Ok to share capacitors between voltage regulators?

Other Hardware General Electronics
1

I need to have two voltage regulators in my circuit; a 5v regulator, and a 3.3v one for a microsd card.

They both require capacitors and the values I used for the 5v seem like they'd work for the 3v and one of them is this huge 47uF electrolytic, so I'm wondering if I can just place them next to each other in the circuit and use the same capacitors for both.

I was also considering supplying the 3.3v regulator with the output from the 5v regulator, but I'm unsure if that might lead to instability. Also, supplying it from the battery would take some stress off the 5V reg allowing me to draw more current from it for other uses. But it would put more stress on the 3v reg which might have to deal with the same 12-16v the 5v might be supplied with.

Here are the datasheets for the two regulators:

2

Hm... I jut realized looking at my schematic that while I might be able to share the 10uF ceramic cap I have on the input from the battery, that 47uF cap is on the 5v line going to ground, so I can't just wire the 3v output from the regulator up to it.

So I guess the 10uF ceramic cap on the input is the only one I need to know about.

[edit]

And I just realized the 10uF is also connected to the 3.3v regulator via the schematic, since both regulators are drawing from the same input. So I guess that answers that question.

3

47uf huge? Well besides that I'm not an expert but it doesn't sound good to me. Capacitor's are fairly cheap it may be worth it to have a working project if you just buy the right ones you need for each voltage regulator.

4

Normally a capacitor is chosen for a reason, you need to try and follow the requirements in the data-sheets, specifically because the values are chosen to prevent the voltage regulators from oscillating and also the distance from the regulator can make a big difference.

Also note the data-sheet uses tantalum close by the regulator & not electrolytic, (electrolytic caps are the bitches of the electronics world), the tantalums have nice high frequency characteristics and a tighter tolerance.

As regards the 3.3 and 5v, do not supply the 3.5v from the 5v regulator, there is no such thing as a free lunch, you would simply increase the dissipation of the 5v regulator and cause additional ripple on its output.

Drive them separately , but stick a reasonable size electrolytic BEFORE them, and a 0.1uf bypass capacitor, generally it is not a good Idea to stick your big storage capacitors AFTER the regulators (capacitors can be considered 'shorts' until they get a charge).

It is unlikely your design is going to fail badly or blow up by the use of the electrolytics, consider that electronics is very much like cooking as regards ingredients, but it is also unlikely to perform with the sort of tolerances one would expect from modern components if you don't follow the manufacturers requirements.

One last point, the higher the working voltage of an electrolytic, the more it 'leaks'.

5

TECman:
47uf huge? Well besides that I'm not an expert but it doesn't sound good to me. Capacitor's are fairly cheap it may be worth it to have a working project if you just buy the right ones you need for each voltage regulator.

Capacitor volume depends on capacitance value and the voltage rating - if you have a 100V 47uF it will be large, if you have a 6V3 47uF it should be quite compact.

The regulator(s) in question may not need such a large capacitor - check the actual datasheet for details.

6

When I said the cap was huge, I meant in size relative to all the other tiny surface mount components I'm putting on this board. The board is smaller than an iPhone and I've got a lot of stuff I'm trying to cram onto it and I'm trying to keep all the components on one side to boot, so space is at a premium.

The 47uF cap is actually a 16V one from Panasonic, and it's around .400" wide:

Normally a capacitor is chosen for a reason, you need to try and follow the requirements in the data-sheets, specifically because the values are chosen to prevent the voltage regulators from oscillating and also the distance from the regulator can make a big difference.

Well, that's why I asked. I've looked at the datasheets. It seemed to me like these values should be okay for the 5V reg, and the 3V reg seems really lax in its requirements:

Input Capacitor (CIN)
A minimum value of 1 μF (over the entire operating temperature range) is required at the input of the LP2985. In
addition, this input capacitor should be located within 1 cm of the input pin and connected to a clean analog
ground. There are no equivalent series resistance (ESR) requirements for this capacitor, and the capacitance
can be increased without limit.
Output Capacitor (COUT)
As an advantage over other regulators, the LP2985 permits the use of low-ESR capacitors at the output,
including ceramic capacitors that can have an ESR as low as 5 mΩ. Tantalum and film capacitors also can be
used if size and cost are not issues. The output capacitor also should be located within 1 cm of the output pin
and be returned to a clean analog ground.
As with other PNP LDOs, stability conditions require the output capacitor to have a minimum capacitance and an
ESR that falls within a certain range.
• Minimum COUT: 2.2 μF (can be increased without limit to improve transient response stability margin)

Here's the layout I am thinking about going with:

7

Also note the data-sheet uses tantalum close by the regulator & not electrolytic, (electrolytic caps are the bitches of the electronics world), the tantalums have nice high frequency characteristics and a tighter tolerance.

The other day I asked in another thread if I had to use tanatlum with my regulators because they're expensive and I wanted to use ceramic, and Grumpymike said I should avoid their use because they can catch fire.

Also, while the diagram shows tantalum, in the section on stability it says this:

The recommended load/decoupling capacitance is a 10uF
tantalum or a 50uF aluminum. These values will assure
stability for the majority of applications.

Capacitors other than tantalum or aluminum can be used at
the adjust pin and the input pin. A 10uF capacitor is a
reasonable value at the input.

As regards the 3.3 and 5v, do not supply the 3.5v from the 5v regulator, there is no such thing as a free lunch, you would simply increase the dissipation of the 5v regulator and cause additional ripple on its output.

I understand that. I was just concerned that the 3.3v reg might overheat if the supply is 12v.

Drive them separately , but stick a reasonable size electrolytic BEFORE them, and a 0.1uf bypass capacitor, generally it is not a good Idea to stick your big storage capacitors AFTER the regulators (capacitors can be considered 'shorts' until they get a charge).

I was just following the datasheet's recommendations.

Also I seem to be getting conflicting information here. On the one hand you're saying it's critical I use the capacitor values in the datasheet for stability, and on the other you're telling me to add two more caps which aren't in the datasheet.

Your suggestion also goes against my own experience with voltage regulators. I've built a number of circuits with only a .1uf on the input and a 10uF on the output of my regulator, and running off a 9v battery. These circuits were driving a couple noisy servos and reading analog inputs and I haven't had any problems. Are you sure you're not being overly cautious? I could understand sticking a larger cap on the input if I was running off something like a wall wart, but that isn't likely with this circuit.

8

The other day I asked in another thread if I had to use tanatlum with my regulators because they're expensive and I wanted to use ceramic, and Grumpymike said I should avoid their use because they can catch fire.

My toaster can also catch fire: tants usually do this when they are abused or used in the wrong way, I can honestly say that one design I worked on shipped over 2 million parts and for that product range I NEVER once saw a tantalum that caught on fire, but that does not mean it cannot happen.

Cost is relative.

Capacitors other than tantalum or aluminum can be used at
the adjust pin and the input pin. A 10uF capacitor is a
reasonable value at the input.

You are not talking 10uf you are talking 47uf, as I said before electronics is like cookery, also be aware an electrolytic can be +-50 - +-50% of the marked value.

I understand that. I was just concerned that the 3.3v reg might overheat if the supply is 12v.

but looking at the data-sheets, it maxes out at 16v input
and the other maxes out at 27 or difference depending on the model, why not use the same range parts (LM1085-3.3) for the 3v3.

You are still within the spec, but ultimately it depends on WHAT load you are driving.

Also I seem to be getting conflicting information here. On the one hand you're saying it's critical I use the capacitor values in the datasheet for stability, and on the other you're telling me to add two more caps which aren't in the datasheet.

The data-sheets cover what is around the regulator for optimum operation of the regulator, my recommendation was related to the fact that I did not know the power supply and cable length, also note that an IC is a 'microcosm' linked to components around it, once you get a few inches of copper between your IC's and power points, you will be glad of a bit of storage near your connector, before you hit the inductive load/Choke that is commonly called a power cable.

Your suggestion also goes against my own experience with voltage regulators. I've built a number of circuits with only a .1uf on the input and a 10uF on the output of my regulator, and running off a 9v battery. These circuits were driving a couple noisy servos and reading analog inputs and I haven't had any problems. Are you sure you're not being overly cautious? I could understand sticking a larger cap on the input if I was running off something like a wall wart, but that isn't likely with this circuit.

Then go with what you know, personal experience trumps free advice (especially with engineers). As regards capacitors after regulators, I'm used to seeing idiots slapping 1000uf or 2200 uf after them, but issues generally start at about 100-220uf.
Also consider that electricity does not just appear at a component, if you are driving a servo, then you will need some storage prior to the regulators, before they hit the power cable.

Are you sure you're not being overly cautious?

I have designed for millions of 'items' shipped, so yes when a simple choice can push your failure rate up .02 of a % and your ass is on the line you do get a bit cautious.

Also looking at your circuit design, route BOTH regulators supply rails (VCC, VSS) separately to the input terminals, ensure when the copper is routed that they are not daisy chained (if this is a single layer pcb)

9

Adding small-value ceramic capacitors in parallel with the electrolytics is common practice to remove high-frequency noise. They work like small low-pass filters. Typical values are 10 - 100 nF (0.01 to 0.1 uF) and don't generally affect the stability of the regulator (the capacitance is within the error margin on the bigger capacitor), but the frequency dependent rejection of electrolytics means that those are not as good at filtering high frequencies, so if you really worry about that 16 MHz clock signal leaking out, an additional ceramic is useful. Note that this is to filter noise from the AVR microprocessor, which the data sheet for the regulator knows nothing about.

One reason to put bigger capacitors on the output side of the regulator is that you can use lower-voltage parts. The difference between a 25V part and a 10V part (or even a 6.3V part) can mean significant space (and cost) savings. But, as already said: capacitors draw a lot of current during start-up, so beware of too high capacitive loads. One way to solve this is to put a smaller electrolytic after the regulator, then an inductor, then a bigger capacitor; the draw of the capacitor will be filtered by the inductor to give it a softer start. It all depends on how much you need to worry about your power.

10

hardcore:
You are not talking 10uf you are talking 47uf, as I said before electronics is like cookery, also be aware an electrolytic can be +-50 - +-50% of the marked value.

The 47uF is on the output of the 5v regulator. That section you quoted is for the input. This is the section dealing with the output:

The recommended load/decoupling capacitance is a 10uF
tantalum or a 50uF aluminum. These values will assure
stability for the majority of applications.

I understand that. I was just concerned that the 3.3v reg might overheat if the supply is 12v.

but looking at the data-sheets, it maxes out at 16v input and the other maxes out at 27 or difference depending on the model, why not use the same range parts (LM1085-3.3) for the 3v3.

The reason I didn't use the LM1085-3.3 is because the board I'm making is the size of a cellphone and I only need the 3.3v to drive a MicroSD card. The LM1085 would be overkill. I only need 50mA. It would also cost more and take up a lot more space on the board.

You are still within the spec, but ultimately it depends on WHAT load you are driving.

For the 3.3v I'll be driving a MicroSD card. For the 5v I'll be driving leds, servos, and a dac connected to a 2W amp.

The data-sheets cover what is around the regulator for optimum operation of the regulator, my recommendation was related to the fact that I did not know the power supply and cable length, also note that an IC is a 'microcosm' linked to components around it, once you get a few inches of copper between your IC's and power points, you will be glad of a bit of storage near your connector, before you hit the inductive load/Choke that is commonly called a power cable.

I see. Well I'll consider adding another 47uF cap if I have space for it in that case. I mean it's possible someone might try to run this off a wall wart. That's not really what it's designed for though. Size is a bigger concern right now.

Your suggestion also goes against my own experience with voltage regulators. I've built a number of circuits with only a .1uf on the input and a 10uF on the output of my regulator, and running off a 9v battery. These circuits were driving a couple noisy servos and reading analog inputs and I haven't had any problems. Are you sure you're not being overly cautious? I could understand sticking a larger cap on the input if I was running off something like a wall wart, but that isn't likely with this circuit.

Then go with what you know, personal experience trumps free advice (especially with engineers).

Well I'm still relatively new at this, so I want to get input from people and make sure I'm not making some horrible mistake. I'd like to save a bundle of cash and have the prototype come back assembled and have it work the first time. :slight_smile:

Also consider that electricity does not just appear at a component, if you are driving a servo, then you will need some storage prior to the regulators, before they hit the power cable.

But is a little 47uF capacitor going to do much if the battery can't supply what the regulator needs for a servo? I honestly don't know, but I am aware of how little a capacitor can store and that a servo may try to draw anything from 100mA to 1A depending on how much load is placed on it. For this reason, I'll be designing my circuit so I can provide a way to drive the servos directly off the batteries or another regulator if necessary. I'm fairly certain 2 small servos will be fine. And I'm pretty confident about 4 if they're not all under load. Beyond that, I'm not really sure what the board will be capable of because there's no precise numbers to go off of with servos.

Are you sure you're not being overly cautious?

I have designed for millions of 'items' shipped, so yes when a simple choice can push your failure rate up .02 of a % and your ass is on the line you do get a bit cautious.

Understood. :slight_smile:

Also looking at your circuit design, route BOTH regulators supply rails (VCC, VSS) separately to the input terminals, ensure when the copper is routed that they are not daisy chained (if this is a single layer pcb)

It's a 2-layer, but I'll keep that in mind for the supply side. The bottom layer will be my ground plane. Also, since the 10uF ceramic caps are cheap and small I may include one for each of the regulators after all instead of what I've done in the schematic.

11

Adding small-value ceramic capacitors in parallel with the electrolytics is common practice to remove high-frequency noise. They work like small low-pass filters. Typical values are 10 - 100 nF (0.01 to 0.1 uF) and don't generally affect the stability of the regulator (the capacitance is within the error margin on the bigger capacitor), but the frequency dependent rejection of electrolytics means that those are not as good at filtering high frequencies, so if you really worry about that 16 MHz clock signal leaking out, an additional ceramic is useful. Note that this is to filter noise from the AVR microprocessor, which the data sheet for the regulator knows nothing about.

I'll have a .1uf bypass cap near the Atmega on its supply pin, and on the supply pins of all my other chips. I assume those will perform the same function as placing one by the regulator.

12

The capacitor on the Atmega helps the Atmega when it sees a transient load spike. However, everything else powered off the 5V rail will also generate some amount of interference. Are you using PWM output for the servo control? LEDs? Other peripherals? Each of those are hooked to the 5V rail somehow, and contribute a little bit of EMI, which you can conveniently filter at the power regulator output, if you really worry about it.
For a battery powered board, it's not clear that I personally would actually worry about that case, but then, I have not shipped 2 million instances of any hardware design :slight_smile:
Also, instead of a 16V part on the output, if size is a problem, consider 10V, or maybe even 8V or 6.3V if you're feeling adventurous. The output is, after all, only 5V...

13

The 16v aluminum ones are actually the same price and size as the 10v version on Digikey:

The only other difference is the ripple current, but I'm not sure exactly what that is yet, or if more or less is better.

I went with the 16v one mainly because some fab houses (Screaming Circuits for example) charge you extra ($10) for each unique part number, and charge extra for cut tape (and Digkey charges $7 to reel them), so on the off chance I needed to use these caps on my input as well, to save myself some money I went for the higher voltage.

The only smaller ones are 6.3v, but the vast majority of sites I've looked at say to derate your caps to 2x the voltage you plan to run them at, so I'd like to stick to 10v or 16v.