Yeah, I want to protect against LEDs being put in reverse. I've done this with signal LED's too much to not really worry about it - but a more expensive LED array wouldn't be on my list to blow up.

Well, that's going to be a tough one to accomplish, because with higher power arrays, the Vf can easily exceed what the max Vr is.

You can only do so much to protect against facepalms. Connecting an expensive high-power array backwards is something one would do just once, unless they are slow learners, gluttons for punishment, or just like seeing smoke.

re: play with it:

Thanks. The problem with the smaller caps are the LED gets hit my a high current on startup. But, I found with a 2.2μF cap across the LED and a 220nF to ground works very well, and prevents the high current surge. You can get 2.2µF 16V caps in 0603.

It's mainly the cap across the LED, because it couples the surge to the current sense resistor. This is basically a variation of the same theme I was talking about earlier in the thread; soft-start problems.

Also, I noticed you dropped the supply to 5V. This isn't going to work, or if it does, the efficiency will be poor. The synch. driver is only rated 6V - 9.5V on the Vcc and Vlogic lines. I'm replacing the supply with 7.5V...

Good catch; I'd just looked at the test fixture that Linear Technology supplied, and they specify 5V for Vlogic and Vcc inputs, which is in contradiction to the datasheet!

You should write them a note, indicating the discrepancy, and ask them which is correct.

... and am probably going to find some high voltage high speed comparator to run off this same line (any suggestions?)

Why not start with an LM111/LM311, and see how that does/doesn't work?

Remember, the LMx11 has an open-collector output; you'll need to use a pull-up resistor.

 

Well, that's going to be a tough one to accomplish, because with higher power arrays, the Vf can easily exceed what the max Vr is.

You can only do so much to protect against facepalms. Connecting an expensive high-power array backwards is something one would do just once, unless they are slow learners, gluttons for punishment, or just like seeing smoke.

Yes, I suppose. I'd like it to be an option, perhaps enabled with a simple switch, or an adjustable pot to set the maximum output.

It's mainly the cap across the LED, because it couples the surge to the current sense resistor. This is basically a variation of the same theme I was talking about earlier in the thread; soft-start problems.

So it's a good idea to make the cap across the LED fairly big but the cap from the sync to ground smaller. This corresponds with my findings..

Good catch; I'd just looked at the test fixture that Linear Technology supplied, and they specify 5V for Vlogic and Vcc inputs, which is in contradiction to the datasheet!

You should write them a note, indicating the discrepancy, and ask them which is correct.

My LTspice is from June 25, 2010, perhaps they have updated it since. If not I will get around to sending an email. The Vlogic supply can be as low as 3V, but the Vcc should be 6V at least. The problem is with lower voltages it can't drive the gates as well - you might get away with logic fets, but no guarantees. I'm fairly sure the datasheet is correct on this matter - the test rig would be based off the datasheet.

Why not start with an LM111/LM311, and see how that does/doesn't work?

Remember, the LMx11 has an open-collector output; you'll need to use a pull-up resistor.

How fast would the LM311 be? With small inductors and capacitors it needs a very fast response time...

 

Yes, I suppose. I'd like it to be an option, perhaps enabled with a simple switch, or an adjustable pot to set the maximum output.

Increased parts count = more stuff to go wrong, more board space, higher cost on the BOM...

So it's a good idea to make the cap across the LED fairly big but the cap from the sync to ground smaller. This corresponds with my findings..

You need to find a happy medium. The cap from output to ground will cut the ripple without delaying start-up excessively, and the cap across the LED prevents overshoot. The larger the inductor, the more likely you'll have overshoot. If the inductor is too small, you'll have power dissipation in the MOSFETs.

My LTspice is from June 25, 2010, perhaps they have updated it since.

Have you updated it using the sync release function? Click on Tools -> sync release.

The Vlogic supply can be as low as 3V, but the Vcc should be 6V at least. The problem is with lower voltages it can't drive the gates as well - you might get away with logic fets, but no guarantees. I'm fairly sure the datasheet is correct on this matter - the test rig would be based off the datasheet.

Check out the LTC4447; it works with a lower Vcc, and doesn't need the diode to feed the boost cap.
If you want to stay with the LTC4442, then go with the LTC4442-1; it's undervolt lockout threshold is higher than the non-1 version.

How fast would the LM311 be? With small inductors and capacitors it needs a very fast response time...

Do I have to read the datasheet to you?

The more I do for you, the less you're going to learn about this stuff.

 

Increased parts count = more stuff to go wrong, more board space, higher cost on the BOM...

More board space, but you can leave the pads empty if you want. And if a dual comparator were used, then there would be very few extra components (a few cents for resistors and capacitors.)

You need to find a happy medium. The cap from output to ground will cut the ripple without delaying start-up excessively, and the cap across the LED prevents overshoot. The larger the inductor, the more likely you'll have overshoot. If the inductor is too small, you'll have power dissipation in the MOSFETs.

I suspect at higher frequencies more power is wasted in the inductor, due to magnetic losses, and in the MOSFETs, due to switching losses.

Have you updated it using the sync release function? Click on Tools -> sync release.

I will try this later.

Check out the LTC4447; it works with a lower Vcc, and doesn't need the diode to feed the boost cap.

I can't find that chip anywhere. There's an incomplete entry on Linear's site. Perhaps it has been discontinued?

Do I have to read the datasheet to you?

The more I do for you, the less you're going to learn about this stuff.

Propagation delay: 200ns. Is this too slow? It's pretty fast, but it's about 50x slower than the current comparator. LTspice has no LM311 in its default install - you wouldn't happen to have one, would you?

 

Download the attached comparators.zip.
Put the .asy files in switchercad\lib\sym\comparators
Put the .sub files in \switchercad\lib\sub
There might be a problem with the single LM339. I'm not going to fiddle around trying to fix it this evening.

[eta]
Odd that they'd discontinue the LTC4447 with no notice on their website.

 

Download the attached comparators.zip.
Put the .asy files in switchercad\lib\sym\comparators
Put the .sub files in \switchercad\lib\sub
There might be a problem with the single LM339. I'm not going to fiddle around trying to fix it this evening.

I don't see any files attached...

 

I don't see any files attached...

dOH!

Ok, here it is:

 

I just tested it with an LM111. It works pretty well - almost indistinguishable from the original, which is good. So it appears a relatively cheap comparator can be used. I was thinking of using an LM393 dual version, because it allows for expansion - either the overvoltage thing I was talking about, or some other feature.

 

Try the LM2903 model; it's basically the same as the LM393 but in automotive temp rating with a lower Vcc range.

 

Try the LM2903 model; it's basically the same as the LM393 but in automotive temp rating with a lower Vcc range.

It works well too.

 

Hmm, for the +7.5V supply I'm probably going to use the LTC3631, unless anyone has any other alternatives. It will probably need a 100u inductor in a fairly big package (maybe 1206.)

Also for the MOSFETs I was considering a dual fet, one option would be Si4946EY. Dual fets however bring the disadvantages of more heat in a smaller package.

 

Actually what I am trying to make is a direct equivalent. Got the drilling done, now to put the other graphic on the other side, then the soldering starts again.

Do you think the price will fall under that of the buckpuck?

 

Hmm, for the +7.5V supply I'm probably going to use the LTC3631, unless anyone has any other alternatives. It will probably need a 100u inductor in a fairly big package (maybe 1206.)

I don't see a need for anything exotic in the logic/Vcc supply. A simple 8.2v Zener with a current limiting resistor and an NPN pass transistor should work OK; it would be cheap and small.

Also for the MOSFETs I was considering a dual fet, one option would be Si4946EY. Dual fets however bring the disadvantages of more heat in a smaller package.

I haven't looked at dual MOSFETs for awhile, and it's a bit too late to look at them this evening.

 

Do you think the price will fall under that of the buckpuck?

It is possible. Some of the parts are almost as hard to get as Buck Pucks. I have the current drawing posted, but I'll be updating them with the current parts list tonight or tomorrow and reposting them.

I finished the PCB tonight. I'll give it the smoke test tomorrow. For the last week I've bee reacquainting myself with home brew PCBs. The results have been very impressive so far. Once you have the procedure down it is much faster than wiring a circuit from scratch.

 

I don't see a need for anything exotic in the logic/Vcc supply. A simple 8.2v Zener with a current limiting resistor and an NPN pass transistor should work OK; it would be cheap and small.

How much power maximum? (35 - 7.5) * 50mA = 1.375W. That would require a heatsink and be quite big. I think the bug reg would be smaller, and be more efficient, but it is more complex.

I haven't looked at dual MOSFETs for awhile, and it's a bit too late to look at them this evening.

Okay, let me know later. I'm going to use this very same MOSFET in one of my projects, but it is for slow-speed on/off control. LTspice has a similar fet but not the exact same one.

 

Rds(on) with that dual MOSFET is kind of high.

Have a look at this one: http://www.vishay.com/docs/73849/si4972dy.pdf
Far lower Rds(on), and total gate charge is roughly half on channel 1, and 1/4 on channel 2.

The big difference is that the Vdss rating is 30v instead of 60v.

When you're selecting MOSFETs, start off by looking at Vdss values that are close to what you need.

You are going to find that most of the time, you will get better Rds(on) and Qg (total gate charge) numbers with the lower Rds(on) ratings. That's because the higher the Vdss rating, the physically thicker the gate channel has to be - which means that it also has to have more surface area, to try to keep the Rds(on) low enough to be reasonable, which jacks up the Qg.

 

Rds(on) with that dual MOSFET is kind of high.

Have a look at this one: http://www.vishay.com/docs/73849/si4972dy.pdf
Far lower Rds(on), and total gate charge is roughly half on channel 1, and 1/4 on channel 2.

The big difference is that the Vdss rating is 30v instead of 60v.

It's probably okay for supplies of more than 30V, unless one of the fets fails closed, which they have a habit of doing, apparently, if abused. In which case it would cause the Vds across the remaining one to meet the supply voltage, which would lead to both fets failing in a catastrophic fashion in a couple of seconds.

LTspice tells me with the current IRF7468 fets, the average power wasted in each is about 150mW. Any idea on how to calculate RMS power? This seems too low to be true.

When you're selecting MOSFETs, start off by looking at Vdss values that are close to what you need.

A Vdss of at least 40V - I want to work with up to 38V supplies.

You are going to find that most of the time, you will get better Rds(on) and Qg (total gate charge) numbers with the lower Rds(on) ratings. That's because the higher the Vdss rating, the physically thicker the gate channel has to be - which means that it also has to have more surface area, to try to keep the Rds(on) low enough to be reasonable, which jacks up the Qg.

I've heard this before. However, the LT4442 is a very powerful driver, capable of charging gate capacitances with up to 1.2A peak on both channels. This is probably enough to allow for a reasonably high Qg, of course not massive, but not too much to be concerned about when you have Qg of 20 and Qg of 25 to decide between, or something like that.

Also for the inductor I was considering a SDR0805-6R8ML (Datasheet). A 6.8μH inductor, Irms(max) = 3.2A, Isat(typ) = 4.1A, Rdc = 0.04Ω, SRF = 33 MHz. I've used SDR0604 inductors - slightly smaller ones - in previous projects, although I haven't got them in production yet. I want this driver to be able to power up to 10W LED's, and the general design to be able to scale up to 100W arrays (obviously with more powerful fets and inductors.) For 10W the inductor needs to be pretty beefy. I figured out with 10W load (2.1A, equiv. to 2x 2.4V LEDs) that the RMS inductor current is about the same as the LED - 2.11A. I might need to increase inductance, but the current doesn't fall with more inductance.

 

Well, then scratch the LM393 off your list; it tops out at 36V absolute maximum supply voltage.

If you're going for such a high input voltage, you'll need some more expensive parts. I don't know why you've settled on 38v, but you'll need to do some more research to find parts that meet all of your criteria.

I also don't know why you're increasing the size (value) of the inductor. The inductor you've chosen would be OK for up to maybe 1.5A; more than that is kind of pushing it.

There aren't any resistors on the gates yet. You'll probably wind up with ringing on them without resistors. Try 5 to 22 Ohms, or at least leave space on your board for them.

Make your traces wide and short as possible.
Don't forget that the comparator will require a 0.1uF/100nF bypass cap across it's supply.

There are plenty of things that will occur with real parts that you won't see in a simulation.

 

Well, then scratch the LM393 off your list; it tops out at 36V absolute maximum supply voltage.

If you're going for such a high input voltage, you'll need some more expensive parts. I don't know why you've settled on 38v, but you'll need to do some more research to find parts that meet all of your criteria.

It would probably operate off the 7.5V supply, there's plenty of output current to accomodate it. The supply is 60V automotive rated (overvoltage lockout), 45V continuous, but the driver is only rated to 38V maximum.

Higher voltages allows for higher output currents, perhaps up to 100W as I said, with bigger inductors+fets.

I also don't know why you're increasing the size (value) of the inductor. The inductor you've chosen would be OK for up to maybe 1.5A; more than that is kind of pushing it.

What is limiting the output? Will the inductor overheat at high current levels, or will it just not work?

There aren't any resistors on the gates yet. You'll probably wind up with ringing on them without resistors. Try 5 to 22 Ohms, or at least leave space on your board for them.

Good idea - I will leave some pads for them, but probably just use zero ohm jumpers.

Make your traces wide and short as possible.
Don't forget that the comparator will require a 0.1uF/100nF bypass cap across it's supply.

Both good suggestions. I've designed a SMPS buck before, but around a custom chip. It was a 1.6 MHz chip, an LM2734X. That allowed me use of small inductors, but at the cost of the layout having to be perfect.

There are plenty of things that will occur with real parts that you won't see in a simulation.

Like the comparators - do these have real-world slew rates, or are they perfect comparators with limited input ranges?

 

It would probably operate off the 7.5V supply, there's plenty of output current to accomodate it. The supply is 60V automotive rated (overvoltage lockout), 45V continuous, but the driver is only rated to 38V maximum.

I don't know what you expect me to say about the above statement; it doesn't make a lot of sense.

Higher voltages allows for higher output currents, perhaps up to 100W as I said, with bigger inductors+fets.

If you want to make this thing somewhat flexible, you might plan on using discrete MOSFETs instead of arrays, as you'll have a much wider selection.

re: inductor

What is limiting the output? Will the inductor overheat at high current levels, or will it just not work?

Did you look at the specs? If you want good performance and long life, then de-rate the components by 50% or more. Irms(max) for that inductor is 3.2A, and it will saturate @4.1A. You do not want to get close to saturating it, or you will see smoke.

See Ronald Dekkers' "Flyback Converters for Dummies" page for a 'scope image of inductor saturation. Power dissipation in the upper MOSFET will skyrocket.

re: gate resistors

Good idea - I will leave some pads for them, but probably just use zero ohm jumpers.

You'll have to see how it does on a test bench. Start out using resistors, and then decrease their values to see what ringing you get.

re:real-world

Like the comparators - do these have real-world slew rates, or are they perfect comparators with limited input ranges?

Why don't you test the model using LTSpice, and see if it agrees with the datasheet?