10V reference needed from my box

[cncjerry]

Hi, I have an HP 5370b that uses 10V ref from an LH0070-1h.  The LH0070 failed and the box won't turn on.  5370b references all voltages to that 10V.  If I apply 10V, the box comes up.  I have parts coming but they can take a while.  I've tested a couple of refs I have, and the issue is the 16.2v regulated that is on that board, cant take much current draw as the 16.2v zener is fed by 315ohm resistor.  I don't want to do much surgery on the board.   The LH0070-1h is a .02% device.  I can live with .1% for now.  I can use a couple of 9v batteries to get ~18V but would need milliamp current draw again.

Here's what I have on-hand:

1) Qty 5 LM399H.
2) a reference module from an HP 3455 volt meter.  This is the A11 board and I think that is an LM399 on it.  It uses 29.5V.  I've tested it and it's stable but I don't have 29.5V handy in the 5370.  Just +/-15, 5.4 and 5v.
3) I have some opamps around but not an lt1001 as in the LM399 reference design I've seen. I have handy: LT1007, LMH6703, lt1677, AD706, op176(?), op(27), TL5532, TL5534, TL082, and probably others.  there is an lm725H on the hp 3455 reference board.

I have lots of  precision resistors, etc.  If I had to drop the power, I have 3045 regulators handy.

So can I put something together with that above or do I wait?

Thanks

Jerry

[Kleinstein]

One could use the LM399 reference without the heater. The reference it self can get away with rather low low current. For the amplifier the AD706 should be OK.
The LT1007 / OP27 and LT1677 are similar and should also work, if the resistors for the gain from 7 to 10 V are relatively low value.
The heater for the LM399 needs quite some current and may thus not be feasible at low power.

[David Hess]

I've tested a couple of refs I have, and the issue is the 16.2v regulated that is on that board, cant take much current draw as the 16.2v zener is fed by 315ohm resistor.

Are no other supply voltage available?

I don't want to do much surgery on the board.   The LH0070-1h is a .02% device.  I can live with .1% for now.  I can use a couple of 9v batteries to get ~18V but would need milliamp current draw again.

Here's what I have on-hand:

1) Qty 5 LM399H.
2) a reference module from an HP 3455 volt meter.  This is the A11 board and I think that is an LM399 on it.  It uses 29.5V.  I've tested it and it's stable but I don't have 29.5V handy in the 5370.  Just +/-15, 5.4 and 5v.
3) I have some opamps around but not an lt1001 as in the LM399 reference design I've seen. I have handy: LT1007, LMH6703, lt1677, AD706, op176(?), op(27), TL5532, TL5534, TL082, and probably others.  there is an lm725H on the hp 3455 reference board.

I have lots of  precision resistors, etc.  If I had to drop the power, I have 3045 regulators handy.

Do you have any trimmer potentiometers?

An LT1007 or OP27 can often replace an LT1001 directly.

I might just divide the regulated 16.2 volts down to 10 volts and buffer it with an operational amplifier follower.

[cncjerry]

16.2V is available all the time, regulated with a 16.2V zener off a 315r dropping resistor.  The other voltages come up when the 10v reference comes up.  Interesting approach to turning on power.  I have trimmers of all types, 10turn type.  I have capacitors of all types. The opamps I mentioned I believe are mostly single rail type.  I also have a 2.5V reference smd device, it is either AD or LT.  I'll post the number.

If the heater on the LM399 takes more than 10ma then that would probably knock the zener out of regulation.

[cncjerry]

one other thing, I tried to use the regulated 16.2 dropped down by the lt3045 but it wasn't very stable.  Setting it precisely was very difficult, I would have to rig up a two stage resistance for the set pin.

[bdunham7]

If you are facing an extended delay in getting the correct replacement part, you might want to consider if a lesser-but-adequate equivalent part might be more readily available as a temporary fix.  I don't know your location, but a US-based eBay seller has these:

https://www.ebay.com/itm/334686294414

[cncjerry]

So I powered the HP 3455 reference model from a Dewalt 20V battery and it came right up to 10.0000x.  I have to put a better meter on it. Using the Dewalt 5A/h battery will give me more than enough time.  I didn't measure the current draw but it has a heater. 

I still have 7 LM399s including the one in the 3455 ref and another ref I have.  I'll have to figure our what to do with them. I assume there is a PC board I can have made around here?  I'm more of a time nut, but I have 2 x 3457, 4 x 3456, 4 x 3455, Keithley 195, 196 and two of those Keithley tan, portable, battery operated type.  I also have a Fluke 1000v calibrator and a couple of DAS units with the KV dividers so I guess I was a volt nut once.

Thanks for the input.

Jerry

[Kleinstein]

16.2V is available all the time, regulated with a 16.2V zener off a 315r dropping resistor.  The other voltages come up when the 10v reference comes up.  Interesting approach to turning on power.  I have trimmers of all types, 10turn type.  I have capacitors of all types. The opamps I mentioned I believe are mostly single rail type.  I also have a 2.5V reference smd device, it is either AD or LT.  I'll post the number.

If the heater on the LM399 takes more than 10ma then that would probably knock the zener out of regulation.

The heater for the LM399 usually needs around 200 mW. So more like 12 mA at 17V.  With quite some extra thermal insulation one may get it below 10 mA, but this would be only after heat up and with the relatively large series resistor it may never get warm.

[cncjerry]

Under normal operation, how warm should an lm399 get?  I'll put a thermometer on it. This one isn't getting very warm but i'm only feeding the circuit 18v whereas it gets 29.5v when installed in a 3455.  It seems like it isn't overly stable but I've only had it on a few hours.  It's been drifting between 10.00000 and 10.00004.  This far exceeds the specs on the device it replaced as it was only .02%

Thx

Jerry

[Kleinstein]

AFAIK the LM399 get some 90° C for the reference itself. The plastic case already provides quite some insulation and at it's surface one should be more like 40-50C. Well warm to the touch but not really hot.
If the power supply can deliver enough current the heat up is quite fast, well under 1 minute. One should see the stabilization quite well in the current consumption going down.
The voltage has little (maybe 1/10 of a degree) effect on the final temperature, but can make a difference in the heat up speed. Some 17 V should not be an issue.
The small residual dependence on the heater voltage is likely due to a change in the temperature.

For the 10 V output the main candidate for settling and drift are the resistors. The 7 V from the LM399 should be quite stable, though the first 10s of hours may show additional drift. Later on has typical random telegraph noise with typical 2 levels some 4 µV separated and jumps on a time scale of something like a minute (can vary between units).

[cncjerry]

I put the 3455 reference back in its metal case.  Pretty stabile, it's only dropped 30uv since this morning when I set it.  I'm using a 3457a NPLC=100 to measure it.  Before putting it in the case, the LM399 didn't feel that warm, definitely above ambient, but not much.  I have a feeling the voltage I'm giving it is limiting the heater.  I'll have to check the schematic, but I'll tell you, it is very stabile for something I grabbed out of my box and just powered up.  30uV over 10e6 for 10 hours isn't bad, have to see if it works its way up as the room increases later during dinner. I could rig a power supply with an LT3045 regulator, and just use it as-is to turn on the 5370B.   This is a couple of factors of magnitude more stabile than what was in there.   I have another reference powering it now, made by Geller? 

On the good news side, I had another standard with one of those LTxxx-10 regulators that I ran out of adjustment on, plus the regulator was bad, fixed that, still had a problem. I noticed that the person who built this had extra resistance in front of and behind the trim pot, so that is working fine now.  I just might go do some volt-nut stuff and see if my fluke calibrator turns on.  That thing has an inner isolated frame. 1000V, here I come!

Thanks for the help all the PMs.

Jerry

[Kleinstein]

For the drift there is a chance that the HP3457 is drifting more than the 3455 reference module. The 30 V range of the 3457 already uses the 1:100 10 M impedance divider at the input. For this a 3 ppm stabilty is already a good values.

[cncjerry]

So the LH0070-1H parts finally arrived a week late.  They were 7 miles away for 5 Days after travelling cross country in two.  16 of them came in a Nat Semi box, box is probably worth as much as the parts,  I paid about $1 per.  Over half are mil-spec.  I wired up one tonight and none of them can touch the HP 3455 reference in accuracy or stability.  I put 4 in parallel and the best I can do without trimming is around 9.9998x on my H3457a.  I might just figure out a way to use the 3455 reference. 

I tried a plot of the jitter test and it seems cleaner but there are other factors (I moved the box and used a different cable).  I am running a test now where you divide down your reference to 1pps and measure that on the start channel and then use a loop of coax to the stop channel.  That seems cleaner with the 3455a reference driving the voltages.

Anyway, 7 more broken pieces of time-nut gear to go (and two scopes to fix.)

Oh. and what do I do with 16 of these references?

Jerry

[cncjerry]

I took four of the Lh0070-1H-MIL references and paralleled the outputs through 100r metal film resistors. I selected four of them and then replaced them one at a time trying to get close to 10V.  I'll measure them all again and write a goal seeking algorithm to see what combo gets me closest to 10V.  Power is supplied from a 20V battery then reduced to 15V using an LT3045 module. I let them cook for a day. 

The room was at 68 degrees when I started the test and dropped to 65 over the course of the evening.  It's hard to read the Octave plot, but setting the spikes aside, the cat was probably sleeping on it, the quad setup has around 50uv P2P of average noise as measured in HIRES with a 3457a meter through a GPIB adapter and a simple python script. Standard deviation is 18.28uv, mean voltage 9.999956.  I need to measure current draw.

I can feed it -15v and set the voltage precisely, but I don't know if the stability warrants it.  The 5370b heats up the entire room when running.  These are interesting references.  I'm going to run the same test using just one of them.

Jerry