handy for making measurements of the effects of variations in core configuration because they didn't varnish the laminations. I can unstack it and restack it however I wish.

I restacked it with all the E laminations together and all the I laminations together. I put the stacks on a flat surface and with a soft mallet, I lined up all the edges and then put a nylon screw through a hole in the laminations to keep everything together.

With all the E's together and placed in the bobbin, I could then take the stack of I's and put them across the ends of the E's so that there was one combined butt joint.

I then clamped the E's and I's together with a few pounds of pressure and measured the power loss with the same setup I described in post #14, applying 120 VAC to the primary with the secondary open circuited..

The DC resistance of the primary is 33.2 ohms.

The first image shows the measurement with the two stacks clamped together. The exciting current is .0787 amps, which gives a copper loss of .0787*.0787*33.2 = .2056 watts. The core loss was 1.65 - .2056 = 1.44 watts.

The second image shows the measurement with the two stacks separated by a .002" piece of nomex, then clamped. The exciting current is .180 amps, which gives a copper loss of 1.076 watts. The core loss was 2.54 - 1.076 = 1.464 watts.

Adding gap to the core substantially increased the exciting current, but hardly changed the core loss at all. This is what I would expect, since the reluctance which the gap adds to the flux path is in air, and air is not a lossy core material.

When the laminations are stacked in interleaved fashion, increasing the gap between an E lam and its mating I lam causes flux which would normally cross the gap to move to the side, and pass through a neighboring lam, increasing the flux density locally there. That would increase loss in the iron locally.

I'll make a pair of measurements on an interleaved stack and post the results.

 

[The Electrician]

This time the laminations were stacked in standard interleaved fashion. The secondary was open circuited, and 120 VAC applied to the primary.

The first image shows the measurements on the interleaved stack; a soft mallet was used to reduce the interleaved butt joints to a minimum. The exciting current was .0542 amps for a copper loss of .0542*.0542*33.2 = .0975 watts. The core loss was 1.77 - .0975 = 1.6725 watts.

The second image shows the measurements after I took a needle and thrust it into each gap on one side of the stack, moving each I lam on that side about .01" away from its mating E lam. The exciting current increased to .174 amps, giving a copper loss of 1.005 watts. The core loss was 2.74 - 1.005 = 1.735 watts.

So, adding all those gaps increased the exciting current from .0542 to .174 amps, and the core loss from 1.6725 to 1.735 watts.

The point of all this is that exciting current is only a very rough indicator of core quality; you really have to measure the true core loss to determine core quality.

 

[kudjung]

Hi There,

Seem like I need to study a lot more about transformer. This transformer was there for a while and I happen to assign to look at the issue. I think we first get this transformer, we only provide them electrical spec. and ask the transformer supplier to design the transformer for us. Look like we might have relying on them too much.

BTW, for the question with welding. I tracked the history. It seem that this transformer was creating humming noise and the supplier put this welding to eliminate the noise.

I'll take a look at all the suggestion.

Thanks once again, very useful.
Kudjung

 

[uart]

Both your current waveforms are ugly. You are driving the iron into saturation.

Yes but that is par for the course with very small transformers like this Bob. Linearity of the magnetizing circuit is not really the same issue as it may be with inductor design.

BTW, for the question with welding. I tracked the history. It seem that this transformer was creating humming noise and the supplier put this welding to eliminate the noise.

The weld could increase the eddy current losses but if it's shallow enough (as looks to be the case) it mightn't be any problem.

Assuming that there are no major differences in the physical construction (as seems to be the case) then my diagnosis is simply that they've used a different type of transformer steel on the new transformers.

The newer steel ("bad" transformer) has the following characteristics relative to the original one :

1. Lower magnetic permeability (slope of the B verus H characteristics). This is why it requires more magnetizing ampere turns.

2. Very similar saturation flux levels as the original (knee of the B/H curve happens at approx the same value of B).

3. The newer steel possibly has a smaller hysteresis loop (lower iron loss) than the original. This might sound counter intuitive given the higher magnetizing current, but notice how the magnetizing current in the newer version has slightly better symmetry between the left and right "shoulders" of the waveform. This actually corresponds to lower hysteresis loss.

Overall I would suggest that the new design is not necessarily "Bad". Yes it has higher magnetizing current and you have to assess whether or not that extra copper loss is indeed an issue for you. You should compare the overall losses of the two, as the new core just might have a lower hysteresis loss (offsetting or partially offsetting the higher copper loss).

Can you make some temperature rise tests of the new and old transformers. Just leave them both running at 130V and no load for about 30 minutes and make some measurement with a contact thermometer on the core. It wouldn't hurt to repeat the test at full load secondary current and again compare temperature rise of the two.

 

[uart]

Hi kudjung. Have you tried measuring the no-load temperature rise of the two transformers yet?