How low is low ESR

[Rusttree]

I'm reading through the Applications Information for Linear's LT4356. On Page 16, it mentions using an electrolytic capacitor with "low ESR" on the output. Price is of some concern on my project, so I'd rather not buy a $2 cap if I can get a $0.20 one that's good enough. But I also don't want to handicap the circuit - especially since this will be in an automotive application.

So what are people's opinion on how low is low enough for a low ESR electrolytic cap? For $2, I can get 27mOhm and for about a fifth the cost I can get 360mOhm (after filtering for the capacitance and voltage rating I need).

Thanks!

 

[ChrisP58]

You will need to calculate what the impact of the ESR is on your actual circuit.

In addition to whatever specific requirements there may be to that PWM, ESR will also affect output noise and I^2-R heating in the capacitor.

Only after you take the numbers into consideration, can you assess the cost/performance trade off.

 

[MrAl]

Hi,

It looks as if that cap ESR has to be low enough to keep the short term surge voltage low enough to protect the DC to DC converter on the output. Lets take a quick look at this.

I'll throw out some numbers just so you can get an idea how this has to work. This should help you decide what you need.

Say we have 10v on the input to the circuit and 10v on the output going to the DC to DC converter. That means we have 10v across CL, the cap in question. Now all of a sudden we get a surge on the input that puts 20 amps through the capacitor for a short time. Now if the ESR is 0.5 ohms that means we'll see a rise in voltage on the output of 20*ESR=10 volts. That means we'll now see 20v at the DC to DC converter. Can it handle 20 volts or just 15 volts? If it cant handle 20 volts but only 15 volts, then we have to limit that voltage to only 5v instead of 10v. This means we'll have to halve that ESR down to 0.25 ohms so that when that 20 amps comes we only see a 5 volt rise across CL.

Thus the lower the ESR the lower the voltage surge will be. In order to really calculate the ESR though you have to know what kind of source you are dealing with on the input as to how much current it is capable of putting out. If it is only a little then a higher ESR will be ok, but if it can put out a lot of current then you need a very low ESR. You might have to do a few tests or estimate what the current get get up to so you have some idea what to go by.

Low ESR caps can be as high as 0.1 ohms but 0.050 or lower is even better.

 

[Rusttree]

Ok, thank you MrAl for the example and detailed description. I'll have to think about the character of the transients I expect to get.

 

[dr pepper]

Another thing to consider is ripple current, a couple of times I've strapped a cap on a pwm circuit to reduce the noise and they ahve exploded due to excess ripple current causing the internal resistance to heat up till the cap goes bang.
Any circuit needs to be designed carefully, if you just throw things together like I sometimes do these things show up.
You can buy esr meters, they are expensive, theres also a project around on the net called an lcm3, its an esr meter that measures down to milliohms and is cheap, sometimes they come up on ebay.

 

[simonbramble]

There is also an issue of stability. When the FET goes into protection mode, you effectively have a linear regulator with a feedback function (via the FB pin). The ESR will affect the stability of the loop. Low ESR in most circuits is always better. Tants or Ceramics are quite cheap (not as cheap as Aluminium Electrolytics), but still within most budgets.

By the way, you might find the LT4363 more suitable. We improved the performance of the 4356 by moving the current sense to the protected side (and hence brought out the 4363). We will not obsolete the 4356 if you are hell bent on using it

 

[Rusttree]

Thanks for the heads up on the LT4363.

That actually leads me to a followup question. Say in the future I'm faced with choosing between two surge suppressors like the LT4356 and LT4363. If I had a Linear representative over my shoulder, he/she could point to the 4363, as you did, and explain why the 4363 is a superior device. As a non-expert, though, I may not have recognized that subtle difference you explained. Obviously the datasheet for the 4356 is written to laud the capabilities of the device, so it's not going to point out its own shortcomings. In this specific example, what resource at linear.com could I have used to recognize that the 4363 is an upgraded and superior replacement for the 4356? Ideally, for example, there would be a chart that shows each Linear device and a link to a replacement part that is considered better. Does such a resource exist?

 

[simonbramble]

Submit a request into the linear website for technical assistance and someone will probably get back to you. The only real difference between the 4356 and the 4363 is the position of the current sense circuitry. This means you can add a scattering of discretes on the front end to make the 4363 capable of withstanding very high transient voltages. It is more difficult trying to protect the differential sense stage (like with the 4356) and still getting good current trip accuracy. Also the 4363 FET drive circuitry is slightly different, but apart from that they are identical