I wish he went back and did the 400 Phillips with the Agilent so that we know if it was the meter. If you want I'll drive 45 min over there and do it because it's killing me.
Really enjoyed your approach to this Dave, the rigs to test electronics are a pure joy to watch develop. Please consider adding any more you do in the future. It's a shame the software let you down finishing the data production but this didn't take away from the wicked problem solving lessons.
Shame I did not see this a year or 2 ago man just discovering your clips and there great. I work for Microsemi , could have had the lads test those on a machine in around 2 mins for 1,000 resistors :P
Hey Dave, I know you sorted it with the import options but for reference, you could use find+replace on your spreadsheet software to get rid of the letters. E.g. search for ' m' (without quotes) and replace with nothing. Great project BTW, and thanks for putting the effort in!
What was the ppm temp coefficient on those things? You seemed to be probing at 28 degrees or so. A 100ppm coefficient would give that change in average resistance if it was that negative in that portion of the range, which it could very well be due to tweaking to get the curve as flat as possible over the entire range
32:00 There may also be multiple production lines giving different distributions and means, causing significant variation between different batches. Only thing you can trust is the datasheet (and not even that after seeing your dodgy digikey video!)
@sepertude Not really. Calibration is all about "confidence" in your readings, and it's possible to get good confidence in that without going to a cal lab. You can use precision resistors like that as a transfer standard.
I once had to test 1k of caps, using a LCR meter. Took 2 days to do, though only about 5 were out of tolerance, failing for high ESR. Not bad for a batch of 40 year old CTS13 tantalum capacitors. I made the 5 failures into little rockets ( reverse voltage of 28V current limited to 10A whilst case lead clamped in vice, will blow the pellet out when the solder melts. If you do not clamp the case then the case disappears leaving the slug behind) and fired them in the office.
Recently you where playing with precision decade boxes and now this there a reason other than just curiosity your doing all this work to establish the tolerance spread of resistors. Oh nice video too. :-)
The test jigs built using pogo pins to make connection to the pcb are also known as a "bed of nails" and can be quite complex. Worked on a system for testing electric car chargers which had high voltage AC/DC present. Must have been 200+ pins for a single board wired into a national instruments system.
@DanFrederiksen Sure, but how many hours would you spend building and tweaking a fully automated jig vs the time saved vs the current 1.5sec / resistor?
The ending was hilarious. If only companies like Agilent (now Keysight - bloody HP keeps rebranding its subsidiaries, hoping consumers will forget the sins of their past) embraced open source at least for software, this bug would have been fixed within hours of Dave uploading the video. Of course, then they wouldn't be able to charge astronomical fees for software "upgrades"...
In case anyone wants to do something like this, I have a bit of advice for an easy way to grab each resistor without any fancy equipment. Drill two holes in a small piece of wood, plastic, foam, anything or anything, then jam minigrabbers into the holes. Done. With just one hand, you depress both minigrabbers at the same time and grab each resistor immediately without any fuss.
+flyguille Perhaps. Dave used crocodile leads and not his test jig to test the calibration resistor at 21:30 even tho he used his test jig to null his meter for zero ohms at 22:39 Can someone tell me if this meter can be calibrated at say the 1kohm mark as well as the zero ohms mark ?
@Joru666 I'd figure the production process has given the nominal value as it turned out, of about .035-0.4% under nominal value. The gap from -0.1 and higher would be because they've been picked to fit 0.1% error range, that would actually make sence to me, correct me if I'm wrong.
@philbx1 You'd still be limited by the response time of the meter and the mechanical movements, so I doubt any such approach would improve the testing time much. Unless you had to measure 100K+ of these things, I think I currently have the optimum diminishing returns approach probably up to 10K-20K, as long as you don't mind doing the work that is!
@Wilfredkazoks The antistatic mat has some resistance on the bottom side, on the top side it's insulated so it doesn't mess with your electronics (imagine that you're servicing something plugged in mains).
26:58 ...or you can use the find and replace feature. Highlight the data set: Find and Replace: Search= "[space]m": Replace= [blank]. There are other ways as well such as using a function to remove a certain number of text positions from a column. Changing the data to a number format is easy as well. You can correct the format manually, or copy another cell that has the correct format, then use "Paste Special," and only paste the cell format. You can also drag the corrected cell and "Paste Special" the format down the whole column ;)
I wonder if some of the negative bias on the readings may come from the agilent meter. Would still be interested to see the carbon film distribution. I haven't any fancy logging meters so if I were to do the same, I'd use a microcontroller connected to serial line to output ADC readings.
@DanFrederiksen Yeah, countless ways to Macgyver a solution to fully automated it. Reliable probing is always a pain though. The devil is always in the detail.
Greg Feneis, Our firm, Quan Tech Laboratories, made test equipment for the industry that did that and MUCH, much more back in the early 60's. Now its over fifty years later and you can't buy that off the shelf? Waaaaah!
I was thinking of having an inverted V, so that however far bent the leads are, the V would capture them. You could have that serrated test pin in the middle. I would have routed or chiselled a piece of wood so the thick part of the resistor has some space, and you could test right close to it. Guess there are a lot of ways to test ...
I would guess the manufacturer has not calibrated his plant recently, or is not worried as all are within a tolerance band. As all the resistors are from the same batch, and were made at the same time, more or less, as you can see from the data, which shows no clear trend related to sample number. They probably were aiming to be no more than half way to the limit, and were not going to change anything whilst it was running well.
Instead of tension terminals, couldn't you use leaf sprung omron switches, which would also trigger the capture? I have a good idea in my head how to do it, but it would be tricky and time consuming to get it all aligned properly, but you could make it so the hump going over the resistor wire triggers the switches, which are hooked in series acting as an and gate, which will trigger the data logger.
@mortaldrumming Countless options. Ebay, Alibaba, or your local electronics store (like Jaycar or Altronics in Oz) who source from the cheapest place they can get it from in China.
Looks like i wont get to sleep tonight (it's 3:30 here in Europe now). First the guys from the geekgroup make a one hour video, now you over half an hour. But HAVE TO WATCH! ;)
Aurelius R said it already: You need to delete the OL as the first data entry, clearly. It's a bug in the Agilent software. I don't know how you didn't correlate that
28:40 can the wires u use cause that offset? it's a pitty that we couldn't see the result of the 5%, i'm interested if it's a gap in the middle of the distribution, that would mean that they use those resistors for a 1%? (sorry for my english)
I really like that multimeter. I think that I have found the next piece of gear I need to purchase. Unless, Dave, you would like to get me an early birthday gift. ;-)
couldnt you have just drug a rig with a connector on each side down the line of resistors and recorded the entire output and then looked at the peaks as the resistance and the valleys as the non- contact ? and then measured the peaks?
From what I've been told by my teacher, If you have, say 5k resistor with 5% tolerance you may be pretty sure it is going to be quite off the nominal value. As he explained it is due to the fact that if the resistor is within lower error value, say 1% it is going to be sold in that range for higher price. That may also be true according to those 1% resistors you measured in the video.
To quote the comedy 'I.T Crowd'. "Have you tried turning it off and on again? " BTW, you know how some smartasses think they have better ways of doing things. Here's mine... Arduino (or whatever) connected to an RC servo with the servo arm shaped like a cam mounted above the pushbutton on the meter. An input wire is connected to a switch on the probe board and this is pressed as the pogo is pushed onto the resistor. Anyway, enough of the crap idea stuff. great patience and resourcefulness Dave!
Dave might have to redo the 1000 measurements because I did not see him using the right tongue angle, or using his tongue at all! Or does tongue angle not matter when measuring, only when trimming?
Hi Dave, Just one thing, the variation might be due to the temperature (and humidity!!) of your workshop, 30ºC is pretty darn hot, what's the humidity like?
Maybe you need to upgrade the meter firmware. A reviewer on Amazon said "The meters were released with substantial bugs. V2.0 firmware was released this month and corrected them. Until 11/11, Agilent will provide a free serial cable to those who ask so new firmware can be uploaded. The process of updating the firmware is straightforward and reliable. "
If you want to interface an old GPIB-able instrument, take a look at the Prologix GPIB to Ethernet interface, it has a telnet like interface etc. that makes interfacing easy.
Suppose you fold the band back and forth to stack them up in the space, say, of 6 cm and then squash each end in a vice to which leads are attached. Then you measure all the resistors at one go to get the average. Wouldn't that tell you something without all the work?
Six Sigma?? Isn't that where you make 6 samples of a product, put them into 6 Mitsubishi Sigmas, attempt to deliver them to a customer, and if the customer receives a product then you know the product is lucky? Especially so if the product is not ruined by oil smoke.
A very useful trick is to plot the data on “normal” paper. You can get log-normal or linear-normal. I don’t think Excel can do that. Don’t know about Open office. This shows true Gaussian samples as a straight line. If the line shows two or more slopes then the sample population is probably made up of more than one group mixed together (made on different days maybe). If the curve as “Z” shape (three slopes) then the population probably started out Gaussian but the manufacturer could not make all parts within tolerance and had to cull the outliers.
It wouldn't matter if the meter had a 50% offset as long as he tests all the resistors with the same meter. He's showing all the resistors compared to each other with a 0.3% resistor as the baseline (which read as 0.9998).
Four rubberized wheels at each side, driven by stepper motor, these v-shaped springs (you know, the ones with a small "coil" in the middle) to press against the resistor as probes... µC with IR diode to control stepping... !? :)
You resetted the propes resistance using some copper on a pcb right? That resistance in thst copper has been subtracted from all the readings. Maybe thats why it undershoots?
Are there any other multimeters that have the same or similar function? That Agilent is looking pretty good for a first time buy of a (better than $50) good multimeter. Any other reccomendations?
can anybody give a clue how to order that box of resistors (and other components) from china? thanks in advance. dave, if you can reply this message, i'd really appreciate it.
Brain fart! I love digital cameras. If you pull a stupid and forget to mark something and you can go back and see what you need to know. What is the MSRP on this meter. I like it a lot. Agilent is like Fluke for being expensive for anything that has lots of bells and whistles.
At 6:33 I couldn't help but notice two little blisters on your middle finger. (Damn hi-res camera...) Out of curiosity, do you know what they are? Warts?
I wish he went back and did the 400 Phillips with the Agilent so that we know if it was the meter. If you want I'll drive 45 min over there and do it because it's killing me.
SidneyCritic ComedyHound do it! I have to know
I went to a few shows and didn't see him, and he doesn't read these, so looks like we will never know.
Yep. They are 50ppm resistors and if they are calibrated at 25°C the offset is explained!
Really enjoyed your approach to this Dave, the rigs to test electronics are a pure joy to watch develop. Please consider adding any more you do in the future. It's a shame the software let you down finishing the data production but this didn't take away from the wicked problem solving lessons.
Im the 29,065 viewer. Learning so much from the videos.
@SajjadBro Yes, it will log any parameter.
Shame I did not see this a year or 2 ago man just discovering your clips and there great. I work for Microsemi , could have had the lads test those on a machine in around 2 mins for 1,000 resistors :P
Hey Dave, I know you sorted it with the import options but for reference, you could use find+replace on your spreadsheet software to get rid of the letters. E.g. search for ' m' (without quotes) and replace with nothing. Great project BTW, and thanks for putting the effort in!
another first prize video. i will never get sick of this show.
@Wilfredkazoks Not possible. The mat is not conductive, only dissipative.
23:22 Ops! Dave wrote the numbers down incorrectly.
Love the FSM shirt and yet another great video. I'm still working through all your videos for my EE education
What was the ppm temp coefficient on those things? You seemed to be probing at 28 degrees or so. A 100ppm coefficient would give that change in average resistance if it was that negative in that portion of the range, which it could very well be due to tweaking to get the curve as flat as possible over the entire range
32:00 There may also be multiple production lines giving different distributions and means, causing significant variation between different batches. Only thing you can trust is the datasheet (and not even that after seeing your dodgy digikey video!)
@sepertude Not really. Calibration is all about "confidence" in your readings, and it's possible to get good confidence in that without going to a cal lab. You can use precision resistors like that as a transfer standard.
EEVblog You need to delete the OL as the first data entry, clearly. It's a bug in the Agilent software. I don't know how you didn't correlate that
@Superfungus0 No, its doesn't. The mat is dissipitive, not conductive. >>Gohms.
Wow. A test jig which doesn't involve hot melt glue? Well done!
I know this is an older video but I still love this kinda stuff :3
300° C ouch... Might have to take your town off my to visit list!
I once had to test 1k of caps, using a LCR meter. Took 2 days to do, though only about 5 were out of tolerance, failing for high ESR. Not bad for a batch of 40 year old CTS13 tantalum capacitors. I made the 5 failures into little rockets ( reverse voltage of 28V current limited to 10A whilst case lead clamped in vice, will blow the pellet out when the solder melts. If you do not clamp the case then the case disappears leaving the slug behind) and fired them in the office.
@Superfungus0 It's not the bandoleer either.
You made that "OneHungLow" sticker, didn't you...
@funkmastertraf No, thanks for asking. But so what if I was?
Recently you where playing with precision decade boxes and now this there a reason other than just curiosity your doing all this work to establish the tolerance spread of resistors. Oh nice video too. :-)
The test jigs built using pogo pins to make connection to the pcb are also known as a "bed of nails" and can be quite complex. Worked on a system for testing electric car chargers which had high voltage AC/DC present. Must have been 200+ pins for a single board wired into a national instruments system.
@DanFrederiksen Sure, but how many hours would you spend building and tweaking a fully automated jig vs the time saved vs the current 1.5sec / resistor?
The ending was hilarious. If only companies like Agilent (now Keysight - bloody HP keeps rebranding its subsidiaries, hoping consumers will forget the sins of their past) embraced open source at least for software, this bug would have been fixed within hours of Dave uploading the video.
Of course, then they wouldn't be able to charge astronomical fees for software "upgrades"...
In case anyone wants to do something like this, I have a bit of advice for an easy way to grab each resistor without any fancy equipment.
Drill two holes in a small piece of wood, plastic, foam, anything or anything, then jam minigrabbers into the holes. Done.
With just one hand, you depress both minigrabbers at the same time and grab each resistor immediately without any fuss.
so, Dave, you got a not calibrated equipment, the average offset means that it is an equipment fault.
flyguille 21:30
+flyguille Perhaps. Dave used crocodile leads and not his test jig to test the calibration resistor at 21:30 even tho he used his test jig to null his meter for zero ohms at 22:39 Can someone tell me if this meter can be calibrated at say the 1kohm mark as well as the zero ohms mark ?
Does it? It wouldn't be the first resistor batch to be biased below the nominal value.
@Joru666 I'd figure the production process has given the nominal value as it turned out, of about .035-0.4% under nominal value. The gap from -0.1 and higher would be because they've been picked to fit 0.1% error range, that would actually make sence to me, correct me if I'm wrong.
@philbx1 You'd still be limited by the response time of the meter and the mechanical movements, so I doubt any such approach would improve the testing time much. Unless you had to measure 100K+ of these things, I think I currently have the optimum diminishing returns approach probably up to 10K-20K, as long as you don't mind doing the work that is!
That was very impressive. I enjoy watching the whole video. You definitely get a thumbs up from me.
am i wrong if i spot an "oscillation" in the resistor value graph?
What's with people complaining about your vocal inflections? I've met plenty of Aussies with that cheerful tone. I like it! :)
Never mind, you've figured it out. I can't believe you don't know how every one of your 136 different multimeters work!
Last two minutes ruined all my impression on that bloody multimeter
^2nd^
Hate mail at Agilent
@GGTrashes Just some skin taken off. Likely from weights at the gym or some such. Haven't had a wart since I was a kid.
@superdau It'll still be there tomorrow!
Looks like you couldn't read your handwriting on that .9966. It was right the first time XD
@EEVblog I am in love with the meter, excellent choice!
(BTW: you do have a good taste for t-shirts ;) )
@Wilfredkazoks The antistatic mat has some resistance on the bottom side, on the top side it's insulated so it doesn't mess with your electronics (imagine that you're servicing something plugged in mains).
why not just use 2 plastic gears on the bottom, 2 metal gears on the top attached to your probes to feed the resisters while testing?
26:58 ...or you can use the find and replace feature. Highlight the data set: Find and Replace: Search= "[space]m": Replace= [blank].
There are other ways as well such as using a function to remove a certain number of text positions from a column. Changing the data to a number format is easy as well. You can correct the format manually, or copy another cell that has the correct format, then use "Paste Special," and only paste the cell format. You can also drag the corrected cell and "Paste Special" the format down the whole column ;)
Upcycle Electronics, He probably uses Windows... can't get common things like that on Windows! (just kidding!)
I wonder if some of the negative bias on the readings may come from the agilent meter.
Would still be interested to see the carbon film distribution.
I haven't any fancy logging meters so if I were to do the same, I'd use a microcontroller connected to serial line to output ADC readings.
Well, he did zero the meter.
I think there is a GPIB to usb adapter.
I find it somewhat comforting that OneHungLow is a brand. At least, I hope it is.
@DanFrederiksen Yeah, countless ways to Macgyver a solution to fully automated it. Reliable probing is always a pain though. The devil is always in the detail.
At 4:10, You need a axial component verifier. Universal auto insertion machines used to have them back in the day.
Greg Feneis, Our firm, Quan Tech Laboratories, made test equipment for the industry that did that and MUCH, much more back in the early 60's. Now its over fifty years later and you can't buy that off the shelf? Waaaaah!
I was thinking of having an inverted V, so that however far bent the leads are, the V would capture them. You could have that serrated test pin in the middle. I would have routed or chiselled a piece of wood so the thick part of the resistor has some space, and you could test right close to it. Guess there are a lot of ways to test ...
I would guess the manufacturer has not calibrated his plant recently, or is not worried as all are within a tolerance band. As all the resistors are from the same batch, and were made at the same time, more or less, as you can see from the data, which shows no clear trend related to sample number. They probably were aiming to be no more than half way to the limit, and were not going to change anything whilst it was running well.
love the FSM t shirt, same views here. and once again a most interesting video.
Instead of tension terminals, couldn't you use leaf sprung omron switches, which would also trigger the capture? I have a good idea in my head how to do it, but it would be tricky and time consuming to get it all aligned properly, but you could make it so the hump going over the resistor wire triggers the switches, which are hooked in series acting as an and gate, which will trigger the data logger.
@mortaldrumming Countless options. Ebay, Alibaba, or your local electronics store (like Jaycar or Altronics in Oz) who source from the cheapest place they can get it from in China.
It should trigger and store every time the reading changes, surely?
@Systemrat2008 No reason, just curiosity!
Looks like i wont get to sleep tonight (it's 3:30 here in Europe now). First the guys from the geekgroup make a one hour video, now you over half an hour. But HAVE TO WATCH! ;)
Aurelius R said it already:
You need to delete the OL as the first data entry, clearly. It's a bug in the Agilent software. I don't know how you didn't correlate that
I am late to this party, but I would tend to agree: it was most probably a garbage character right at the start of the data file.
28:40 can the wires u use cause that offset?
it's a pitty that we couldn't see the result of the 5%, i'm interested if it's a gap in the middle of the distribution, that would mean that they use those resistors for a 1%? (sorry for my english)
I really like that multimeter. I think that I have found the next piece of gear I need to purchase. Unless, Dave, you would like to get me an early birthday gift. ;-)
@Films4You The last one had OL first, and it loaded fine.
couldnt you have just drug a rig with a connector on each side down the line of resistors and recorded the entire output and then looked at the peaks as the resistance and the valleys as the non- contact ? and then measured the peaks?
From what I've been told by my teacher, If you have, say 5k resistor with 5% tolerance you may be pretty sure it is going to be quite off the nominal value. As he explained it is due to the fact that if the resistor is within lower error value, say 1% it is going to be sold in that range for higher price. That may also be true according to those 1% resistors you measured in the video.
This could be a fact for older times. But today it is much cheaper to make different lines for different tolerances.
The sharp serrated ones are sharp and serrated.
Does anybody know why all resistors in carbon are under 1k? It seems to be something the manufacturer wants but I don't know why. Thanks
Good job!
@pikuorguk lol me too. Can't find them anywhere though...
To quote the comedy 'I.T Crowd'. "Have you tried turning it off and on again? "
BTW, you know how some smartasses think they have better ways of doing things. Here's mine...
Arduino (or whatever) connected to an RC servo with the servo arm shaped like a cam mounted above the pushbutton on the meter.
An input wire is connected to a switch on the probe board and this is pressed as the pogo is pushed onto the resistor.
Anyway, enough of the crap idea stuff. great patience and resourcefulness Dave!
@TheETGShow I agree, have you tried that?
There are also usb -> hpib/gpib converters. I have worked with one made by Agilent (Propably not cheap though)
Dave might have to redo the 1000 measurements because I did not see him using the right tongue angle, or using his tongue at all! Or does tongue angle not matter when measuring, only when trimming?
Hi Dave,
Just one thing, the variation might be due to the temperature (and humidity!!) of your workshop, 30ºC is pretty darn hot, what's the humidity like?
Maybe you need to upgrade the meter firmware. A reviewer on Amazon said "The meters were released with substantial bugs. V2.0 firmware was released this month and corrected them. Until 11/11, Agilent will provide a free serial cable to those who ask so new firmware can be uploaded. The process of updating the firmware is straightforward and reliable.
"
If you want to interface an old GPIB-able instrument, take a look at the Prologix GPIB to Ethernet interface, it has a telnet like interface etc. that makes interfacing easy.
Suppose you fold the band back and forth to stack them up in the space, say, of 6 cm and then squash each end in a vice to which leads are attached. Then you measure all the resistors at one go to get the average. Wouldn't that tell you something without all the work?
Six Sigma?? Isn't that where you make 6 samples of a product, put them into 6 Mitsubishi Sigmas, attempt to deliver them to a customer, and if the customer receives a product then you know the product is lucky?
Especially so if the product is not ruined by oil smoke.
OMG... so bad that the joke is good.
Dave,
What about rebooting your PC to reset and initialize the COM port ?
@eevblog, I second mortaldrummings comment, where can i get cheapo onehunglow resistors by the thousand?
A very useful trick is to plot the data on “normal” paper. You can get log-normal or linear-normal. I don’t think Excel can do that. Don’t know about Open office. This shows true Gaussian samples as a straight line. If the line shows two or more slopes then the sample population is probably made up of more than one group mixed together (made on different days maybe). If the curve as “Z” shape (three slopes) then the population probably started out Gaussian but the manufacturer could not make all parts within tolerance and had to cull the outliers.
It wouldn't matter if the meter had a 50% offset as long as he tests all the resistors with the same meter. He's showing all the resistors compared to each other with a 0.3% resistor as the baseline (which read as 0.9998).
23:20 Dave didn't record the original reading correctly -- there's the human error he is trying to eliminate
He notised it - "Bloody humans" :)
Four rubberized wheels at each side, driven by stepper motor, these v-shaped springs (you know, the ones with a small "coil" in the middle) to press against the resistor as probes... µC with IR diode to control stepping... !? :)
You resetted the propes resistance using some copper on a pcb right? That resistance in thst copper has been subtracted from all the readings. Maybe thats why it undershoots?
Resistance of that PCB trace would be in miliOhms, so not affecting the experiment.
bar10005 the undershoot is not big either..
The manufacturers likely didn't have a good 1K reference.
Nice one again dave ;)
Are there any other multimeters that have the same or similar function? That Agilent is looking pretty good for a first time buy of a (better than $50) good multimeter. Any other reccomendations?
OMG OneHungLow really exist! I was thinking its Dave's myth!
So "OneHungLow" is a real company?
25:03
So you're the one who delayed the Platinum Chip's delivery!
can anybody give a clue how to order that box of resistors (and other components) from china? thanks in advance. dave, if you can reply this message, i'd really appreciate it.
The 0.9966 was 0.9966 before, he must have written the 6 as a 0.
Does anyone know where can I order OneHungLow resistors? Google didn't provide anything usefull
It means cheapest of the cheap Chinese electronics
Does One Hung Low actually exist as a brand, or did you ask them to print that on the label?
One hung low printed on that label - I don't believe it !!
man, I was hoping you would throw the meter at something.
Can someone pleas post a link to the shop in which I can buy these resistors. What's the price of this resistors?
Dave dont you have to use calibration certificate to see how accurate meter is on that range and than correct results in Excel?
.9966 was correct the both times at ~23:12
Brain fart! I love digital cameras. If you pull a stupid and forget to mark something and you can go back and see what you need to know. What is the MSRP on this meter. I like it a lot. Agilent is like Fluke for being expensive for anything that has lots of bells and whistles.
At 6:33 I couldn't help but notice two little blisters on your middle finger. (Damn hi-res camera...) Out of curiosity, do you know what they are? Warts?
hmmm how much did that box of 1000 resistors cost from china?
"Even Stevie Wonder can probe these things" epic dude!!!