Meta42 power supply design question

[Wodgy]

Lately I've been reading a bit about power supply design, and I realized that I don't quite understand how the Meta42 virtual ground works.

According to its spec sheet, the TLE2426 rail splitter is designed to handle (typically) only 20 mA of current. Using power-hungry op-amps and buffers, isn't it likely that the actual Meta42 circuit will draw more current than this?

If that's true, wouldn't it be better to just omit the TLE2426 and replace it with a pair of resistors, and stack the virtual ground buffer so that it could handle more current?

There's probably something here I'm missing, so any light you guys can shed on this would be appreciated!

 

[KurtW]

On the META42 board the TLE2426 rail splitter drives the EL2001 buffer, which takes little currrent to drive but then gives you a 100mA output capability which should be more than enough unless you're using very hungry output buffers like the EL2009.

 

[Wodgy]

Since the buffer is necessary, wouldn't it be simpler to omit the TLE2426 entirely?

 

[Voodoochile]

You could, but then you'd need a pair of resistors to divide the voltage to feed the buffer input. Not a big deal in itself.

But, the railsplitter is more energy-efficient than the resistor network, which would be key on a portable.

 

[morsel]

The TLE is cheap, smaller than two resistors, and you don't have to go matching precision resistors or adjusting trimpots since it has two laser trimmed resistors inside.

 

[Wodgy]

Thanks for the replies, everyone -- you've answered my questions admirably.

One last question comes to mind (I'm still learning about all this), I noticed that the DIP version of the TLE2426 has a noise-reduction pin. This implies that it introduces a small amount of noise that's worth compensating. Do you think that using two carefully matched metal-film resistors would introduce a lower amount of noise than the TLE2426? Or are laser-trimmed resistors inherently less noisy than full size resistors?

 

[morsel]

It doesn't matter. FWIW, I never use DIP TLEs.

 

[Voodoochile]

I can't say for sure, but I would doubt that a pair of resistors would be quieter than the TLE.

 

[tangent]

Quickly to expand on a point that hasn't been made quite strongly enough: a 4.7K resistor divider with a 9V battery draws about 1mA, and that goes up as the battery voltage goes up or the resistance goes down. The TLE2426 only draws 0.2 mA. I guess you could increase the value of the divider resistors to bring it down to 0.2mA, but then you miss out on the second benefit of the TO-92 TLE2426, that it's smaller than even a pair of tombstoned resistors. And that's also the main reason we don't use the DIP-8 version of the TLE2426: because that eliminates the space advantage, especially considering that you need to add a CNR cap to take advantage of the DIP version's features. And space ain't cheap on the META42. It's all at a premium.

 

[Arzela]

Hi all,

Forgive my ignorance in electronics,
but could someone tell me if an elantec
buffer/ resistor network would serve
to keep the rails equidistant from ground when the circuit is under load? If so, how
does the resistor divider/ buffer topology compare to the TLE2426/ buffer
topology in this regard?

Also, what is the relationship
between current draw and the
value of the resistor divider?
Is it linear? What are the tradeoffs
if one uses larger resistor values?

I'm rather new to this hobby...I'm
sure this will be the first of many
questions....

 

[Arzela]

Quote:

Originally posted by Arzela Hi all, .... Also, what is the relationship between current draw and the value of the resistor divider? Is it linear? ....

V=RI...

Duh.....Basic Physics
was 11 years ago for me!

 

[aos]

I personally use 2 resistors and a 10nF bypass cap because I had very low noise regulator in front of buffer and didn't want to screw that up. Higher values of resistors generate more noise. But in my case there was another benefit of resistors - since BUF634 didn't like being loaded with 70mA and didn't split evenly (or even close), I adjusted one of the resistors so I have 8.6k and 10k resistors, and with them I do get kind of even split. Gain of a buffer is dependent on current and it's not 1, can drop to 0.9 or even 0.8. At which point you don't have even split at all.

 

[tangent]

aos is right: there's no law that says you've got to use a TLE2426 instead of a pair of resistors. Resistors take up more space, they're not precisely trimmed (an advantage in aos's case) and they draw more power unless you use large values. But they do work, despite all of these disadvantages.