PPM 412 DC Voltage Reference (teardown)

[pigrew]

The PPM 412 is an English DC voltage reference, with internal battery power and an integrated 6-digit KVD. My unit was manufactured in 1982, seems to be very well used (or even abused), and was dead on arrival for the sum of $130 US dollars.

I was unable to find ANY information on this unit, possibly since the company's name is PPM (most common word in the English language?). It could be the "Pulse Power and Measurement" company, but their website mostly talks about EMC issues related to directed energy weapons, and not voltage references.

As the unit is still in repair, I've don't yet have any good measurement data yet.

Circuits

I've reverse-engineered the main PCB (see attached), and can describe most of the circuit functions:

The AC power supply is simply a bridge rectifier with a filtering capacitor, driving a battery-charging current source (10 mA or 30 mA, depending on the rear panel's switch). A stack of 14 NiCd cells is used to create a split rail (negative two cells and positive 12 cells). An op-amp is used to blink a battery LED when the battery voltage is less than a threshold set by VR1. A bodge of a big film capacitor was added between GUARD and EARTH in order to help with EMC. As received, the batteries could not hold a charge, so the reference didn't function at all.

A second DC rail (I'm calling it VR) is created as a zener-regulator buffered by a BJT. The breakdown of the Zeners is a bit higher than the battery's nominal voltage, so it's not really regulating most of the time. It's more of a voltage-limiting circuit to protect the op-amps.

The reference amplifier uses OP07 amplifiers for both stages. The first op-amp sets the Zener reference's pull-up voltage to about 11V using a pair of precision resistors to set the gain. A set of five resistors (including a multi-turn potentiometer VR2) are used to set the Zener's current. The reference voltage 6.276703 is printed on the rear panel, however I'm not sure how it compares to 1990 Volts (what did England use in the 80's?). (Speaking of which: the kg, mole, ampere, and Kelvin are expected to be changed in a vote next week...) I've never actually seen the Zener reference, it's mounted in a cute little plastic box, epoxyed to the PCB. This first amplifier receives its feedback from the sense terminal routed to the output negative lug. The OP07's offset terminals are left open, but there is a guard ring around the inputs connected to GUARD (different than the circuit common and different than earth). As the input terminals are biased to V_ref, I don't know why they didn't connect the guard to the reference voltage. Also, they only guarded the top of the PCB and not the bottom.

The output of the first amplifier is routed to the range switch (which optionally divides it by 10), applying either 6.2xxx or 0.62xxx volts to the KVD. The KVD has 6 digits. The bottom four are fixed resistors while the top three decades have trim-pots. The top decade is composed of a series of 9.99k resistors with the small trimmer in series.

The KVD's output is then routed to the positive terminal of the second OP07. This amplifier's offset trim is connected to a front-panel trim-pot to set the overall amplifier offset. Also in an attempt to reduce the input bias current, a Widlar current source, divided down by another BJT's current gain beta is connected to the positive terminal. I assume this is supposed to match the temperature characteristics of the input BJT of the OP07. The PCB can be reconfigured to provide either a negative or a positive current source. I wonder if the temperature curve of the Widlar source could match the Op07's input bias current to make the circuit even more stable. The voltage across R21 can be measured to determine if the currents are properly balanced. This op-amp also has a pair of precision resistors to amplify the KVD's output to be 10.99999 V full scale.

Repairs
(Some have been completed, some are to be completed)

• NiCd batteries are dead. They've been replaced with sub-C NiMH batteries. A 30 mA float charge should be safe.
• Front panel neon line indicator flickers. Replaced bulb with A1B neon and 150 kohm resistor
• Two front panel DPDT switches were making bad contact. Why did they use two different brands/models? Were they previously replaced? Locking range switch replaced with C&K 7201K12CGE2, KVD top-range switch replaced with NKK M2022SS1G01. Both are gold-plated brass.
• Front-panel outputs were broken off in shipping. Cry. Replacing with BU-P3770 CuTe binding posts.
• Electrolytic capacitors have been replaced with 105C models. Extra holes were drilled to mount radial cap in place of old axial 1mF 63V condenser.
• A few of the copper solder-mount turrets had cold solder joints. I reflowed them, hopefully not heating up the precious resistors too much.
• With the bad switches, the output jumped between 50 and 150 ppm high. Hopefully this will be fixed, and I'll adjust the three front-panel trim-pots if needed.

Ideas for improvements

• Add a line filter to the AC mains input.
• How stable is an unheated temperature compensated zener in comparison to a LM399???
• Apply deoxit to the KVD switches (if needed)
• Verify that the voltage across R21 is zero (implying that the op-amp's input bias current is properly compensated)
• Add sense binding posts to the front-panel (probably not needed since the output current is so limited)
• The design doesn't always regulate VR. Perhaps one of the Zener's could be replaced with a lower voltage so that it more often regulates the applied bias?
• Verify the Zener voltage is the voltage on the back panel.
• Verify linearity of KVD, and adjust as needed.
• The PCB has a spot for adding a divide by 10 and divide by 100 network (for inputs of 100 and 1000 volts). Perhaps these need to be populated.

EDIT: Updated schematic

[pigrew]

(photos)

(I've done some repairs already in these photos, and the RF shields are removed. The 33uF capacitors are new, and some binding posts were installed while I wait for the CuTe ones to be delivered)

[pigrew]

After replacing the front panel switches, the output became much more stable, and closer to nominal. The reference measures about 6.276599 V, 17 ppm below the recorded value. I guess this is acceptable drift over 35 years… The 1V and 10V outputs are both about 30 ppm high.
Note that the switch I ordered had heavily corroded terminals, I couldn’t solder to it (see photo). I think that their gold plating was too thin, allowing the nickel to oxidize. ☹

Based on availability from Arrow, I ended up with C&K 7201K2CBE as the range switch (through hole pins instead of solder lugs) and NKK M2022SS1G01 for the KVD top range switch. I always find it confusing as to the order of the washers to use for mounting the switch. I used: switch->nut->lock-washer->locking ring->PANEL->nut.

I found a few minor mistakes in the schematic I drew (linked in the first page), so I’ve fixed them, and re-uploaded it.
Since I have no calibration history, and it was in such rough shape when I received it, I plan to do a somewhat full adjustment to it according to the following plan.

My adjustment plan:

• Add fuses (both positive and negative), 300 mA or so, to battery PCB.
• Add AC line filter near AC inlet (FN 402-0.5/02).
• (Don’t adjust Vref bias current, assume it is OK)
• Verify input bias cancelation of M2.
  
  • Remove LINK1.
  • Charge C13 to about 3V. (repeat at 0V and 6V?)
  • Observe drift of output (positive or negative)? over 10 minutes or so.
  • Tweak R40 to null output voltage drift of amplifier output.
  • Replace LINK1.
• Verify range switch load compensation
  
  • Directly measure Vref (from back panel) with range in 1V and 10V
  • Adjust R30 until the Vref switch does not effect Vref.
• Adjust DC offset (center pot) of M2 such that 0V (1V range) outputs 0V.
• Adjust overall gain (left pot) until 10.99999 is accurate (as measured by DMM).
• Adjust 1V gain (right pot) until 1.099999 is accurate (as measured by DMM).
• Check KVD linearity (maybe adjust?)
  
  • KVD linearity verified using “reference ratio” mode of Solartron 7081, comparing 412’s output to its voltage reference (so 7081’s 10V calibration shouldn’t have an effect).

[pigrew]

Last night, I used my house as a thermal chamber to measure the drift of the voltage reference (after I've tweaked it a bit).

I logged using a few different brands of equipment (HP, Fluke, and Solartron, PicoTech) the room temperature, 10V output, and the ratio between the 10V output and the zener reference voltage.

The Solartron was used to measure the ratio between the output and the reference voltage (using its ratio mode). Over a 4 C temperature change, the ratio changed by about 0.4 ppm/C. As this is a ratio measurement, I have high confidence in this figure.

Using a HP34401A, Fluke 8505A, and Keithley 2000, I logged the 10V output voltage. I didn't see any RTN noise (good thing), but the output changed by almost 1.8 ppm/C. This is much more than the datasheet-value of 0.5 ppm/C of a LM399 reference used in the K2000 and the HP34401A, so I think it's real. The next step would be to directly log the reference voltage, to avoid any effect of the amplifiers/ratio resistors.

What's your suggestion? Do I blindly change the Zener-pullup resistance and do overnight logging to see if the TC gets better? Re-adjust the pull-ups such that the zener voltage is the number indicated on the nameplate? As the TC is positive, I need to reduce the current?

EDIT: I won't touch it for now. 2ppm/C is similar to the specifications of other calibrators with unheated zeners. I also need to repeat they measurement in the summer, with a temperature closer to 23C. The zero-TC point could be right around there. The other possibility would be to ovenize the zener, but I'd be better off just starting from scratch, I think.

---

The second OP07's input bias is strongly dependent on the input common mode voltage. I measured a range of about -0.5 to 1.0 nA bias when sweeping the voltage from 0 to 6V. I ended up nulling the bias current when in the middle of the range. Is there a replacement op-amp that someone would suggest? I think I'd need to bodge a modern op-amp into the TO-99 footprint. It's supply voltages are about -2.5V and 14V, outputs 0.5 to 11.5 V.