DE/EU Simple LM399 Reference Board, improved second Version

[Mr. Scram]

Finally got around to crunching the numbers & plotting the results! One vertical division on the volts axis is one LSD of the 34401A meter display. The meter had been on since about 11:00. At 12:15 or so I went into the lab and turned up the thermostat. I started logging both voltage and temperature at that time, but didn't turn the reference on until about 12:35. The thermostat timer turns the heating off at 18:00. Although the board was just sat on the bench, without a case or cover, it doesn't seem much affected by the room temperature.

There are a few spikes (e.g. around 17:50 & at 20:29) where I was at the bench. Some of them directly relate to turning other equipment and lights on & off. I made an initial zero test, with power off, and during this I observed that the board - specifically the input protection MOSFET - seems to be light sensitive. A bright desk light just above it gave a few microvolts output from the reference. Presumably the DS diode in the MOSFET was acting as a photodiode, and although the epoxy package is black, it's very thin. It is possible the offset voltage of the output amplifier may also be slightly photosenstive. Voltnuts may want to remove the LED before putting it in a screened, light-tight box!

I love that even a fairly "mundane" multimeter like the 34401A can show the presence of people in a room by measuring the small deviations they bring about in other electronics. The diode finding is also very interesting!

[RoadRunner]

Mine arrived safely this morning and is currently being tested. I'm using an HP 34401A which was last calibrated by Keysight on 8th January, and has a history of very good accuracy on the 10V range. At last calibration, it read 10.00002 V from a 10.0 V source (with an uncertainty of 0.000076 V). Allowing a couple of hours for the meter to warm up, then turning on the reference, gives me readings of around 10.00003V after 2 hours. It's varying a bit, but that is probably with room temperature. I'm logging that, and I will analyse the results later. The specification accuracy for the meter (10V range, 90 day) is 0.00025 V, so it is doing much better than it needs to!

Finally got around to crunching the numbers & plotting the results! One vertical division on the volts axis is one LSD of the 34401A meter display. The meter had been on since about 11:00. At 12:15 or so I went into the lab and turned up the thermostat. I started logging both voltage and temperature at that time, but didn't turn the reference on until about 12:35. The thermostat timer turns the heating off at 18:00. Although the board was just sat on the bench, without a case or cover, it doesn't seem much affected by the room temperature.

There are a few spikes (e.g. around 17:50 & at 20:29) where I was at the bench. Some of them directly relate to turning other equipment and lights on & off. I made an initial zero test, with power off, and during this I observed that the board - specifically the input protection MOSFET - seems to be light sensitive. A bright desk light just above it gave a few microvolts output from the reference. Presumably the DS diode in the MOSFET was acting as a photodiode, and although the epoxy package is black, it's very thin. It is possible the offset voltage of the output amplifier may also be slightly photosenstive. Voltnuts may want to remove the LED before putting it in a screened, light-tight box!

The diode finding is interesting to me as well, i will myself look into this. will share if i can come up with some results.

[DC1MC]

Mine arrived safely this morning and is currently being tested. I'm using an HP 34401A which was last calibrated by Keysight on 8th January, and has a history of very good accuracy on the 10V range. At last calibration, it read 10.00002 V from a 10.0 V source (with an uncertainty of 0.000076 V). Allowing a couple of hours for the meter to warm up, then turning on the reference, gives me readings of around 10.00003V after 2 hours. It's varying a bit, but that is probably with room temperature. I'm logging that, and I will analyse the results later. The specification accuracy for the meter (10V range, 90 day) is 0.00025 V, so it is doing much better than it needs to!

Finally got around to crunching the numbers & plotting the results! One vertical division on the volts axis is one LSD of the 34401A meter display. The meter had been on since about 11:00. At 12:15 or so I went into the lab and turned up the thermostat. I started logging both voltage and temperature at that time, but didn't turn the reference on until about 12:35. The thermostat timer turns the heating off at 18:00. Although the board was just sat on the bench, without a case or cover, it doesn't seem much affected by the room temperature.

There are a few spikes (e.g. around 17:50 & at 20:29) where I was at the bench. Some of them directly relate to turning other equipment and lights on & off. I made an initial zero test, with power off, and during this I observed that the board - specifically the input protection MOSFET - seems to be light sensitive. A bright desk light just above it gave a few microvolts output from the reference. Presumably the DS diode in the MOSFET was acting as a photodiode, and although the epoxy package is black, it's very thin. It is possible the offset voltage of the output amplifier may also be slightly photosenstive. Voltnuts may want to remove the LED before putting it in a screened, light-tight box!

The diode finding is interesting to me as well, i will myself look into this. will share if i can come up with some results.

No so long ago there was RPi2 that was actually fully rebooting because of a "transparent" package of the power regulator:
https://www.alphr.com/raspberry-pi-2/1000375/why-a-camera-flash-will-reboot-your-raspberry-pi-2

It seem that the quality of the IC packaging went down together with the price

 Cheers,
 DC1MC

[cowasaki]

I would like one of these too. What is the size of the board. I’ll read the thread fully later, I take it the power requirements are in the thread.

[RoadRunner]

I would like one of these too. What is the size of the board. I’ll read the thread fully later, I take it the power requirements are in the thread.

You got one, size is 48x100mm.

[not1xor1]

I would like one of these too. What is the size of the board. I’ll read the thread fully later, I take it the power requirements are in the thread.

You got one, size is 48x100mm.

So you selected the wrong country in your account profile... obviously you are from Flatland 

Apart of jokes do you plan to go on selling these boards for some more time?
I would like to buy one in 3-4 months (I have to move stuff around because of home renovation, HEPA filtered ventilation system).

[toastedcrumpets]

Power requirements are low, it takes 0.15A at 15V initially while the heater is getting everything up to temperature, then it settles below 0.025A  for me. You may get different results depending on air temperature and the level of insulation you have.

[RoadRunner]

I would like one of these too. What is the size of the board. I’ll read the thread fully later, I take it the power requirements are in the thread.

You got one, size is 48x100mm.

So you selected the wrong country in your account profile... obviously you are from Flatland 

Apart of jokes do you plan to go on selling these boards for some more time?
I would like to buy one in 3-4 months (I have to move stuff around because of home renovation, HEPA filtered ventilation system).

I do have only limited time in hand to do extra stuff , so as long as something else more important does not come up it shoud be available,  3-4 Months is long time to say something.

[jpb]

It arrived today - thank you RoadRunner.

I've not insulated it but I've fed it through an LT3042 board I got from ebay to give it a low noise 15V supply.
My 2015 is dead on and my 2000 is reading between 2 and 3 in the last digit.

I'm very pleased though I'm not sure I like the power sockets which seem to require a fat probe to be stuck in rather than the thin wires I was trying to use - perhaps I'm using them wrong.

Sorry for the out-of-focus shot of the board.

Edit : the 2000 now seems steady at 2 and occasionally down to 1 in the last digit.

[cowasaki]

Hoping mine arrives tomorrow.  I have spare 20v AC transformers and a batch of triple DC output PCBs I designed for another project.  Basically the transformer output is rectified then an electrolytic cap for positive and another for negative rails, space for a small cap per output, switching and output.  The two main rails are to 7812/7912 regs then a 7805 giving the three outputs.  If I build one with a 7815 and forget the neg rail it should give a perfectly stable 15v DC.  How good does the supply voltage have to be?  Is this up to it?  If so I have some nice cases so will build it up tomorrow.

[jpb]

I don't know how good the supply should be.

I just used a (modified) LT3042 board because they are cheap on ebay (I went for a UK supplier which was £12 because I was in a hurry but the ones direct from China are less than £6). I desoldered the 33K surface mount resistor and used a Vishay 150k also off ebay. Probably way over the top but as the thing is a reference I thought it was worth over engineering.

I intend to use a supply like you suggest but make it 18V instead of 15V and use it to feed the LT3042 board.

[jpb]

As a postscript to my earlier post, I now find my meter probe is stuck in the connector! I managed to pull out three of them but not the fourth.

So be careful not to use probes as connectors - just straight thick wire I guess.

I must say I do not like these push connectors, I will probably have to desolder at least one to retrieve my probe. Going on the Wago web site I see that there is a special tool needed to remove wires.

RoadRunner - was there a particular reason for selecting these connectors?

EDIT : after pushing down hard on the white button and pulling hard on the probe it has now come out (though I did bend it in the process) So problem solved.
I intend putting the board in a box with terminals so might connect these with thick copper wire (I hope this doesn't lead to emf issues arising from different metals).

[cowasaki]

I don't know how good the supply should be.

I just used a (modified) LT3042 board because they are cheap on ebay (I went for a UK supplier which was £12 because I was in a hurry but the ones direct from China are less than £6). I desoldered the 33K surface mount resistor and used a Vishay 150k also off ebay. Probably way over the top but as the thing is a reference I thought it was worth over engineering.

I intend to use a supply like you suggest but make it 18V instead of 15V and use it to feed the LT3042 board.

I designed the board and only actually wanted 1 with a spare but when I went to put the order in it was 80p more for 10 so I went with that and they sent 11 anyway.  Somewhere I have about 7 or 8 taped together with a piece of masking tape.  If I find them soon enough you can have one for the postage.  The BOM is 4 x diodes, 1 x regulator, 1 x electrolytic cap (maybe 1000-2200) and I have a 100nF plus an LED and to 4 way headers.  If I find the boards soon enough I'll chuck all the bits in other than the electrolytic and regulator for you but obviously you will need a transformer giving maybe 3v more than your reg.


EDIT:  Just added the schematic below and image of the board.  With the USB isolation section empty this is the PSU I'll be using with the negative rail gone, a 7815 instead of the 7812 and no 7805

[jpb]

I don't know how good the supply should be.

I just used a (modified) LT3042 board because they are cheap on ebay (I went for a UK supplier which was £12 because I was in a hurry but the ones direct from China are less than £6). I desoldered the 33K surface mount resistor and used a Vishay 150k also off ebay. Probably way over the top but as the thing is a reference I thought it was worth over engineering.

I intend to use a supply like you suggest but make it 18V instead of 15V and use it to feed the LT3042 board.

I designed the board and only actually wanted 1 with a spare but when I went to put the order in it was 80p more for 10 so I went with that and they sent 11 anyway.  Somewhere I have about 7 or 8 taped together with a piece of masking tape.  If I find them soon enough you can have one for the postage.  The BOM is 4 x diodes, 1 x regulator, 1 x electrolytic cap (maybe 1000-2200) and I have a 100nF plus an LED and to 4 way headers.  If I find the boards soon enough I'll chuck all the bits in other than the electrolytic and regulator for you but obviously you will need a transformer giving maybe 3v more than your reg.

Thank you for the very kind offer, but I'm probably OK as I have an external (unregulated linear) supply  which is basically a transformer plus diodes and electrolytic capacitor I was planning to use and then I'll just mount my 18V regulator in the box (that is my current plan - use the box as a heat sink).

[cowasaki]

My board arrived this morning.

I'm about 1/3 of a mV out at 10v on the Keithley 2100 about the same the other way on the PM2525. (I had to swap the power leads round).

[jpb]

My board arrived this morning.

I'm about 1/3 of a mV out at 10v on the Keithley 2100 about the same the other way on the PM2525. (I had to swap the power leads round).

I haven't checked the specs on a Keithley 2100, but that difference seems quite large. Was everything warmed up for a while?

The closeness of my results I think was slightly fortuitous as my study was quite warm (from equipment running in various experiments I'm doing) and I think RoadRunner calibrates at quite a warm 24C (that is from memory - perhaps I have that wrong) but I would still expect the variation to be a few 10s of microVolts not 300 or so.

[cowasaki]

My board arrived this morning.

I'm about 1/3 of a mV out at 10v on the Keithley 2100 about the same the other way on the PM2525. (I had to swap the power leads round).

I haven't checked the specs on a Keithley 2100, but that difference seems quite large. Was everything warmed up for a while?

The closeness of my results I think was slightly fortuitous as my study was quite warm (from equipment running in various experiments I'm doing) and I think RoadRunner calibrates at quite a warm 24C (that is from memory - perhaps I have that wrong) but I would still expect the variation to be a few 10s of microVolts not 300 or so.

It was on for hours.  I've had it on again tonight and it's 9.99954 now

So my only other reference is a £4 10v ebay reference so show here 2nd we get..

So 2100 reads 9.99954 & 10.00590
The PM2525 reads  10.002 & 10.010
The UT60E reads 10.00 & 10.01
The UNI-T 4102C reads 9.99 & 10.00

Now we have 9.99965  when I switched it on cold it read 10.00118

It is just sat on the desk connected to my bench power supply which is set to 15.00123 (using the 2100)

[cowasaki]

According to my 4 wire PT100 probe on the 2100 the temperature is 27.197 or 27.1 with the infra-red thermometer

The temp of the LM399AH is 29.7 using the probe or 29.9 using the thermometer

[RoadRunner]

My board arrived this morning.

I'm about 1/3 of a mV out at 10v on the Keithley 2100 about the same the other way on the PM2525. (I had to swap the power leads round).

I haven't checked the specs on a Keithley 2100, but that difference seems quite large. Was everything warmed up for a while?

The closeness of my results I think was slightly fortuitous as my study was quite warm (from equipment running in various experiments I'm doing) and I think RoadRunner calibrates at quite a warm 24C (that is from memory - perhaps I have that wrong) but I would still expect the variation to be a few 10s of microVolts not 300 or so.

Yes it seems to be quite large, Most of people i get feedback from had variation with there insturment around +/- 100 micro volts, Since toastedcrumpets pointed out ,now During calibration i cover board with 3 CM think foam, while ambieant temprature around 22C.

Regards

[cowasaki]

My board arrived this morning.

I'm about 1/3 of a mV out at 10v on the Keithley 2100 about the same the other way on the PM2525. (I had to swap the power leads round).

I haven't checked the specs on a Keithley 2100, but that difference seems quite large. Was everything warmed up for a while?

The closeness of my results I think was slightly fortuitous as my study was quite warm (from equipment running in various experiments I'm doing) and I think RoadRunner calibrates at quite a warm 24C (that is from memory - perhaps I have that wrong) but I would still expect the variation to be a few 10s of microVolts not 300 or so.

Yes it seems to be quite large, Most of people i get feedback from had variation with there insturment around +/- 100 micro volts, Since toastedcrumpets pointed out ,now During calibration i cover board with 3 CM think foam, while ambieant temprature around 22C.

Regards

So what should I do with it as best practice? Cover it in foam in a sealed box? I can do that but wouldn’t the temperature be higher. Is my temp the reason for the low figure? The device showed a higher figure earlier before the heat increased.

Thanks.

[kado]

Hi,

i have two boards and realised, that the 10 V output voltage varies in the 100s uV if the 15V input change in 100s mV below the 15V checkpoint.

Karsten


Gesendet von iPhone mit Tapatalk

[RoadRunner]

Hi,

i have two boards and realised, that the 10 V output voltage varies in the 100s uV if the 15V input change in 100s mV below the 15V checkpoint.

Karsten


Gesendet von iPhone mit Tapatalk

Thanks for pointing, yes power supply need to be as close as possible to 15V.

Regards

[DC1MC]

Hi,

i have two boards and realised, that the 10 V output voltage varies in the 100s uV if the 15V input change in 100s mV below the 15V checkpoint.

Karsten

Is the reverse true as well, if the 15V increases over 15V the variation goes as well ?

 DC1MC

[cowasaki]

Hi,

i have two boards and realised, that the 10 V output voltage varies in the 100s uV if the 15V input change in 100s mV below the 15V checkpoint.

Karsten


Gesendet von iPhone mit Tapatalk

Thanks for pointing, yes power supply need to be as close as possible to 15V.

Regards

Well the 10V board also makes a good thermometer

Attached is the board when the temperature is 24.5C and the voltage is 15.0+- 0.05V

10uV out which is much more like what I expected.  I'm going to build a new 15v board and wrap the ref board and stick it in it's own box but with my reference resistors as well

EDIT: Just read that....  So is it better to be 0.1V over than a little under?  OR as close as possible..... and just HOW close?

ALSO....  Is it safe to wrap it up in foam?  I don't want to damage it by overheating the board or anything like that.

Thanks


This is now in the range of ..... which is out, the reference or the DMM? :-)

[kado]

Tune down slowly your 15V feeding source and you will get a exact 10.00000 reading. If thats the way to go with a calibration source i am not sure.
Was myself astonished about that behaviour.

Karsten


Gesendet von iPhone mit Tapatalk