The probe multiplier has been set to 100x, and the 1/f noise amplifier has a gain of 10,000x, so just mentally drop down one SI unit class in your mind when looking at these scope shots (i.e., if it says mV, that's actually uV).
You must be wrong, 1/f noise in the order of a few hundred µVpp on a LTZ1000 is far to much. My references show 1/f noise <1.6ppm peak-peak for the 10V output.
-branadic-
My mistake, that should have been two unit classes (i.e., if it shows mV, that's actually nV).
I finished 9x LTZ1000 references (8628, 8716, 8830, 3x 9015, 3x 1811) with currently buffered zener output voltage only. Another one dated 1811 was finallized before and is with Greg since the beginning of the year.
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg2178047/#msg2178047I'm still not sure if NOMCA16032001 is the right way to go to boost output voltages to 10V. Need to perform some measurements as well as T.C. measurements and most likely T.C. optimization on the references themself. But they are all powered on since 24.02.2019, measurements have started and will be updated continuously.
What I can say for now, the 3 specimen manufactured in 9015 have a large difference in zener voltage, while all three specimen from 1811 are much closer to each other. Noise still looks a little high and I have to improve my power supply cabling to minimize noise pick up, I know that it's not perfect.
-branadic-
It's the noise of the buffered output.
-branadic-
I finished 9x LTZ1000 references ...
If at any time you should be willing to sell an assembled, tested and characterized 10v boxed reference, I would like to buy one. I am not overly concerned about a bit noise, TC and drift, so I will be happy with a medium performer.
I like this "low-voodoo" no-psychedelic-colors & no-whistles-and-bells design. Thanks.
First 265h of data with the references continuously running with 9x LTZ1000CH in the design of the group buy, plus 3x LTZ1000CH based on the design of Andreas plus my 2DW233 reference in an oven. Shown is the difference from initial value in volts plus standard deviation plus temperature.
-branadic-
Update:
Results after 1000h, I changed from Prema 5017 SC to R6581D on 31.03.2019 which results in lower standard deviation of readings by meter, but also accelerated measurements as I get a reading within 4s instead of 10s. The values shown are the average of 10 readings per reference.
2DW233 reference is still drifting downwards, but shows quite large std, need to investigate on that.
So let's start the next 1000h and hopefully get higher resolution results with the "new" meter.
-branadic-
Beside measurement for monitoring aging I today started measuring the different references for 1 hour to detect abnomal behavior and immediately found LTZ1000 #1 (date code 1986) to be a jumper (popcorn noise), while #2 to #9 are looking fine.
-branadic-
Hi All,
Can someone measure the pin pitch on these old vintage parts.
I suspect the pitch for say pre 2000 is around 6.2mm rather than 5.08mm as stated in the current package.
This has been puzzling me for some time and I am sure this is the case.
Rob
Can't confirm that. All the references (1986, 1987, 1988, 1992, 2018) fitted into the same footprint on the board without any issue. Maybe your specimen is a counterfeit part?
-branadic-
Hi,
Can't confirm that. All the references (1986, 1987, 1988, 1992, 2018) fitted into the same footprint on the board without any issue. Maybe your specimen is a counterfeit part?
The 6.2mm pitch spacing refers to calculated spacing taken from photographs of the 03458-66509 reference board for the HP 3458 multimeter.
So I have thought of 3 possibilities:
- The legs were offset (bent) outwards to fit the holes - this seems to be the case as it is clear all the legs are bent outwards on the solder side.
- The holes in the pads were say 1mm diameter, allowing for the legs (0.45mm dia.) to fit on the inside edge of the holes.
- The photo is not linear and while the pitch of known parts in the centre of photo can be used to check spacing of close by parts, other parts near the edge of the board will appear to have greater spacing - after close inspection of the printed photo, it does appear to be very linear.
Only a side photo of a vintage 03458-66509 board would answer this puzzle.
Thanks
Rob
Can't confirm that. All the references (1986, 1987, 1988, 1992, 2018) fitted into the same footprint on the board without any issue.
Same here, all LTZs I've seen/bought are same pin pitch.
Midi,
Now that it has been established the vintage parts have the same pitch, then it must be a slight angling of the legs.
You can see on the solder side the pads have centres which are greater than the span of 2 pins of 8 Pin op amp and 2 of the socket holes.
Could you take a side photo of the ref board to show the legs?
Hopefully we should see a slight angle and this might have been done to assist with manual soldering of the part into place.
Regards Rob
Sorry, pictures were taken before I read this post.
It had a bend cap that was fixed and it is now back in my meter.
Not shure why this is so important, nobody has confirmed change in pitch...
Hi Midi,
The layout is important to me as I am designing a replica copy of the original 03458-66509 board but without actually owning one, so it all has to be done from photos like yours.
From the tests carried out on the vintage LTZ1000 parts, is there a conclusion that the vintage parts produce more noise than the modern parts?
Also, could it be these vintage parts were part of a larger set that were all tested / sorted and these were the parts that were not as good.
I wonder if the heater power is linked to this low frequency noise, so the harder the heater is driven (for the higher die temperatures) the more the die temperature fluctuates and creates a greater swing in the noise level. If the radiated and conducted heat loss is kept low, then perhaps the noise swing will reduce.
I know that it has been said running the LTZ1000 at a reduce temperature will dramatically improve the long term stability.
Running at a higher temp. accelerates the aging process towards long term stability, but at the same time reduces its life span.
Rob
Some update... I changend from Prema 5017 to R6581D in march 2019 which got calibrated in june 2019 at Metrology Meeting 2019, thus the big jump in the data. I moved 3458A from work to my home in august 2019, to get additional datapoints, that are not shown here, but used LTZ-1000-1, -2 and -3 to verify that this succeeded without additional changes/jumps. So experiement is still running and I take measurements monthly.
-branadic-
Sorry that I have to ask:
the #1-#9 references are all with 7V output and the LTZ1000-1 through LTZ1000-3 are measured on the 10V output?
Or why is the jump larger for the lowest 2 devices?
Is there some linearity problem?
with best regards
Andreas
That is correct Andreas, #1-9# are the raw 7.xV output while 1-3 are 10V output. Furthermore 1-3 are individually t.c. trimmed, while #1-#9 only have an initial 402k resistor installed.
-branadic-