0
\$\begingroup\$

I would like to know your perspective on the following:

When measuring 24VDC both the lower test limit (LTL) and the upper test limit (UTL) are defined by either product, safety or 'common' specifications (the latter are chosen limits by company standards). Lets say for example that the 24VDC is tested passed if it measures between 21.6V and 26.4V, i.e. 10% tolerance.

The measured values can be represented by this matrix for 6 measured values at 6 different points in time: [23.2, 22.95, 23.4, 24.0, 23.5, 23.0]

The 'chosen' limits mark one result as FAIL, the 22.95V.

Employee A:

Lower the LTL to 22V, but keep the UTL at 25V to protect the applied product. Results in asymmetrical limits

Employee B:

Lower the LTL to 22V, but also raise the UTL to 26V to balance the tolerance.
Results in possible product failure (critical limit may be 25V)

Both approaches are approved by the customer. The product using this 24VDC is not known.

How would you deal with this situation?

\$\endgroup\$
  • 1
    \$\begingroup\$ Hold on, if your limits are ±10%, 22.95V would be a pass. I'm not sure I understand what your question is. \$\endgroup\$ – Hearth May 10 '17 at 14:59
  • \$\begingroup\$ My bad. (lets just say, a little less than the LTL). The tolerances don't really matter. The point is lowering the LTL because of measurements close to this limit. The power supply is a subsystem of a larger system. The reason the measurement is close to the LTL is probably due to cable lengths. I am just wondering If we should just lower the LTL and keep the UTL as is. If the voltage drop of the internal cables may result in a drop it would make sense that the output voltage is 'less' than 24V if the output of the PSU is an ideal 24V and not 24.somewhatV. \$\endgroup\$ – Weaverworm May 12 '17 at 7:39
0
\$\begingroup\$

Go with Employee A. Look closely at the idea of firing Employee B. The critical factor is "Results in possible product failure".

Why is there any doubt? And what sort of customer will approve such a decision? The term "clueless" comes to mind, but I'd be interested in other opinions.

There is simply nothing sacred about having symmetrical tolerances. If you must have such, specify 23.5 volts nominal, with +/- 6.4% tolerance.

|improve this answer|||||
\$\endgroup\$
  • \$\begingroup\$ By lowering the LTL you indicate the expected voltage also drops to lets say 23.5V instead of 24V (symmetrical limits). In some cases where the product specifics are well documented asymmetrical limits do make sense. 24V supply (UTL: 25V, LTL: 21V, the product may fail at 26V but not at 21V or even 19V). I think the proper argument will be with the product or system connected to this power supply. This sets the importance of the limits... \$\endgroup\$ – Weaverworm May 12 '17 at 7:46
0
\$\begingroup\$

I agree with Beast.

There are specs for Load regulation not mentioned.

-clearly this PSU is 24.0 at one time perhaps with no load and -4.3% load regulation ( which might not even be max load) is excessive in my PS spec experience and should be ~2% with total error budget of +-10% to include temp, drift, aging, ripple, load reg. cable loss, line V sensitivity vs Pd and temp rise of hot spot.

Or perhaps your test methods are flawed.

Furthermore if taking a population of PSU readings i would calulate the 3 sigma error and compare with UTL and LTL and this is related to quality metric Cpk and iF Cpk is not >>1 then I would report findings with PSU supplier to correct the regulation problem and disqualify them until fixed.

If you have no such spec to include all these sources of error voltage, then don't fire the guy who wrote the poor specs but educate him on error budget, standard deviation , Cpk and especially Load regulation error.

Stick with +-1 V until you have a better spec. Overvoltage depends on customer design and if they say +-10% then there is no need to worry, but this is worst case including all sources of error. There ought to be OCP, OTP, OVP and UVP protection also with fault indicator, which only costs pennies to add to the spec.

Personal experience.

Hammond, Brown, Lambda, Powerone, Shindengen,

The best product is not just who makes it, but who writes the best spec. in your company and due diligently performs DVT to verify these results.

  • with ongoing reliability tests due to unforeseen OEM transfers to Mexico or Asia with hidden flaws in the transfer after qualified.

    Been there done that.

PowerOne was a classic example of failures in Mexico but Ok in San Diego for Hipot failures with DC grounded . Their QA mgr refused to believe me until I pointed out a dozen process control flaws.

|improve this answer|||||
\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.