Check capacitor dielectric

[Raveninghorde]

I am fairly sure the answer is no.

Is there an easy way to check a capacitor's dielectric?

I have some pre production buck converters that aren't stable where as
the prototypes were OK. The input and output caps are specified as 4u7
50V X7R. The caps on the boards look suspiciously thin for the value.
1210s but only about 1.0mm thick. I bet someone has used Y5V to save
money.

 

[Raveninghorde]

Measure its voltage:capacitance curve. That's what will matter here.

ESR might matter too. A fairly simple setup could measure both.

What voltage are they running at?

John

24V input, 8.4V output.

 

[Raveninghorde]

A Y5V might lose 40% of its capacitance at 17% (8.4/50) of its rated
voltage. Add a little temperature effect and you could be down over
half. Freeze spray the caps (-60% delta-C or so for Y5V) and see how
that affects stability.

The input side cap could be down over 80%, but that wouldn't be as
important to loop stability.

Measure them!

John

As a quick check we replaced the 4u7s and more than halved the output
ripple. I'll measure them tomorrow.

The problem isn't cured so the next step is to replace all capacitors
that affect stability and check them.

 

[Raveninghorde]

For a given brand, the firs thing is to get/derive volume VS
capacitance for the various formulations (NPO, X7R, Y5U, etc) and see
which type fits what you have; should stand out like a sore thumb.
Next way is measure capacitance VS temperature for curve matching.

They sent me the data on the capacitor and it is Murata and should be
2.5mm thick. They are claiming Miurata also do a 1mm thick version
and I have told them I do not believe them.

 

[nospam]

I am fairly sure the answer is no.

Is there an easy way to check a capacitor's dielectric?

I have some pre production buck converters that aren't stable where as
the prototypes were OK. The input and output caps are specified as 4u7
50V X7R. The caps on the boards look suspiciously thin for the value.
1210s but only about 1.0mm thick. I bet someone has used Y5V to save
money.

High value compact ceramics capacitance is very voltage dependant which is
probably your problem (assuming it is not temperature related).

Measure the capacitance under changing bias and compare with what you think
is a good one. If you don't have a meter capable then you could lash up
something with a signal generator and scope/meter to ring it with an
inductor under bias.

 

[Raveninghorde]

I am fairly sure the answer is no.

Is there an easy way to check a capacitor's dielectric?

I have some pre production buck converters that aren't stable where as
the prototypes were OK. The input and output caps are specified as 4u7
50V X7R. The caps on the boards look suspiciously thin for the value.
1210s but only about 1.0mm thick. I bet someone has used Y5V to save
money.

The 4u7 caps were Y5V but that wasn't the cause of the instability
although the supplier has been castigated. They claim Murata sent the
wrong part on the reel, not that I believe them.

I had replaced an axial 4.7uH inductor with a surface mount part and
improved the power tracking to reduce pcb losses. The better inductor
and reduced damping were enough to cause a problem.

Fortunately linking out the inductor fixes the problem. The spectrum
analyser shows no crap getting out of the board even without the
inductor. It looks like the multilayer ground planes and 7 caps from
47n to 4u7 have killed any conducted RF.