Right , here we go. I just got a Hewlett-Packard 6632A power supply in with remarkable 'roots'...
Jep. Call it 'returns on my annual tax-dollar investments'.
The machine is squeaky-clean inside. Not a speck of dust... probably sat in a cleanroom all its life.
Pulling off the cover gives us this view :
A massive hunk-of-iron transformer , custom designed, overspec'd , side-welded so there is no 'humming' of loose core plates, and terminated with fast-on connectors. Switchable 110 / 220. And a one-piece double-sided circuit board holding the entire control and power system of the supply.
Top left we find the 'we-do-not-kid-around-with-fake-caps' tank capacitors and a bridge rectifier constructed from heatsinked TO220 diodes...
Note , just left of the heatsinked diodes: C124 and C127. These are axial ceramic capacitors. A very uncommon sight , apart from high-quality equipment. These are AVX caps. Just left of C127 is C159. This is a wet-tantalum capacitor. Those are also relatively rare, apart from military/avionics and high-end computing stuff.
These do not suffer from the catastrophic failure modes common to the classical dry tantalum caps. Tantalum caps ( dry type) are notorious for spontaneously combusting , burning holes in the board... ) even when you treated them perfectly fine they may still, for unknown reason, decide to commit hairy-scary and unleash the unholy flames of hell on your circuit board.Actually the reasons are known. typically they have been mistreated during assembly. heated too long , subject to too high ripple currents , or they have eveolved little cracks in the skin allow moisture to seep in... and then they go poof... It is commonly known that all electronic components work internally on smoke... once the smoke comes out it's game over...
The machine is GPIB, as you'd expect from any well-behaved test equipment, ( none of the 'Universally Stinking Bus' stuff or that twisted lan muck here ) controllable and has a dual processor system based around two microcontrollers. One cpu does the GPIB through an intel 8291 and HP driver asic ( basically 75162's in a single box with some additional protection ) and talks to the main cpu through an optical link. The outguard cpu drives the display and scans the keyboard as well. So the inguard cpu can focus on controlling the supply and taking readings.
The 'outguard' earth-referenced logic:
The 'inguard' system logic: This has a little E2Prom (93cx6 type , microwire interface, predecessor of SPI) that stores calibration parameters. It sits just left of the CPU.
Notice the crystal bottom right of the CPU : you find a silver circuit trace running around it. You see the same kind of silver traces around R32 and the dac , and R32 and U23. the same goes for U22 and the other dac and opamp.
The trace around the crystal (it's actually a resonator) is simply tied to ground to shield the clock frequency from rippling through in the analog domain. The traces around the opamp and resistors are not tied to ground but to the output of the dac and the input of the opamp. This forms a 'guard' and prevents leakage and stray capacitance of making the circuit unstable or oscillating. The incoming signal sees a high impedance source. The guard' ring' around it does not sit at ground level , but at the same , but buffered, incoming level. Since both the guard and the incoming signal are the same level the stray capacitance doesn't matter. There is no voltage delta anyway to induce a current.
The bridging. Two optocouplers provide galvanic isolation between earth ground and the supply output.
The 12 bit DAC's to control the voltage and current , as well as perform a readback ( they mak a SAR using a DAC and a comparator ) are CERDIL's... not exactly the cheapest flavor around ... you can clearly see the guard traces ( silver ) . the reason for removing the soldermask over the guard is to remove that pathway as well. Soldermask has a specific dielectric constant that is worse than air. By removing the soldermask you make a lower leakage capacitance across the guard ring.
Notice again the axial capacitors (AVX)
When it comes to sensing the current they don't muck around either ... Kelvin resistors.. the 4 wire resistor has two power connections and two sense connections. Notice on the left bottom pin that there is again a guard ring present. For some reason they forgot to open the soldermask ( i looked carefully and it is a guard ring ... )
The regulation loop is purely analog and uses precision parts such as glazed resistors and 0-tempco resistors (the dark blue resistors in the middle)
And the output stage speaks for itself ... we do not mess around here either.
There is a thermal overload switch on the bottom heatsink as this holds the downprogrammer that lets the supply sink current if needed.
The output is a real push-pull stage just like you would find in an opamp.
So this machine will generate the required output voltage , no matter what. If you try to force 12 volts on its output while it is set for 5 volts it will just bare its teeth , growl at you and pull that 12 volts down ( within the limits of its current setting of course ).
The output stage is powered asymetrically. The topside is fed from a 35 volts DC rail while the negative ( the 'load' side) is fed from a -8 volts rail. So there are limits as to how hard it can 'pull' down but it is sufficient to stay within its spec'd operational range.
These things are built like a tank and virtually indestructable.
There is a variant ( option code) that has the terminals on the front panel and the front actually has the holes pre-drilled but covered with the foil. The machine has separate sense lines and a switch that can control the slew-rate of the output.
I'm in doubt if i will drill the holes and put the terminals on the front ( i have real hp terminal blocks ) ... it would mean having to destroy the stickers .... and those are pretty unique ...
Home
Help
Search
Login
Register
