KA278RA05 LDO Linear Reg -Users?

[digsys]

Looking to see if anyone has ACTUALLY used one of these (not just read the specs).
There was a very short thread back in 2013 - www.eevblog.com/forum/chat/ka278ra05-7805-lm317/  but no help there unfortunately.
I was looking for a replacement for the LT1083/4/5 series, and this seemed to have perfect specs. Very low DO, Insulated case, 2A, 35Vin max, damn cheap price !
Even the quirky (weird) VSet arrangement doesn't bother me, only doing fixed outputs. Simple, right, so I whipped up a few PCBs, based on many of the LM317 / LT1085 etc layouts
that I have, over dozens of years.
Then the poo hit the fan. Vop vs temp is terrible, a few degrees rise on the case, and the OP+ shoots past 2%. 20deg rise, and OP goes into thermal
runaway? (internal?) and suddenly Vop = Vin !! And it's well heatsinked ! Then on some cold starts, it latches up !? It's just a simple Linear reg, 1 Vset resistor and all combinations of
RF caps IP and OP, Low ESRs, 10-50mA min loads etc. Nothing helps. 2 different PCB layouts. Got me beat.
Supplied from Digikey, Mouser and Farnell so they're not fakes. I have a couple 100+ if anyone (pref local) wants to play with them. I'm doing 2 new alternate layouts, as a final attempt.
Anyone have non-standard ideas?
PS: http://www.fairchildsemi.com/ds/KA/KA278R05C.pdf

[David Hess]

Watch out for the junction to case thermal resistance which is higher on the TO-220 overmold package.  This should not be a problem in a low dropout application.

I am almost certain that your problem has to do with oscillation.  Double check the input and output decoupling.  The datasheet does not discuss it in detail but if the output capacitor ESR is too low, it will likely cause problems.  Just as a test, add a pair of standard 100uF aluminum electrolytic capacitors directly across the input and output to ground and see if the situation improves.

[digsys]

Thanks for the reply, but I've tried all that 20 times. Scope OP is clean as a whistle. Never seen a linear reg that finicky?, IF that is the issue.

[KhronX]

Might this issue be related with Figure 40 (on page 14 of the datasheet)? Ie. the tempco of the internal resistors 

[digsys]

.. Might this issue be related with Figure 40 (on page 14 of the datasheet)? Ie. the tempco of the internal resistors   

Yeah, I saw that, and it did seem a bit excessive ! IF that is so, who would design a regulator that bad !!, let alone what I am seeing. I'm doing another set of layouts to see if I can work it out.

[David Hess]

Thanks for the reply, but I've tried all that 20 times. Scope OP is clean as a whistle. Never seen a linear reg that finicky?, IF that is the issue.

Low dropout regulators are that finicky.

What is your oscilloscope's bandwidth?  I have not seen it with an integrated regulator but the oscillation frequency might be outside the oscilloscope's frequency response.

Does the behavior change with a light resistive load on the output?

What happens if you bypass the adjustment pin to the output with a small capacitor?  Excess stray capacitance to ground at the adjustment pin will cause problems although they should have shown up on your oscilloscope as oscillation.

[digsys]

Low dropout regulators are that finicky.

Never had any problems with the LT108X series, and this just seems a bit more than plain finicky.

What is your oscilloscope's bandwidth?  I have not seen it with an integrated regulator but the oscillation frequency might be outside the oscilloscope's frequency response.   

I only used a palin 100MHz bench one. I do have a couple 500MHz LeCroys here, but it just doesn't "feel" like oscillation.

Does the behavior change with a light resistive load on the output?

Behaviour directly corresponds to thermal temp of package (using a Flir), and that is WELL below the maximums they state. And even taking in consideration temp settling time.

What happens if you bypass the adjustment pin to the output with a small capacitor?  Excess stray capacitance to ground at the adjustment pin will cause problems although they should have shown up on your oscilloscope as oscillation.

Yup, have more stuff to try.
They couldn't possibly sell a regulator this bad.?? Thanks for the input. I'll update if I find it.

[David Hess]

What is your oscilloscope's bandwidth?  I have not seen it with an integrated regulator but the oscillation frequency might be outside the oscilloscope's frequency response.   

I only used a palin 100MHz bench one. I do have a couple 500MHz LeCroys here, but it just doesn't "feel" like oscillation.

It seems like an oscillation to me but you should have seen it with a 100 MHz oscilloscope.

They couldn't possibly sell a regulator this bad.?? Thanks for the input. I'll update if I find it.

I would not expect Fairchild or On Semiconductor to do so.  Could the part be a counterfeit or damaged?

[digsys]

I would not expect Fairchild or On Semiconductor to do so.  Could the part be a counterfeit or damaged?

Exxacerly ! Plus I always buy from reputable suppliers. These (~200) I got from Digikey, Farnell and Mouser. All the same.
I have noticed an odd start-up latch issue? Anyway, I found a NEW best friend :-)  LT3086 - Not-isolated and much dearer but MUCH nicer specs.

Not to give up entirely - I am happy to send a few pieces, plus a usable PCB to anyone who wants to experiment. FOC.
PS: Must have a history and capable of testing.

[digsys]

Might this issue be related with Figure 40 (on page 14 of the datasheet)? Ie. the tempco of the internal resistors

Had another look at that - They say there is a fixed 1.8K Internal resistor R1 (+~100R to gnd, (R2+R4) to get 24V). By calculations 24x~2.0K = 290mW of heat !!
Seems like a stupid amount of heat just to set V !!? Well, I've sent ON (Fairchild) all my findings, just have to wait to see what they say.
- Hopefully NOT stupid replies like "what colour caps you using" .. "did you check you DMM". No joke, I've had dumb answers like that in the past.

Offer still stands - 2-3 regs + a PCB to play with FOC, if anyone wants to have a play.

[David Hess]

In fixed operation, the high temperature coefficient of the internal resistors does not matter because they track each other and only the ratio matters.

Figure 39 has the values of R1 and R2 swapped.  The top internal resistor is 600 ohms.

They obviously did not intend the output voltage to be raised by a large amount and this will indeed cause excessive power dissipation in the lower internal resistor.

[digsys]

They obviously did not intend the output voltage to be raised by a large amount ...

Hard to mis-interpret this - :-)  Output adjustable 1.27 - 32V  We'll see if they can explain it.