EEVblog Electronics Community Forum
Electronics => Metrology => Topic started by: VintageNut on November 08, 2016, 01:29:48 am
-
I bought this PCB to make a lab reference. LT1013 date code 1993. Headers for mounting arrived today. Mounted on a nice ground plane perf board.
A 2461 SMU is supplying 18V for the ref board.
DMM7510 is measuring the reference output.
I will let this settle for a few days and take some more data and post it.
-
the thermal isolation cap is missing :-(
-
Add some styrofoam or similar insulation on the top and bottom and tighten it up with a tie strip through the two holes.
This should increase the stability a lot.
I did not have this on my first LTZ1000A based PCB either.
But it would also be interesting to see the differences on your 7510 with and without thermal insulation.
-
Yes. It needs a metal box with foam insulation. All of the wiring is unshielded as well.
Warmup was only a few minutes. I connected the DMM before I powered it up. Something like 2 to 3 minutes to stabilize.
After clearing the DMM buffer, short term instability is less than 1uV.
My house is running no HVAC for the last three weeks. I am enjoying the short period where no air conditioning is required and no furnace is required. So, the temperature inside swings+/- 5 deg F
-
Hello VintageNut,
styrofoam is no long term solution, as the oven is getting 95°C hot.
The original cup isolator is a special sulfone plastic, withstanding > 100°C.
Also, it is necessary to isolate the bottom / soldering side.
Otherwise, you could directly pimp the circuit, to let it run on 65°C..55°C by paralleling 100k..75k to the 15k BMF resistor.
Then, styrofoam would be sufficient.
I don't know, what you plan to do in the end with this PCB..
Frank
-
After clearing the DMM buffer, short term instability is less than 1uV.
My house is running no HVAC for the last three weeks. I am enjoying the short period where no air conditioning is required and no furnace is required. So, the temperature inside swings+/- 5 deg F
My lab has horrible temp control and can easily swing > 5F :palm:
What filter settings are you using, rolling 10 or ?
-
I am using 10NPLC. Here is a long time graph. I have never worked with an LTZ1000 before. This looks like popcorn noise to me.
-
This is how my LTZ looks like on the DMM7510 after almost 130000 readings
But I had to move it away from any kind of disturbances like WiFi or DECT phones and so on.
And where these measurements are taking it was till not perfect.
-
This is how my LTZ looks like on the DMM7510 after almost 130000 readings
But I had to move it away from any kind of disturbances like WiFi or DECT phones and so on.
And where these measurements are taking it was till not perfect.
28 uVp-p. That is similar to what I see in my LTZ if you throw away the large jumps.
-
I decided to try to shield some from noise. I placed the ref board in an empty DIY audo amplifier enclosure.
This is much better. I wonder if the large spike is from the DMM7510 which is almost on top of the ref board.
-
The LTZ reference measured is about the same class as the reference inside the DMM7510. So you see the sum of noise from both refs and rest of the DMM (ADC circuit). The spikes are likely from the DMM, maybe external RF noise or grid coupled "noise" spikes or thermal EMF from someone opening the door.
To really check for noise / stability one would need a second more stable (or at least as stable) reference and measure the difference with higher resolution.
The easier part may be noise in something like the 0.1-10 Hz range - this could be done with a specialized filter (AC coupling, not necessary the 0.1 -10 Hz band filtering) and amplifier.
-
What I am going to do is try to run the ref board from a battery that is placed inside the DIY audio amp enclosure.
Can the 3458A ref board be run at a voltage lower than 18V?
-
Finally I can post, phew. :phew:
VintageNut
You can use A9 PCBA with anything from 10VDC to 20VDC. I usually test all LTZ modules with +15, just easy to get by with 7815.
No need for negative voltage either.
-
Finally I can post, phew. :phew:
VintageNut
You can use A9 PCBA with anything from 10VDC to 20VDC. I usually test all LTZ modules with +15, just easy to get by with 7815.
No need for negative voltage either.
That is exactly what I need to know. I will try some batteries and see how that goes. I am suspecting that this particular LTZ may be a bad one that was pulled and discarded.
-
The negative supply serves as a sort of passive current compensation..
You'll see that, if you sum up all different currents, like through the zener, and both transistors, and compare that value to the current through that strange dummy resistor to negative supply.
Anyhow, this is not needed, and gives no big advantage, as the circuit is an exact copy from the datasheet.
Therefore, you can run it from any single positive supply, like TiN already wrote.
Again, before attempting to make precision measurements, proper thermal and electromagnetic shielding is required.. this board is sitting inside a double shield in the 3458A..
And your measurements already showed many spikes from external sources (SMPSU?, AC mains line switching?) and maybe popcorn noise or influence from air draught (these longer fluctuations).
Frank
-
The negative supply serves as a sort of passive current compensation..
You'll see that, if you sum up all different currents, like through the zener, and both transistors, and compare that value to the current through that strange dummy resistor to negative supply.
Anyhow, this is not needed, and gives no big advantage, as the circuit is an exact copy of the datasheet.
Therefore, you can run it from any single positive supply, like TiN already wrote.
Again, before attempting to make precision measurements, proper thermal and electromagnetic shielding is required.. this board is sitting inside a double shield in the 3458A..
And your measurements already showed many spikes from external sources (SMPSU?, AC mains line switching?) and maybe popcorn noise or influence from air draught (these longer fluctuations).
Frank
I agree. The SMU voltage is probably not clean enough to power the LTZ1000. The board is inside a heavy metal enclosure but not mounted well and not thermally insulated enough. Battery supply is next experiment. More work to do. Just getting started.
Thanks for the comments!
-
Ah yes, another hint from the "Super LTZ .." thread.
These short spikes were always visible, in mostly all stand alone / DIY solutions, because the original circuit is extremely sensitive to external noise.
Especially both BE diodes like to pick up and rectify any external noise, causing these spikes and in turn, shifts of the reference voltage.
Andreas has made big improvement on this sensitivity, by adding a bunch of blocking capacitors on these crucial points..
Both of my own DIY LTZ1000 references still showed some spikes, despite a proper shield and grounding, but simply adding two capacitors even improved that behaviour.. please search on the latest pages of this blog, I think I have indicated, what to do very easily.. also going for your reference..
The fully equipped circuit of Andreas is even more immune, and has even less disturbances left.
Frank
-
Ah yes, another hint from the "Super LTZ .." thread.
These short spikes were always visible, in mostly all stand alone / DIY solutions, because the original circuit is extremely sensitive to external noise.
Especially both BE diodes like to pick up and rectify any external noise, causing these spikes and in turn, shifts of the reference voltage.
Andreas has made big improvement on this sensitivity, by adding a bunch of blocking capacitors on these crucial points..
Both of my own DIY LTZ1000 references still showed some spikes, despite a proper shield and grounding, but simply adding two capacitors even improved that behaviour.. please search on the latest pages of this blog, I think I have indicated, what to do very easily.. also going for your reference..
The fully equipped circuit of Andreas is even more immune, and has even less disturbances left.
Frank
Excellent. I will read that thread.
-
I posted something in September about my new LTZ #5, but can't find the post about the improvement of my older LTZ1000 references.
Excuse me, I think I have to look inside the box again, which capacitors I really added. Afterwards, the sensitivity to a SMPSU was completely gone.
Frank
Edit: I added 100nF capacitors, each in parallel to the 120 Ohm Zener resistor and the 1k temperature divider resistor.
This improved RF disturbance a lot.
-
Hello,
the most important change is the capacitor between Pin 6 + Pin 7 on the LTZ1000 (heater setpoint).
The other changes can partly not be done without further cirquit changes if you do not want to get oscillations.
See also thread on volt-nuts:
https://www.febo.com/pipermail/volt-nuts/2010-October/000578.html (https://www.febo.com/pipermail/volt-nuts/2010-October/000578.html)
with best regards
Andreas
-
I did not see any difference from these capacitors however from test result on my PCBA (https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg1002367/#msg1002367).
-
Hello Illya,
try to measure the heater voltage (pin 1 of LTZ or base/emitter of heater transistor).
without capacitor I had about 6 mVpp noise measured on oscilloscope in average.
(I used a 20 MHz analog scope at that time so sorry no pics available).
But there were sporadic jumps up to -40mV (asymmetrical).
With C11(+C12) the noise was reduced to 3mVpp.
The sporadic jumps after this change +/-5mV.
When you do the calculation 5.4V at pin 1 for the heater gives (-0.7V) 4.7V heater voltage.
With around 400 Ohms heater resistance = 55mW
with -40mV jump (4.66V) 54mW
1 mW difference at 400K/W gives 0.4 K temperature difference (statically).
Times 50ppm/K for the unstabilized zener gives up to 20 ppm change (which is only dampened by thermal time constant).
So believe my: I do not want to have any unnecessary noise / jumps in the heater cirquit.
with all capacitors installed according to my cirquit diagram in LTZ1000 thread
I measured wideband (10kHz) heater noise on LTZ#5.
I used a 10uF foil capacitor for AC-coupling to go have a 0.16 Hz lower bandwidth limit.
With best regards
Andreas
-
I will have to get my LNA amplifier assembled and tested first to do this test.
My only scope have too high noisefloor to do this kind of measurements without preamp.
-
I will have to get my LNA amplifier assembled and tested first to do this test.
My only scope have too high noisefloor to do this kind of measurements without preamp.
What scope do you have?
-
Not a suitable one (https://www.eevblog.com/forum/testgear/tek-csa7404-repair-project/) :) Lowest range is 20mV with P6245 probes.
-
Not a suitable one (https://www.eevblog.com/forum/testgear/tek-csa7404-repair-project/) :) Lowest range is 20mV with P6245 probes.
10X probes are for high frequency cowboys. Use coax, change the probe multiplier setting to 1X and now you have 10X better flicker noise measurement capability. I am certain you already knew that.
Your scope goes for $2k - $3k on ebay. It was a big deal in its day.
-
That trick does not work here, because scope has high-frequency custom TekConnect ports, not the usual BNC's. This means only DC coupling and only 50 Ohm input.