A Test Bench Power Supply
zero volts, this voltage has to be
cancelled. This is where Z1 comes
into play (a precision 1.225 volt Zener)
by developing an almost exact voltage
of opposite polarity. Its circuit consists
of negative half wave rectifier D3, filtered by C2 and current limited by R4.
Now, when P1 and P2 are at zero,
a negative 1.225 volt level is presented at the current source point and
adding this voltage to the positive
1.25 volt constant of the regulator
equals zero volts on the output or
nearly so. On my unit, I have 0.015
volts and 0.019 volts, respectively, on
the variable outputs at “zero” setting.
Going back to the voltage adjusting
pots P1,P2, their values are
determined by two things: the desired
output voltage of the regulator and
the constant current source. Again, I’ll
use this unit as an example.
The constant current was set at
5. 2 mA as previously described. Also,
since I am taking the low end of the
adjustment string to a -1.225 volts
which almost nullifies the regulator’s
constant of +1.25, the output voltage
will track the exact voltage drop
across P1 and P2. For a maximum output voltage of 20.0 volts, we need a
20.0 volt drop across P1,P2 and —
according to Ohms law — R=E/I =
20/0.0052 = 3850 ohms total.
From past experience, I knew that
a single turn pot would not give a
smooth enough voltage adjustment
for this great of range. It would need
about 10 times as much travel for
smooth transitioning. I was faced with
two choices: a 10 turn adjustment pot
or two pots with a 10:1 ratio. A 10
turn pot is too expensive and time-consuming for repeated adjustments,
so the combo unit was the clear
choice here. These pots should have a
resistance ratio of about 10:1. I chose
5K and 500 ohms based on the resistance calculated. In practice, the fine
pot is normally left at its midpoint and
then tweaked after the coarse pot has
set the approximate output desired.
This now allows for a fine adjust of
±1.25 volts — a very optimum span.
With the fine adust pot (500 ohm)
set at its midpoint and putting 250
ohms into the resistance string, this is
subtracted from the 3,850 ohm total
previously calculated and leaves us
with 3,600 ohms to add for P2. Since
the nearest standard pot is 5K, it will
have to be shunted to attain the value
we need. This is accomplished with
R5 (12K) and is calculated based on a
5,000 ohm pot. Bear in mind, most
pots have a ±20% tolerance and R5
will have to be adjusted as required
(hence, the asterisk on R5).
With P2 (coarse) at maximum and
P1 (fine) in its mid position, the voltage
output will be 20.0 volts, and P1 will
allow for a fine adjust of ±1.25 volts
throughout the total range. With P1
and P2 set at minimum, the voltage
output will be zero or nearly so. One
additional note is that there is a small
error current flowing through the P1,P2
string and into the regulator’s adj. pin.
This is approximately 50 µA and has
been omitted from prior calculations
for simplification and because it is of
little consequence here.
C3 is added for input bypassing
and stability insurance. D5 is a protection diode and will shunt any external
capacitive discharge around the
regulator rather than through it; in any
event, causing a more positive voltage
on the regulator output than its input.
C4 gives much improved ripple rejection and D6 prevents unintentional C4
discharge through the IC, again
shunting these currents around the
regulator via D5. C5 and C6 are added
for improved transient response.
As for the LCD panel meters, I
selected the smallest I could find (but
still maintain adequate character
height for readability) due to limited
front panel space. They are all 200.0
millivolt meters. The voltmeters are set
up for 20.00 volt operation by proper
decimal location and input scaling.
R7, R8 are the scaling resistors and P4
is added for calibration. These trimmers are small and cheap and are
much more accessible than the panel
meter’s internal calibration pot.
The current metering presented a
small problem. I wanted to measure 0-
1 amps in milliamp steps. This equates
to 100 µA/mA since this is a basic
200.0 mv voltmeter (no decimal display
is used here). The shunt required for
this conversion is 0.1 ohms (R1). I
could not add this into the positive
output line as there is no provision to
keep it in the regulator’s feedback loop.
This would have presented an
increasing regulator load current error.
For that and other reasons, I did not
want it in the negative output either.
However, it could be added ahead of
the regulator where its voltage drop
would be of no consequence. That
location presented one other problem
though, and that was, even with no
load current, there is still the regulator
quiescent current of 5. 2 mA flowing
through this point. This had to be dealt
with as even with a 5 mA load current,
the meter would show over 10. 2 mA
— a 100% error at this point.
The solution was to offset the
quiescent current and that is the
function of R2, R3, P3. Since 1 mA is
read for every 100 µV developed
across the shunt, a negative 520 µV
offset had to be applied to the panel
meter’s positive input lead. The values
of those components provide just
that, so that at no load the quiescent
current is exactly cancelled out giving
a meter reading of 0.0 mA.
Even though the raw B+ voltage
sags at full load and introduces an
error in this divider string, it only
amounted to a 1 mA error at the full
load of 1,000 mA. This error is far
beyond the panel meter’s accuracy
and can be ignored. The second
variable supply is not shown in detail
on the schematic because it is a
carbon copy of the first one. Just bear
in mind, you will need double the
components shown to complete both
supplies. A diode (D4) to power a 14V
LED panel light is attached to one of
these supplies to indicate DC power
on/off. You will notice I have incorporated a four pole-single throw switch
for all DC supplies and the indicator
light. When I switch off DC power to
a circuit, I want just that — no power!
By merely switching off AC power,
there is a bleeding off DC voltage
present at the panel jacks. This could
hang on a long time for light loads.
March 2007 43