How to make a burn-in test?

[JAY X]

Hi!

After building anything, a preamp, compressor, eq, we all test them to see if they meet the specs. Ok so far. But i am also aware of a 24h burn-in test, (maybe the name is not correct), the gear is put into an ¿Oven? At a certain temperature to simulate continous usage for 24h. As i don't know the details of such tests, i ask here... ¿With music/signal playing? Just turn on, and let it heat?...how?

Thanks for your advise!!
Jay x.

 

[swpaskett]

Burn-in is really any long term test. It doesn't necessarily require baking the unit; sometimes just powering it on and letting it sit idle for a while, then checking to see if it still works, whatever that means to the tester, is considered burn-in.
The test conditions are whatever you decide they should be. Moet people have heard of the bathtub curve, where initially failure rates are high as parts with latent defects fail early (infant mortality), then the failure rate drops and levels off with a slight upward tilt during the useful life of the product, and finally old age, where parts are at the end of their useful life and the failure rate climbs again. The idea of burn-in is to take a unit past infant mortality, thus increasing the reliability as seen by the consumer.
Elevating the ambient temperature can hasten infant mortality, thus shortening the burn-in timetable. How long one does this is often dictated by the buyer, or by the reliability engineers who have been around the technology enough to know what breaks and how. Temperature and pressure cycling and thermal vacuum testing bring on infant mortality even faster.
Everyone's needs are different so it is hard to say how long and at what temperature burn-in should be performed. For small production runs a four hour soak at the highest operating temperature might be sufficient. Run a full functional test before and after the soak, at temperature if possible.

 

[JohnRoberts]

It has been a while since I was responsible for manufacturing QA (quality assurance) and the testing can vary with the SKUs. For example Power amps were routinely tested making high power, including surviving shorted outputs etc. Low power rack-mount SKUs had easier factory hurdles to clear.

Even last century, when I was last involved in large scale production testing mostly used computerized board level ATE (automated test equipment). Circuit boards were tested before installation into chassis because they are easier to repair. Test performance is monitored and adjusted as needed. Patterns of failure get fed back to the factory making the sub assemblies.

Burn-in in general is an attempt to fast forward product life to capture infant failures (maybe search bathtub curve and MTBF). Depending on the weak links, marginal solder joints, or loose hardware, can be stressed by multiple thermal cycles, like heat the sku up to operating temperature, cool down, repeat. Back in the 80s my small company put rack mount products into a burn-in rack overnight. The rack used on/off clock timers to accomplish multiple full thermal cycles during each overnight period. Of course technology and weakest links have evolved over the decades.

FWIW I wrote a piece about this for my "Audio Mythology" magazine column back in the early 1980s. Solid-state device failures are not significantly a time at elevated temperature phenomenon, more like never melt the silicon ever (one exception was metal migration inside some fine line ICs over time but not a widespread problem). OTOH electrolytic capacitor life is definitely a time at elevated temperature phenomenon as heat accelerates electrolyte loss.

Coincidentally I recall while I was writing that column, the incandescent light bulb in my desk lamp burned out when I first turned it on.

JR

 

[pucho812]

At my old work place we would do a burn-in test on new gear as well as any repair. We were dealing with a combo of tube and discrete electronics. Now mind you by the time tubes were installed, the tubes themselves had already had their own burn in and testing for failure and to find ones that were up to spec. The units had already been on and off several times, but were not calibrated yet if new or recalibrated if repaired.
For the tubes, their burn in consisted of running on a jig which ran power through them at the B+ and heater voltages required. They had their own little circuit in case we wanted to pop in with a DMM or other test device and check. We usually didn't have to because by this stage tubes were already tested and selected. This was as JR said to capture infant failures. I did however catch prior to burn in a tube labeled a 12AX7 that tested like a 12AY7 and had the inside design of a 12AY7, mistakes happen.

For the gear once built or repaired. We had a rack that we would run the units on. They would basically stay powered for 24 hours. The units would have a +4dBU sine wave at 1K going through them and we could monitor the output of any unit at any time. We usually did just a visual check to make sure lights were on, and the meter was resting close to 0VU. Again this was to do a check for any infant failure such as one of the repair techs rushing a solder job leading to faulty joints, while we don't want it to it can happen.
Durning the burn in once in a while found a faulty tube. Again even though it went through a rigorous process already, it can happen.
After burn in, we would calibrate and the units would go off into the world.

I will say that we hardly ever found an issue while doing the burn in stage of production or repair. This is because we were very focused and payed attention to detail.
I would say in a typical year where we made some 300 units of a model each quarter, so figure around 1200 units that were burned in of a model, I may have found 1 or 2 that were not 100% right and this was due to stuffing the wrong resistor value. thankfully the wrong values were usually an extra 0 or decimal point from the right value, so like 1K instead of 10K.

 

[JAY X]

Hi!

Very interesting, now I have a better idea how to do such tests!!

Jay x