Buck converter questions (integrated inductor?)

I am adding a buck converter SMPS IC to my PIC serial data translator/transceiver in order to be accept a source voltages of 12v or 28v DC, output 5V for IC's @about 50 to 60mA load. PIC is running on a 20mhz crystal. Attached is the schematic (The 7805 will be replaced by the SMPS).

It's my first time using a buck converter. I've read about them, done research, some shopping and now have a few questions. This device is for somebody else so i value reliability over component price. I can't have issues about brown-outs or EMI once it's shipped, obviously. I'd like to minimize footprint but that's a lower priority (but excludes ready made 20$ smps boards).

1. It seems the major hassles with buck converters are related to the inductor and pcb EMI due to traces. I've searched for "integrated inductors" to mitigate this and found that TI and Enpirion make some but they only accept input voltages up to 5.5v.

Anyone knows of integrated inductor buck converters accepting from 12v to 28v input? I might have missed them. Or are there other solutions without inductors? Preferably thru hole (haven't done SMD yet but could always solder it i suppose).

2. How much ripple can my design tolerate? Worst case is the recommended max / min IC supply voltages (taking into account the 4.35v max brownout on the pic)? Obviously i will try to minimize ripple as much as possible.

3. If i understand correctly, in designs using inductors, ripple is minimized by using low esr output caps and higher freq switching?

4. About inductor selection, is there anything i should know about or look out for when choosing for SMPS? AFAIK the specs i should care about are inductance value, max current and and physical size. I've never shopped for inductors before and maybe used one once like 20 years ago.

5. What is the effect of inductor resistance on SMPS? For my design, is this important and if so how do i evaluate the max allowable resistance?

I've shopped around and read other threads recommending specific ic's for their designs but if you know one in particular (stable, reliable) i'd appreciate the suggestion.

Thanks!

You will have to use a "real" inductor. Here is a typical buck converter...

Check out this Guide to Selecting Inductors for Switching Regulators

Have a look at the design tool (Webench) here: **broken link removed**, just type in your requirements and it will spit out a design for you.

Gordito said:

I am adding a buck converter SMPS IC to my PIC serial data translator/transceiver in order to be accept a source voltages of 12v or 28v DC, output 5V for IC's @about 50 to 60mA load. PIC is running on a 20mhz crystal. Attached is the schematic (The 7805 will be replaced by the SMPS).

It's my first time using a buck converter. I've read about them, done research, some shopping and now have a few questions. This device is for somebody else so i value reliability over component price. I can't have issues about brown-outs or EMI once it's shipped, obviously. I'd like to minimize footprint but that's a lower priority (but excludes ready made 20$ smps boards).

1. It seems the major hassles with buck converters are related to the inductor and pcb EMI due to traces. I've searched for "integrated inductors" to mitigate this and found that TI and Enpirion make some but they only accept input voltages up to 5.5v.

Anyone knows of integrated inductor buck converters accepting from 12v to 28v input? I might have missed them. Or are there other solutions without inductors? Preferably thru hole (haven't done SMD yet but could always solder it i suppose).

2. How much ripple can my design tolerate? Worst case is the recommended max / min IC supply voltages (taking into account the 4.35v max brownout on the pic)? Obviously i will try to minimize ripple as much as possible.

3. If i understand correctly, in designs using inductors, ripple is minimized by using low esr output caps and higher freq switching?

4. About inductor selection, is there anything i should know about or look out for when choosing for SMPS? AFAIK the specs i should care about are inductance value, max current and and physical size. I've never shopped for inductors before and maybe used one once like 20 years ago.

5. What is the effect of inductor resistance on SMPS? For my design, is this important and if so how do i evaluate the max allowable resistance?

I've shopped around and read other threads recommending specific ic's for their designs but if you know one in particular (stable, reliable) i'd appreciate the suggestion.

Thanks!

Click to expand...

Hello there,


After reading your post i see you dont have too much experience working with switch mode power supplies, so my recommendation for you would be to use one of the National Semiconductor "Simple Switchers". These come in a variety of ratings so you can pick what you need by visiting their site. They dont call them "simple" for nothing either, being so simple they are almost down right boring Just a tiny attention to layout (as per data sheet) and you 're there.

Yes output cap ESR (and even input cap ESR) play a role in the converter, but sometimes too low of an ESR in the output cap can cause more instability due to the nature of high Q LC networks in feedback loops. You can probably get away with something in the 0.1 ohm range, both input and output. You also have to check the ripple current rating to make sure it will hold up in your design.

The inductor series resistance affects the efficiency as well as the ability of the converter to attain the required output, but i doubt you will have a problem here with the low current you seem to need. Stay under 1 ohm and you're probably ok. It's only when you get into more high powered circuits that you need to be much more aware of the series resistance because of it's impact on efficiency. On the plus side, more series resistance makes the circuit more stable.

The inductor inductance affects ripple current which in turn also affects output ripple voltage, but if you work with around 100uH with a 50kHz converter you're probably ok there too. If you want really really low ripple, add a small post LC filter with a hand wound 2uH air core inductor and 100uf low ESR cap. It's very surprising what a post filter can do for little cost and small additional weight. It's important though to NOT connect the feedback line to the output of the post filter, but to keep the feedback at the regular output cap instead and use the post filter as a passive filter in line with the output, after the feedback connection.

The 'catch' diode is fairly important. It must be a high speed diode like a Schottky for example. If you try to use a run of the mill 1N4002 or similar diode you may not even get it to work at all.
If i rem right a 1 amp Schottky is 1N5817 and they are fairly cheap too, but check the voltage rating to be better than your app calls for.

Many of the Simple Switchers are already compensated too, so you dont have to play around with a little compensation network all day to get it stable.

STM Linear has some dip package switchers. I used a **broken link removed** 5.1V reference design, for a PIC project once. It was kind of klunky looking, but seems to work just fine.

Good info!

Thank you all for taking the time to reply and educate me. Especially Mr Al.

With the info, tips and tools provided i believe i can now properly select the correct design and components. The inductor selection guide is a godsend for an inductor newb like me.

I'll post an update in a few days once it's done.

Woaa the WEBENCH tool is wonderful.

Question: How important is Soft Start? From what i found in the L4962 datasheet it seems the output voltage could overshoot when the SMPS is turned on?

"Output overcurrents at switch on are prevented by the soft start function. (snip)
The output stage is thus re-enabled and the output
voltage rises under control of the soft start network.
If the overload condition is still present the limiter
will trigger again when the threshold current is
reached. The average short circuit current is limited
to a safe value by the dead time introduced by the
soft start network.

21 out of 41 initial designs have this function... is this something i need, considering only half of the ICs/ designs in WEBENCH with my params have it?

Hi again,


Soft start is a good idea when the output voltage can not be allowed to go over the set point even for a short time. I would say that if you power a PIC at 5v then you should use soft start.
What happens in the circuit when you first turn it on is the output isnt there yet (0v) and so there is no feedback. The internal circuit uses the feedback to measure the output and make corrections to the duty cycle to 'fix' the output voltage to the correct level (comparing it to a reference voltage). Since the output initially looks like 0v the internal circuit thinks the output is too low and so will ramp the duty cycle up fairly quick. This causes the output to shoot up (according to other circuit constants too). The output gets to the correct set point, but there is a delay for the internal circuit to recognize this and so it keeps ramping up the output (even though it is now past the set point). Eventually the internal circuit sees this new over voltage condition and cuts back the duty cycle, in which case the voltage dips low again, and then the internal circuit sees it low again so it increases the duty cycle yet again, and these cycles repeat for a short time. Eventually, after each up and down cycle the output voltage gets closer and closer to the actual set point until finally it actually settles at the correct set point. This process takes some time to complete and so the output swings up and down for a while, and unfortunately the 'ups' are going to be above the set point and that is not good for some voltage critical loads.
Using a soft start mechanism, the initial voltage upswing is slowed down to give the internal circuit time to properly deal with the initial 0v output so there is only a tiny over shoot until the circuit settles to the final correct output voltage. This is extremely important for some loads as they would simply blow up without it. The PIC may or may not blow out seeing as they can take up to 6.5v i think, but that's not a good idea anyway so some slow start is always a good idea. It usually doesnt take too much to get some sort of slow start built in, like maybe an extra capacitor or something, but it depends on the actual controller IC being used so you have to refer to the data sheet.

Thanks for clearing that up for me, i will get one with soft start. The power on reset for the PIC is 72ms so no problem there with the soft start period.

I checked the heat dissipation and for 0.1W the junction temp (ambient = 30 C) is around 42 C so no heat sink required if my understanding is correct (absolute max is 150 C and the datasheet graphs show that for that temp all seems good).

EDIT: I am doing through-hole components rather than SMD... if i go with equivalent values (i'll try to stick with the same manufacturers) can i expect it to still work as designed or i need to account somehow for the longer traces? Should i also order some (very) slightly lower value caps, for example, to account extra capacitance due to traces?

EDIT2: hmm i think i'll try my hand at surface mount using the parts numbers from the BOM, gonna waste less time that trying to find equivalent through hole components.

Make sure you follow any layout recommendations as closely as possible for the circuit you are using. That's critical for proper operation of a switching regulator.

Gordito said:

Thanks for clearing that up for me, i will get one with soft start. The power on reset for the PIC is 72ms so no problem there with the soft start period.

I checked the heat dissipation and for 0.1W the junction temp (ambient = 30 C) is around 42 C so no heat sink required if my understanding is correct (absolute max is 150 C and the datasheet graphs show that for that temp all seems good).

EDIT: I am doing through-hole components rather than SMD... if i go with equivalent values (i'll try to stick with the same manufacturers) can i expect it to still work as designed or i need to account somehow for the longer traces? Should i also order some (very) slightly lower value caps, for example, to account extra capacitance due to traces?

EDIT2: hmm i think i'll try my hand at surface mount using the parts numbers from the BOM, gonna waste less time that trying to find equivalent through hole components.

Click to expand...

Hi,


Im not exactly sure what you mean by "longer traces". What you do about it depends on where they are located and what pins they connect to. In general, you create a common connection point for all the grounds and also try to keep the feedback run as short as possible.
The capacitors you will most likely be using will be very much higher in value than the stray capacitance so i dont think you have to worry about that. Sometimes in the compensation network this might become an issue, but you can use an IC chip that does not require a compensation network and that takes care of that
If you let me know what chip you would like to work with perhaps i can offer some better ideas, and there may be other folks here that have worked with that chip that might have experience with it too.

MrAl said:

The capacitors you will most likely be using will be very much higher in value than the stray capacitance so i dont think you have to worry about that.

Click to expand...

Ah, that's what i meant by 'longer traces', stray capacitance (because the through-hole components are larger than SMD). Actually i would presume the biggest difference between a through hole and SMD implementation would be the ESR values for caps? If the layout is done as correctly as possible.

Since it's my first time doing a SMPS and wanting it to work reliably i have decided to follow the WEBENCH design doc EXACTLY - I ordered the exact SMD parts (only Digikey had them all BTW) and will copy the PCB layout given in the application notes for the **broken link removed**

I've attached the WEBENCH doc in case anyone is curious what design came out of it. At this time i prefer learning to solder SMDs rather than debugging a substituted component SMPS

Gordito said:

Ah, that's what i meant by 'longer traces', stray capacitance (because the through-hole components are larger than SMD). Actually i would presume the biggest difference between a through hole and SMD implementation would be the ESR values for caps? If the layout is done as correctly as possible.

Since it's my first time doing a SMPS and wanting it to work reliably i have decided to follow the WEBENCH design doc EXACTLY - I ordered the exact SMD parts (only Digikey had them all BTW) and will copy the PCB layout given in the application notes for the **broken link removed**

I've attached the WEBENCH doc in case anyone is curious what design came out of it. At this time i prefer learning to solder SMDs rather than debugging a substituted component SMPS

Click to expand...

Hi again,

Whether it be SMD or through hole parts the manufacturer should show the ESR on the data sheet.

I took a look at the parts list. How did you end up with an inductor as large as 680uH?
Did you have to enter the desired max ripple voltage on the output?

MrAl said:

I took a look at the parts list. How did you end up with an inductor as large as 680uH?
Did you have to enter the desired max ripple voltage on the output?

Click to expand...

Hi, i'm looking at the WEBENCH tool now and the only numeric inputs are Vin min, Vin max, Vout, Iout and Ambient temp.

Then the resulting designs can be filtered according to various characteristics, one being the Vout ripple. I didn't touch that (left it at max) and choose a design based on low part count and smaller footprint which happened to be the one with highest efficiency.

The tool doesn't seem to permit changing the selected design's inductor value and watching the effects on the characteristics or component values, which would've been educational for me (need to do that in a sim then).

Is 680uH a large inductor value in order to get 16.724mV ripple, when compared to other designs?

Gordito said:

Hi, i'm looking at the WEBENCH tool now and the only numeric inputs are Vin min, Vin max, Vout, Iout and Ambient temp.

Then the resulting designs can be filtered according to various characteristics, one being the Vout ripple. I didn't touch that (left it at max) and choose a design based on low part count and smaller footprint which happened to be the one with highest efficiency.

The tool doesn't seem to permit changing the selected design's inductor value and watching the effects on the characteristics or component values, which would've been educational for me (need to do that in a sim then).

Is 680uH a large inductor value in order to get 16.724mV ripple, when compared to other designs?

Click to expand...

Hi again,

Maybe you can state what values you input to the program and we can go from there, such as input voltage, output current, etc.
I mentioned the 680uH because the more typical value at around 250kHz would be 50uH to 250uH, but that's not to say that 680uH wont work.
Are you still looking for 50ma output, or higher or lower? And your input is 28v ?

It's kinda on the report:

Vin min 11V
Vin max 29V
Vout 5V
Iout 70mA

I agree that the inductor value seems high for a 350mW buck.

hmm i put 70ma Iout thinking i should enter a 'worst case' current, but now i'm thinking that buck converters probably don't work that way, i.e. they should be designed with the normal average current expected (around 50ma) and the design incorporates some output current tolerance? Although in my case the current will be constant.

Perhaps the higher inductor value is required to compensate this IC's lower part count design?

This might be getting a bit off-topic but when using the WEBENCH tool (or other generated SMD designs), do you guys trust it enough to skip prototype testing and directly design the final pcb and solder the parts rather than 'somehow' testing it (i suppose on a little SMD pcb with a center area for the IC)? I am not all clear on the workflow... Usually i breadboard everything before but that's all through-hole.

Thanks again to everyone participating in the thread, most informative

Read this...

indulis said:

Read this...

Click to expand...

Thanks, it complements and clarifies other info i found.

Using the equation therein I calculated the inductor value and it gives me 621 uH, a difference of -8.67% compared to the WEBENCH value so it's not too off.

LIR = (Ioutpp / Ioutmax) = 16.649mA / 0.07 = 0.237842857
L = (VinMax - Vout) * (Vout/VinMax) * (1/Fsw) * (1/(LIR* IoutMax))
L = (29 - 5) * (5 / 29) * (1 / 400k) * (1 / (0.237842857 * 0.07)
L = 0.000621348

Playing around with Vout values confirms that worst case for inductor value (in order to compensate for capacitor ripple) is when Vout = Vin / 2. In my case that would make it a 1.09 mH inductor at Vout = 14.5v