From: Winfield Hill
Subject: Re: Simple adjustable 0 - 400V regulator, with flaw
Date: 20 Jan 2003 05:55:40 -0800
Organization: Rowland Institute
X-Newsreader: Direct Read News 4.20
> Thomas wrote:
>> Winfield Hill wrote:
>>> Very good Thomas, however I think your drawing needs a
>>> wee bit of work.
>>>. +480V ___ HIP5600
>>>. ---| |---+----------------------+----(O) HV out
>>>. |___| | 1.5K
>>>. | | ,-o->o |
>>>. | | | o--------+
>>>. | | | |
>>>. | | | adjust 500k
>>>. '---- | -[rs]--+-o->o-----> pot
>>>. | o--, |
>>>. 10uF === fixed | |
>>>. 450V | | |
>>>. | 380k |
>> More complicated switch. Hmm. One alternative increases quiescent
>> current to get rid of the extra contact: an extra 1.5k resistor
>> from HV out to the hot end of the 380k.
>> Another, slightly offensive one but cheap:
>>. +480V ___ HIP5600
>>. ---| |----------+------+---------(O) HV out
>>. |___| | 1.5K
>>. | | |
>>. +---|>|------+ +------,
>>. | 5V zener | | |
>>. | | 500k |
>>. '---[ Rs ]-- | --> pot 380k
>>. | | |
>>. 10uF === | |
>>. 450V | | o
>>. | `------o<-o
>> With the switch in the upper position, the pot is a series resistance
>> to the control/reference input to the HIP. It may not like that, but
>> this is real cheap.
I added a protective zener. A few other changes would be required...
> Time to read the specs before making everything high impeadance.
> The minimum load is 0.755 to 0.770 mA, depending on temperature and
> input voltage. Ratings say to expect 1.00mA minimum load.
The data sheet is confusing on this issue. At one spot in applications
it states the mimimum current is 0.5mA, and elsewhere shows a 450uA fixed
current source as an app. At another spot it says 1.0mA (and many of the
spec'd parameters are at 1.0mA). The bias current is specified as 0.4mA
min, 0.6mA max, and the curves show about 440uA. I've been running the
chip at about 0.8mA in this design. However, they state 1mA min on the
front page, so I'll think about changing my design to match that.
> The ADJ current is 0.053 to 0.075mA, depending on temperature. Ratings
> say to expect 0.080mA maximum ADJ current.
> In the first diagram, the output voltage will vary by up to 5.5 volts
> with temperature when the pot is centered. Drift could be annoying even
> if it's below the accuracy of mechanical adjustment.
What junction-temperature change did you assume? In my application the
adjustment resistors are the only load (the output is driving some beam-
steering electrodes in a vacuum), and the temperature change at say 300V
out is about 4.5 degrees (the '5600 has a small heat sink). This creates
only about +0.8uA or +1% maximum change in the ADJ pin current.
(BTW, dT = 4.5C also creates a -0.18% change in Vref. In my original
circuit the ADJ current was 8% of the 800uA total, so the net effect on
the developed I_ref is +0.08% + -0.18% = -0.1% change, which is fine.)
A big issue is the effect an added I*R drop has on the nominal 1.18V REF
voltage, it'll clearly increase this value substantially. The '5600 spec
for V_ref is +/-10%, and +/-23% for the ADJ current. If we limited the
I*R increase to say 40% over Vref we can keep the 10% tolerance. Such an
approach would dictate a series Rs of under 40% * 1.18V / 65uA = 7.2k, so
we could add a say 6.81k resistor and redefine the effective Vref = 1.62V,
or if Rs was 3.3k, then Vref = 1.39 volts. Etc.
> In the second diagram, the output will vary by as much as 40 volts when
> in the "fixed" mode. Might as well get rid of the fixed mode and paint
> a dot outside the adjustment knob where it's 300V.
Let's see, with 1.5k as shown and 1.62V as above, we get I = 1.08mA, so
the 300V setting resistance would be 277k. Thus the 0.8uA bias-current
change from dT = 4.5C would result in a 0.8uA*277k + 0.8uA*6.8k*277k/1.5k
= 1.22V output-voltage change. All this assumes Rs = 6.8k was used, but
any value well over 1.5k, such as 3.3k, would provide a good benefit.
So it seems the idea could fly, and provide a reduced current pulse to
the pot. However somehow this configuration still isn't very appealing
to me, in part because it doesn't solve the voltage jump during switching,
as nospam's circuit does (but his requires my favored, unusual LND150).