Some linear regulators are marketed as "adjustable", but all linear regulators create a specific voltage in relation to their ground pins. Since the ground pin voltage can be offset from the circuit ground, e.g. by using a Zener diode or a feedback voltage divider, can't we adjust any linear regulator this way?
Yes, you can; but if the regulator's ground current changes significantly (or drops too low), the zener may not regulate as well as is needed and output accuracy will be affected.
In addition, the additional ground impedance (from the zener) may tend to create instability.
Some regulators may have difficulty starting up (into a load at 0 V) if the regulator's ground pin is higher than the output.
For simple regulators like LM7805, using a zener usually works.
An adjustable regulator is one with an exposed feedback node so that the output voltage can be set by the designer using an external resistor divider.
Contrast this with a fixed regulator. A fixed regulator does not use an external voltage divider (it may have one internally) and does not have an exposed feedback node.
If you choose to redefine what "adjustable" means you may be able to say that all regulators are adjustable. But under the current meaning, many regulator have a fixed output voltage. Because the distinction between "adjustable" and "fixed" is useful, I believe the existing terminology is useful and should be preserved.
The technique of "floating" a regulator above ground is sometimes used, but that is quite separate from the issue of whether the regulator is fixed or adjustable.
Most regulators do NOT pass their bias or quiescent current to the output pin. They pass their quiescent current to the GND pin. So if you were to use the GND pin as the nexus of a voltage divider, the quiescent current would create an error in the voltage found there. And the quiescent current may be variable (I am not sure). So I don't think this is advisable for the average linear regulator.
I downloaded the datasheet from here: https://www.ti.com/lit/ds/symlink/lm317.pdf
The circuit that caused the question was based on an adjustable version of an LDO, the AP2202-ADJ. It's datasheet schematic looks like this:
Now the question was whether we could not take the non-adjustable version and do this:
The idea is that the regulator strives to regulate the difference between GND and Vout to be a fixed voltage, so that through R1 a constant current must flow, and thus we get an "inverse voltage divider", because that same current flows through R2, and we thus can adjust to a higher output voltage, defined by the ratio of R1+R2 to R2. Nice! We get the absolute values of R1, R2 to play with, but their ratio is locked.
Small problem: This is a loaded voltage divider. The GND pin of the AP2202 carries the quiescent/operational current. The datasheet tells us how much that might vary:
Between the minimum and maximum output current, there's a range of 95 to 1900 µA (2500 µA if considering the whole temperature range). It mainly depends on output load current (Table on page 15).
So, assuming we don't want that to affect our output voltage by more than 1% (this regulator guarantees 1 % accuracy, so we're making it twice as bad), so we need the current through R2 to be in the order of 100 times the current difference of ca. 1800 µA to make the additional current not change the output voltage significantly. That's a whopping 180 mA to get that output stable!
Sadly, this regulator is only specified up to 150 mA. So, that solution doesn't work.
Now, you bring up a Zener to replace R2, which of course is nice, because then the voltage becomes less dependent on the current. If we keep the current through R1 at say, four or ten times the ground pin current variation, the resulting 10% change in current through the Zener will not affect the voltage across the Zener diode much. So, that might work here. Small caveats:
- Zeners are excellent thermometers, more than they are constant voltage generators. Solution would be to replace with a temperature-compensated precision voltage reference. (The venerable TL431 comes to mind, most models can sink 20 mA or more.)
- Don't forget that this is a very simple regulator, with nothing but a pass transistor. That means it can pull the output voltage up, not down. The only thing that can pull the output voltage down is the actual load (and now our voltage divider) discharging the output capacitor. While it's but little, reducing the load current reduces ground current, which in turn reduces the voltage across the Zener, will make the regulator turn off the pass transistor, even if just a little, reducing the current. I don't have any information on the dynamics on that process, but this might lead to exacerbated voltage downswings one the load current drops. It's unlikely this leads to instability, but I'm pretty certain it will have interesting effects on the ability to absorb switching noise from a digital circuit supplied with this.
It's not like they just changed the name of the pin. If you look at the internal schematics you'll see that the output node of the LM317 is pretty much what in an LM7805 would be the ground node. This is why the quiescent current goes through the output as others have pointed out.
When you put a resistor or zener in the ground path of a fixed regulator you aren't really 'adjusting' it, you're just adding an offset. If you measure from the ground pin to the output it will still be regulating at it's rated voltage (assuming it's getting enough current through the ground pin).
From reading your question and some comments, it appears you're getting confused by the terminology.
A fixed regulator IC is designed to operate in the configuration shown in its datasheet's typical circuit. To operate, it only relies on components shown in that circuit and/or specified in the datasheet text. Then it will meet the specifications of its datasheet. The IC and those components form the fixed regulator circuit.
An adjustable regulator is the same. But its datasheet will show and specify the components for its output adjustment. The IC and those components form the adjustable regulator circuit.
So you must separate the term for the component from the term for the circuit. Adding extra parts around a fixed regulator can make a new, adjustable regulator circuit.
But the IC hasn't stopped being a fixed regulator. It still adjusts its output current to maintain a fixed voltage between its GND and VOUT terminals (or equiv).
So forcing external conditions on a fixed regulator, using circuitry the IC designers did not expect or require, doesn't change what that part is.
A fixed resistor doesn't become 'adjustable' when there's a potentiometer in parallel with it. The fixed resistor becomes part of an adjustable resistance circuit with two components in.
Not comparing floating regulators (like lm317) with common fixed voltage regulators (like 78xx), lets compare the ST and TI datasheets for the LM1117, which is offered with fixed and adjustable versions and has a basic topology more similar to the lm317.
TI diagram (fixed and adjustable are shown):
ST diagram (fixed - there is no access to the node between the feedback resistors):
The difference is clear: the feedback voltage divider is inside the package. So, it is not a simple difference of marking ADJ or not.
For this IC, you could add a resistor between the GND pin of the fixed version of the IC and the GND of your circuit to only increase the output voltage. Variations of the quiescent current (e.g. with temperature) will cause variations at the output voltage.
LM317 is a 3-terminal IC advertised as an "adjustable regulator". It still adjusts its output current to maintain a fixed voltage between its GND and VOUT terminals (or equiv). (that voltage being 1.25V)
Yes, if you connect the ADJ pin to the GND of the circuit and load the output with a resistor which guarantees the minimum output current. What is not obvious at the first glance: if you use the voltage divider to adjust the output voltage above the 1.25V, you can't use resistors with large values, otherwise changes in quiescent current would represent large changes to the output voltage.
Regardless of the method used to raise the potential of the GND pin of the fixed version (resistor, Zener, shunt regulator), you are on your own to find out the effect of this modification to the stability of the control loop and drift of the output voltage.