Yeah, great anecdotal recount, but no real info on how to solve the problem. His approach was trial and error which happen to work. No indication of the current limit used or just what "slow' means in terms of ramping up the voltage(ms, secs, minutes?). It's not clear to me why the voltage needs to be ramped up. If the current is limited that should be adequate to prevent damage and allow the tantalum cap to self heal.
Designs that use esr for stabilization is not engineering. If you need a specific esr for dampin, use a low esr cap with a resistor in series.
This also allows you to tune the damping to the exact value needed, and to change it when the environment changes.
You must have some gigantic capacitor bank with a lot of joules to blow a simple 1/4 Watt resistor.
Even so, a tantalum would probably short and blow the power supply.
Better to not use tants in the first place. Use polymers.
I sometimes put a small resistor on the input side of a reg. That lets us measure the current without compromising regulation. And allows for a bit of noise filtering. It can shift some dissipation too.
The eternal dilemma: where is the current going?
The X-chapters has some great data on resistor exploding.
That's not useful, it's risky in these days of parts shortages.
I could repeat my trick for stabilizing 317 and 1117 regs with low-esr loads.
But small tantalums work great on those regs.
X-chapters, p 27. We blew up some big wirewounds too. Flames and fun.
Any part will blow up if you don't understand it and abuse it.
Nope. 220 uF @ 15V is 50 mJ. With a 100-milliohm resistor, that will be delivered in ~20 microseconds, which is 2.5 kW peak.
Nah, switchers nearly all have cycle-by-cycle current limiting and stuff.
Cheers
Phil Hobbs
As I said in another post stabilizing a regulator is not a precision job. You just need to stay well away from the stability boundaries.
Sure, that's why I do it sometimes. It's a royal PITA to make sure the resistor will survive, though, and it does take extra board space.
Cheers
Phil Hobbs
If one were looking for extreme values, maybe. There are usually several suitable ones to choose from IME. After all, there's lots of daylight between 5 milliohms from two parallelled MLCCs and 5 ohms from a wet aluminum.
Sure, your 10 nF lead capacitor. Works great unless you're making 1.5V or something.
With switchers, where the FB pin is ground-referenced, I've been known to increase the noise gain the way you do with a decomp op amp (series RC from summing junction to noninverting input).
Yup.
Cheers
Phil Hobbs
You can stack LM1117s to make 1.25, 2.5, 3.75 ... with no resistors!
Yeah, but to take the pcb with it requires a tantalum.
That depends on the load. You can also have situations where the ringing is high Q. I had a pll once where the ringing was 5V p-p on a 5V line. This required careful adjustment to obtain the exact resistance to damp the ringing.
I still don't see how you can have enough joules to kill a resistor, especially on a board so small there's not enough room.
How do you guarantee the minimum 5mA load current (adjustable version) without load resistors on each?
If you know that the range of resistors that will ensure stability is wide enough, you can use a copper trace on the PCB, which is at least not going to be out of stock, and probably fairly robust to pulses. I recall needing about 20 squares of metal on one board I did, which in spite of being a compact board was not inconvenient.
I have a bunch of wet tantalum caps that I desoldered from a DEC PDP (11?) computer as a child, and though they are older than me, none of them have corroded, nor have the ones I reused ever failed in any way, whereas I have had several failed-short dipped tantalums.
(I hate to think of the present value of some of the vintage electronics I scrapped for parts, especially early HP desktop calculators, paper tape teletypes etc, though in my defence it would've gone to the dump anyway without my intervention, and much of it was not working).
1oz copper is about 500 uohms/square. A couple hundred mohms, to stabilize an LM317, would get messy.
Whan you ask for 1 oz copper, you're lucky to get anything close.
Vintage wet-slugs, the CSR13 types, had silver cans! A truckload of them would be valuable.
Yep. Some regulators require some ESR but in the tens of mOhms, then it is more attractive.
IIRC, the purple trace is a quite good Nichicon ALU. Organic polymer was a decade or 2 to the right, wet tantalum one or two decades to the left. The rise was _always_ 30 dB/decade, so it must have a well defined mechanism.
0 dB is 1 nV/rtHz referred to the input of the amplifier.The wet slug tantalums are hermetic with glass feed troughs. They don't get incontinent.
Gerhard
Have something to add? Share your thoughts — no account required.
Ask the community — no account required