TiN I would be happy to take that deal!
CM, I would also be happy to contribute funds to the effort. Sadly, I have no parts to offer...
Sounds like a "stone soup" reference, I like it! One created by and owned by the community.
kj7e's VHD200 13k/1k is as good as it gets for R4/R5 -- fantastic!
I can supply:
- LTZ1000ACH, used, not yet tested for popcorn noise
- R1: Vishay RNC90Z 120R (2ppm/C)
- R2, R3: Vishay RNC90Y 69.8K (5ppm/C)
TiN just messaged me and suggested the cheapest route might be for me to collect the parts and ship them together to him, then he can assemble, calibrate and ship it back to us. He's out of town for the holidays, will be back Jan 7, and can ship it back to us in February.
Indeed, that was also idea of the club, no ? This would hopefully boost quality of future rotate runs in 2018, until somebody donates portable JJA
.
I'd prefer hermetic 100-120ohm for good stability though.
February is due because it would be waste to calibrate any LTZ1000 reference without pre-aging for 200-400 hours first.
Wow, lots of activity all of a sudden. What else is needed for this reference? I can chip in as well, CM.
For what we are trying to accomplish here, lets buy the best possible components that TiN suggests (i.e. hermetic resistors). It will probably be cheapest and quickest to place a single order and have it shipped directly to TiN.
Cellular, how is your logging going?
For what we are trying to accomplish here, lets buy the best possible components that TiN suggests (i.e. hermetic resistors). It will probably be cheapest and quickest to place a single order and have it shipped directly to TiN.
Sounds good. However, we will have to wait for new vishay resistors to be made.
Cellular, how is your logging going?
I used this evening to make some more tweaks to muxr's plotting script, here are the results so far: (attached)
I was a bit disappointed -- the voltage is wandering around by quite a bit. Also, we had a huge cold front roll in through Austin, and my A/C was set to "cool", so the temperature in my apartment left regulation until I came home from work and kicked on the heater. This, of course, caused a huge dive in humidity. So, not ideal conditions, but interesting data!
By way of comparison, here is an updated set of plots from Nov 4th, just before I mailed the reference out to nikonoid. You can see how much better the reference was behaved! (note that the std-dev is 4x smaller)
for reference, here's the current set of scripts I'm using to log and plot data:
https://gist.github.com/cellularmitosis/11e73bfc15f64f3d3f3c9a02e3d112ad
Hello,
for me it looks that there is something defect. (noise increased by a factor of 5).
Perhaps the reference falling out of the socket? Or something with the wiring?
with best regards
Andreas
CM, this is very strange. After a quick initial warmup it was much more stable than this. You can download my sessions from the server.
One difference that I noticed is humidity. My environment is very dry (around 30%). Your humidity is much higher at around 75%. Humidity can take a while to sip into reference and resistors and humidity can also affect things dimensionally.
I also ran the reference inside of shielded tray that protected it from noise and air drafts.
Another thing is that RH (Relative humidity with relative being a key word) is temperature dependent. Same concentration of water in air results in different RH percentage at different temperatures. Maybe someone more knowledgeable can weight in on what is more important in affecting reference, RH% or concentration of water in air.
Yeah, I agree this is strange, and something must be wrong with the setup. Perhaps copper oxide has snuck up on me?
Last night I swapped over to the alternate cable (using the Pomona gold-plated grabbers and some nickel-plated banana plugs), but overnight the data still looked pretty noisy.
This morning before I left for work I hooked up one of the taobao LM399 boards (
https://www.eevblog.com/forum/metrology/cheap-ebay-ad584-voltage-references-my-experiences/msg1274955/#msg1274955). Previously I've seen these be stable within 10uV for several hours, so that should be a good test.
I also realized my room placement isn't ideal -- the meter is near a sliding glass door, and it looks like the temperature regulation in that area isn't as tight as near the AC thermostat.
I have also recently acquired a Keithley 2015, so I can throw that into the mix for a day and see what the data looks like.
Ok, I think it's an issue with the SVR-T itself.
First, here's a graph of the SVR-T using the alternate cable. The standard deviation is 54uV, so no improvement there.
Second, a graph of the LM399 board. The standard deviation is 3uV, and the only variable which changed was which board was connected to the meter.
Time for some further investigation!
EDIT: Wow, take a look at the first humidity graph and guess what time I took my shower this morning
Once again, the 'myth' of hermetic resistors vs humidity, yes hermetic is essentially immune to humidity at a very high price point. Nobody of significance, Keysight, Fluke, ect., are using hermetic resistors in the LTZ circuits because it isn't necessary. But if you want to spend a whole lot more money on hermetics to find this out, be my guest, it isn't my money you're spending. There are many LTZ circuits that have gone into far more difficult environments than a 3458A and they don't use hermetic resistors, they work just fine. Yes some resistors are very sensitive to humidity but well made resistors should be no problem in LTZ circuits.
Check your circuits, you've likely got other factors that are out weighing humidity effects and if humidity is causing a problem, you should take the time to find out just what is being sensitive to it and not just grabbing at straws.
Could it have been an appliance? The first thing I do when I get up is preheat the water in the Keurig. If you have an electric water heater, that could explain noise at that time frame.
Lights, dimmers, and all other electrified items in the house are suspect.
It will be interesting if you see a pattern.
CM, I reviewed my sessions and the last one (Log 7) was the noisiest, with few spikes as high as 60uV. My data was recorded at 10 NPLC, so you expect it to be somewhat noisy. The Log 6 was better than 7, but worse than any other recording.
There was nothing special happening to reference between experiments, so I am not sure why last one shows the most noise. I remember I had short power outages during the week of logging, but I am not sure on what day they took place. The power supply I was using for it was Hantec lab supply. It is not HP quality, but good enough for the purpose.
By the way, on Standard Deviation measure is only applicable to noise when statistical mean of your measurements is stable and not changing. Therefore StDev over large sessions with temperature variations is fairly meaningless. You would need to isolate much shorter subsections of your session and calculate StDev over them.
nikonoid, interesting that a smaller version of the noise showed up before shipping. What a mystery!
I tried reseating the temperature compensation jumper, and also reseated the reference IC, but it looks like that hasn't helped.
Ideas for further debugging:
- remove the temperature compensation jumper
- remove the IC from the board and put it into a temporary protoboard circuit
- throw the unit into the freezer and thermally shock it a few times
- solder leads directly to the output terminals (rather than using grabbers)
any other ideas I might try?
EDIT: adding a zoomed-in 10-minute window
I took the AD587 off of the board, placed it into a machine socket, and clipped 15V and the meter leads to it, and I am shocked at how stable and low noise the output is. And this is without any capacitors or other circuitry at all! Just the IC in a socket!
Here's ~20 minutes of data (attached).
EDIT: adding photo of connections
Those alligator clips bother me.
3 hours in, this looks pretty well behaved. If you flip and reverse the voltage graph, it looks like you can see some correlation with the temperature graph.
Now I'm curious which component on the board is causing the noise. I can run without the temperature compensation jumper, which will eliminate several components.
Ideas for further debugging:
- remove the temperature compensation jumper
any other ideas I might try?
I was wondering if that little 2-pin jumper could have developed a poor contact?
CM, I am glad to hear that chip itself is doing good. I think this is the only part that is difficult to replace. I feel bad the board started misbehaving while I had it. The whole time it was not a subject to any stress, electrically or mechanically. I did not remove the chip or the jumper.
Good luck with troubleshooting. Let me know if I can somehow help.
nikonoid, please don't feel bad! I am sure it isn't your fault, and anyways, I adopted the viewpoint that this board would serve as "training wheels" for the community, before we moved on to something more serious like an LM399 or LTZ1000.
I reflowed all of the solder joints and gave the board a once over with some IPA. This seems to have helped a bit, but the performance is nowhere near as good as the bare IC.
I actually just thought of one other variable -- I added a piece of electrical tape across the output solder pads, to reduce the likelihood that the bare copper clips would accidentally short across both output pads. I wonder if the tape adhesive is acting an intermittent conductive path, causing the strange output?
I wonder if the tape adhesive is acting an intermittent conductive path, causing the strange output?
Only one way to find out! I would generally use kapton tape for this application.
Horowitz and Hill *do* mention that electrical tape is evil and ought never to be used. Ever.
I can go ahead and place an order for a 120 ohm VHP202Z for our community LTZ1000 reference.
http://www.vishaypg.com/docs/63120/hzseries.pdfHere's what I was thinking of ordering: Y6071120R000F0L
That's Y6071 (VHP202Z), 120R000 (120 Ohm), F (1% tolerance), 0 (lead finish), L (bulk packaging).
If anyone else would like one for themselves, let me know and I can add it to this order and mail it out to you. The last time I ordered some VHP202Z they were about $28 each. Texas Components was able to deliver them in about two weeks.