23:40 many years ago I had an internship at Vishay, and one thing that surprised me was that next to their automatic assembly lines, they have a special department to hand-manufacture and hand-validate passive electronic components. The application for these is not surprising however: it's for satellites, where you need zero fault tolerace.
@@InTimeTraveller you can design a circuit to be fault tolerant, but you can also personally not be tolerant of faults. Zero fault tolerance is correct in this case, as the users tolerate not faults. It's also not possible to even remotely approach infinite fault tolerance.
I worked in the defence industry 30 years ago and even then we could order resistors with exact resistance to 6 d.p. Somewhere I have a 1.000000 ohm resistor encapsulated in plastic as a key ring.
What is the resistance of that resistor's leads? When resistances are low like this, the only real way to achieve the accuracy that you cite is to use 4-terminal (aka "Kelvin") force-sense leads.
Excellent, thankyou Marco. I designed the 7000 series for Datron which was then acquired by Wavetek and then Fluke. In fact the use of the Thin film networks preceded Wavetek and was introduced by us at Datron in the belief that inherent process ratio accuracy was probably more reliable and certainly cheaper than selection by measurement. Thin film also has an advantage over foil because it avoids the epoxy bond between the foil and substrate which is very humidity sensitive. My experience of the ADR after two years working with it (and the thin film networks) is good. Lower noise, especially in the 1/f region sub below 1 Hz is excellent and best measured with two ADR references back-to back. Tytpically half the noise. I have also been amazed by the process repeatability that ADI achieve, far tighter than the LTZ part, at least within two batches that I've seen. Oh, and I agree with the advantages of biasing to run near zero chip TC. All of the Datron Wavetek products did this. Anyway keep up the good humoured work on this and don't let it take over your life!
Thank you John, this means a lot! I wish I could have found more creative solutions myself, but the boxes equipped with your techniques are by far the best performers.
I was just scrolling through the comments and noticed you were able to reduce the noise of these devices by using them back to back? If you don't mind, can you give me a clue how can back to back diodes remove half the noise? as i thought using 2 in series you'd be able to only reduce the noise by 1.4? I am currently using 4 references stacked and divide down to get half the noise but perhaps I am missing something?
90% if the time I have no real clue what you are talking about, yet I still watched almost an hour video of you testing something I didn't know existed, nor was needed, or that there was a part almost as old as I am that had done the job previously. But I learn something every time.
As somone that has no need for an ultra precise 10v reference and is happy with an onboard software calibration that as far as I know it just works. I think Im going to loose an hour learning about something I will not need to know but think it will be worth it for all the German humor. Thank you and may your success continue.
I worked at Vishay a bit, and while I didn't work on these resistors in particular, the numbers they're breaking are done according to taylor series approximations so they can be customized pretty easily. The "software" just does the taylor series math for you. Its pretty neat, and sometimes done with lasers rather than by hand.
It's lovely to see you back at work on custom hardware instead of reviewing tools. I knew you wouldn't be satisfied with the previous result of your home-built LTZ standard. Nothing like making 12 different versions to figure out what variables actually matter!
Talking about those guys having their initials on the die, I actually had a very tiny caricature of myself on a device that was used "quite often"... It may have been removed in later generations, because the production guys had trouble with design rule tests on the layout, but, it was kind of fun to know that a tiny representation of me was in maybe a couple of million devices... Fun days! Now, back to precision!
Hey Marco, quick tip when using SCPI commands: Instead of using a time.sleep function between commands, use the "*OPC?" command (operation complete?) so your instrument will return a message when it's done with the previous command :)
Does that also work over ethernet? Had the problem with some Keithley that it only returns an ACK immediately and doesn't delay a response (in general a more "special" instrument though, so could be just that instrument)
@@sophyx9 As far as I know, yes. I have used it over ethernet, usb and gpib with different instruments from different brands and it always worked. It is a general SCPI command like RST or CLR
@@sophyx9 Yes it does. I have used *OPC? with ancient devices to brand new ones that aren't even released yet. So far there hasn't been a single device that it doesn't work with. The code execution can be made as fast as the devices allow with it.
One can also use *WAI; to perform another operation (e.g. a read-out of a register or value) and receive its result once all pending commands up to *WAI are finished. scpi-99: "After executing *WAI the device must execute no further commands or queries until the No Operation Pending flag is TRUE, or receipt of a dcas message, or a power on."
Marco, you are a masochist building so many voltage references, and I love it. Please let us know if any of them decide to misbehave and have to be given some tough love to get them back on the straight and narrow. PPMs are not overrated. They're essential. Anyone who says you only need only one voltage reference is lying to themselves and should not be trusted.
I cannot explain why but this is so pleasing for me to watch! I know how much work goes into making a good 10 minute video, so thanks for making a one hour masterpiece. PPM love from Denmark
I am just amazed how well you can match the bell curve of peak precision hightech electronics knowledge to the ambient-dependent trend of funny commentary. Absolutely jaw-dropping..
Marco, Thanks for all of this (and supporting @Hygens Optics!). For those of us hobbyists that have that rare occasion to need a voltage reference to reference check/calibrate their Siglent and Rigol bench DVMs and don't feel like the cost of sending in for a professional cal.
As usual with a PPMs video I understood all of the words and even some of the sentences but I think it's going to take a few watches to understand half of what's going on. None the less a very watchable video.:) PS congratulations on getting close to your goal.
Marco, from what I saw on your 3458A hook up you'd benefit from some metrology grade cables, i.e. PTFE silver plated copper or such like. Anyways, great 1hr video that obviously took weeks to make. Nice one!
years ago, before I found Marco's channel, I thought PPM stood for Parts Per Million. but now, as I listen to him use words I understand but have never applied to anything this precise, I know what it truly means: Phriends Phorever, Man.
I’m finally starting my electrical engineering program at Purdue University, and it’s so cool to better understand your videos after going to my classes!
I really appreciate the explanation about how you generated the KiCAD b-roll. It makes a lot of sense, and it was really obvious once you explained it.
Jeezes, this is some incredible work! Both research and storytelling is on point. Can't wait for my next order from PCBWay, I'll add this to the basket.
Did you check if the gold ground bar actually makes contact with the aluminum in the case? The anodization on that aluminum is probably 10-15 micrometers thick and very insulative. Of course 15 microns of Al2O3 with such a large area creates a relatively high capacitance capacitor if there is no puncturing of the oxide film. For at least noise in the +30 MHz range it might actually be a relatively low impedance path to ground.
Great work! I can only dream of having the money to do this type of stuff but just watching is fun. I have been mulling various ways of hacking together cheap reference standards for many years.
my favorite German on UA-cam. thanks for sharing man! you're partly responsible for my getting back in to electronics, but thinking about it on a different level. also made me realize, my level is very low, like could I build the 9 digit multimeter from CERN? maybe... but it would be a giant struggle, you make it look easy.
I'd love to get an in depth video analyzing the die shape and the reasons behind that, and another video on your detail design process of the stuff you are showing us on kicad.
If you print your cup upside down (so the flat bottom of the cup is the last thing printed) with dense but light supports you'll be able to print it ok. The surface that seals to the PCB would want a light sand to bring it back to flatness as it'll be wavy after support removal but that's easier than gluing multiple parts together and should help prevent any PPMs leaking out the join. If you print it at a bit of an angle as well the force for releasing off the film will be dramatically reduced. It will of course add to the print time somewhat but not normally too bad. I got myself the same rack from reichelt(sp) a while ago. Looking forward to making a similar module for it ;-) my goal is to put my workbench tools into subrack modules (eventually) and to learn a bunch by making those tools.
Shout-outs for various specific boards: ADRmu #1 - 23:0441:35 includes VFR resistor array "golden boi" - about at data-sheet noise performance LTZmu #2 - 42:58 - pimped-out resistors - 1.6 uV p-p ADRmu #4 - 45:40 - specialized circuit trial unit - maybe popcorn noise ADRmu #13 - 21:10 ADRmu #9 - 26:26 includes the best ADR1000 (closest to 10volt output) - 1.22 uV p-p LTZmu #3 - 28:2944:20 - low cost resistors with tempco. fine tuning ADRmu #7 - 30:45 monolithic resistor array with tempco. fine tuning (modify pin #5 -> R2 60k to 20k) 47 Celcius set-point (20 ppb final tempco.) - best low noise unit ADRmu #8 - 30:56 cooked pre-aged at 140 Celcius ADRmu #9 - 46:53 factory new without burn-in and no tempco. fine tuning - to be used to verify the spec. sheet drift ADRmu #107 - 47:50 - no artificial burn-in and mid-grade resistors and tempco. fine tuning ADRmu #12 - 49:04 - jumper for raising oven temperature temporarily
Looking at your temperature sweep thing...we did a little thing in one application, using a few temperature sensors, inside the enclosure, to look out for differentials, and had the option to add resistor "heaters" on the board, at determined points, to equalize the temperature inside the enclosure during constant use... The lab environmental temperature doesn't affect the internal *distribution* very much, so that one can minimize thermocouple effects to pretty darned small...or, so I thought...
Can't get over how well you integrate ads in your videos, I was just as interested in the sponsored segment as the rest of the video. @pcbway whatever you're paying this guy, it isn't enough! Fantastic marketing!
Thanks Marco for this detailed video and revealing all of the minefields of traps and problems in trying to make a complete practical voltage standard. I've really learned a lot watching your videos and am impressed with all your attention to detail and hard-work. I especially like your discussions of tradeoffs! I have 2 731B's, an HP-735, a Fluke 332D and a 343A calibrator along with several 3455A and a 3456A and by mixing and matching individuals against each other, and looking at their references with a null-meter, they ALL drift with room-temperature! I was very disappointed, that at least one of them couldn't be used as the God-Box! So, I may have to get one of your set-ups just so I can have some point of stability, and can accurately characterise the rest. I want to replace the reference in the calibrators and the DMM's since a xxx1000 would be much better than what is in them. I guess what I'm dancing around is, that I can now fully understand the expense and outlay for something like a 3458A just to stop the ppm-madness!
I always had an issue of measuring 0.1hz to 10hz with a scope and measuring the peak to peak, then using the crest factor. Noise is statistical so you can get lots of different values. Rms would be better, but noise density would be best. When I played with the ltz1000 a few years ago, I didn't cut the leads and soldered directly to the end of the leads. This helps thermal isolation and allows for mechanical relief. Lastly, keep in mind that the resistors also drift. You should try burning the resistors and swapping out a new zener. You will definitely see a difference. Loved the video.
@Bibarco Yes, you can use a scope or the analog discovery 2 to do the fft. You will need a lot a gain to overcome the noise floor. Typically, they use an low frequency high pass to not saturate the input. Then we are back to a similar issue.The issue of measuring reference noise is the DC component. Using 2 channels and cross correlation to cancel out the un-corrilated noise will work, but it will take a lot of samples.
The 6 layer approach could use internal planes to form interplane capacitance. This capacitance along with different decoupling caps near the ICs could combine to produce a very high frequency filter with a reduction in inductance. To put it another way, the interplane capacitance is like you own capacitor design by setting the board dielectric and the prepreg thickness, it is very effective, and is one reason you may want to add additional plane layers.
@@benjaminbawkon8040 The comment was targeted towards people in the comments and not the author of the video, as to why you might want to increase the layer count from 4 to 6 in addition to heat equalization across the board.
Loved the video, I've been missing your campaign for PPMs. Mad idea: what about a single module with three references fed from the same source voltages on the same board? Make a kind of 'consensus source' where the oddities of each source are averaged out.
Similar in spirit to Guildline 4410 (Statronics VS4)? The ADR1000 is just so much more expensive than LM329 are (were, they seem to be obsolete), but also much lower noise.
Why no whole board Peltier like the CERN voltmeter? Even only working when on external power should be useful and likely much cheaper than fancy resistors
15:00 Dang it, I thought you had a script that took the kicad files and processed them into mouse movements. Though I did wonder how you selected parts for placing... Assumed it was on a different screen.
Huh? That’s a standard (if mostly obsolete for normal uses) IC package. Op-amps, voltage regulators, etc. all used to ship in those, before DIP took over.
Marco, would you have time and interest in making a video showing possibilities of how to eliminate Power Supply noise? I mean the audible noises! There is a horde of people on the forums who, like me, change power supplies for LED lighting and even cell phone chargers because of the damn noise of the sources. Thanks.
Cool video, taught me a lot about the rh1021 bmh-10 and the DLA variant from analog. I think it uses a similar internal design with the "buried zener" and I never knew much about what that meant.
Hey, Marco, do you have any 0.01% 10 M resistors hanging around? I seem to have destroyed one in my 500,000 count meter...I was thinking about "building" a 10 Mohn with something like 9.8 M and trimming with 100-200 K, but...10 M seems to be hard to find. BTW, I do have a couple of 1 M and 100 K in 0.01%, if you'd find that useful to have around?
There are some decent soviet glass-encased microwire resistors on eBay sometimes, lots of data in here www.eevblog.com/forum/metrology/i-plan-to-make-an-100-meg-standard-resistor/
27:22 to err is human . I'm still mad that I connected my battery in reverse on my UPS and there wasn't any protection against polarity inversion, even half a second has enough to explode the mosfets, and that was $10 each in mosfets. (lol, entire new used unit without the batteries is like $100 ) whatever, I was just trying to fix my old UPS while I waited for the newer one, lead times are horrible
me again...i was wondering about case grounding at the board edge...if the case is aluminum, its own oxide layer is a pain, and anodizing even worse...maybe phosphor bronze contact springs, mechanically-coupled to bare aluminum, and these spring contacts contact the board edge? but, hey, what do i know?
Neat. Maybe it's my background coming into play, but I wonder if using something like a, gasp, digital control circuit would do for things like the heater. It might seem like blasphemy, but when you're talking MHz a PWM signal should be pretty accurate as a divider. The other thought was to use a frequency standard to somehow enhance the accuracy. However, I haven't done the research to figure out what the drift or tempco of the different options there are. I don't need anything near this accurate, but it's a fun thought experiment, and fun to watch you working on it.
Is that part cool just because it has “WTF 001 XD” printed on it? 😂 Also I was like “wow I wish I was that competent with keyboard shortcuts” and then Reps says “I’ve just been playing video of deleting traces in reverse” 😅😂
Better noise: Absolutely! Better stability: Yes we could probably slow the material aging processes a lot Environmental immunity and compatibility with this relatively low effort implementation: Not so sure ... Either way xdevs.com is planning something like this and I can't wait to hear about the results :) xdevs.com/doc/xDevs.com/cryotest/cryoltz.jpg
Let me make a random guess, do you have a Poweredge R710 or similar running in the background? The fans in the background sounds is exactly as mines haha
23:40 many years ago I had an internship at Vishay, and one thing that surprised me was that next to their automatic assembly lines, they have a special department to hand-manufacture and hand-validate passive electronic components. The application for these is not surprising however: it's for satellites, where you need zero fault tolerace.
we use these resistors, those are not cheap
I am working in a research facility in Germany. We build satellites. And we use exactly those resistors :)
In satellites technically you need infinite fault tolerance but I get your point 😝
@@InTimeTraveller you can design a circuit to be fault tolerant, but you can also personally not be tolerant of faults. Zero fault tolerance is correct in this case, as the users tolerate not faults. It's also not possible to even remotely approach infinite fault tolerance.
Oh nice to see you here. Yeah, aerospace is just another level.
I worked in the defence industry 30 years ago and even then we could order resistors with exact resistance to 6 d.p. Somewhere I have a 1.000000 ohm resistor encapsulated in plastic as a key ring.
I wonder what it's drift would be.
I inherited a pile of super precision resistors a while back. So expensive. Some custom milspec job. Wish I had data sheet.
DEFENSE = GOVERNMENT, no wonder they could squander a super high precision resistor on a lost key ring…
What is the resistance of that resistor's leads? When resistances are low like this, the only real way to achieve the accuracy that you cite is to use 4-terminal (aka "Kelvin") force-sense leads.
i have some 0,005% "general radio" resistors from a military "ballistic computer", 1000 and 1M ...
That isolation slot joke is rough 🤣
Makes me wonder what the magic smoke released from it would smell like. I'm thinking almonds
Not a shape usually associated with stability😬
@@friedtomatoes4946 LOL
Hakenkreuzisolationsschlitz.
Congrats on being just as reactionary and incorrect as he is. That is a manji at worst.
shorinjiryublog.files.wordpress.com/2019/04/manji11.jpg
Excellent, thankyou Marco. I designed the 7000 series for Datron which was then acquired by Wavetek and then Fluke. In fact the use of the Thin film networks preceded Wavetek and was introduced by us at Datron in the belief that inherent process ratio accuracy was probably more reliable and certainly cheaper than selection by measurement. Thin film also has an advantage over foil because it avoids the epoxy bond between the foil and substrate which is very humidity sensitive. My experience of the ADR after two years working with it (and the thin film networks) is good. Lower noise, especially in the 1/f region sub below 1 Hz is excellent and best measured with two ADR references back-to back. Tytpically half the noise. I have also been amazed by the process repeatability that ADI achieve, far tighter than the LTZ part, at least within two batches that I've seen. Oh, and I agree with the advantages of biasing to run near zero chip TC. All of the Datron Wavetek products did this. Anyway keep up the good humoured work on this and don't let it take over your life!
Thank you John, this means a lot! I wish I could have found more creative solutions myself, but the boxes equipped with your techniques are by far the best performers.
I was just scrolling through the comments and noticed you were able to reduce the noise of these devices by using them back to back? If you don't mind, can you give me a clue how can back to back diodes remove half the noise? as i thought using 2 in series you'd be able to only reduce the noise by 1.4? I am currently using 4 references stacked and divide down to get half the noise but perhaps I am missing something?
90% if the time I have no real clue what you are talking about, yet I still watched almost an hour video of you testing something I didn't know existed, nor was needed, or that there was a part almost as old as I am that had done the job previously. But I learn something every time.
As somone that has no need for an ultra precise 10v reference and is happy with an onboard software calibration that as far as I know it just works. I think Im going to loose an hour learning about something I will not need to know but think it will be worth it for all the German humor. Thank you and may your success continue.
It isn't precise, it is stable, as you can see most aren't even 10.0V , they will just retain that number for a very very long time.
Ah yes, Marco with a hour long video never disappoints :)
You know you're using a niche part when using it 5-6 years after release still makes you an early adopter
I worked at Vishay a bit, and while I didn't work on these resistors in particular, the numbers they're breaking are done according to taylor series approximations so they can be customized pretty easily. The "software" just does the taylor series math for you. Its pretty neat, and sometimes done with lasers rather than by hand.
It's lovely to see you back at work on custom hardware instead of reviewing tools. I knew you wouldn't be satisfied with the previous result of your home-built LTZ standard. Nothing like making 12 different versions to figure out what variables actually matter!
I dont even know that much about electronics, yet still watch new Marco Reps videos as soon as I notice them.
Same, I'm here for the soothing voice and dry jokes. I might even learn something!
Same.
The sprinkling passives was the best part. Glad you took the time lol.
I tried not to laugh, but did not have the capacity to resist.
I will see my self out.
Talking about those guys having their initials on the die, I actually had a very tiny caricature of myself on a device that was used "quite often"... It may have been removed in later generations, because the production guys had trouble with design rule tests on the layout, but, it was kind of fun to know that a tiny representation of me was in maybe a couple of million devices... Fun days!
Now, back to precision!
well, now i want the part number
lp5523
@@derkarhu5079 Couldn't find any pics, do you have one?
@@xThirdOpsxLP5523
@@derkarhu5079 TI 9 channel LED driver? Your face has been lighting up people's lives around the world!
Hey Marco, quick tip when using SCPI commands: Instead of using a time.sleep function between commands, use the "*OPC?" command (operation complete?) so your instrument will return a message when it's done with the previous command :)
Does that also work over ethernet? Had the problem with some Keithley that it only returns an ACK immediately and doesn't delay a response (in general a more "special" instrument though, so could be just that instrument)
@@sophyx9 As far as I know, yes. I have used it over ethernet, usb and gpib with different instruments from different brands and it always worked. It is a general SCPI command like RST or CLR
@@sophyx9 Yes it does. I have used *OPC? with ancient devices to brand new ones that aren't even released yet. So far there hasn't been a single device that it doesn't work with. The code execution can be made as fast as the devices allow with it.
like I said, a more "special" instrument, Keithley 4200-SCS and also not the newest version (software and hardware).. :D
One can also use *WAI; to perform another operation (e.g. a read-out of a register or value) and receive its result once all pending commands up to *WAI are finished.
scpi-99: "After executing *WAI the device must execute no further commands or queries until the No Operation Pending flag is TRUE, or receipt of a dcas message, or a power on."
Any day with a new Marco Reps video is a damn good day.
It is immensely satisfying watching SMD components soldered when paste is used.
Marco, you are a masochist building so many voltage references, and I love it. Please let us know if any of them decide to misbehave and have to be given some tough love to get them back on the straight and narrow. PPMs are not overrated. They're essential. Anyone who says you only need only one voltage reference is lying to themselves and should not be trusted.
I cannot explain why but this is so pleasing for me to watch! I know how much work goes into making a good 10 minute video, so thanks for making a one hour masterpiece.
PPM love from Denmark
I am just amazed how well you can match the bell curve of peak precision hightech electronics knowledge to the ambient-dependent trend of funny commentary. Absolutely jaw-dropping..
14:54 take it from an professional PCB designer. Your method is how 90% of motherboards are designed. Also copper ground floods gooood.
when he just casually whips out the 25,000 usd oscilloscope
Marco, Thanks for all of this (and supporting @Hygens Optics!). For those of us hobbyists that have that rare occasion to need a voltage reference to reference check/calibrate their Siglent and Rigol bench DVMs and don't feel like the cost of sending in for a professional cal.
Signs of life! I figured the PPMs had you chained to a radiator somewhere. I always enjoy your videos, Marco.
As usual with a PPMs video I understood all of the words and even some of the sentences but I think it's going to take a few watches to understand half of what's going on. None the less a very watchable video.:)
PS congratulations on getting close to your goal.
That's a very good description of what watching these videos is like.
Marco, from what I saw on your 3458A hook up you'd benefit from some metrology grade cables, i.e. PTFE silver plated copper or such like. Anyways, great 1hr video that obviously took weeks to make. Nice one!
Sure, although I heard Virgin-blessed Negative-Oxygen Cryo Copper treated with Baby Jesus Tears is better.
PTFE is (generally) good, but silver has slightly worse thermal EMF than pure copper. His cables are quite fine.
Well, this is a documentary rather than just a Volt Nut video !!
years ago, before I found Marco's channel, I thought PPM stood for Parts Per Million. but now, as I listen to him use words I understand but have never applied to anything this precise, I know what it truly means: Phriends Phorever, Man.
I’m finally starting my electrical engineering program at Purdue University, and it’s so cool to better understand your videos after going to my classes!
Fantastic! I love the level of detail for this wonderful and *very* rare topic!
I really appreciate the explanation about how you generated the KiCAD b-roll. It makes a lot of sense, and it was really obvious once you explained it.
This has to be the best collection of EE word salad I have ever heard! Bravo!
Not sure I have ever seen somebody put this much work into a video. Really amazing, thanks for sharing
your voice is the most soothing and enamoring thing i have ever heard, i can never stop watching your videos
Jeezes, this is some incredible work! Both research and storytelling is on point. Can't wait for my next order from PCBWay, I'll add this to the basket.
"I did it this way because it was the LEAST popular option."
Man... You crack me up!
Keep up the Excellent Videos!
That sprinkle of passives
must've been a tedious cleanup 🤣
Did you check if the gold ground bar actually makes contact with the aluminum in the case? The anodization on that aluminum is probably 10-15 micrometers thick and very insulative. Of course 15 microns of Al2O3 with such a large area creates a relatively high capacitance capacitor if there is no puncturing of the oxide film. For at least noise in the +30 MHz range it might actually be a relatively low impedance path to ground.
Glad to see you put those caps to good use!
Great work! I can only dream of having the money to do this type of stuff but just watching is fun. I have been mulling various ways of hacking together cheap reference standards for many years.
my favorite German on UA-cam. thanks for sharing man! you're partly responsible for my getting back in to electronics, but thinking about it on a different level. also made me realize, my level is very low, like could I build the 9 digit multimeter from CERN? maybe... but it would be a giant struggle, you make it look easy.
I'd love to get an in depth video analyzing the die shape and the reasons behind that, and another video on your detail design process of the stuff you are showing us on kicad.
your one of the better youtubers man thanks for making open source amazing things keep it up dude
If you print your cup upside down (so the flat bottom of the cup is the last thing printed) with dense but light supports you'll be able to print it ok. The surface that seals to the PCB would want a light sand to bring it back to flatness as it'll be wavy after support removal but that's easier than gluing multiple parts together and should help prevent any PPMs leaking out the join. If you print it at a bit of an angle as well the force for releasing off the film will be dramatically reduced. It will of course add to the print time somewhat but not normally too bad.
I got myself the same rack from reichelt(sp) a while ago. Looking forward to making a similar module for it ;-) my goal is to put my workbench tools into subrack modules (eventually) and to learn a bunch by making those tools.
1 hour of our beloved MArco Reps is a godsend right now
"It's not pitchfork time yet" I love your dry wit.
Shout-outs for various specific boards:
ADRmu #1 - 23:04 41:35 includes VFR resistor array "golden boi" - about at data-sheet noise performance
LTZmu #2 - 42:58 - pimped-out resistors - 1.6 uV p-p
ADRmu #4 - 45:40 - specialized circuit trial unit - maybe popcorn noise
ADRmu #13 - 21:10
ADRmu #9 - 26:26 includes the best ADR1000 (closest to 10volt output) - 1.22 uV p-p
LTZmu #3 - 28:29 44:20 - low cost resistors with tempco. fine tuning
ADRmu #7 - 30:45 monolithic resistor array with tempco. fine tuning (modify pin #5 -> R2 60k to 20k) 47 Celcius set-point (20 ppb final tempco.) - best low noise unit
ADRmu #8 - 30:56 cooked pre-aged at 140 Celcius
ADRmu #9 - 46:53 factory new without burn-in and no tempco. fine tuning - to be used to verify the spec. sheet drift
ADRmu #107 - 47:50 - no artificial burn-in and mid-grade resistors and tempco. fine tuning
ADRmu #12 - 49:04 - jumper for raising oven temperature temporarily
Videos like this give me an ultimate satisfaction so thanks for that Marco.
I was just thinking about how I was running low on good dry wit, and here you are!
Looking at your temperature sweep thing...we did a little thing in one application, using a few temperature sensors, inside the enclosure, to look out for differentials, and had the option to add resistor "heaters" on the board, at determined points, to equalize the temperature inside the enclosure during constant use... The lab environmental temperature doesn't affect the internal *distribution* very much, so that one can minimize thermocouple effects to pretty darned small...or, so I thought...
As always 1 hour well spent....cheers.
Wooooow long time no see and you came back with voltage
reference!
The hour just flew by... fascinating... As always.
Can't get over how well you integrate ads in your videos, I was just as interested in the sponsored segment as the rest of the video. @pcbway whatever you're paying this guy, it isn't enough! Fantastic marketing!
I almost always skip ads, but with this video I watched it too.
They have a great product offering, didn't know about custom flex boards until his videos and seeing is believing.
Thanks Marco for this detailed video and revealing all of the minefields of traps and problems in trying to make a complete practical voltage standard. I've really learned a lot watching your videos and am impressed with all your attention to detail and hard-work. I especially like your discussions of tradeoffs!
I have 2 731B's, an HP-735, a Fluke 332D and a 343A calibrator along with several 3455A and a 3456A and by mixing and matching individuals against each other, and looking at their references with a null-meter, they ALL drift with room-temperature! I was very disappointed, that at least one of them couldn't be used as the God-Box! So, I may have to get one of your set-ups just so I can have some point of stability, and can accurately characterise the rest. I want to replace the reference in the calibrators and the DMM's since a xxx1000 would be much better than what is in them.
I guess what I'm dancing around is, that I can now fully understand the expense and outlay for something like a 3458A just to stop the ppm-madness!
I always had an issue of measuring 0.1hz to 10hz with a scope and measuring the peak to peak, then using the crest factor. Noise is statistical so you can get lots of different values. Rms would be better, but noise density would be best. When I played with the ltz1000 a few years ago, I didn't cut the leads and soldered directly to the end of the leads. This helps thermal isolation and allows for mechanical relief. Lastly, keep in mind that the resistors also drift. You should try burning the resistors and swapping out a new zener. You will definitely see a difference. Loved the video.
An Analog discovery II can measure frequency response as low as you have time. Normally the Amplitude is measured with an fft to be immune to noise.
@Bibarco Yes, you can use a scope or the analog discovery 2 to do the fft. You will need a lot a gain to overcome the noise floor. Typically, they use an low frequency high pass to not saturate the input. Then we are back to a similar issue.The issue of measuring reference noise is the DC component. Using 2 channels and cross correlation to cancel out the un-corrilated noise will work, but it will take a lot of samples.
The 6 layer approach could use internal planes to form interplane capacitance. This capacitance along with different decoupling caps near the ICs could combine to produce a very high frequency filter with a reduction in inductance. To put it another way, the interplane capacitance is like you own capacitor design by setting the board dielectric and the prepreg thickness, it is very effective, and is one reason you may want to add additional plane layers.
I'm pretty sure Mr. Reps is familiar with interplane capacitance. Lol
@@benjaminbawkon8040 The comment was targeted towards people in the comments and not the author of the video, as to why you might want to increase the layer count from 4 to 6 in addition to heat equalization across the board.
Very happy about long videos!
The PPM whisperer is back and he sounds like he's about to evolve into PPB whisperer!
Loved the video, I've been missing your campaign for PPMs.
Mad idea: what about a single module with three references fed from the same source voltages on the same board? Make a kind of 'consensus source' where the oddities of each source are averaged out.
Similar in spirit to Guildline 4410 (Statronics VS4)? The ADR1000 is just so much more expensive than LM329 are (were, they seem to be obsolete), but also much lower noise.
@1:52 theres a picture that really looks like a real city landscape. are you sure thats circuits? that looks so weird!
thx for sharing!
Why no whole board Peltier like the CERN voltmeter? Even only working when on external power should be useful and likely much cheaper than fancy resistors
Loved how at start he showed circuits and cities :D
Honey wake up, a new Reps video jut dropped!
Endlich ein neues Video!! Geilo!
Earlier today, I wondered when another video would be ready. And here we are!
This channel is a reference for the electronics community. 😀
You might be interested in ALD1108E. If I'm reading the datasheet right you can use them as stable, electrically controlled trimmers.
15:00 Dang it, I thought you had a script that took the kicad files and processed them into mouse movements. Though I did wonder how you selected parts for placing... Assumed it was on a different screen.
Always a huge pleasure to see a new video from you, and this one was squarely in the "things I am obsessed about" target zone. Marvellous stuff.
37:20 Why can't SLA printer not handle this case? Is there an explanation, why this shape is not possible?
I just looked at the Marco Reps merch. There is an EMU however there is nothing with PPM. 😟
I absolutely love the vacuum tube pin out. I guess the LTZ1000 was a drop in replacement for a tube?
Huh? That’s a standard (if mostly obsolete for normal uses) IC package. Op-amps, voltage regulators, etc. all used to ship in those, before DIP took over.
How the hell did I watch this video for an hour, not understand a thing and still enjoy it?
You need to get up to around 200ºC or higher to release the trapped charge.
Marco, would you have time and interest in making a video showing possibilities of how to eliminate Power Supply noise? I mean the audible noises!
There is a horde of people on the forums who, like me, change power supplies for LED lighting and even cell phone chargers because of the damn noise of the sources.
Thanks.
Why not chill the chips instead of heating? An ice bath is crazy temp stable. Why not use multiple chips to offset any swings in a single chip?
Ah, just what I needed today. Some stable ppms for an unstable world.
With your username and channel icon, I was sad to see that you have no videos posted. Where are teh lasers?
Anyone can name all the cool parts at 6:07?
Cool video, taught me a lot about the rh1021 bmh-10 and the DLA variant from analog. I think it uses a similar internal design with the "buried zener" and I never knew much about what that meant.
Have you ever considered potting one to keep control humidity?
Marco back!!🤩🤩
What orientation did you 3D print the cup? I thought if it was sideway, it would work.
Aaarrrghhhh ... ah hour..... we are blessed
What is the surface-mount reference you showed at 0:25?
I think the patent you are talking about has already expired.
Probably LTC6655
Hey, Marco, do you have any 0.01% 10 M resistors hanging around? I seem to have destroyed one in my 500,000 count meter...I was thinking about "building" a 10 Mohn with something like 9.8 M and trimming with 100-200 K, but...10 M seems to be hard to find.
BTW, I do have a couple of 1 M and 100 K in 0.01%, if you'd find that useful to have around?
There are some decent soviet glass-encased microwire resistors on eBay sometimes, lots of data in here www.eevblog.com/forum/metrology/i-plan-to-make-an-100-meg-standard-resistor/
27:22 to err is human . I'm still mad that I connected my battery in reverse on my UPS and there wasn't any protection against polarity inversion, even half a second has enough to explode the mosfets, and that was $10 each in mosfets. (lol, entire new used unit without the batteries is like $100 )
whatever, I was just trying to fix my old UPS while I waited for the newer one, lead times are horrible
me again...i was wondering about case grounding at the board edge...if the case is aluminum, its own oxide layer is a pain, and anodizing even worse...maybe phosphor bronze contact springs, mechanically-coupled to bare aluminum, and these spring contacts contact the board edge?
but, hey, what do i know?
I dont know why, but I watched this about 6 times, something about stable voltage references is calming
Neat. Maybe it's my background coming into play, but I wonder if using something like a, gasp, digital control circuit would do for things like the heater. It might seem like blasphemy, but when you're talking MHz a PWM signal should be pretty accurate as a divider. The other thought was to use a frequency standard to somehow enhance the accuracy. However, I haven't done the research to figure out what the drift or tempco of the different options there are.
I don't need anything near this accurate, but it's a fun thought experiment, and fun to watch you working on it.
Is that part cool just because it has “WTF 001 XD” printed on it? 😂
Also I was like “wow I wish I was that competent with keyboard shortcuts” and then Reps says “I’ve just been playing video of deleting traces in reverse” 😅😂
5:00 "Auf der Heide blüht ein kleines Blümelein" 🎶
laught so hard at this part x'D
Dear, did you analyzed the spectrum of the noise in ADR1000?
We need more videos man, love your stuff 🤠
This is so fascinating man! I only wish for even more videos :D
I don't really know anything about this stuff, but I enjoy watching anyways.
Does anyone know what reflow plate he uses at 30:42 ?
Interesting, I always thought HP gave LTC the design for ltz1000, dunno where i read that.
Wouldn't you get better noise and stability by cooling it down with a regulated TEC rather than a heater?
Better noise: Absolutely!
Better stability: Yes we could probably slow the material aging processes a lot
Environmental immunity and compatibility with this relatively low effort implementation: Not so sure ...
Either way xdevs.com is planning something like this and I can't wait to hear about the results :) xdevs.com/doc/xDevs.com/cryotest/cryoltz.jpg
Ah of course xdevs would be already on it. interesting
Let me make a random guess, do you have a Poweredge R710 or similar running in the background? The fans in the background sounds is exactly as mines haha
Greight video, thanks Marco!
Please tell me, what software you use for making schematics, simulating, and pcb design?