Both DMMs are much more noisy and less stable (about 10 times) than the LTZ1000 circuit. The 34401As internal LM399 may produce these 1..2ppm steps. The bigger glitches may be produced by the LTZ1000 circuit, caused by RF disturbances, from switch mode power supplies in your lab. Watch out for LED lighting.. xdevs circuit lacks additional capacitors, which suppress most of these disturbances. I added 2 capacitors to the LT circuit, and things improved a lot, already. So please consult eevblog, the Super LTZ1000 thread for details on Andreas' circuit.
Looks to me that the glitches are related to your presence in the room, this has happened to me measuring very low noise resistors, I would suggest shielding all cabling. Thanks a lot for the video...great info!
Agreed. I was measuring a 50K ohm super precision resistor with similar flying leads using a Keithley 2450 SMU. Just walking past the setup, about 3 feet away, affected the readings with similar unexplained drops in measurements. The good thing with the keithley is you can see the drops on screen as it plots each measurement in real time. I have since purchased triaxial cables and appropriate shielding to repeat the measurements.
Watch the video and am thrilled. In the scene with the housing and the insulation, I realize that a metal housing only makes electrical sense and serves a purpose if it is also grounded to earth. And zipp, zapp zerapp in the next moment he puts a ground cable to the housing. I had to grin. Good work. Thank you.
I have noticed that radiated noise effects very small value uV values, I had to turn off as much switchmode equipment as I could in my lab when using my Fluke 343A DC calibrator ( looks the same as yours except it has one more knob for the uV range)
Too bad you didn't monitor more things: Lab power supply, room lighting, time-of-day, power line changes, etc. I'm not sure I would connect the two DMMs together for this; one at a time might be better. Please let us know what you did to get access to a better meter, etc. GREAT video!
Great video and very informative about the components. I too would like to build one of these circuits. Like you said, with any testing, you need a device that is more accurate than the device you are testing to rule out errors in the test equipment. Perhaps a better reference meter would show a much flatter line - but these type of meter are very expensive. I would be interested to know where you purchased your Vishay Foil Resistors from, because Digikey does not stock these and require a minimum order quantity of 500. Perhaps if you make up 2 of these boards and use each to cancel each other out, then measure the small difference in voltage. This is provided each is very close in output voltage. So you could have a series of boards and read the voltage difference off each pair combination. With a series of boards, it might be found that one specific board has a lower drift than the others and this would become the main reference. Keep up the good work!
Not sure yet. Perhaps volt-nuttery will prevail, and I will build more. I am thinking about 10V standard, and I am not sure whether to build it based on this reference, or live this one alone as a compact and portable unit, and build another one with 10V scaling amplifier.
Will never ever pay 17 dollars per board...no way. this is crazy. Other chinese manufacturers will offer the same design board at a fraction of the cost. Why make oshpark rich for no reason ??? No tnx.
Good job, now build few more boards so you can use 34401A as null-meter between two LTZs, then you can easily see if there are unstability/jumps between LTZs. You mounted LTZ with quite some spacing from PCB. I had random 0.5-0.6ppm jumps when I did similar on one of my boards, and moving LTZ chip flush to the board + adding additional cap on bottom side fixed those jumps. I've also tried extra caps like Dr.Frank/Andreas use, but saw no difference in measurements, in _my specific_ environment. Also I noticed you used same cap on position C4? Reference design has 22nF cap there, not 100nF.
FeedbackLoop I've heard that any extra airflow by the ltz1000 messes with the thermal effects since the board helps keep the reference a constant temp by adding thermal mass as well as preventing the introduction of stray air currents.
I wonder if that voltage reference device is monolithic on the inside. Maybe the "heater" is a resistor on the die itself? Amazing stuff! Than you so much for sharing! Great data analysis!
So, if the actual voltage is not important at all, why is this good as a voltage reference? I understand that it drifts very little with temperature, but how can it serve for calibration if the voltage is arbitrarily around 7.1 to 7.2V?
Depends on a use case. One way is to have a raw reference with a sticker on it with the expected value (7.10148 in my case). Another way is to add a scaling buffer to bring the output voltage to a convenient value such as 10V. But that is another challenge to design such a buffer with low drift and low noise.
Oh, it has been years back. I think it was a part of some soldering toolkit like this: www.ebay.com/itm/Elenco-SE-1-Solder-Ease-4-Piece-Kit-Brush-Scraper-Probe-and-Heatsink-/331478325103 I think a copper alligator clip would also do the business, but this special clip is narrower.
I am not sure about this particular design. I saw different designs with slots routed in the PCB in a spiral fashion to isolate the reference thermally and from mechanical stresses in the rest of the board. It is controversial and unclear to what extent such tricks help. Let's say, the industry standard 8.5-digit metrology-grade meter HP 3458A doesn't have such things on its LTZ1000 reference board.
Diabolical Artificer I believe those exposed areas are to radiate heat away from the traces. To avoid thermocouples, you must avoid any temperature drop across a solder junction, and the only other place for the temperature drop to occur would be along the trace or wire.
I've read those are for heat sinks for the reasons the other commenters mentioned, the 34401 dmm also uses a socket to avoid mechanical stresses in it's voltage reference though it is a different kind (lm399 vs ltz1000).
Very nice video. Its my first time here. I think the UPS could cause a glitch or some changes in the measurments. I am using an APC UPS for my network equipment and this kind of UPS is using an "autotransformer" to step up for example an input of 200V to my setting of 230V. If some time passed by and the input is close to the setpoint (in my case its flucuating between 227V and 229V) the UPS enteres an "ECO" mode and a relay bypasses the autotransformer (which is the backup step up transformer, too). I am not 100% sure because the ECO-mode despripction in the APC manual isn't quite technical. Even if the reference is extremly stable, both instruments could be affectet by the switch over and of course to every change in their input?!
Let's see... Measures between 7.10146 and 7.10147 after a few minutes of warming up. Something must have drifted a bit in three years. Probably both my HP 34401A meter and the reference, the meter being more likely. Even the best meters can drift a few ppm in three years.
The main requirements for such a reference is long-term stability and temperature stability. But the actual voltage is not important. It happens to be somewhere in the 7V - 7.2V range. Of course, it would be nice to have some precise value as well, but that is an extra requirement, which usually is achieved by some sort of additional circuit to scale the output of this reference, say, to 10V. The scaling factor can be trimmed in some way to get the desired value. This scaling circuit is a separate challenge to achieve its long-term stability, temperature stability and low noise.
first of all do us all a favour.. do not use aluminium case for better TC if you need a cooler or heatsink then don't use such a material that quickly coresponds to temp changes (aluminium) otherwise to acheive a better stability.. use plastic case and foam.. . not plastic foam styrofoam
Both DMMs are much more noisy and less stable (about 10 times) than the LTZ1000 circuit.
The 34401As internal LM399 may produce these 1..2ppm steps.
The bigger glitches may be produced by the LTZ1000 circuit, caused by RF disturbances, from switch mode power supplies in your lab. Watch out for LED lighting..
xdevs circuit lacks additional capacitors, which suppress most of these disturbances. I added 2 capacitors to the LT circuit, and things improved a lot, already. So please consult eevblog, the Super LTZ1000 thread for details on Andreas' circuit.
Looks to me that the glitches are related to your presence in the room, this has happened to me measuring very low noise resistors, I would suggest shielding all cabling. Thanks a lot for the video...great info!
Agreed. I was measuring a 50K ohm super precision resistor with similar flying leads using a Keithley 2450 SMU. Just walking past the setup, about 3 feet away, affected the readings with similar unexplained drops in measurements. The good thing with the keithley is you can see the drops on screen as it plots each measurement in real time. I have since purchased triaxial cables and appropriate shielding to repeat the measurements.
Watch the video and am thrilled. In the scene with the housing and the insulation, I realize that a metal housing only makes electrical sense and serves a purpose if it is also grounded to earth. And zipp, zapp zerapp in the next moment he puts a ground cable to the housing. I had to grin. Good work. Thank you.
That was a REALLY nice project. Thanks for the upload.
LTZ1000 PCBs always look cool.
I have noticed that radiated noise effects very small value uV values, I had to turn off as much switchmode equipment as I could in my lab when using my Fluke 343A DC calibrator ( looks the same as yours except it has one more knob for the uV range)
Pretty cool, what software did you use for those charts ?
Too bad you didn't monitor more things: Lab power supply, room lighting, time-of-day, power line changes, etc. I'm not sure I would connect the two DMMs together for this; one at a time might be better. Please let us know what you did to get access to a better meter, etc. GREAT video!
Great video! What software are you using to display and navigate the graphs?
Octave
Great video and very informative about the components. I too would like to build one of these circuits. Like you said, with any testing, you need a device that is more accurate than the device you are testing to rule out errors in the test equipment. Perhaps a better reference meter would show a much flatter line - but these type of meter are very expensive. I would be interested to know where you purchased your Vishay Foil Resistors from, because Digikey does not stock these and require a minimum order quantity of 500. Perhaps if you make up 2 of these boards and use each to cancel each other out, then measure the small difference in voltage. This is provided each is very close in output voltage. So you could have a series of boards and read the voltage difference off each pair combination. With a series of boards, it might be found that one specific board has a lower drift than the others and this would become the main reference. Keep up the good work!
I got Vishay resistors from eBay.
Good video, thanks for this. I am on the Volt-nut and Time-Nut path myself now :-)
Very good work
awesome! want to get rid of one of those extra pcbs? i really want to make one of these but dont want to order 3 boards:)
Not sure yet. Perhaps volt-nuttery will prevail, and I will build more. I am thinking about 10V standard, and I am not sure whether to build it based on this reference, or live this one alone as a compact and portable unit, and build another one with 10V scaling amplifier.
Will never ever pay 17 dollars per board...no way. this is crazy. Other chinese manufacturers will offer the same design board at a fraction of the cost. Why make oshpark rich for no reason ??? No tnx.
indeed use some polarity protection on boards like this with relatively costly components, a simple diode can save your day.
Good job, now build few more boards so you can use 34401A as null-meter between two LTZs, then you can easily see if there are unstability/jumps between LTZs.
You mounted LTZ with quite some spacing from PCB. I had random 0.5-0.6ppm jumps when I did similar on one of my boards, and moving LTZ chip flush to the board + adding additional cap on bottom side fixed those jumps. I've also tried extra caps like Dr.Frank/Andreas use, but saw no difference in measurements, in _my specific_ environment.
Also I noticed you used same cap on position C4? Reference design has 22nF cap there, not 100nF.
I used two stacked 10nF Panasonic film caps as C4.
Why would some space under LTZ lead to jumps?
Where did you put your additional cap on the bottom side?
FeedbackLoop I've heard that any extra airflow by the ltz1000 messes with the thermal effects since the board helps keep the reference a constant temp by adding thermal mass as well as preventing the introduction of stray air currents.
Great job, thanks!
I wonder if that voltage reference device is monolithic on the inside. Maybe the "heater" is a resistor on the die itself?
Amazing stuff! Than you so much for sharing! Great data analysis!
Something like that.
You may want to read this thread (very long and detailed):
www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/
@@feedback-loopthank you!
@@Tadesan Oh, it took you 5 years to reply 😀
So, if the actual voltage is not important at all, why is this good as a voltage reference? I understand that it drifts very little with temperature, but how can it serve for calibration if the voltage is arbitrarily around 7.1 to 7.2V?
Depends on a use case. One way is to have a raw reference with a sticker on it with the expected value (7.10148 in my case). Another way is to add a scaling buffer to bring the output voltage to a convenient value such as 10V. But that is another challenge to design such a buffer with low drift and low noise.
@@feedback-loop Thanks! So, it means that you should measure it first with a super accurate voltmeter?
Yes. It effectively becomes your transfer standard, so that later you can check other meters against it.
Hello,
where did you find the little grip which help you to divert the heat transfer?
Oh, it has been years back.
I think it was a part of some soldering toolkit like this:
www.ebay.com/itm/Elenco-SE-1-Solder-Ease-4-Piece-Kit-Brush-Scraper-Probe-and-Heatsink-/331478325103
I think a copper alligator clip would also do the business, but this special clip is narrower.
Thank you for reference link
Interesting video. What is the purpose of those circular traces on the PCB around the LTZ1000?
I am not sure about this particular design. I saw different designs with slots routed in the PCB in a spiral fashion to isolate the reference thermally and from mechanical stresses in the rest of the board. It is controversial and unclear to what extent such tricks help. Let's say, the industry standard 8.5-digit metrology-grade meter HP 3458A doesn't have such things on its LTZ1000 reference board.
Diabolical Artificer I believe those exposed areas are to radiate heat away from the traces. To avoid thermocouples, you must avoid any temperature drop across a solder junction, and the only other place for the temperature drop to occur would be along the trace or wire.
I've read those are for heat sinks for the reasons the other commenters mentioned, the 34401 dmm also uses a socket to avoid mechanical stresses in it's voltage reference though it is a different kind (lm399 vs ltz1000).
Very nice video. Its my first time here. I think the UPS could cause a glitch or some changes in the measurments. I am using an APC UPS for my network equipment and this kind of UPS is using an "autotransformer" to step up for example an input of 200V to my setting of 230V. If some time passed by and the input is close to the setpoint (in my case its flucuating between 227V and 229V) the UPS enteres an "ECO" mode and a relay bypasses the autotransformer (which is the backup step up transformer, too). I am not 100% sure because the ECO-mode despripction in the APC manual isn't quite technical. Even if the reference is extremly stable, both instruments could be affectet by the switch over and of course to every change in their input?!
I would like to see the histograms..
Is that ceramic-filled laminate?
The board? No just a regular PCB.
Hi, what software are you using for the graphs?
got it, octave!
How's the reference doing please ?...cheers.
still between 7.10146 and 7.10147
Could cosmic rays cause such glitches i wonder?
Mustafa Yasir Aydın no
Great video...7.10148...is it still the same ?
Let's see... Measures between 7.10146 and 7.10147 after a few minutes of warming up. Something must have drifted a bit in three years. Probably both my HP 34401A meter and the reference, the meter being more likely. Even the best meters can drift a few ppm in three years.
excuseme, why accuracy is not mentioned?
The main requirements for such a reference is long-term stability and temperature stability. But the actual voltage is not important. It happens to be somewhere in the 7V - 7.2V range. Of course, it would be nice to have some precise value as well, but that is an extra requirement, which usually is achieved by some sort of additional circuit to scale the output of this reference, say, to 10V. The scaling factor can be trimmed in some way to get the desired value. This scaling circuit is a separate challenge to achieve its long-term stability, temperature stability and low noise.
Got it. Thank you very much
круто
ты русский?
first of all do us all a favour.. do not use aluminium case for better TC if you need a cooler or heatsink then don't use such a material that quickly coresponds to temp changes (aluminium) otherwise to acheive a better stability.. use plastic case and foam.. . not plastic foam styrofoam
Уж лучше на Русском