Regarding the bandgap reference and minimum loading, that is a good pickup on the different path. [...]
What do you mean by that?
But in practice it works.
That I'm sure it will, it does work, but we are treading on unspecified ground now, and this is often ill-advised, especially when no thorough knowledge exists of how and why it works (which, quite frankly, seems to be the case here).
And the resistor network does not have to be very low impedance.
It will when you want the voltage to be exact and stable. You will, for example, need an equivalent resistance (R1//R2) of 25 Ohms to force a regulator with a 4mA spread in bias current to stay within 100mV. If you use more, the output voltage will be way more dependent on the bias current. I have now assumed a 4mA spread, but the minimum was not specified, so the actual spread is likely to be larger. To be fair, with just a 0,333V bump needed this is only a minor issue, requiring about a 400/25 Ohm divider to get it right (exact figures are slightly more difficult to calculate, but these can be).
Forget OP requirement and needs for the part in synthesizer. I was commenting in regards to earlier comments about putting a diode or a single resistor on the ground pin of a 7805 to raise the voltage.
Ok, that was not completely obvious to me, sorry about that!
![Embarrassed :-[](https://www.eevblog.com/forum/Smileys/default/xembarrassed.gif.pagespeed.ic.zcm3EmNu9T.png)
Diode is going to be much less heat stable and blind luck to get exact voltage. Single resistor I don't see how that works, at all. Using feedback resistor and voltage divider, it works very well.[/quote]The single resistor is quite simple, but as you said yourself: "But in practice, it works." I would like to add it is actually quite simple. The (relatively stable) bias current flows out of the Gnd pin, and will cause a voltage drop over the resistor. As said, the bias current has a wide and incompletely specified range, but is likely to be rather stable and can be tweaked to a specific regulator. The resistor does amplify the changes in bias currents, so using this for high voltage "bumps" would be ill advised.
BTW, just because engineer specifically design/optimize and marketer specifically advertise things, it doesn't mean it won't work another way.
That is true. It will also likely not die the second you exceed the absolute maximum input voltage. Or oscillate if you leave out all caps. But only relying on guaranteed specs makes it less likely for a design to break when you switch manufacturer or the manufacturer switches process, or tolerances conspire against you. That is good engineering. Of course you can do your own characterization, but that would require extensive testing over input voltage range, temperature and whatever may be important. That is a lot more work than just calculating your circuit for worst case datasheet values. How much resources are you saving by exceeding guaranteed limits?
Well, I think it will die (slightly) over max voltage. Maximum voltage specs are a usually quite a hard limit of the transistor geometry. It will cause depletion areas of the transistor to hit the next area of doping, introducing new charge carriers from other parts of the IC, essentially causing the junction that is supposed to be isolating to conduct, with often catastophic consequences. The failure mode of overvoltage is not always a simple one, but never ever rely on having any time at all when overvolting any piece of silicon. Oscillation will occur in LM337s, but LM317 are specified to be able to run without output capacitor. Input capacitors are required as soon as the inductance of the feed wire starts getting an issue. This happens remarkably quickly, especially at high currents. It is capable of very high frequency oscillations (into the low 100s of kHz), even when moderately loaded.
The marketing argument might work for consumer applications, but most silicon is specifically designed for industrial use and the specs are often quite reliable and complete. I must admit there are a few quirks still present, like always characterising the maximum power as the one you can dissipate at T
C=25°C, which is quite inplausible to actually achieve. It does give you a metric to compare it with other manufacturers.
When I read that, I thought he either didn't have as much input voltage as he thought, or he did something wrong. But it never fails someone brings up minimum load on the 317, tolerances, any other gotcha that is printed somewhere in a datasheet.
The manufacturers continue to print the misleading sample circuit with an LM117, despite the LM117 having been removed from the datasheet because they discontinued it. So it is an easy mistake to make. And troubleshooting is about excluding things that are easy to test first, and replacing resistors is one of the easiest tests you can do.
If I remember correctly the only difference between the LM117 and the LM317 is its temperature range and accuracy figures, for most means and purposes they are the same. I believe it was a military spec unit and cost quite a bit more too. There might not have been a sufficiently large market for it. The army might be switching to switchmode too...
There is no one going to design a product doing what I just described. This does not improve the circuit (it has drawbacks I already mentioned in addition to what crumble pointed out). There are already products that are just as cheap.
But if you have an lm7805 in your hand and you want it to output 10 or 20V, you can do it the right way, and it will work, and it is not a hack. It is using a (integrated) circuit as a tool to achieve what you want.
So you're saying nobody is going to design that way, but you can still do it won't be wrong doing so? Sounds contradictory to me... Additionally, if you have a 5V regulator the issue with voltage variations will be quite some larger when you want a significant voltage bump like up to 20V. Please do your calculations, what kind of a resistor divider will you need if a 1mA variation in bias current will only yield a 16mV change in output voltage (please note the 1mA is the specified change in bias current as specified for the µA7805 and the 16mV is the voltage regulation of an LM317 as specified at 0,04%/V line regulation at 25°C). You
can use a 7805 for this, but it will be a bodge and is only advisable for rather tolerant loads.
Max input voltage of an lm317 is stated in the datasheet as 35V. Tim has used them at over 100V. If you understand how it works you can sometimes do things with circuits that aren't covered in the datasheet.
Regarding lm317 minimum output, I don't care which manufacturer you use. If your input voltage is static, the output load is static, the output voltage is going to be correct, even with below minimum load.
I think you're just missing some basic knowledge on how this thing works. The max input voltage of the LM317 is not specified, only the V
i-V
o. The circuitry is connected between those pins and will only see this voltage. As long as the voltage drop over the LM317 is less than 35V (or 37V, specs seems to differ slightly) you're ok. If the output get short circuited or the thermal protection kicks in you're in trouble. It will then lower the output voltage and the voltage over the regulator will exceed the max voltage and the unit
will die, no matter how you hope and pray. A normal LM317 (there is a 57V specced LM317HV) will not tolerate 100V between any two of its pins. Higher voltage rating will require more silicon to accomodate larger depletion layers and it will cost you money. You can get a 125V/700mA
TL783 for just a slightly higher price (but there's more, just do a search for one on Mouser or Digikey or so).
The minimum output current for the LM317 is usually catered for by using the recommended resistor divider network with a 120 Ohm between V
O and Adj. It will cause a current of ~10mA to flow, which is specified as being the upper limit in bias current, and no current will need to be drawn by any connected circuitry. That is why the minimum load current is explicitly in the datasheet of the LM317 and not in the µA7800 series datasheet (which coincidentally does not specify anything at all about its use under 5mA, which is kind of a fail). When using larger values for the top resistor you will at one start to have issues with voltages floating up, but the spread in bias currents is rather large and might not be an issue with every LM317 unit.