The LM399 needs ~20mins minimum to get it 'ready for use', but saying that can take months to stabilize from new. If you power up continuously for a few days you should see some better stability. Take a note of room temp after 30mins, then after a few days get the room to same temp and check the drift. PS. Remove the flux from the board.
Many, many years ago, I used a "zero offset, zero drift" choppers stabilized opamp for a reference board, and guarded traces, and... Lots of fun to be had with this stuff! Fun to watch others do it, and lots of comments!
Resistors have thermal drift. It's best to use ratiometric networks where the they track and the parts are isothermal. Great idea for a board, though. Real confidence builder to have a precision source. ( That ratiometric trick is good for setting absolute gains with an op amp, too. Chopper stabilized op amps will get you damn near zero drift, as well. )
Reminds me of a +/- 10V ref I had to come up with for a test set back in the early 90's . I used an AD581 and an opamp to give me the inversion. It worked well but we'd leave it stabilize for a while too before doing any test. 👍
Nice little reference board. But after many decades of designing and fixing PCB’s learned to despise rectangular and green PCB’s. Any I design now are routed to the edge of traces, giving a complex shape. The current colour is yellow. No additional costs and the PCB’s look sharp.
if its a new part it should stay on for at least a couple of month to stabilize. also depends on TC of the resistors you used. Pots usually have high TC. It's not so much about the accuracy of your resistors as it is about their low TC.
I think since the pot is used as a voltage divider its TC shouldn't matter much, however the other resistors in the feedback network should have a closely matched TC for good stability vs temperature.
@@victorman2227 One could use Vishay Low TCR SMD Thin Film Resistors, ± 2 ppm/°C, 0.01%, skip the potentiometer, and document the full burn-in reading, such as 10.083412 instead of 10.000000. Nothing wrong with having an non-integer calibrator. Also, using the highest precision OP Amp and skip the socketing. In multimeters, I noticed the LM399 is mounted a distance above the board, leaving longer leads, instead of creating any interruptions in the PCB substrate.
@@pault6533 Yep, the only need for a round number is that it soothes the heart. For example in a piece of test equipment its possible to digitally store a constant of what the voltage is and not have to add any potentiometers. Another alternative way to adjust the output could be placing a number of resistors in parallel with increasing values, after measuring we can snip out the resistors that will change the voltage to the desired value. Higher value resistors will have a smaller impact on the resultant parallel resistance. I once made a 0.05% 5V supply to power an arduino doing some A-D in a power supply. It had to stay within 0.05% in 20-50C range (16 ppm/C max?). A discrete TC zener diode tested to 2ppm/C in 10C range (like the 1N821 but 9V) and OP07 opamp with wirewound resistors made a 10.xxxxV reference, which was then divided with a potentiometer to an accurate 5.000V which was the reference for a linear regulator. I didn't have a temperature chamber to check the tempco but after a day of the PSU being stressed at full load the voltage stayed within limits (it got rather hot inside). This is an order of magnitude lower than the precision that you are talking about though.
I like to learn something new everyday. I started early on this Saturday morning. You mentioned increasing the voltage with the op amp. Up until today, I didn't think this was possible! I didn't know that an op amp could be used as a DC to DC convertor, which I thought could only be done with a switch mode power supply. Great video series BTW!
OK, I just read up on the op amp subject. I forgot about that the voltage supply rail must be higher than the output voltage. So, not really a DC to DC convertor.
It's not a switching DC DC converter. It has a higher power source, 15V from outside. Its resistor feedback gives it a total gain of ~1.4 so it converts the ~7V of the zener to 10V at the output. Although you could say that most linear regulators are at their heart a little reference voltage (usually 1.25V), a simple opamp, and a pass transistor. Outside resistors set the gain of the opamp to drive the pass transistor so that you get your desired stablized voltage at the output of the regulator. Ofc extra circuitry gets put on the die for other features.
@@ivolol Thank you for the explanation! When I heard him say "boost the voltage," my whole understanding of power supplies became in question! I get it now, the high power source is the answer.
good video demonstration on how bad 10V reference can get with poor component design. multimeter's internal LM399 temperature drift is down to couple ppm there are cheap and simple Bandgap references around ~50ppm
My LM399 circuit is rock stable. But I used only low ppm resistors (no trimmer), an LTC2057 opamp, and the LM399 that I bought on ebay was supposedly used already.
I reckon the drift is caused by the trimmer, as it is a mechanical device and susceptible to mechanical shocks or temperature gradients, unlike fixed precision resistors. I think adding a bunch of parallel low tempco resistors during the calibration step could be a fail-proof approach.
You would want to do a "burn in" on it to get it to a stable point. I've seen some people saying that you need to burn it in for years at a time for it to be completely stable.
Yeah. Use it for fifty-some years at an ambient temp of 120C, using 120% of the maximum current. It will burn in and then burn out. After the burn out phase it will be the most stable device ever 😉
When i see a potentiometer that doesn't have 'buffer' resistors I'm usually unimpressed with the design. Is that a proper term? I named them end stops, but I come from machine background.
Lots of input about aging it in. Please give us updates. Missed seeing you melt it together. I'm still searching for the multistep controller to use for oven.
Hi. Great video! I've asked this bore and I promise this is the ladt time I ask. What kind of camera, lens, and software/computer do you use? I am wanting to build a video microscope for electronic board assembly. Thanks!
I dont know what op-amp youre using, but could it be the offset drift of that opamp. Probabily changing a few uV when small temperature changes. Id think its important to use a chopper stabilized op-amp :)
You can graph the stabilization of the voltage using a log graph for time. Much of the drift may be in the buffer circuit, especially SMD resistors. What is the lowest impedance voltmeter you can calibrate before the voltage is depressed by the current draw?
You made fun of the LM399 reference... You used an op amp and some resistors and you want it to be stable...... In the schematic from which you were inspired, AN-184, Figure 2, Texas Instruments wrote there that a stability of 20ppm is obtained at the output. You had to at least try the circuit in Figure 4, next page, if you wanted to be close to the stability of the LM399! There is also a note about the resistors involved... they must be 1ppm TC... Imagine what the potentiometer must be like! Don't expect too much from what you did... Like you said, it's just for fun!
Is that one of the multi-volt precision reference units just to the left of your Keithly? I have one of those, and I went and got it before watching part 1, just to see if that used the same LM ..... but I couldn't get a clean view through the plastic.
@pault6533 ...very nice. 10mA is better than 5mA. Sounds like a better unity gain follower could be installed. May nothing break into oscillation...and AC ripple be in micro volts regardless of the amount of current drawn. Maybe IMSAI will run a little oven or output voltage vs. temperature on his nifty PCB. While a military type temperature span ...that is a little extreme..it be nice to see Vout when say oven temp is varied from 32 deg. F to 140 deg. F ? Sort of like what might be in the typical interior of an automotive application. I can see this reference in like my AD converter...for fuel injection standalone controller or the for little Chinadoll 2 cycle pile FI- I need to get back too.. Thanks Kind regards Fred
The LM399 needs ~20mins minimum to get it 'ready for use', but saying that can take months to stabilize from new. If you power up continuously for a few days you should see some better stability. Take a note of room temp after 30mins, then after a few days get the room to same temp and check the drift.
PS. Remove the flux from the board.
Many, many years ago, I used a "zero offset, zero drift" choppers stabilized opamp for a reference board, and guarded traces, and... Lots of fun to be had with this stuff! Fun to watch others do it, and lots of comments!
Most likely, the LM399 is more stabile, than the opamp circuit.
If it‘s a chopper you might be fine. But I fear the resistors and pots usually have a much higher drift than the 399 (once it‘s burned in).
Resistors have thermal drift. It's best to use ratiometric networks where the they track and the parts are isothermal. Great idea for a board, though. Real confidence builder to have a precision source. ( That ratiometric trick is good for setting absolute gains with an op amp, too. Chopper stabilized op amps will get you damn near zero drift, as well. )
Reminds me of a +/- 10V ref I had to come up with for a test set back in the early 90's . I used an AD581 and an opamp to give me the inversion. It worked well but we'd leave it stabilize for a while too before doing any test. 👍
Nice little reference board. But after many decades of designing and fixing PCB’s learned to despise rectangular and green PCB’s. Any I design now are routed to the edge of traces, giving a complex shape. The current colour is yellow. No additional costs and the PCB’s look sharp.
I always always at least round the corners. Man, I hate handling PCBs with square corners.
The cutout is mainly to reduce the mechanical stress on the pins, that's why it is angled 90° usually
if its a new part it should stay on for at least a couple of month to stabilize. also depends on TC of the resistors you used. Pots usually have high TC. It's not so much about the accuracy of your resistors as it is about their low TC.
I think since the pot is used as a voltage divider its TC shouldn't matter much, however the other resistors in the feedback network should have a closely matched TC for good stability vs temperature.
@@victorman2227 One could use Vishay Low TCR SMD Thin Film Resistors, ± 2 ppm/°C, 0.01%, skip the potentiometer, and document the full burn-in reading, such as 10.083412 instead of 10.000000. Nothing wrong with having an non-integer calibrator. Also, using the highest precision OP Amp and skip the socketing. In multimeters, I noticed the LM399 is mounted a distance above the board, leaving longer leads, instead of creating any interruptions in the PCB substrate.
@@pault6533 Yep, the only need for a round number is that it soothes the heart. For example in a piece of test equipment its possible to digitally store a constant of what the voltage is and not have to add any potentiometers. Another alternative way to adjust the output could be placing a number of resistors in parallel with increasing values, after measuring we can snip out the resistors that will change the voltage to the desired value. Higher value resistors will have a smaller impact on the resultant parallel resistance.
I once made a 0.05% 5V supply to power an arduino doing some A-D in a power supply. It had to stay within 0.05% in 20-50C range (16 ppm/C max?). A discrete TC zener diode tested to 2ppm/C in 10C range (like the 1N821 but 9V) and OP07 opamp with wirewound resistors made a 10.xxxxV reference, which was then divided with a potentiometer to an accurate 5.000V which was the reference for a linear regulator. I didn't have a temperature chamber to check the tempco but after a day of the PSU being stressed at full load the voltage stayed within limits (it got rather hot inside). This is an order of magnitude lower than the precision that you are talking about though.
I like to learn something new everyday. I started early on this Saturday morning. You mentioned increasing the voltage with the op amp. Up until today, I didn't think this was possible! I didn't know that an op amp could be used as a DC to DC convertor, which I thought could only be done with a switch mode power supply. Great video series BTW!
OK, I just read up on the op amp subject. I forgot about that the voltage supply rail must be higher than the output voltage. So, not really a DC to DC convertor.
It's not a switching DC DC converter. It has a higher power source, 15V from outside. Its resistor feedback gives it a total gain of ~1.4 so it converts the ~7V of the zener to 10V at the output.
Although you could say that most linear regulators are at their heart a little reference voltage (usually 1.25V), a simple opamp, and a pass transistor. Outside resistors set the gain of the opamp to drive the pass transistor so that you get your desired stablized voltage at the output of the regulator. Ofc extra circuitry gets put on the die for other features.
@@ivolol Thank you for the explanation! When I heard him say "boost the voltage," my whole understanding of power supplies became in question! I get it now, the high power source is the answer.
good video demonstration on how bad 10V reference can get with poor component design. multimeter's internal LM399 temperature drift is down to couple ppm
there are cheap and simple Bandgap references around ~50ppm
My LM399 circuit is rock stable. But I used only low ppm resistors (no trimmer), an LTC2057 opamp, and the LM399 that I bought on ebay was supposedly used already.
I reckon the drift is caused by the trimmer, as it is a mechanical device and susceptible to mechanical shocks or temperature gradients, unlike fixed precision resistors. I think adding a bunch of parallel low tempco resistors during the calibration step could be a fail-proof approach.
You would want to do a "burn in" on it to get it to a stable point. I've seen some people saying that you need to burn it in for years at a time for it to be completely stable.
Yeah. Use it for fifty-some years at an ambient temp of 120C, using 120% of the maximum current. It will burn in and then burn out. After the burn out phase it will be the most stable device ever 😉
When i see a potentiometer that doesn't have 'buffer' resistors I'm usually unimpressed with the design. Is that a proper term? I named them end stops, but I come from machine background.
Lots of input about aging it in. Please give us updates.
Missed seeing you melt it together. I'm still searching for the multistep controller to use for oven.
Hi. Great video! I've asked this bore and I promise this is the ladt time I ask. What kind of camera, lens, and software/computer do you use? I am wanting to build a video microscope for electronic board assembly. Thanks!
I dont know what op-amp youre using, but could it be the offset drift of that opamp. Probabily changing a few uV when small temperature changes.
Id think its important to use a chopper stabilized op-amp :)
Ooooooooooh - the ppm goodness!
You can graph the stabilization of the voltage using a log graph for time. Much of the drift may be in the buffer circuit, especially SMD resistors. What is the lowest impedance voltmeter you can calibrate before the voltage is depressed by the current draw?
Thanks for sharing
I'm pretty sure the 2015 would be using an old LM399 to measure the new one, right? :D
You made fun of the LM399 reference... You used an op amp and some resistors and you want it to be stable......
In the schematic from which you were inspired, AN-184, Figure 2, Texas Instruments wrote there that a stability of 20ppm is obtained at the output. You had to at least try the circuit in Figure 4, next page, if you wanted to be close to the stability of the LM399! There is also a note about the resistors involved... they must be 1ppm TC... Imagine what the potentiometer must be like! Don't expect too much from what you did... Like you said, it's just for fun!
And, funny thing, the "TI app note" was done by National Semiconductor, in the days before the "monopolization" of a bunch of companies...
Is that one of the multi-volt precision reference units just to the left of your Keithly? I have one of those, and I went and got it before watching part 1, just to see if that used the same LM ..... but I couldn't get a clean view through the plastic.
ua-cam.com/video/vqCbuzUk_yk/v-deo.htmlsi=AsleHkm07KQQU2IX
@@IMSAIGuy yep, that's the one! Thanks for the link!
What do you use(wire?) for your power hook up terminals? Also, where can I source the banana plugs and mini clips you use? Thanks for the videos!
ua-cam.com/video/JUBgeCZ4uH8/v-deo.htmlsi=Vr28O6tfnffmLRxM
@@IMSAIGuy Thank you! What kind of metal/wire do you make your power loops / attachment points on the bread boards with?
@@945-b2m 22 gauge solid
latest watch tower have pcb color
Nice reference board. Lots of purists.Okay now load it with 20 mA. What does the voltage do?
I did some testing on mine using OP177. The voltage was stable until 10 mA
@pault6533 ...very nice. 10mA is better than 5mA. Sounds like a better unity gain follower could be installed. May nothing break into oscillation...and AC ripple be in micro volts regardless of the amount of current drawn.
Maybe IMSAI will run a little oven or output voltage vs. temperature on his nifty PCB. While a military type temperature span ...that is a little extreme..it be nice to see Vout when say oven temp is varied from 32 deg. F to 140 deg. F ? Sort of like what might be in the typical interior of an automotive application.
I can see this reference in like my AD converter...for fuel injection standalone controller or the for little Chinadoll 2 cycle pile FI- I need to get back too..
Thanks
Kind regards
Fred
Of course current drawn to be like no more than 20 mA...?
Enjoy
Is the PCB available on PCBWay ?
www.pcbway.com/project/shareproject/LM399_10_000V_Voltage_Standard_a2dd87a5.html
@@IMSAIGuy thanks !
Ok, where does one use a 10v reference?