Buck Puck Equivalent

Thread Starter

Wendy

Joined Mar 24, 2008
23,415
OK, I've finished drawing this sucker, which took longer than usual (added some changes to my PaintCAD because of it).

Anyhow, this is only a proposed design. The MOSFET sets the minimum voltage, with a digital type it can go down to 5V, with a conventional it goes down to 10V.

I went with negitive logic on a lot of it to favor a N-channel MOSFET.

Any comments or flaws noted?



OK, I noticed the chip designations. This will be fixed in a future revision.
 

Attachments

THE_RB

Joined Feb 11, 2008
5,438
I'm curious why you would go for such a high parts-count design? You've got 8pin IC and a 14pin IC etc. And dual rail supplies?

You could have used a 34063 chip (8pin) with a couple of external parts. That chip will also will drive lowish power leds (say up to 1W) without needing a FET!

Or if you like tinkering there are some great 2 and 3 transistor efficient switching constant current LED drivers here;
http://www.romanblack.com/smps/a05.htm
 

Thread Starter

Wendy

Joined Mar 24, 2008
23,415
Actually, no dual rail supplies. No ground, just plus and minus, which is different than my usual format. Since this circuit is open ended (it can wind up anywhere in any application) I didn't want to use a ground symbol. Have a suggestion for another way to show it?

As to parts count, there are chips out there that have all the parts integrated (for that matter I suspect the puck buck does), but if you're going to build a SMPS type current regulator that is pretty much a block diagram. The existing design will go up to 1.2A, with tweaking it can do more. I suspect L1, C2, and R2 are pretty much standard.

When I get a chance I'll look up your link.

This was an excersize to understand what might be in a PuckBuck. It answers the question why they didn't connect the LED to one power supply lead or another.
 
Last edited:

Thread Starter

Wendy

Joined Mar 24, 2008
23,415
I saw his SMPS design previous, it is interesting. I'm having a little trouble understanding the logic on the transistors and how they're switching. That is a topic for another thread however.
 

SgtWookie

Joined Jul 17, 2007
22,230
OK, so let's finish up this thread.

1) You're trying to use an LM358 to sense near Vcc. Although the LM358's common mode input range extends to ground, it can't "see" anything above +V -1.5 at room temperature, or within +V -2 across the temp range. Since you're trying to sense at near +V, you'd need an IC that had rail-to-rail inputs.

2) The LM358 has a unity gain bandwith of 1.1MHz on a clear day with a tailwind while going down a steep hill, but usually more like 0.7MHz. While that might be OK if you were using it at unity gain, you're using it open loop; not just once, but twice, as a comparator. I've been rather vocal in the past about abusing opamps as comparators, so I'll skip almost that whole thing - but just keep in mind that a comparator will run circles around an opamp in open loop.

3) The lower half of the 556 feeding into the 2nd half of the LM358 - I'm curious as to how you expected this to work? You're running the 1st half of the LM358 in open-loop, and that implies the output will be driven to the rails. The voltage on C1 (thus the noninverting input) will vary between 1/3 +V and 2/3 +V, while the inverting input will vary between nearly ground and +V -2v or so. Since that part isn't really doing what you thought it would, all it's doing is adding a huge delay in response to the control loop signal.

4) Using 1/2 of a 556 as a Schmitt trigger - well, sure you can do that. I don't know offhand what the propagation delay will be through it, though. One thing about switchers that's critical is minimal propagation delays in the control loop. From the time the control loop receives an input to the time it switches off the MOSFET, our poor little LED will have been blasted with maximum current for quite some time.

5) L1 is much too low a value for this slow control loop. Unfortunately, the control loop is so slow that you would need a value of inductance so large it would require a small pickup truck to haul the inductor around in.
 

Thread Starter

Wendy

Joined Mar 24, 2008
23,415
Ouch.

Pin 1 of the 358 is an analog point, that will be linear. It doesn't go rail to rail, doesn't have to, since it is compared to a rough triangle waveform that is 1/3 to 2/3 of total power supply voltage. Outside that range the feedback loop becomes digital.

The second 555 is specifically to handle the lack of rail to rail ability of the second half of the 358, and serves a simple driver for the MOSFET.

So in both cases the LM358 outputs only needs operate in the 1/3 to 2/3 voltage range of the power supply, rail to rail was never an issue.

I thinking this doesn't have to be high speed, I don't believe it has to be (and I could be wrong), since we are talking a relatively steady state for an LED. I am concerned about the power up surge, which could be an LED popper. That was the only comment you gave that raises my concern, but even with high speed components it could be a problem. Since this uses a fundamental PWM I was thinking it didn't need to be that fast, but I'm probably wrong on that issue.

The PWM oscillator is at 50Khz, the low pass filter formed by L1 and C1 is somewhere around 7.4Khz. The original PuckBuck does all of this with a coil that fits in the total footprint around a quarter, and only slightly thicker.

So it boils down to a basic problem (other than whether it will work or not), is there a way to soft start this circuit?

If you were doing this, how would you it?
 
Last edited:

Thread Starter

Wendy

Joined Mar 24, 2008
23,415
I'm curious why you would go for such a high parts-count design? You've got 8pin IC and a 14pin IC etc. And dual rail supplies?

You could have used a 34063 chip (8pin) with a couple of external parts. That chip will also will drive lowish power leds (say up to 1W) without needing a FET!

Or if you like tinkering there are some great 2 and 3 transistor efficient switching constant current LED drivers here;
http://www.romanblack.com/smps/a05.htm
Thanks for the heads up on the 34063, I downloaded the datasheet, it is an interesting chippie. That datasheet didn't have schematics on how to use it as a constant current source at first glance, but it looks doable.
 

Attachments

SgtWookie

Joined Jul 17, 2007
22,230
I was assuming you actually wanted feedback as to what I was thinking, and maybe some simulation results. Well, it's both. I thought to myself, this doesn't look so good, but I'll simulate it anyway. The simulation was worse than what I was thinking.

Pin 1 of the 358 is an analog point, that will be linear.
Do you mean the left LM358? If so, no; you're running the opamp open-loop, so it will be at one rail or the other. After a very long and slow transition, that is. The only point where it will be linear is during the transition.
It doesn't go rail to rail, doesn't have to, since it is compared to a rough triangle waveform that is 1/3 to 2/3 of total power supply voltage.
That's one of the problems; the only time it'll be in that range is during transition.

If +V is 12v and -V is ground (0v), the triangle waveform will be between 4v and 8v.

The LM358 output will be either under 1v or about 11v.

Outside that range the feedback loop becomes digital.
But you're running what should be a digital feedback loop through amplifiers designed for linear operation with internal compensation. Comparators aren't internally compensated, which makes them lousy for use as opamps, but superb (and fast) for use as comparators.

The second 555 is specifically to handle the lack of rail to rail ability of the second half of the 358, and serves a simple driver for the MOSFET.
I see. Well, some 555's can oscillate at up to 2MHz, the majority can get up to around 500kHz, so that's not too bad. Still, that does add to the propagation delay in the control loop.

So in both cases the LM358 outputs only needs operate in the 1/3 to 2/3 voltage range of the power supply, rail to rail was never an issue.
Well, they're in constant saturation because they're being run open loop. It takes a long time for them to come out of deep saturation; they weren't all that fast to begin with.

I (was) thinking this doesn't have to be high speed, I don't believe it has to be (and I could be wrong), since we are talking a relatively steady state for an LED.
You want to maintain a relatively stable current of 1A across a more-or-less fixed voltage drop using a small inductor; 10uH. That will require a very fast control loop. Without doing any analysis, a SWAG tells me you'd need around 500kHz switching speed or better; most probably the latter. The slower the switching speed, the larger the inductor you'll need. A single LM358 could not keep up with that, even running at unity gain.

I am concerned about the power up surge, which could be an LED popper. That is the only legitimate comment I saw from you, but even with high speed components it could be a problem. Since this uses a fundimental PWM I was thinking it didn't need to be that fast, but I'm probably wrong on that issue.
Gee, my feelings are now bruised. :( ;)

Really, I'm giving you my honest opinion, backed up by what I've observed in simulations of your circuit, and tried to lace it with a bit of humor, too. You're one heck of a great contributor around here, and I had no intention of hurting your feelings or bruising your ego.

But as I mentioned in the other thread, trying to engineer a switcher from numerous IC's, particularly those as old and slow as LM358's and 555/556 timers is bound to be a frustrating and less than fruitful endeavor.

The PWM oscillator is at 50Khz, the low pass filter formed by L1 and C1 is somewhere around 7.4Khz. The original PuckBuck does all of this with a coil that fits in the total footprint around a quarter, and only slightly thicker.
The PWM oscillator (left half of the 556) actually oscillates at 40kHz in simulation, using ideal components. However, since the LM358 is running open-loop, the PWM oscillator has no effect.

So it boils down to a basic problem (other than whether it will work or not), is there a way to soft-start this circuit?
I'd played around with something vaguely similar awhile back, but I was using a high-speed opamp feeding the control voltage to a 555 timer. Startup was finicky.

If you were doing this, how would you it?
I'd use a current-mode switching PWM regulator IC that was specifically made for the task. A UC2842 would be a good candidate.

These types of things are the 555's of the switching power supply world. There is a practically bewildering assortment of these available, each with their own strengths and weaknesses; sort of like the differences between CMOS and bjt 555 timers, only far more diverse.

There are people on here with vast background knowledge and experience in power supplies; they know more than the both of us (and another several dozen or so thrown in) ever will.

Power supplies are really a pretty specialized field. In order to get really good at them, you would have to immerse yourself in them for a considerable period of time.

I'm not anywhere near that good at it. But I can read a datasheet, plug my numbers into the formulas, and get pretty close to what the datasheet claims.
 

Thread Starter

Wendy

Joined Mar 24, 2008
23,415
My feeling aren't hurt, but you sounded angry in post #5, which did puzzle me some. I thought the point of the 1st op amp operating rail to rail was pretty far off base too, which made me wonder how long you looked at it. I do plan on using this op amp input pretty near a rail, I'm under the impression it can handle it.

I'm not actually trying to design something I would call practical, I've seen too many designs that were better. I do want to make something that will work. If it works I'll have the theory down. I think the PWM should transfer through the second op amp pretty much intact, we'll see. I'm using the second op amp as a comparitor because I needed one and there it was. The 1st op amp should work OK as used. I used the 556 for the MOSFET for pretty much the same reason, it was convenent and it wouldn't keep the MOSFET in the linear reagon like the op amp would.

I think I've solved the soft start issue using C3, as long as the LED is not plugged in after powerup.



I'm thinking as a test I'll use a conventional LED biased to 20ma. If this circuit doesn't pop a wimpy and cheap LED (and regulates the current) then it works. I'll also use a variable resistor to check the range.

You can almost tell when I'm thinking of breadboarding something when I include power pins and pin numbers.
 

Attachments

SgtWookie

Joined Jul 17, 2007
22,230
My feeling aren't hurt, but you sounded angry in post #5, which did puzzle me some.
That's the problem with posting on forums; no body language or tonal inflection to help decide whether someone's kidding around or serious about it.

I can fairly say that I was somewhat annoyed when I made that post, as I really didn't want to wade through this subject; I consider myself a neophyte in power supply design. There are many experts on the subject; I profess myself to not be among them. However, I did attempt to lighten the mood a bit by injecting a bit of humor here and there.
I thought the point of the 1st op amp operating rail to rail was pretty far off base too, which made me wonder how long you looked at it. I do plan on using this op amp input pretty near a rail, I'm under the impression it can handle it.
It can handle sensing inputs down to ground. However, don't count on it sensing within 2v of the positive rail.

The output will go within 200mV or less of the negative rail, but even with a light load won't go more than about +V -1v.

I'm not actually trying to design something I would call practical, I've seen too many designs that were better. I do want to make something that will work. If it works I'll have the theory down.
I understand. However, these switchers have to have really super-fast response times. If you're using 1970's technology on the front end, you'll never get there.

I think the PWM should transfer through the second op amp pretty much intact, we'll see. I'm using the second op amp as a comparator because I needed one and there it was.
Bill,
I've simulated it already. I spent a couple of hours playing around with it last night when I should've been asleep, and had to suffer my spouses' wrath due to it. It won't work as you've drawn it. I've already described why it won't work.

The 1st op amp should work OK as used.
Maybe in a really, really low-speed application. This isn't a low speed application, and it's running open-loop, which makes the response time really terrible.

I used the 556 for the MOSFET for pretty much the same reason, it was convenient and it wouldn't keep the MOSFET in the linear reagon like the op amp would.
Yes, I understand that.

I think I've solved the soft start issue using C3, as long as the LED is not plugged in after powerup.
C3/R3 delays the control loop even more.



I'm thinking as a test I'll use a conventional LED biased to 20ma. If this circuit doesn't pop a wimpy and cheap LED (and regulates the current) then it works. I'll also use a variable resistor to check the range.
Enjoy the fireworks :)

You can almost tell when I'm thinking of breadboarding something when I include power pins and pin numbers.
Yep!

I'm just trying to save you (and perhaps others) from a melt-down.

Trust me, your results won't be pretty. But I know you'll have to try it.
 

Thread Starter

Wendy

Joined Mar 24, 2008
23,415
I didn't know about the +V side input problem on the LM358. You mentioned it (in clear concise english), but I thought you were talking about the output (I had spotted the Darlingtons). I may change course and go with a P-channel to allow the setpoint to be near the negative rail instead of the positive side. I've never used a 358 before, does it show?

*Sounds of schematic paper being crumpled*

I planned on C3 delaying the loop, that is the softstart. I wouldn't do this with a dynamic circuit that varies a lot, but I'm thinking in terms of a slow drift on the LED.
 

SgtWookie

Joined Jul 17, 2007
22,230
I didn't know about the +V side input problem on the LM358.
I suspected it, but I had to review datasheets to be certain.

I basically regurgitated what the datasheet said, only in easy-to-read terms for the benefit of the n00bs on the forums.

You mentioned it (in clear concise english), but I thought you were talking about the output (I had spotted the Darlingtons).
They suffer from similar limitations.
I may change course and go with a P-channel to allow the setpoint to be near the negative rail instead of the positive side. I've never used a 358 before, does it show?
With each "new" part comes a few bruises, destroyed brain cells and a bit of smoke...

*Sounds of schematic paper being crumpled*
I can only imagine the gigantic piles of scrap paper generated by budding power supply engineers.

I planned on C3 delaying the loop, that is the softstart. I wouldn't do this with a dynamic circuit that varies a lot, but I'm thinking in terms of a slow drift on the LED.
I know. However, the R3/C3 network adds even more of a phase delay in the feedback loop that's already running at a snail's pace. The 'soft start' simply makes the situation much worse.

The control loop portion wouldn't work well even for a low-frequency audio circuit - and for this application, you need a response time several orders of magnitude lower than exists in the schematic you drew up.
 

THE_RB

Joined Feb 11, 2008
5,438
It's extremely easy to do constant current with a 34063 because you only need one resistor between pins 6 and 7 to give current limiting (limit = 0.3v).

AND if you don't need to regulate voltage you can dispense with the entire voltage detect system.

Also for LED currents up to 1.5 amps you can use the 34063 internal switch, so all you need is;
* 34063
* current sense resistor
* schottky diode
* inductor
* in and out caps

That's only 6 parts in total for a constant current buck SMPS. (edit) whoops you need a 470pF cap on pin 3 too, so that's 7 parts total)

But personally i still think 2 transistors beats it. ;)
 

Thread Starter

Wendy

Joined Mar 24, 2008
23,415
Thanks. I'll try to pin the design down where I'm comfortable with it and start showing it to the intermediate users.

I'll probably to bug the guys on the other site and see if I can wrap my head around that design.

It will also end up in my cookbook, which I've barely started, in my blog.
 

THE_RB

Joined Feb 11, 2008
5,438
Ahh! It's for your 555 cookbook? :)

I've done a number of SMPS circuits using a 555, usually you need to add a transistor as the voltage or current detect. So you need a 555, a transistor as a detector and the output switch FET/transistor etc.
 

Thread Starter

Wendy

Joined Mar 24, 2008
23,415
Actually, long term it will be an everything cookbook. At least that is my goal. I mentioned it in the feedback and suggestions, everyone thought it was a good idea, but VB isn't quite up to it... yet. The biggest showstopper is the lack of any editing for the albums, you take em as it presents em. For a cookbook to be useful it has to be organized in a logical manner.

The new blog I've started is in it's infancy. I originally started as a test, then it occurred to me that this could work. It might be possible to expand it long term, writing this post some new ideas occurred to me.

For those who haven't had the pleasure of using one, cookbooks are durn useful to have kicking around, an excellent technical resource of the first magnitude.

I'd like to see those other circuits. I'm never too proud to use what works.
 
Top