LM337L stability

Yes, I have had them oscillate.

Two options:

1:put a 0.47 Ohm resistor between the regulator circuit and the load+capacitors (but take the resistor for the adjust pin straight to the Vout pin of the regulator, not the load/capacitor side of the 0.47Ohm

OR:

2: put a 100uF tantalum (or some similarly large aluminium electrolytic) in parallel with the MLCCs

Chris

We have a 4.7u tantalum at the regulator but out in the circuits there are two 10u ceramics too. Something on the board is getting goofy at

-25C, and I thought it might be the 337. If we ECO it, I guess we should use a bigger tant and smaller ceramics, just in case.

The 337 is an LDO-like circuit, with the output being the collector of an NPN, which can be tricky. 317s and 1117s output from emitter followers.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Or, if you can tolerate it, a longer, thinner trace to the loads might add enough resistance.

From what I recall, The 317L is better than the 337L in that it doesn't oscillate with MLCCs at room temperature, but still rings badly (and has very little or no line rejection at the ringing frequency) with 10uF ceramic capacitors only.

A little MOSFET with the gate driven from a square wave with a drain resistor to the power rail allows you to test the load rejection, which should show up any stability problems.

I have noticed that the datasheets for older parts are often vague enough that it would be impossible to guarantee that any circuit using them would work in every "worst case". They specify that regulators require "some" ESR for stability, for example. Many old op-amps only guarantee performance at +/-15V supply though they are widely used with other voltages. If one has the philosophy of only using parts within their guaranteed ratings, then this pretty much rules out most of the parts with reliable long-term availability or reasonable price. I wonder what people do in high-reliability industries.

Chris

** Using op-amps at lower supply voltages means tolerating less GBW and slew rate. Both can be addressed if you use a higher performance type to compensate. ** How so ?

Operating parts within maker's ratings is standard practice, so is deliberate de-rating to improve reliability.

** Use parts that are certified for such use and within maker's ratings.

... Phil

I'm not saying you can't design reliable circuits using older parts, I'm just saying many of the older datasheets do not contain enough information to "prove" (in a paperwork sense, not a practical sense) that the circuits will work. I agree that you can choose a better part, but often even if a part is ten times better than it needs to be, it is only a typical specification so on paper it is not guaranteed to work. More practically, many regulators often don't specify a minimum ESR value for the capacitors, but (as we have found out, now that 10uF ceramic capacitors are common) they do require some unspecified amount of ESR or otherwise they oscillate.

"Within their guaranteed ratings" was a poor choice of words on my part, what I mean is "relying only on guaranteed specifications". With many parts, many of the parameters are only guaranteed under a couple of operating conditions (e.g. nominal supply voltage and 25 degrees C etc.) and there are often no guaranteed specifications under other conditions.

If you want a part with multiple vendors then that rules out most of the recent op-amps for example, as the recent ones are usually not second-sourced, nor really cheap. For older op-amps like e.g. LM358, few of the specifications are guaranteed except at a 5.0V supply voltage, so if your supply voltage is 12V for example, then the input bias current specification is not strictly guaranteed to apply, nor is the maximum output swing VOH guaranteed. We can figure out what it is likely to be, but I could not prove that it must always be so on all future units.

Maybe I should have said highly-regulated industries rather than high-reliability. I can imagine that in some products one might have to prove on paper that the device parameters would be good enough for the circuit to work, over all operating conditions, not just the one or two points that are guaranteed by the chip manufacturer.

I guess if one paid TI enough money, they would make a special version of the LM358 with a guaranteed VOH specification when the supply is 12V for example. Or one could find some LT part which guaranteed all of its specs under all of the permitted operating conditions - they are good about that.

I am fortunate that the things I design only have to *actually* work. As an exercise, I have from time to time tried to design circuits that I can prove could not possibly fail except if one of the components were outside one of its *guaranteed* specifications. I have found it to be surprisingly difficult to do.

One circumstance where I do insist on a guaranteed specification is LDO regulators. When choosing an LDO, if they cannot put in writing what range of capacitor ESR will result in stable operation, then I assume that they do not know the answer, or are ashamed of it.

Chris

** I think expecting any data sheet to do that is not realistic. ** Where a spec is crucial, 100% testing of the part is an option.

I have often seen op-amps, transistors and FETs with coloured paint dots on them to show they have passed parameter testing while some manufacturers use house numbers on semis to show they are "selected" types.

** See above. ** Think that is where using "selected" parts becomes essential - buy enough and makers will do the selecting for you.

** I bet it is.

... Phil

>

Yes, come to think of it, I have seen a lot of those specified in the manuals of older test equipment that I have tried to repair.

Chris

Vaguely remember hearing about an appnote floating about somewhere that specifically addresses the issue of poles and zeroes.

Apparently this can be a problem when decoupling 3-terminal regulators with MLCCs.

Did Ti copy that technical document from a 1980s parts catalog template? It all looks familiar. I haven't seen that "simple adjustable switching regulator" schematic in decades, but I'm pretty sure it makes the designers of modern switching chips (70% to 90% efficiency from 100 microamps to 1.5 amps) cringe.

==========

In addition to higher performance than fixed regulators, the offers full overload protection. Included on the chip are current limit, thermal overload protection and safe area protection. All overload protection circuitry remains fully functional even if the adjustment terminal is disconnected.

Normally, only a single 1?F solid tantalum output capacitor is needed unless the device is situated more than 6 inches from the input filter capacitors, in which case an input bypass is needed. A larger output capacitor can be added to improve transient response. The adjustment terminal can be bypassed to achieve very high ripple rejection ratios which are difficult to achieve with standard 3-terminal regulators.

Besides replacing fixed regulators, the is useful in a wide variety of other applications. Since the regulator is "floating" and sees only the input-to- output differential voltage, supplies of several hundred volts can be regulated as long as the maximum input-to-output differential is not exceeded.

Also, it makes an especially simple adjustable switching regulator, a programmable output regulator, or by connecting a fixed resistor between the adjustment and output, the can be used as a precision current regulator. Supplies with electronic shutdown can be achieved by clamping the adjustment terminal to ground which programs the output to 1.2V where most loads draw little current.

--
I will not see posts from astraweb, theremailer, dizum, or google 
because they host Usenet flooders.

They have a pnp output stage and apparently that puts a pole or zero (I forget) in the wrong place and they will oscillate with a range of output capacitance. More or less capacitance would fix the problem as would making an RC filter with the cap