I plan to use 317/337 for an application with a load current between 10 -- 20 mA. My initial guess was that for such small loads, the 317L would probably be better than the higher-current versions like 317T. However, comparing the datasheets (brand ON in this case) I find almost no important difference between them. Load and line regulation are the same. Noise is the same. They even have the same minimum load current. The only striking difference is that ripple rejection vs. frequency seems better for the L version, being almost flat up to around 10 kHz, and thus better. On the other hand, it is not clear if that diagram assumes the use of Cadj or not, which I might not be able to use in my application. Also the L version will of course get a bit hotter, although I won't go near the max Pd spec, which will influence some parameters.
It seems the L version is the better choice for ripple rejection and lower price, but are ther are factors I am missing? Of course different brands will behave differently too.
Yes, I guess I eventually will have to measure them and compare anyway, but I would be interested in opinions and experiences.
It seems the L version is the better choice for ripple rejection and lower price, but are ther are factors I am missing? Of course different brands will behave differently too.
Yes, I guess I eventually will have to measure them and compare anyway, but I would be interested in opinions and experiences.
For those loads, the L version should be fine. Just make sure that you have all of your capacitors as shown in the application data. Do however, resist putting the reverse diode across the regulator. This has been known to rectify RF and cause errors.
Maybe I was unclear in my question. I know the load itself will be no problem. The question was rather whether the L version might perhaps perform better than, for instance, the T version for such small loads, or if it is maybe the other way around, that the T version performs better even att small loads?
Thanks for the hint about he diode. I am not sure yet if I could safely omit it, but I'll consider it.
Thanks for the hint about he diode. I am not sure yet if I could safely omit it, but I'll consider it.
Christer said:
both the L and T versions like to be loaded.
as for the rectification effects -- i have not experienced this with an NPN regulator --
I would bet a sizeable number of centimes that they are essentially the same chip inside the package.
jackinnj said:
Meaning the L version would perform better for loads within its capability range? That was my initial assumption. I will probably already have to add dummy load resistors just to be safely above the 10 mA minimum current.
cliff said:
Just different packages and different current limits? Possibly, but the representative schematics of the two are quite different, and they at least seem to differ considerably in ripple performance.
cliff said:
where do you get centimes anymore? last time I got a centime in France it was 1967. still using ha'pennies, tuppence, crowns, and shillings over there?
back to the LM317L (or 317M) you have to take into account the difference between Vin and Vout, multiplied by the various currents, elsewise the device seeks to protect itself. the 317M's will do 1/2 amp and they are easy to heatsink as they are D-Pak or something similar.
jackinnj said:
I know, and the L version will be sufficient, no doubt. The load current will be lower than the required minimum for the regulator, so I will have to add resistors. The question is which version will perform best at such small loads. I haven't found any previous discussion on this topic, and it seems the L version is seldomly used, at least when performance is discussed. But that might be because most applications require more current or more power dissipation?
One thing that might matter is that the T version has a thermally better package, but in this case the load current will be almost constant, so there won't likely be any changes in Tj anyway.
Maybe there will anyway be more differences between different brands than between the L and T models?
I would guess the basic control circuitry is the same... while the large device has larger drive transistors. This might explain the difference in HF rejection... less base current required.
Check the die dimensions... there IS a significant difference.
🙂
Check the die dimensions... there IS a significant difference.
🙂
Makes sense Poobah. However, I just checked a few more data sheets (yes I should have done that earlier) and it seems not so simple either. The ON LM317M seems to behave exactly like their L version in ripple rejection vs. frequency. However, when checking some other brands, both LT317 and Samsung KA317 show a much better ripple behaviour than ONs 317T, altough not quite as good as the M and L models. Then 78xx series from ST performs even better than all of the above. Not sure I can use any of those, though.
Oh well, it seem these regulators are unpredictable and differ a lot between brands and models, or maybe the manufacturers just measure differently?
Oh well, it seem these regulators are unpredictable and differ a lot between brands and models, or maybe the manufacturers just measure differently?
poobah said:
Yes, in this case I will have plenty of headroom for passive filtering before the regulators, so I suppose load regulation is really the most interesting parameter. They don't show and such diagrams vs. frequency though. I think LT at least showed the Zout vs. f.
Can't tell you how many times I've made mysterious problems disappear with a little bit of resistive decoupling.
🙂
🙂
Christer said:
The datasheet 240R between out & adj is sized to give 1.25/240 = 5mA minimum load, which works for the average 317, but there's no problem using 120R to ensure 10mA min. 😉
A 317 calculator.
cpemma said:
Yes, that is the obvious solution as long as it does not violate any limitations in the datasheet. I hadn't bothered to check that yet, since there are still other things have to decide upon before. I see no reason to go below 10 mA, though since that is the datasheet spec. for the worst case. However, relating back to my original question one might also ask if performance would improve from further dummy loading? It wouldn't be surprising, I think, if the performance increases when the DC part of the load current increases compared to the AC part. Oh well, I will have to set up a test rig for this some day.
Christer,
I'll bet your right about increasing load currents. I know that many topologies have their smallest phase and gain margin in the feedback loops when the load is at a minimum. I wouldn't care to speculate where the best point is though. I am fairly certain that input rejection falls as load increases though... the old trade-off again. I HATE comprimise...
😀
I'll bet your right about increasing load currents. I know that many topologies have their smallest phase and gain margin in the feedback loops when the load is at a minimum. I wouldn't care to speculate where the best point is though. I am fairly certain that input rejection falls as load increases though... the old trade-off again. I HATE comprimise...
😀
- Status
- Not open for further replies.