Spot Checking your Reflow Oven – Part I

Reflow ovens have come a long way, they are much more stable, better controlled, well built and less expensive.   Nevertheless, the need for profiling has not been eliminated, since after all the reflow oven is just a means to an end.  You still need to dial in your oven set points to achieve a desired profile and what oven set point temperatures you choose is not what you see on your PCB.  You may be surprised how many line operators worldwide do not make this connection!

Once you have developed numerous profile recipes, many of you would only like to spot check the reflow oven to see if there has been any change.  The logic goes, if I have developed robust profiles and I can check my oven periodically to confirm there hasn’t been any change, I can therefore infer that my profiles are still valid?  Well, yes and no.   “Infer” is the action word here that can lead to trouble if you are not careful.   Let me give you an example.

For fun I ran some popular oven verification tools you can find on the market to see if they could catch changes that I knew would throw my profiles out-of-spec.   What I did first was to profile an actual PCB, followed by my oven verification fixture.   Running the fixture soon after my profiles gives me a baseline to compare against for future spot checks.  I then took a small clip on desk fan and blew into the entrance of my oven tunnel, knowing that it would change the thermal dynamics inside my process.   I made sure my fan was low enough not to trigger an oven alarm, but just enough airflow that based on my experience would in fact change the process environment inside my oven.

Here is what I found.   My profiled board width was ~12″ wide so a sizable mass.  I used a fixture of the same width, since changing my belt width is not practical.  When I ran the fixture I was given oven statistics to compare against my future spot check.  No problem so far.  When I ran the fixture a second time (fan a blowing), I found a 1 degree delta change.  Now I think you would agree 1 degree doesn’t seem like much at all.   Of course I am not looking at a profile, rather machine data so here is that word, I “inferred” that my process was still fine.   Wouldn’t you?  Now intuitively, I knew this fan blowing up the tunnel of my oven is creating process changes perhaps more than 1 degree.  So how do you know, well you run a profile!   Sure enough when I ran a profile, I did not see any longer just a  1 degree delta.  In fact my profile was out of spec.

So what happen?  I can only assume the fixture which I was using that weighed in at close to 7 lbs, must be absorbing some of the changes inside the oven, masking what is really going on.  Keep in mind the fixture is not designed to profile, rather it is designed to detect  measurable change to my process that would impact my profile.   The mass itself must be factoring into my ability to detect change.  Again this is just my observations and my conclusions are subjective, but I can say that it did not capture what is the whole point of this exercise.   So I guess the mass of your vehicle is important and it is not all just relative.  Any thoughts out there, please post your comments.

Well I guess it is back to profiling, or is it?  Stay tuned for part II as I explore other options.

Share

Stop Destroying PCBs in profiling your reflow process

2009 Presentation at SMTA Long Island on how to use software tools that avoid destroying your PCBs during the profiling process.

To view the complete video series (click here).

To subscribe to my Podcast for iTunes (click here).

Share

What to do with Zig-Zagging TC Readings?

What’s wrong with this picture?

Profile Lose TC

Well if you have ever used Kapton tape to attach a thermocouple, you have certainly seen your share of profiles like this!

So what, it is a perfectly good profile, right?  Yes, but no.  I had a customer who was using KIC’s Navigation (auto prediction) to help create a better “deep in-spec” profile.  The only problem, they were trying to optimize on a TC reading that was bouncing literally all over their PCB.   Navigator is an awesome tool, but it can only work with what you feed it.  If you feed it garbage, it will give you garbage.  In their case, it was trying to find them a new solution where literally every time the board was run the bouncing TC that was attached (or I should say was not very well attached) with Kapton was giving false readings.    Navigator would give a different solution based on what the TC was reading at that given time.   It is like try to put post-it notes on the ocean.

Solution is very simple, eliminate the TC reading from your graph.  You can easily do this with the profile you just ran.  Look what happens, you go from a far out-of-spec of 126% PWI to a far in-spec of 48% PWI.

Profile Lose TC2

So you saved your hard work this time, but you are after all one thermocouple reading short.  You added that TC to your profile for a reason.  Next go around, do yourself a favor and use a better material for attachment, such a conductive double side aluminum tape, which by the way, a recent study from RIT proves it a superior attachment method aside from sticking to your PCB much better.

Share

Across the Belt Uniformity and Reflow Profiling

I am often asked the question about how to handle components that are close to the outer edge of a PCB.   Today a question came in on Circuitnet to highlight this problem:

Title: Issues with BGA Components Near PCB Edges

What issues are we likely to see when we place BGA components very close to PCB edges?

What impact might it have on reliability?

Will equipment (screening, placement, reflow, etc.) require modification?

T. B.

I leave it to the screen printer, pick and place and reflow oven guys to answer the equipment part of the equation, but I can answer how one can determine with a profile if your BGA is getting what it needs as well as how other aspects of your PCB are impacted.

Across the Belt Uniformity:

There can be anywhere from a 2 – 5+C variation in temperature across the belt.  For example, BTU uses this homemade fixture to test for uniformity.  The idea is fairly simple.  With a set of type K calibrated thermocouples, you can easily monitor 6 TCs across the belt.  You want obviously to see the least amount of variation (if you were wondering the front TC is for measuring air temperature which is also used for automatic mapping of the profile to the oven zones with KIC2000 software).

BTU tool

Profiling for Reflow:

BGAs typically require more heat to reach their peak temperatures than smaller massed components like electrolytic capacitors.   For example, your BGA might have a peak temperature of 245C.

PCB2

While your electrolytic capacitors cannot tolerate as high as a peak temperature, therefore you want to set their maximum peak temperature lower, for example to 235C (this is just a relative example).

PCB3

With KIC2000 software, you can define each component in isolation.  If the BGA is off on the edge, I might need to bump up even further my peak temperature spec since in many reflow ovens, the outer edge near the rail is the coolest.  This is why you see some ovens with heat tape running along the rails!  Keep in mind of course as you crank up your oven to reach higher temps to reflow your outer edge BGAs, everything else on your board is also going to be impacted.   More the reason you better be hooking up thermocouples to temperature sensitive components to ensure they do not get fried while you are focusing your attention on your BGAs.  Profiling software that can “balance” the board is a must.  If there ever was a case where software can help solve complex problems in profiling, here you go!

I had a webinar back in July talking about BGA profiling.  There is also a video that illustrates what I explained above.  You can find this in an earlier posting:  http://profilingguru.com/reflow/profiling-bga-webinar/

Share

Thermocouple Attachment Results are in!

The Rochester Institute of Technology under the guidance of Dr. S. Manian Ramkumar Ph.D. just conducted (October 2009) the most comprehensive study to date on thermocouple attachment methods.  Part I of II was to determine the most accurate and reliable method of thermocouple attachment.  Part II that has yet to be released is to determine the best attachment methods for BGAs, with the goal of seeing if there are reliable non-destructive methodologies, so stay tuned.

Results in a nutshell:

Aluminum Tape out performed all materials even Kapton! In an ideal word, the best attachment method of a thermocouple to a component is what I like to call a naked TC.  Aluminum double sided conductive tape was the closest thing to having nothing at all to attach the thermocouple.  Kapton tape is less responsive (deflecting and insulating heat), never mind if you have ever seen a saw-tooth TC plotted on a profile you know it has a very hard time staying in place on your PCB.   Additionally, High Temperature Solder which I have always considered the gold standard, is the least accurate or responsive.  When you get to the critical peak temperature of your profile, high temperature solder is sluggish to respond to the rapid change in temperatures, thus distorting your readings. As Phil Zarrow and Jim Hall discuss in Board Talk, “mass” on your thermocouple is not your friend.  Phil Zarrow:

any measurement method, the key element is to get the thermocouple in good contact with what you are trying to measure and to do it in a way that does not modify the area with a lot of extra mass or material that is going to give you an inaccurate reading….

Bingo!  This is actually what this study shows, now with the numbers to back it up.

Study Methodology:

The study looked at:

  1. Aluminum Tape
  2. Kapton Tape
  3. Chemtronics – CircuitWorks CW2400- Two Part Epoxy
  4. High Temperature Solder
  5. Loctite – 382 Instant Adhesive

The study used a KIC Explorer with standard type K thermocouples.  Multiple runs of a substrate coupon (62 mils thick plain copper coated with silver) was routed into 12 uniform 0.24″ isolated sections.

Test bank2Three identical test coupons were used and run multiple times.  KIC’s air-TC was utilized as the control to which each thermocouple was measured as the coupon traveled through all heated zones.

A total of three boards were used, running each board through twice, allowing the internal temperature of the KIC device to drop below 40 degrees C before rerunning the profile.

The tape attach methods were measured uniformly for each RTD connection, using a dial caliper, while the high temperature solder and epoxy quantities for attach were found to be visually uniform.

Mean

This graph indicates the mean temperature differential that was noticed within the oven for the various attach methods. The readings are based upon the complete profile starting at room temperature and ending at the peak temperature. The data from the cool down zone was eliminated from the analysis.

The graph shows the mean differential and the 95% confidence interval for each attach method. The Aluminum tape had the least differential (-0.48) followed by Kapton Tape, Loctite Adhesive, CW-2400 and then HT-Solder. The Confidence intervals among most of the attach methods do not overlap except Kapton and Loctite, indicating that the means of the attach methods are significant. Significant differences exist between the methods except between Kapton and Loctite as there is overlap. Clearly Aluminum tape outperforms all of the other methods.

Zone Differences

The thermocouples seem to behave similarly within each of the zones of the oven. Zone 6, where the soldering takes place or the peak temperature is reached, the thermocouple attach methods show a much higher temperature than the air temperature, indicating that the PCBs have attained much higher temperatures than the air. A closer examination of ZONE 6 reinforces the selection of Loctite or Aluminum Tape for Phase III of this project.

Conclusion:

When considering accuracy, repeatability and responsiveness, Aluminum Tape is a winner.   There are of course advantages and disadvantages to each material.  For example one can argue you can re profile a PCB set up with high temperature solder, but considering that the mass of the solder distorts your readings, this study even brings into question this bedrock of thermocouple attachment.  Never mind high temp solder destroys your PCB as well as there is little control over the size of the blog from TC to TC and board to board.   Also don’t forget every time you profile the same board again it loses some mass, which will be the focus of more blogs to come.

Share

Are you profiling bare boards or bricks?

No one of course reflows bare boards, so why would you profile one?  For that same matter, no one sells bricks, so why do you put one through your reflow oven?

Profiling Bare Boards:

Today I came across CM doing exactly this.  They were processing networking boards.   They were  just too complex and expensive to profile, so the solution instead of finding a scrap board or some other reasonable substitute was to profile it as a bare board.  I guess the thinking was it is better than nothing, but can anyone honestly say that a bare board comes close to representing a true production board?  After all wouldn’t you agree profiling modern boards with mixed components, higher densities and micro-BGAs are already a challenge and to think profiling a bare board would yield any reliable results is a stretch?

Profiling Bricks:

So if this is such a terrible solution, what about putting a brick through your reflow oven?   A brick, come on Brian, who does this?  Well what do you think you are doing when you take one of the many fixtures available on the market that are used for characterizing an oven and using it to profile?    I bet if you melted them down (with profiler included) they aren’t far off in mass from a brick.   Consider the following attributes of a large mass:

  1. A large mass will behave differently than a production board.
  2. A large mass acts like a heat sink and will cause the oven to react differently compared to when a production board is run through the reflow oven.
  3. A large mass will result in a change to airflow due to its larger size as compared to the production board.

Now notice I included the profiler as part of the mass.  Many fixtures further add mass by adding a two pound weight to the fixture!   Now don’t get me wrong, these fixture do give you a relative measurement from week to week or month to month as to changes in the oven, but they do not tell you if your product is in spec nor provide a thermal profile.   Changes in the oven do not neatly correlate to changes in one’s profile.  After all how can they?   Chaos theory came out of the field of thermal dynamics, nothing neat about it.  Just like a bare board is no substitute to a populated PCB, a brick is also no substitute.

Don’t take my word for it, hear it from the oven manufacturers themselves.  http://profilingguru.com/reflow/standard-calibration-tool-for-reflow-process/

Here is a quote from Fred Dimock of BTU:

Oven manufacturers normally use custom designed test fixtures to simulate a board but their real purpose is to measure uniformity across the oven and confirm that the oven is working correctly.

….I have personally seen companies place unrealistic performance specifications on reflow oven testing with (fixtures) boards that have little to do with actual production needs. For example, we once were required to show that an oven could reproduce an inspect ramp soak spike profile on two 12 X 18 inch aluminum sheets that were 0.040 and 0.080 inches thick without changing any recipe parameters….

….From a surface mount manufacturing point of view – single board oven performance testing has little benefit. The real answers are to use actual boards with TCs placed on the critical components….

Both solutions profiling bare boards and bricks are inadequate.  Make matters worse if you do both such as I saw with this CM, the results are only compounded.  In other words, you are developing a profile based on an unpopulated board and afterwords taking measurements with a thermal mass that does not in anyway represent how your oven will in fact react to a production board.  This is classic garbage in, garbage out.autofocus

Now there are alternative solutions that don’t require the destruction of a production board in the process.   Many of the automated systems will create accurate virtual representations of production boards without the need to attach a single thermocouple.   There are also some brilliant software solutions that allow you to create accurate profiles without the need to run a profile.  http://profilingguru.com/category/reflow/automation/

Share

Why are you replacing BGAs?

There is a great post today in Circuitnet titled “BGA Replacement Limits,” that can be found under Circuitmart.  Panelists answer the following question:

How many times can a BGA component be replaced at the same location on the same PCB and retain reliability?

Mark McMeen of STI Electronics suggests that the answer may be as little as two times!

…most companies err on the cautious side and only replace twice at the same location after the initial build which is normally 2 thermal cycles for top and bottomside reflow thermal cycles.

I think a broader question needs to be asked, why are you replacing BGAs in the first place?  In my experience, often the answer is due to poor reflow profiling.  Often there is nothing wrong with the oven, PCB or BGA.   Why is it so hard to properly profile a BGA?  I believe the reason is most folks don’t have the option of placing a thermocouple underneath the BGA nor sacrificing a board in drilling a hole on the underside for TC placement.   In the old days, you could get away with snaking a TCs under the BGA, but with micro BGAs this is just not an option.  So what do people do?  They stick a TC on top of the BGA or along side it.  Many do nothing at all which is kind of scary and wind up asking question like how many times can I redo my board.

To go to show how hot of topic this is, I held a series of webinars a couple months ago with a turnout in the hundreds.  I shared some ideas, here is an abridged 8 min version of the session for those of you that missed it. Part of the answer is proper TC attachment which by the way is currently under study at RIT to see the most reliable method as well as determine if there is a non destructive methods that is both valid and repeatable.

The other part of the equation is profiling your PCB not only for your BGAs but also those components that cannot tolerate as high of temperatures. I’ve seen plenty of manufacturers so focused on a $500 BGA, ignoring pretty much what else is going on with other components on their PCB.   Certainly having the ability to define separate specifications, for example a peak temp for a DIP while addressing the special needs of your BGAs will lead to fewer BGAs having to be reworked in the first place.

After all, which is better, to treat the symthoms or the root cause?

Share

Thermocouple Attachment Discussion

Phil Zarrow and Jim Hall of ITM Consulting have a very good piece on TC attachment on Board Talk hosted by Circuitmart.com.

In their first session, they talk about permanent TC attachment, such as high temp solder and epoxy (click here for a link to their recording).  Yours truly left a comment with the boys:

We are a big fan of conductive Aluminum tape, used along with Kapton for strain relief like you mention in your podcast. We talk about high temp solder and epoxy which can work also, but like you said you got to be careful of mass. Lot of times we see unequal amounts applied per TC that can throw your readings. What is your take on aluminum tape, realizing of course it is a non permanent solution?

Well, they came back with a terrific response (click here for a link to their recording), where they make a clear distinction between destructive vs. non destructive methods.  Non destructive methods are often the only option, since customers cannot sacrifice a board for profiling.

Phil goes on to say:

any measurement method, the key element is to get the thermocouple in good contact with what you are trying to measure and to do it in a way that does not modify the area with a lot of extra mass or material that is going to give you an inaccurate reading….

Phil talks about using for example Kapton as a strain relief to ensure there are no stresses on the point of TC attachment.  I’ve been saying for years to use techniques such as window paning where you apply Kapton around the boarder of your aluminum tape to help keep your TC secure if profiling more than once the same PCB.  Make sure not to put Kapton over the bead since Kapton can behave as an insulator.

I think Phil makes a great point on emphasizing the “size” of the tape you are using.  Again you don’t want the material’s mass to become an issue.  So the name of the game is don’t go overboard.  Personally I prefer a 1/4″ square piece of aluminum tape along with 1/4″ Kapton.

Jim Hall makes also an excellent point that the same goes for “destructive” methods when using high temp solder and epoxy.  You don’t want to overdo it, or the mass can effect your readings.   I would add further that you need to be very careful that the mass be equal from TC to TC.   It has been my long held belief that the blob of epoxy or solder if of unequal amounts TC to TC, PCB set up to PCB set up will add variability into your process.  Just keep your materials to a minimum to get the job done.

Many of these assertions are currently under review by an RIT study.  Hope to have results as early as the end of this month.  KIC conducted a study 10 years ago on all the materials mentioned (click here for the report).  Since a decade has past, one could assume materials have improved therefore warranting a second look.  Stay tuned!

Share

Non Destructive BGA Profiling Test #1

I am currently investigating a non destructive method of BGA profiling that is reliable.  Here are the results of my first test.

Set Up:

Four thermocouples are attached to the same BGA (TOP, SIDE, INSIDE and BOTTOM surface), as pictured below.  Conductive aluminium double sided tape is used along with Kapton.  A KIC Explorer is the profiler.

To see more on Thermocouple attachment visit my post:  http://profilingguru.com/tcs/thermocouple-attachment/

A hole was drilled out to attach the INSIDE TC.

pic1

pic2

Results:

Two tests were run, the first was running the board on the belt followed by running the same board on the chain/tab conveyor.

sample1

As you can see the delta for ramp and peak is the greatest, while soak is minimal.  The inside TC runs the hottest and the underside bottom TC follows fairly closely the behavior of the inside TC.

sample2

This second profile was run on the belt with the same board but for a different BGA.   Again we see similar behavior, where the INSIDE and BOTTOM TCs exhibit similar behavior.

sample3

This third profile was running the same board and same BGA as in the second example but this time on the chain/tab.   Interestingly, all TCs were a good predictor of the INSIDE TC except when getting to the cooling zone.  The BOTTOM TC was only a good predictor of the INSIDE TC.

Share

BGA Profiling Webinar Recording

The following presentation was first held as a 30 min BGA Profiling webinar in July 2009, with over 120 participants.  Due to its popularity an abridged 8 min version was created.

To subscribe to my Podcast for iTunes (click here).

Share