I don't think this was a heatsink issue, the conductance was plain not suitable. Off to buy a replacement.

 

900mA is too much for your 640mA fet.

But in the three seconds of its life did it keep good regulation?

 

Who knows, the DVMs did not have a chance to settle down. My suspicion is not.

I've changed my mind about using a conventional MOSFET too. The circuit is specifically designed to give the gate a logic level voltage, it would be too much hassle to change over.

I think the FET had 300ma on it's side, since that was the input current. The output current is a cumulation of the capacitor/inductor/schottky interaction.

So I just need to get the replacements on order.

 

Bill,
Have a look at this one:
http://www.mouser.com/ProductDetail...=sGAEpiMZZMtCrm2fS1SYQpifKMb8GHwHS1wfYfxeAYg=
Logic level, Vdss=-20, Rds(on)=50m when Vgss=-4.5, Qg=25nC, Id=-24A, $1.31 each.

They have much cheaper MOSFETs in SMT/SMD, but I'm assuming you wanted a TO-220 package. Mouser really doesn't have any other viable options in a TO-220 package; cost and gate charge go up radically.

 

Looks good. Thanks. Gotta love a conductance of 0.005Ω. The amperage looks good for future projects too (like my 6.5A power supply).

 

Keep in mind the Vdss rating; only -20v. That means you won't be able to get to your original goal of 24v for the supply, but it'll work OK at lower voltages.

I'd also look for a Zener that regulates with a lower current; as I mentioned before, the 1N4733 just won't work very well due to it's high current requirement.

This one would be just about perfect, and at $0.03/ea it's darn cheap:
http://www.mouser.com/ProductDetail...GAEpiMZZMstCHp3EWKGl7RSz/WB%2byfUnVQsXENBRZQ=

 

What do you think will happen with a 1N4733A? Slow turn on of the MOSFET?

I've already ordered the MOSFETs, unfortunately.

 

What do you think will happen with a 1N4733A? Slow turn on of the MOSFET?

I've already ordered the MOSFETs, unfortunately.

The comparator will not be able to sink the 49mA required to keep the zener in regulation; or if it does, it will be damaged.

 

OK, I have the new MOSFETs in, but before I go there I need to go over some things (IE, the zener).

In this schematic the zener is solely to prevent putting too much voltage on the MOSFET gate. As such it isn't that critical, as long as the minimum saturation voltage is met.

The 1N4733a will not do that, but a large zener voltage will in the same family. Remember, part of the design goal here is off the shelf parts. I had to order the MOSFETs special, and I would have ordered the zeners if I had thought of it on time.

Here is an experiment I ran with the 1N4733a. The odd selection of numbers for currents is for logarithmic graphing. I replaced the resistor as needed to select between ranges.

..........Voltage
Current....Drop
100 µa....3.25V
178 µa....3.44V
316 µa....3.66V
562 µa....3.89V
1.00 ma...4.10V
1.78 ma...4.32V
3.16 ma...4.52V
5.62 ma...4.69V
10.0 ma...4.83V
17.8 ma...4.94V
31.6 ma...5.02V
56.2 ma...5.09V
100 ma....5.16V

Basically the minimum voltage drop has to turn the MOSFET on all the way, and we're cool. I am figuring this is at 5V for the MOSFET, and no more than 10V across the gate source at maximum voltage.

Anything wrong with this argument?

I originally stated the target voltages are 6V to 24V (maybe 30V). I'm going to have to look at that again, but those are the numbers I'm striving for. With the voltage divider formed by R4/R5 I don't think it will do it, I'm going to have a second look at those values while I'm looking at the zener circuit. I'll set up an experiment similar to above only with the R4/R5 values included and see what the zener does.

After typing all of the above I'm rethinking these specs, 6V to 10V across the gate/source. A zener of 9.1V (1N4739A) would do this nicely. I'm aiming for a 6V - 30V power supply, so I'll test it to 15V, then after all the main tests I'll go for the 30V spec in steps.

 

..Power
.Supply...Zener
Voltage..Voltage..Current
.6.0 V...5.22 V...0.4 ma
,7.0 V...6.09 V...0.5 ma
,8.0 V...6.95 V...0.6 ma
,9.0 V...7.94 V...0.7 ma
10.0 V...8.87 V...0.8 ma
11.0 V...9.16 V...1.3 ma
12.0 V...9.17 V...2.1 ma
13.0 V...9.18 V...3.0 ma
14.0 V...9.19 V...3.8 ma
15.0 V...9.21 V...4.6 ma

I'm probably going to try the old MOSFET one more time with the new drive circuitry to give it a fair chance.

 

Looks good. Thanks. Gotta love a conductance of 0.005Ω. The amperage looks good for future projects too (like my 6.5A power supply).

Rds(on) is specified as 0.05 (50m Ohm) when Vgs=-4.5v; 0.07 (70m Ohm) when Vgs=-2.7v.

 

Good enough, but we can't plan on every logic pMOSFET being as hand picked as these I bought. The driver changes will work with either extreme of MOSFET, maybe even non-logic types (if the power supply is adjusted). I'm wondering if the driver contributed to the other pMOSFET failure, probably not, but I'll adjust the specs a bit more open ended and see.

If I were using non logic types I'd probably eliminate the zener and R4 all together, and adjust R5 accordingly.

The 9.1V units seem a bit tighter and wider than the 5.1V units. I've heard the higher voltages tend to work better. If we went much lower than 5.1V I'd be tempted to use LEDs. Can't be much worse.

How's it going Wookie? Me, off to a DeMolay advisors meeting. It never stops.

I'll redo some of the experiments in post #150 to reduce the range. It occurs to me that a narrower range may speed up switching times. 4.5V will be my set minimum, max is as low as I can get.

 

OK, I have the new MOSFETs in, but before I go there I need to go over some things (IE, the zener).

In this schematic the zener is solely to prevent putting too much voltage on the MOSFET gate. As such it isn't that critical, as long as the minimum saturation voltage is met.

My objection to it's selection is that in the datasheet for the Zener, Izt (Zener test current) is specified as 49mA; and the LM339's output is only capable of sinking ~1/10 that much current before its' saturation voltage becomes excessive. Getting Vgs to -5v quickly ensures that the Rds(on) will be minimal. Now, this particular MOSFET will turn on suitably if Vgs gets to ~-3v - so the 1N4733 might work, but one would have to test them to see what the actual breakdown voltage was for the current going through it. It's a case of using a part outside of it's specifications; one might work and nine more might not.

I am a stickler about going by what's in the datasheets. If one follows the datasheet specifications for a given part, they're pretty much guaranteed that it will work properly, and reliably. The further one deviates from the specifications, the more "iffy" the design becomes.

The 1N4733a will not do that, but a large zener voltage will in the same family. Remember, part of the design goal here is off the shelf parts. I had to order the MOSFETs special, and I would have ordered the zeners if I had thought of it on time.

Well, the Zener that I linked to IS an off-the-shelf part (and dirt cheap, too); but you won't find it at a local Radio Shack store.

If you really want an off-the shelf solution, one could use a "rubber Zener" instead; two resistors and a transistor. 1k from base to emitter, and 5k-10k from collector to base, depending on the transistor used.

Here is an experiment I ran with the 1N4733a. The odd selection of numbers for currents is for logarithmic graphing. I replaced the resistor as needed to select between ranges.

..........Voltage
Current....Drop
100 µa....3.25V
178 µa....3.44V
316 µa....3.66V
562 µa....3.89V
1.00 ma...4.10V
1.78 ma...4.32V
3.16 ma...4.52V
5.62 ma...4.69V
10.0 ma...4.83V
17.8 ma...4.94V
31.6 ma...5.02V
56.2 ma...5.09V
100 ma....5.16V

Basically the minimum voltage drop has to turn the MOSFET on all the way, and we're cool. I am figuring this is at 5V for the MOSFET, and no more than 10V across the gate source at maximum voltage.

Anything wrong with this argument?

Did you test it over temperature, too?
What happens if the Zener is in boiling water?

I originally stated the target voltages are 6V to 24V (maybe 30V). I'm going to have to look at that again, but those are the numbers I'm striving for.

Max Vds for that MOSFET is 20v, which I mentioned before. You should keep the input voltage under 20 unless you want to kill the MOSFET. There simply were not any other viable options from that distributor in a TO-220 package.

With the voltage divider formed by R4/R5 I don't think it will do it, I'm going to have a second look at those values while I'm looking at the zener circuit. I'll set up an experiment similar to above only with the R4/R5 values included and see what the zener does.

After typing all of the above I'm rethinking these specs, 6V to 10V across the gate/source. A zener of 9.1V (1N4739A) would do this nicely. I'm aiming for a 6V - 30V power supply, so I'll test it to 15V, then after all the main tests I'll go for the 30V spec in steps.

A 1N4739 has much too high of an Izt specification (28mA). People will have to test each of them at the current which will be flowing through them, as iin this application t's just too far out of the specs.

The BZX79 series Zeners use 5mA for Izt up to 24v, which is just peachy. I simply don't know why you are hung up on using the 1N4728A-1N4752A series Zeners, as they just aren't very appropriate.

Going to a higher-voltage Zener means that R5's value will be more critical when changing Vcc.

 

About the family of zeners, its simple enough. I have 4 separate outlets in this town, not one sells anything other than the 1N47XX series, or larger. I don't know it, but I think the NTE series is also the 1N47XX series, which makes that 6 local vendors. Part of the issue is about availability.

I've never used anything else over 30+ years, over an extremely wide range of conditions, and have never, ever had any problem with using them, current specs aside. This includes both ends of that current spec, the only real number I've had to pay attention to is the wattage. I have used them in zero crossing circuits a lot, where the resistor dropped all the voltage and the zener was fundamentally off.

Then there is the application. The only reason we have them is to keep the power supply from exceeding the gate source breakdown voltage (which is 8V rereading the specs). They simply aren't that critical, especially in the low end of the power supply range where the resistor voltage divider is doing all the work. If you don't exceed 9V for the power supply you could leave them off and not notice. You could also replace them with a forward biased red and blue LED, which would provide good visual feedback on how the circuit is working (and it would never exceed the current specs of the LEDs). Looking at the new MOSFETs specs I don't really have to change anything, but I'm bumping the zener up to 5.6V.

When you first started talking about the zeners I though it might have been some other spec (still could be an issue, I'll take oscope readings), such as frequency response. There may be other issues I'm not considering.

I have a full stock of these diodes, similar to my ¼W resistors, starting at 3.6V and ending at 18V, and before that I still had a junk box full of them. They are standard as it can get. The only diodes I have off the base are bigger in current spec (higher wattages), not less.

When we publish the final design I'll recommend the other zener when the hobbiest buys their MOSFETs, but for the moment I'll use what I have. I'll be including all the data, including the parts I used. Again, the reason is I don't feel it is that critical overall, unlike the MOSFET specs. I will continue doing basic experiments to verify this though. I may even build a version using two LEDs.

I had missed the pMOSFETs 20V drain source breakdown specs. I would have preferred 30V since it sets the upper voltage limit, but now I have the parts I'm not going to worry about it. 15VDC upper power supply limit it is, maybe 18VDC, it is still good for some automotive power supplied applications such as map lights or whatever.

I spent this week end drawing for the 555, LEDs article, other than the zener experiments. The LED buck project is very close to being finished, but the fat lady has yet to sing. I'm thinking of making the circuit board too, and testing it to destruction (it will be easier to repair). It's not like I actually need this project... It has been an interesting learning experience though, one of my major motivators.

When it is lighting a pseudo LED I'll be picking your brains for test protocols. Maybe we can get the zeners to fail some other way, and the 0.05Ω resistor isn't that robust either. Chewing gum and duct tape anyone?

 

OK, I have the drawing updated. With luck they will work, I get back with you.

The 3rd drawing (not displayed) is a sheet to make transfers.

 

Good work Bill. Wait to see your work.

I am just thinking for testing the supply you could probably just get away with your multimeter on the 10A range with maybe a 10W 10 ohm resistor in series to protect both the meter and the supply.

 

My first test, the one that smoked the first transistor, was a ammeter as a load. It is theoretically a constant current source, so it should be able to take it. We'll see.

The second one, using a circuit board, I'll play with some component values. For example, the 4.7Ω will become a 10Ω to see if it affects performance.

Since the logic level MOSFET is fairly limited on voltage I'll use my power supplies limit of 15V as the max voltage.

One of the things I did a little differently is I beefed up the traces carrying the 0.7A current.

 

My first test, the one that smoked the first transistor, was a ammeter as a load. It is theoretically a constant current source, so it should be able to take it. We'll see.

The second one, using a circuit board, I'll play with some component values. For example, the 4.7Ω will become a 10Ω to see if it affects performance.

Since the logic level MOSFET is fairly limited on voltage I'll use my power supplies limit of 15V as the max voltage.

One of the things I did a little differently is I beefed up the traces carrying the 0.7A current.

I suggest you put a small heatsink on that fet and monitor its temperature. It is rated to 175°C so it should be able to get pretty hot before it toasts, but running it at 175°C for a long time will significantly shorten its lifespan.

 

I have calibrated fingertips, and the reverse part number brands to prove it. No knowing yet, but I bet it stays cool as a cucumber.

 

Good deal, Bill.

Do you have a commercially wound 100uH inductor?

If not, refer back to the packing list of the stuff I sent you.
Item 41, the T42 (OD=0.42") bright blue toroids. 8 turns of AWG-20 or AWG-22 should give you around 100uH-120uH; no taping required as they're coated. They're not really consistent, but they are pretty small so they'll fit your board layout.

You could also use item #39, the T50 (1/2") flat blue toroids with 9 turns for 100uH, 10 turns for 120uH.

It doesn't matter which way you wind them. Just try to spread the turns out pretty evenly around the toroid; if the turns are bunched up in one location, that tends to kill the Q factor of the inductor (increased loss). 1 turn = 1 pass through the center of the toroid. If the T42's get warm, swap to the T50's.

Item #40's (T50 light blue) with 8 or 9 turns would also work.