Ultralow Noise Tester: 9V Battery vs. 7805 vs. LTZ1000

this video has been supported by jlc pcb hey guys last time we defied the shortage of black nitrile gloves and fought briefly against the omnipresent malicious influence of temperatures on all things precise in spite of that handicap well the shortage of gloves persists but today we are fighting a completely different kind of omnipresent malicious influence on all things precise noise we had a few brushes with noise before it's an important metric for judging power supplies and dacs after all but so far it used to be as easy as plugging in a scope and measuring rms or peak to peak values directly things are not so easy anymore nowadays there are switching power supplies out there whose noise you can't even distinguish anymore from that of linear ones and it gets worse there are use cases where esoteric effects like popcorn noise in semiconductors mata it just so happens that those are precisely the use cases that i'm interested in i want to know how noisy my voltage references are for example to find out i need a test setup whose own noise doesn't overshadow that of my device under test in particular it needs to be a lot less noisy than the ultra precision voltage reference ltz-1000 in this case good luck with that here's why that matters these voltage references are usually used in analog to digital converters if we were to equip a voltmeter with a 70 millivolt peak-to-peak noise v-ref which we've just measured then its least significant digits would become random and meaningless essentially turning the meter into an equivalent of the 5 hardware store special nothing wrong with a five dollar hardware store special but i like my digits intact all eight and a half of them so back to the task at hand low noise measurement naturally i'm not the first one to be interested in this there is for example linear's application note 124 by analog virtuoso jim williams he came up with a sophisticated design based on a cookie tin containing a discrete j fat amplifier and a hermetic low leakage wet tantalum slug capacitor which would cost about 700 bucks today then there are numerous designs by eevblog forum members where mostly modern integrated op-amps are doing all the hard work that jim williams had to do by hand everybody agrees though what's needed is an 80 decibel gain preamp with a 0.1 to 10 hertz band pass filter that will make even the tiniest of esoteric effects visible and we're going to get results that are directly comparable to the 0.1 to 10 hertz noise specs that are given in all respectable component data sheets this is what 0.1 to 10 hertz is going to look like when we are done just as a morsel of interestingness in between all of these boring screen captures and this is why the 0.1 to 10 hertz band is the most interesting and the most important to have under control in my design i want to try and use a brand new chopper stabilized operational amplifier the ada4523 has only just been released in may 2020 and it's an absolute low noise champion theoretically usable even below 0.1 hertz with about 4 nanovolt per square root hertz in my region of interest it would be good enough on its own but just because i can i'm going to use four of them in parallel averaging their outputs will only attenuate the unwanted uncorrelated op amp noise by almost one half while leaving my measurement signal untouched these four op-amps are configured for times 10000 gain their amplified and averaged output signal goes into a fourth order selen key low-pass filter which will strip away all the high frequency content above 10 hertz i'm turning all that into a pretty good-looking pcb layout if i may say so myself with multiple options for a few critical components like the input capacitors and jlcpcb is turning that into a picture-perfect reality their infamous 5 boards for 2 plus shipping offer is ideally suited for this kind of experimental project it allows you to make improvements or correct mistakes that are guaranteed to be on every first generation pcb ever at negligible cost for a simple two layer board like this their service is also amazingly fast which is in many cases even more important than a low price i had these boards in my hands only four business days after uploading the files there are four layer boards up to 100 by 100 millimeter have also gotten really cheap so much so that i ended up with an entire catalog of future circuit board projects even though the board is shaped to fit inside a metallic fissure enclosure and to make electrical contact with the exposed edges i still made an effort to fit all the sensitive high gain circuitry underneath an additional internal shield speaking of the devil oh whoops never mind those those are destined for another circuit my bad speaking of the devil here the ada-4523s are being deployed to not screw up the placement too much i'm using a fantastic keycat plugin called ibom it generates a single javascript soaked html file with an interactive map of your board as per manual pcb assembly protocol i'm going to solder everything haphazardly because it's just an unimportant first prototype right then i'm going to reconsider because maybe this is interesting enough for a youtube video after all no kidding i'm not sure how palatable this low tech stuff is going to be after we've been looking at exciting lasers and powerful cnc machines so if you have an opinion about this let me know either way this msop op-amp is not that critical it only has to do the low-pass filtering at unity gain far away from the things that usually give op-amps a hard time like nano vaults high frequencies or its power supply rails i think i chose an op2177 for this position or it volunteered who knows unfortunately resistors capacitors and all components in general are noise sources on their own the circuit might not work that well without components though so we've got to make the best out of it first and foremost at all temperatures other than 0 kelvin resistors emit thermal noise that depends only on resistance value and temperature so there's little we can do about it other than cryo cooling we can however reduce contact noise which depends on geometry and composition what that boils down to essentially is carbon resistors bad no surprise there and thin film and wire wounds good in this area i'm using thin film precision resistors exclusively and they are dirt cheap so no problem there's also shot noise which depends on current flowing through a resistor but there is hardly any current flowing here anywhere so it shouldn't matter too much so far miniware's new ts80p soldering iron has been doing pretty well for itself in spite of lead-free solder and an unsuitable cone-shaped tip but still to clean up my poor manual soldering without stressing the precision parts too much i'm putting the board on 150 degrees c hot plate and i'm pushing the joints over the edge of reflowing with a minimal amount of hot air the tsatp is called that because it works with usbc power delivery now making it twice as powerful as the original ts80 and about half as powerful as their ts100 that makes it warm up a lot faster and allows it to tackle more challenging tasks than i would have dared approach with its quick charge predecessor this emi shield is still quite challenging though both in terms of finger heat resistance and soldering iron power oh yeah getting hot my friend franz was kind enough to take some x-ray images of my favorite soldering iron tips revealing what we already suspected while looking at the original ts80 its copper tip is almost as large as those of the big boys it's a heat reservoir that is charged off-duty and can be dumped all at once into a joint that makes it perform better than i would have expected from purely looking at the 30 watt rating however large ground planes can still drain that reservoir quickly and then it's hard for the heater to catch up obviously the tsatp can't and it isn't meant to replace a full-size stationary station but i still like it as a clever little mobile tool for taking down to the cnc machine or elsewhere these chunky bnc connectors are taking even the higher power pace cartridges a solid minute or so to bring up the temperature so there's no shame in calling and a little help for a nicer looking fillet all right now our low noise amplifier is approaching completion all that's missing are one or two input caps the ones for which jim williams used one of these hermetic tantalum cans unfortunately i don't have any of those handy at the moment and i'm not too keen on buying either so let's see if we can get away with something cheaper the main purpose for this input cap is to provide ac coupling we don't want to amplify the entire 7vdc in or whatever reference we are measuring by times 10 000 that would be pretty catastrophic as a matter of fact we only want to see the nanovolt ac content on top the capacitance value dictates which frequencies are attenuated how much to let point one hertz through it takes a fairly large capacitance value of 1000 microfarad now all caps have leakage current even the multi thousand dollar hermetic tantalum ones we have to get rid of that by giving it a path to ground which is what that 2k resistor is for careful though even a super low capacitor leakage current like 10 nano ampere for example flowing through a 2k ohm resistor will cause a 20 micro volt dc offset behind our input cap there is no simple way around it that offset voltage will get amplified by our high gain op-amps this gives us the criteria based on which we can select the input cap candidates i need an extremely low leakage current i need approximately 1000 microfarad and i would like it to fit inside of this enclosure please [Music] dexter what do you think do we have any options left it looks like we do even though it has been a pretty dry year with too little electrolytes in the ground i suspect that with twice our usual detection proficiency we'll find some older plants though oh we are on the right track and the score all right now to desolder those without starting a wildfire ooh calling this relic an older plant is quite an understatement this is a special find deserving of a little reward the root of that cluster is actually a hermetic tantalum cap what an unbelievable coincidence judging by their surface patina i reckon these are at least 47 years old they have a lower than specified capacitance but otherwise they seem fine more than fine i only just learned about the supreme qualities of tantrum caps other than being planned obsolescence enforcers leakage current however the property that i'm most interested in for this project not that great originally i had 5 of these vintage german hermetic tantalum caps i let them all soak for 24 hours and then i selected the one with the highest leakage current of over 20 nano ampere at 10 volt 4. a happy little autopsy [Applause] as it turns out these guys are actually made from many smaller axial tantalum caps some parameters like esr benefit from this construction strongly overall leakage current however is the sum of all individual leakage currents and that is me these are wet tantalum slugs though just not as big as the few thousand dollar parts here's another special capacitor which isn't native to germany unfortunately it's a leaded 470 microfarad mlcc and i actually had enough hope in this candidate to import a set of two for this outrageous price because how on earth could a solid-state ceramic cap have any leakage current at all well after a similarly long soaking time overnight the pure sub nano ampere leakage results looked great unfortunately there seems to be a major temperature coefficient one or two centimeter proximity to my finger makes a major difference be it a thermocouple effect or growing and shrinking of the ceramic layers i wanted to give him a try anyway for a one-off piece of test equipment the price would have been tolerable and to the fit would have been beautiful but with a shorted input where otherwise i would expect to see my noise floor it behaves like a slow roller coaster enough of this foolishness i honestly expected microphonic pickup by those huge mlccs to be a bigger problem but i guess it's mostly just filtered out squeezing the enclosure deforming the pcb and putting a force on the mlcc leads is very noticeable of course but i still think the roller coaster issue is a thermal one those chopper amplifiers are actually drawing six milliampere each draining the round batteries i had planned before within minutes and are turning that energy into heat which goes along with air convection and all that kind of fun here's the solution to that input capacitor problem it's a lot of normal aluminium electrolytic caps as it turns out their leakage current varies a lot out of approximately 31 000 microfarad caps from different series by the most popular manufacturers i was able to select four specimen with less than 10 nano ampere after 24 hours of soaking the only pattern i was able to recognize is a dependence on voltage rating the more distance you can put between your expected dc voltage and the cap rating the better for my primary interest in 7 and 10 volt references 50 and 63 volt capacitors worked best without becoming too huge or too expensive when charging an electrolytic cap for the first time it forms a thin isolating oxide layer by means of anodization that process stops itself very elegantly just as the oxide layer's breakdown voltage reaches the capacitor voltage then over the course of hours or even a day that oxide layer gets polarized that effect is called dielectric absorption and only when that is done does the capacitor reach its lowest leakage current afterwards whenever the cap is confronted with a different voltage it's carefully grown oxide layer and its polarization have to be reformed to some degree that's an unavoidable inconvenience that will ultimately impact the way my lna is used by the way if you don't have a nano ampere meter yet you most likely do in many cases you can just connect a voltmeter in series and measure voltage over your input impedance to get a good idea ok so by selecting the best out of many i got a good cap now it's too big but the revision 2 was already necessary before because of two battery related mistakes i'll do that later though this video is already way too long and i want to finish it by showing you some results there is no power switch yet so i have to start the thing with two jumper links then it takes a minute or two to wake up catch its breath and reveal the most important parameter its own noise floor with a shorted input with a halfway open enclosure it looks like it's in the 100 nano volt peak-to-peak neighborhood my scope probe setup doesn't let me configure such a high gain directly so all readings have to be divided by 10. to test if the gain is correct and the filters are working i'm using an arbitrary waveform generator and a 1 mega ohm 100 ohm voltage divider well we are breaching vna territory now but as far as i can tell with my simple tools the gain indeed looks perfect between 0.1 and say 8 hertz then we slowly start to taper off that's about as good a filter response as i had expected from such a simple circuit first i'm going to try and measure 9 volt battery noise because i'd like to use those as a power source for all the other candidates there is a reason why these aren't our highest performance voltage references today they have a significant temperature coefficient i'm sitting perfectly still trying not to breathe but a gentle whiff of fresh air through an open window absolutely makes my readings plummet not sure if this is a meaningful measurement at all i vaguely recall seeing some better results on the internet this is a 7805 linear voltage regulator an absolute staple to electronics everywhere and for as long as i can remember in this test setup it's got plenty of high quality capacitance to support it and it's fed from a ripple-free 9-volt battery the caps i've used here exceed the datasheet recommendation by far but it's great to see what a fantastic 0.1 to 10 hertz noise performance one can get from such a cheap and common part even more so under load i'm genuinely impressed this thing the lt 3042 is regarded as one of if not the overall lowest noise linear voltage regulator on the market in the arcane arts of 0.1 to 10 hertz noise however it is not that great they aren't boasting about it in the data sheet so it's probably just not designed for that i put it into an absolute shielding fortress to get to 10 microvolt peak to peak we are homing into what started my interest in noise to begin with apart from this guy and his merspaw posts on discord this lt 1031 is a buffered buried zener diode based 10 volt reference and it's a straightforward sanity check for my preamp it's good three to four microvolt peak to peak whereas the data sheet states a maximum of six okay now it's getting interesting this is one of my two kx ltz boards the one that was powered on the longest the one that brings back fond memories of the good times where we had large public things like makerfaire hanover where i had that reference measured at the stand of the physical technician it's not doing well though unfortunately 20 to 30 microvolt peak to peak is at least an order of magnitude too much that's what i get for treating that one with maximum care i hope that's just a dead op amp or the heater transistor and not the carefully aged ltz itself maximum care maximum care maximum gear this is the other bare ltz board which i've treated disrespectfully played around with used as a video prop multiple times hell i even touched it without gloves but its noise looks almost good it's configured for 5 millimeter current so i'm expecting no more than one microvolt peak to peak noise with a minimum of thermal insulation with a minimum of thermal insulation over the metal can i'm getting that and more yep 800 nano volt peak to peak noise that's my high score for today my setup has no magnetic or electromagnetic shielding worth mentioning so that might be an opportunity for improvement honorable mentions i don't think i've introduced you to the newest addition to my bench yet for now i hope it's sufficient to say that advantaged rs6581t is also based on an ltz and its resistance measurement current is derived from it so of course it's a great low noise performer as well we could go all out and have the oscilloscope's fft function calculate a noise power density spectrum but i don't really see why right now other than that looking really cool last but not least the bench dominating the almost bench breaking fluke 5700a multifunction calibrator set to 10 volt dc of course it performs precisely as predicted in their extended product specifications document 2 micro volt peak to peak and some change holy moly originally this video was meant to be about the tsatp no idea how that could have escalated to 25 minutes of noise rambling well at least i got that out of my system now the preamp will come in handy soon i'll fix the battery thing before committing it all to github then it'll be buried underneath affiliate links in the video description alright thanks for enduring see you later you

Channel: Marco Reps

Views: 171,387

Rating: 4.9485531 out of 5

Keywords: marco, reps, low, noise, amplifier, preamp, lna, frequency, ltz1000, oscilloscope, rms, peak, band, metrology, adc, multimeter, arduino, raspberry, pi, fluke, keithley, solartron, analog, linear, regulator, jfet, hermetic, tantalum, capacitor, pcb, soldering, shielding, tektronix, chopper, kicad, jlcpcb, ts100, ts80p, ts80, microscope, reflow, smd, mlcc, source, measure, unit, smu, nanoamperemeter, impedance, selection, electrolytic, battery, lt3042, lt7805, ltz

Id: XpbDMo8an5w

Length: 24min 46sec (1486 seconds)

Published: Mon Aug 17 2020