Correct capacitor choice - OT

[Christopher4187]

Hi,

I have a design which is a simple alarm system with two relays and the maximum spike current is about 200mA. I am using a 5 volt regulator (digikey# 296-12290-1-ND) but I can't figure what is the best choice of capcitors to use. On the output side, I have a .1uF for the electronics but I also have a 1500uF for transients in the circuit. My question is more about the input capacitor. What size but also what type of capacitor should I use. For this specific voltage regulator, TI says they used a .33uF on the input but they don't mention what type. Does the type really matter? As a general rule when making the power supply circuit, should a common design be used for all types of circuits? Thanks in advance for the help.

Chris

[mister_e]

Usually, the ouput one don't need to be as this big because the regulator have to do his job. Usually 10uF tantanlum + 0.1uF Ceramic are enough and must be sit as close as you can of your regulator output. NOW, depending of your circuit, you may need something bigger close to some device to reduce the noise comming from XYZ device. Let's say you have some 'high current driver' you may need to add some capacitor of different value close to them... also called bypass capacitor. Usually they are listed in the datasheet.

Generally, i use 10uF Tantalum + 0.1 uf Ceramic at the regulator output, 10uF tantalum + 0.1 uF ceramic close to the PIC +0.1 uF ceramic close to each and every IC in the circuit + a/some bigger(s) one(s) if needed for some specific reason. Capacitor are there to avoid your ICs to generate noise on the supply line. As they are called bypass capacitor, they will bypass the noise where they're install. Closer to your IC they will be, less are the chance to introduce noise in the rest of your circuit. In conjunction with a good groound plane (or 2 if analog is involved) you reduce your chance to generate noise on the supply AND generate 'RF' (EMI) noise.

At the regulator input... same thing apply 10uF Tantalum+0.1uF ceramic installed really close to the regulator is a good practice. Depending where your DC is comming from, you may need to add a bigger capacitor to reduce the ripple. If it's comming from a DC wall-mount transformer, they're already some capacitor mounted inside but adding something in a range of 100-1000 uF maybe of help. Better if you use some LC filter. Even a RC filter is really great too. A low value resistor in serie (let's say 4.7 - 10 ohm) + 470 uF may kill a lot of ripple... in fact better noise rejection than using a simple 470uF capacitor. Also, the low value resistor may react as a fuse.. no it's not a fuse but may act as.

You can do a search on googgle with 'Ground boucing' 'Ground plane' 'Bypass capacitor' and you'll find more info that you need.

[mister_e]

Just few other thing. Misplacement of you component, bad PCB design may and will ruine your life when they're not correctly done. Supply line must be bigger than the signal one and, IMHO, routed at first with the capacitor stuff. Once it's done and correctly done, you add your signal line one the top/bottom (and middles if you use multilayer PCB).

NEVER trust any autorouter CAD if your design is critical... Even with the best one and all the according restriction/direction defined, it may screw up your life. Well that what i'd discover after ages of PCB design.

I'm sure you don't want to do some PCB modification later because it was bad designed. For a single project it's ok, for a hundred/thousand/million unit.... ARRRRGH!

[Christopher4187]

Thanks for the reply. The PCB is grounded and the PIC sits within 1/2" of the regulator. I think I have the output portion of the capacitor issue down now but more about the input side. If, for example, I don't have anything on the input side....the car battery is going straight to the regulator, can this damage the regulator? Is the only negative effect noise on the line? I was thinking of adding a 10uF tant. and a .1uf on the input and output, do you think this would be good? The regulator is rated for 500mA and the peak current of the board is 200mA. The nominal current is 10mA and the peak is only seen when the alarm is activated, which hopefully should be never. Thanks again for the reply.

Chris

[mister_e]

Where this peak of 200mA comming from. A 5 volt driven siren?

In car application where, at least, alternator inject a huge amount of noise in the 12Volt line, i would suggest a rc or LC filter. But even 1000 uF and up should work great... plus the 10uF tantalum + 0.1 uF.

Some use 47Uf electrolytic but they're far to be as efficient as tantalum are.

....the car battery is going straight to the regulator, can this damage the regulator?

No this is why they're made for. But my biggest concern will be about the quiescent current of your regulator. The one you suggest need 5-6 ma all the time. It's not too bad but be sure your PIC is most of the time in sleep mode and your external hardware don't add too much in the current need.

10mA is still fair for a car battery, but you really should reduce it. 78L05 will need less quiescent current BUT is really limited in current drive.

There's some other linear regulator who need less quiescent current but can drive more mA. LP2960 is a great one, widely use by myself, seems to be great as now.

Tell us more about your circuit and we will come up with a better solution.

[Christopher4187]

OK, I've taken some measurements of the circuit and here is what I found. The board is in production already (low quantities - see picture) so I measured at the fuse and my measurements were a bit different than I wrote before because I was measuring board current instead of the 5 volt regulator current. The nominal current of the regulator is ~7mA. When the alarm is in standby mode, the current will go from ~7mA to 15mA because of an LED. When the alarm activated, the current will go from ~15mA to 60mA. I looked at the regulator you suggessted but it's too expensive. The one that I want to use is $.31 for quantities of 100. However, if there are less external components required, I would think about using it because I pay $.25 for each smt component on the board. I can't use a SMPS because I have a wireless IC on the board but the entire board has a ground plane so noise should not be a problem. I currently have 30 units sold with no problems noted but I am crossing my fingers. I would like to improve the power supply circuit and I don't think I need that large cap on the output. Let me know your thoughts.

Thanks,

Chris

[mister_e]

Look you PM.

[Ron Marcus]

Many high frequency transient spikes will go right through the regulator. A .1 or even .01 uF capacitor will help, but a series chip choke type inductor will give you the best gain for the money. I have been working with switching regulators for battery boost, and they were reeking havoc with a RF receiver built into the app.. I isolated the regulator circuit using a chip inductor on B+ and tied the grounds together through another chip inductor. The high frequency noise dropped by a factor of almost 20. It meant 800 foot range as opposed to 200.

[Christopher4187]

Did you simply use a oscilloscope to measure the noise?

[Christopher4187]

What does "look you PM" mean?

[mister_e]

Private Message... top of this page at the right corner bellow your name.