From: Winfield Hill
Subject: Re: Adjustable current source
Date: 18 Dec 2002 05:17:10 -0800
Organization: Rowland Institute
X-Newsreader: Direct Read News 4.20
>Winfield Hill wrote:
> And for my next trick... an adjustable version.
> Can you give us an idea of how to calculate the noise and ripple on
> the output?
Although its purpose is a robust high-power current source, I'd still
expect my open-loop circuit to be reasonably quiet at high currents,
although not in the same league as precision quiet-reference opamp-
controlled closed-loop current sources. We can do some estimating.
The LM317's reference noise is 30ppm (rms % of Vout, 10Hz to 10kHz).
Ditto for the LM317L, LM338, LT1083, LT1084 and LT1085. (But wait,
the LM350 data sheet says 10ppm - that needs further investigation!)
My current-source circuit uses a sense resistor selected for full-
scale current, e.g. 1.25-ohms for 1.0A with an LM317, so the noise
contribution should be 30ppm of 1A, or 30uA rms noise 10Hz to 10kHz.
In my design the 5V reference also plays a role, in subtracting from
the LM317's reference (see below). If we assume the 5V supply has a
10ppm noise level (typical 7805), we can neglect this contribution.
As for ripple, that's spec'd at -60 to -65dB for the LM317, given of
course as a Vout/Vin measurement. I'm not sure how to use this data.
> Would a 10uF capacitor from the LM317 adjust pin to ground help
> reduce the noise? If so, what would it do to the loop response in
> your circuit?
. | | Rs 0 to 1.0A current source
. Vin ---| LM317 |-x- 1.25 ohms --+---->
. RAW |_______| |
. | R2 250 R2 R3
. | | Iout = Vp -------
. '--------------------+ Rs R1 R4
. 0 to 5V |
. Vp -- R4 5k --+- R3 1k -, | provided,
. program | __ | Q1 | R2 R3 1.25V
. '--|- \ | |/ ----- = -----
. | >--|---| R1 R4 5.0V
. +5V -- R4 5k --+--|+_/ | |\V
. Vcc | | |
. R3 LM358 '-----+
. 1k LT1013 |
. | R1 200
. gnd |
If you were to add such a capacitor in this design, it would serve
to lower the output noise, measured as a voltage, to 30ppm of 1.25V
or 38uV rms. For a resistive load dropping say 7.5V, this would be
a 6x noise reduction at full current, even more at reduced currents.
But it would also slow the current-source's response to load changes
at high frequencies, that is to say, reduce its source impedance.
Of course an LM317 current source isn't high-Z at high frequencies
compared to a conventional opamp C-S, etc., anyway. As for loop
response, the circuit is open loop, so there'd be no loop-stability
problem from such a capacitor.
My circuit does suffer from a relatively-high voltage-programming
error, an error that's a constant fraction of the full-scale current,
which can be seen by examining the full formula.
. R2 R3 1.25 R2 R3
. Iout = Vp -------- + ---- - 5 --------
. Rs R1 R4 Rs Rs R1 R4
Note that the last two terms drop out if R2*R3/R1*R4 = 1.25/5, like
I selected in the circuit above, but the tolerance in this 1.25V/5V
ratio depends on the accuracy of the LM317 and 5V references. An
LM317 is typically about 1%. Assuming an accurate 5.0V, this would
still be a fixed 10mA zero error in a 1A max current source. Adding
a zero-current-setting trimpot to the 5V-side R3-R4 divider would be
a good idea. The adjustment would of course be made when the LM317
is cool. It'd also be wise to add a 5 to 10mA current sink at point
x to provide an independent power path for the LM317 at low outputs.
While my circuit may not be a good candidate for precision currents,
it features a robust thermally-limited output element and may find
a place in high-current automated battery-charging experiments, etc.
If one uses an LT1083, the circuit could be used up to 7.5A.
One last note. As it stands, the C-S's voltage-compliance range is
VinRAW - 3V down to 2.4V. But if the two points marked "gnd" in the
drawing are returned to -2.5V, etc., the C-S will operate down to 0V.