What is this schematic symbol? Arrow with a circle around it

[cdevidal]

It's found on page 1 of the datasheet for the LM386 audio amplifier schematic. I doubt it's a current source, which I've only ever seen depicted as a disconnected arrow inside a circle. But I'm open to correction on that.

[SiliconWizard]

It's a current source.

[golden_labels]

As SiliconWizard said, but don’t treat it as a specific discrete component. For integrated circuits schematics depict behavior, not the actual implementation. In practice it’s likely an output of a current mirror.

[sparkydog]

As SiliconWizard said, but don’t treat it as a specific discrete component. For integrated circuits schematics depict behavior, not the actual implementation. In practice it’s likely an output of a current mirror.

...or an LED driver. Or it could be a discrete component, e.g. an Onesmi NSI series CCR. Not saying it's likely, but it's possible...

[golden_labels]

LM386 is a 8-pin PDIP/SOIC package. I’m pretty certain it doesn’t contain a discrete component. In particular not a DPAK package, which is physically larger.

[tooki]

As SiliconWizard said, but don’t treat it as a specific discrete component. For integrated circuits schematics depict behavior, not the actual implementation. In practice it’s likely an output of a current mirror.

...or an LED driver. Or it could be a discrete component, e.g. an Onesmi NSI series CCR. Not saying it's likely, but it's possible...

Yeah, no. Zero percent chance it’s a discrete component here.

[T3sl4co1l]

LM386 is one of those parts old enough to 1. have a fairly representative schematic in the datasheet already, but even more so, 2. have been decapped and reverse-engineered; you can inspect the physical transistors on die at your leisure. 

I don't remember it exactly, but a typical implementation in those days, was lateral PNP current mirrors.  The "lateral" detail is important because it's a cheap way to make poor quality PNPs -- hFE is quite low, which has consequences for how the mirror is designed.  For sourcing bias current into other (mostly NPN based) circuits, they're perfectly suited.

Related: http://www.righto.com/2020/06/reverse-engineering-and-comparing-two.html
I don't actually find a die photo offhand, but this isn't too surprising as the subject is very poorly represented by Google Search.

Tim