Low height output capacitor for LDO. Tantalum or...

[mahi]

I'm designing a simple Atmel ATtiny microcontroller circuit for use in an older car. Power consumption is just microamps when idle (this is important because the circuit is powered permanently) up to a couple of tens of milliamps for short moments when active.

For the voltage regulator I chose the LM2936-5.0 (datasheet). This is an automotive LDO which provides protection against the hostile 12V system, has a very low quiescent current and is well available (even from hobby-oriented stores like Conrad which may be important if someone wants to build the circuit for themselves).

I want to squeeze the circuit in an existing module which puts some constraints on the design: The module is mounted in the engine bay which means a temperature range from freezing cold in winter to almost cooking in summer and the components may not be taller than 4 mm because the board will reside in a narrow space behind the existing PCB.

This is not an issue for the (SMD) voltage regulator, microcontroller, transistors and passives, but the LM2936-5.0 requires an output capacitor of at least 10µF and I don't think I can squeeze an electrolytic can in there. Besides, given the temperature conditions I'm not sure an electrolytic capacitor is the best choice for this to begin with.

Right now I'm using a 10µF / 16V tantalum capacitor. Works fine in my prototype. Great temperature stability and it should be well derated for the entire temperature range (LDO output voltage is 5V). Since the circuit is permanently powered I don't think the inrush current is much of an issue.

But... judging by capacitor topics on this forum and elsewhere, using tantalum capacitors is almost considered a crime nowadays. According to some it's just a matter of time before they explode.

That's exaggerated of course but still I'm a bit worried about my choice. Especially if others decide to build the circuit for their car.

I was wondering what others think about this.

• A) 10µF / 16V tantalum capacitors are fine. Just make sure to use a reputable brand
• B) I'd rather go with tantalum polymer (10µF / 10V or perhaps even just 6.3V). Due to the lower ESR a series resistor is mandatory in combination with the LM2936
• C) Avoid tantalum like the plague. Go with X7R MLCC and a series resistor (what value would you suggest considering the poor stability of these capacitors in comparison to tantalum ones?)
  
• D) Forget about the LM2936 and use another LDO that is designed to work with pretty much any MLCC/ESR (suggestions on something that is targeted at automotive applications, outputs up to around 50mA, has very low quiescent current, comes in a hand-solderable package and is widely available in low quantities?)

Just to be clear: I'm not interested in mass production or cost optimization. This is just hobby-level stuff but obviously I don't want it to fail prematurely. I want it to last for many years.

[texaspyro]

I use a lot of LM2936's.   They are very nice devices.  I think there is a version that goes to 60V input.

I'd go with a 1 ohm series resistor and a 22uF 25V MLCC cap in a large footprint package.  MLCC capacitance drops with increasing voltage and the effect is larger in smaller footprint packages.  A "high-density" MLCC operated near its max voltage can have less than 1/10th its rated capacitance.

I have some 22uF/25V MLCC caps in a 1210 package that seem to work well without a series resistor with all the LDOs I've tried them with... I don't know who made them... and you can't have any... I', running low 

[Someone]

You'll be running down at the low current so just a little trace resistance between the regulator output and the capacitor is sufficient.

[DaJMasta]

You can get 10uF MLCCs as 0805s with 10V rating (though it's a price increase over the 6.3V flavors), so you could certainly do it with a pair of 4.7s or three 3.3s or something similar.  There should be a number of choices available among MLCCs that will fit the bill, and in the adverse environmental conditions, I think the MLCCs have the edge over the other choices again - very little self heating and no electrolyte to leak.  You may still want to consider something to deal with thermal expansion or vibration better, but there are hundreds of options 1206 sized or smaller, 10V rated or larger, intended for automotive use 10+uF MLCC caps on digikey starting at a little over 3 cents each if you're buying a whole reel.

[David Hess]

I would be satisfied with any of those solutions although I would use a 25 volt instead of 16 volt solid tantalum.  Given the environment, I would be very careful about environmental conditions like contamination, vibration, and temperature coefficient of expansion.

[helius]

For a very small increment in cost you can specify solid polymer tantalum instead of the old kind. They do not ignite at all, even if they are blown by reverse polarity. They also do not require any derating (although some derating is desirable for reliability).
See https://www.eevblog.com/forum/chat/capacitors-explained-by-james-lewis-of-kemet/msg1209543/

[slugrustle]

A Taiyo Yuden EMJ325KB7226KMHP looks like a good fit. 22uF, +/-10% initial, 16V, X7R (+/-15% from -55°C through 125°C), soft termination, AEC-Q200 qualified, 1210 package, 2.5mm +/-0.3mm thick. It only loses ~30% of capacitance due to DC bias at 5V, so even with temperature tolerance and initial tolerance, you're still over the required 10uF from -55°C through 125°C.

Taiyo Yuden product page: http://ds.yuden.co.jp/TYCOMPAS/or/detail.do?productNo=EMJ325KB7226KMHP&dataUnit=M
Digi-Key: https://www.digikey.com/product-detail/en/taiyo-yuden/EMJ325KB7226KMHP/587-5657-1-ND/


Another option is to put 3 TDK CGA6P3X8R1E475K250AE in parallel. This is 4.7uF +/-10% initial, 25V, X8R (+/-15% from -55°C through 150°C), soft termination, AEC-Q200 qualified, 1206 package, 2.5mm +/-0.3mm thick. It has virtually no loss of capacitance due to DC bias near 5V. With three in parallel, you'll have at least 10uF from -55°C through 150°C.

TDK product page: https://product.tdk.com/en/search/capacitor/ceramic/mlcc/info?part_no=CGA6P3X8R1E475K250AE
Digi-Key: https://www.digikey.com/product-detail/en/tdk-corporation/CGA6P3X8R1E475K250AE/445-173765-1-ND/


Since the 5V LM2936 requires at least 0.3ohms ESR, I would go with the Susumu KRL1632E-M-R330-F-T5 in series with either capacitor option, as it is at least 0.3245ohms worst-case. It is 0.33ohms nominal, +/-1% initial tolerance, +/-50ppm/°C, 1206 case, 0.75W, AEC-Q200 qualified, -55°C to 155°C rated.
https://www.digikey.com/product-detail/en/susumu/KRL1632E-M-R330-F-T5/408-1588-1-ND/

[mahi]

Thanks everyone for your feedback!

So far MLCC seems to be in the lead. Solid tantalum really is out of favor. And there I was thinking it would be the duck's guts for my application

slugrustle: Thanks for the recommendations and the mentioning of AEC-Q200. I've been looking for stuff that said automotive in the application notes but I was not aware there was a specific qualification for it.

Given the environment, I would be very careful about environmental conditions like contamination, vibration, and temperature coefficient of expansion.

Thank you for your considerations! The final board will get conformal coating. As for vibrations I hope the choice for low-profile and low-mass components will be sufficient. The board is relatively small (about 25 x 40 mm in the current design), double-sided, and the largest component is a SOIC14 microcontroller. To be honest, I don't see much measures against vibrations and thermal expansion on the original boards in the car and they have been working for 25 years without issue (apart from dried electrolytic capacitors in a few modules).

[helius]

MLCCs are subject to fracture and delamination if they are strained, which can be prevented by careful placement on the board. Sometimes internal slots will even be used to direct strain away from them. There are also special series of MLCCs with "flex termination" or "fail open" features to help with the problem.
http://www.kemet.com/Lists/FileStore/Flex%20MItigation%20Solutions%20Rev%203%20%20PHub%20copy%207.30.14.pdf
https://product.tdk.com/info/en/products/capacitor/ceramic/mlcc/technote/solution/mlcc02/index.html

[slugrustle]

helius makes a good point. Both of the caps that I recommended have soft termination.

[Siwastaja]

Large size MLCCs in this kind of environment are even more prone to randomly blowing up and burning your board than tantalums. Definitely do check this out in great detail. If you end up using them, this requires combined knowledge in mechanical design and soldering processes. Soft termination MLCC is a better option, but even that doesn't solve the problem 100%.

Tantalum with good voltage derating and current limiting (the regulator) might be a good choice, but do check the extra derating required by the high temperatures and/or buy a part specified for high temperatures.

[slugrustle]

If cost isn't an issue and you want to avoid both MLCC and tantalum, you could go with a film capacitor and series resistor.

Two Rubycon 16ST685MC24532 in parallel would get you at least 10uF over the initial tolerance of +/-20%. These are 6.8uF, 16VDC, 1812 case, 1.8mm thick, rated -55C to 125C. $12.50 for two.
https://www.digikey.com/product-detail/en/rubycon/16ST685MC24532/1189-1799-1-ND/

The 10uF Rubycon 16ST106MC44532 is in a 2.6mm thick 1812 case and "only" $7.05, but it has an initial tolerance of +/-20%. Does the LM2396 datasheet indicate if it accounts for capacitor tolerance? What is the convention on that?
https://www.digikey.com/product-detail/en/rubycon/16ST106MC44532/1189-1792-1-ND/

[mahi]

If cost isn't an issue and you want to avoid both MLCC and tantalum, you could go with a film capacitor and series resistor.

Like I said before, I'm not aiming for mass production or the lowest possible cost. I want it to be a reliable quality product but that does also not mean I'd go with the most perfect and most expensive solution. I'm just looking for advice of people with more knowledge so I can make an educated guess of what is good practice in situations like this (and it seems opinions vary).

The hardware and software will be open source. For my own board I'll probably stick with the solid tantalum capacitor (although I will replace it with an even higher voltage rating). After all, alternatives were given but no one voted against solid tantalum.

Nevertheless I'm going to update the PCB layout to accept a wider range of capacitor footprints, add a series resistor and provide a list of capacitors that should do the job. That way people can decide for themselves.

The 10uF Rubycon 16ST106MC44532 is in a 2.6mm thick 1812 case and "only" $7.05, but it has an initial tolerance of +/-20%. Does the LM2396 datasheet indicate if it accounts for capacitor tolerance? What is the convention on that?

Strangely, the tolerance on that 10µF capacitance is nowhere to be found in the datasheet. The number is a bit too round to be an absolute limit of the voltage regulator, though. I take it that means you can use a 10µF capacitor with standard tolerances (?) as long as those are met at the output voltage and over the entire operating temperature range. Most applications I've seen use a plain 10µF electrolytic capacitor but obviously they do not have the same space and temperature constraints.

It doesn't happen very often even in automotive applications, unless the board itself is in a very awkward position, such as directly attached to mechanical structures that tend to bend or move.
Ta caps don't fail without overvolt, excessive inrush or reverse voltage. Ta caps are among the most reliable caps at automotive temperature.

I don't think there's much chance of mechanical stress. The board is small and low-mass. The module enclosure it will reside in is loosely mounted to the car's bodywork so there are no direct engine vibrations and even bodywork vibrations would be somewhat dampened.

Anyway, I'm glad to see several people here would not mind using a solid tantalum capacitor for this application. Based on datasheets I found them to be perfectly suited but after reading several topics I got the impression they should be avoided at all cost. It's good to know alternatives are available and I will take these into account, but the original design was not wrong and I guess that's what counts most for me at the moment.

Just for reference: The original PCBs in the car are full of solid tantalum capacitors and they have been working fine for 25 years! So as long as you go with quality parts that are properly dimensioned for the application, I think there's little to worry about...

[Conrad Hoffman]

Everybody's terrified of tantalum caps, but an awful lot of them have been working just fine for many decades. With proper derating, which you've done, the tant should be fine. One desirable property of tantalums is they have consistent and predictable esr, something necessary for consistent and predictable regulator performance. MLCC parts have come a long ways, but the high-k dielectrics are still terrible in terms of voltage coefficient. Derate and measure at the operating voltage just to be sure of what you're getting and then make sure you have some series resistance.

[helius]

The bit about tantalum caps being reliable I won't dispute, but you should pay particular attention to using them in automotive applications because of something called a "load dump" which is caused by loose battery contacts. When the alternator is running full steam and there is a momentary disconnection of the battery, up to 200 V may be present on the vehicle power buss for half a second! All automotive electronics need to use protection from these and other voltage transients and you should make sure that your tantalums are suitably rated for any spikes the protection lets through.
http://www.electronicdesign.com/power/eliminate-those-automotive-load-dump-circuit-protection-headaches

[mahi]

helius: The tantalum capacitor is after the input protection and the voltage regulator so it should not be affected by load dumps.

I just pulled a module from the car and traced its power supply for the microcontroller:



I'm surprised the tantalum capacitor is only 10V. Maybe they were better quality 25 years ago?

Due to a thick layer of conformal coating I could not read the markings on diode D2. It looks like your average 1W glass zener but it might be a TVS diode.

I neither have the knowledge nor the equipment to confirm whether this circuit meets standards like ISO-7637-2 or ISO-16750-2 but clearly the manufacturer thought it was sufficient and their decision is backed by 25 years of operation.

The input protection is actually very similar to other examples I found online and what I applied to my own circuit. I guess if it works for them, it'll work for me as well.

[Siwastaja]

High voltage MLCCs smaller than 0805 or low voltage MLCCs smaller than 1210 don't crack easily, so don't get scared by the flex, bend, crack and explode shit.
It doesn't happen very often even in automotive applications, unless the board itself is in a very awkward position, such as directly attached to mechanical structures that tend to bend or move.

It's easy to say so (I did the same), until you actually get personally hit. For me, it was a simple one-off application, in a nice lab environment. Absolutely no mechanical flex, no mounting holes nearby. No big temperature differencials or cycles during use. Quality brand (AVX IIRC). 1210 X7R 10uF 50V part on a 10V bus. The only thing I did wrong was using too much heat during manual soldering using an iron, since the caps were retrofitted between copper fill areas on a PCB, possibly too close to each other in parallel. After months of use, one suddenly started glowing bright yellow for some time, producing quite a lot of smoke, mostly from the PCB.

I have seen too many MLCC cracking failures (personally/physically, or documentation thereof) to ignore the usual advice. Actually I think it's paramount to really follow the advice when using MLCCs, it's not scaremongering! They do crack easily, and are only reliable if the usual manufacturer notes are indeed followed.

The reason why the failures are not occurring all the time everywhere, is exactly because the designers, in most cases, seem to know what they are doing, and do follow the instructions, and possibly use soft terminated parts - especially in automotive.