Chris, are you under UNIX or Windows on that PC?
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
(for those unfamiliar, Joe is the designer of the SVR-T)
Chris, are you under UNIX or Windows on that PC?
Windows 7 Pro, 32 bit.
Nikonoid, I like that temperature - voltage correlation graph. I need to learn how to make those!
Sad to hear that the SVR-T board is probably toast (sort of). This also gives us direction for any subsequent traveling standard. The nature of its use is that it will be connected and disconnected many, many times by many people. Which is a recipe for error no matter the skill level and intentions of those involved. So we should probably incorporate idiot proofing to the extent possible. Definitely a diode bridge on power inputs. Outputs are tougher since protection techniques will generally compromise performance, but we can probably come up with some ideas.
This comment does not point fingers at any individual. As I understand the data history the noise increase does not correspond to a setup change. Perhaps the cause is a power spike, ESD or some variant of the purple plague. So protection will not eliminate all possibility of damage or degradation, just reduce the chances.
There may also still be merit in circulating the SVR-T with trim circuit disconnected. It appears that the device has pretty good performance on its own.
Yeah I should reiterate that I am not upset about this at all. The big win is that this cal club has a lot of interest and participation, and a few broken eggs is no big deal at all. And, I would much rather have a mishap early on than have it happen when years of calibration history are at stake.
Yeah, I don’t think it is toast, but rather I now need to use an external op amp for the offset trim and temperature compensation circuitry. That will be an enjoyable design exercise! However that also invalidates any history we have thus far, so this is a good time to make a break and switch to another reference.
In terms of the things we’ve learned so far, input protection is a good one. Good binding posts and having everyone use the same cables are another big one. A shielded enclosure is another. In general, doing everything we can to make everyone’s setup as identical as possible.
...perhaps we can overcome the danger of exposing the unbuffered output by coming up with a ritual to follow. In the airline industry, using checklists in time of emergency has revolutionized their safety records, and now doctors are starting to adapt that practice to avoid operating room mishaps (see the book “Checklist Manifesto”). We might have a ritual where you always start by connecting the buffered output first, and then as the last step you swap one lead from the buffered to the unbuffered output post, to minimize the chance of an output short, etc.
Nikonoid, I like that temperature - voltage correlation graph. I need to learn how to make those!
Cellular, this is very easy. It is called XY plot or scatter plot. Excel, as well as any other platform, has them.
Since voltage is pretty noisy in a way that is not correlated to temperature, it is beneficial to run voltage readings though pretty heavy averaging filter before plotting it against temp. In Excel for each cell you can just average 20 previous: R23 =AVERAGE(I4:I23)
You can even make Excel plot trend line for you and display formula for trend line. Slope coefficient will give you the best PPM/C approximation and R^2 will be indicative of the quality of correlation (100% excellent, 30% BAD).
QuoteThere may also still be merit in circulating the SVR-T with trim circuit disconnected. It appears that the device has pretty good performance on its own.
Interesting idea. Could a new trim circuit be made for it?
QuoteThere may also still be merit in circulating the SVR-T with trim circuit disconnected. It appears that the device has pretty good performance on its own.
Interesting idea. Could a new trim circuit be made for it?
My understanding of Joe's post is that it's the trim circuit on the chip itself that is damaged, so nothing can be done, other than running with the trim circuit disconnected and accepting the uncorrected tempco.