Author Topic: LM399 based 10 V reference  (Read 536623 times)

gmac34 and 3 Guests are viewing this topic.

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: LM399 based 10 V reference
« Reply #600 on: November 09, 2017, 09:04:15 pm »
Hello Branadic,

what are those large voltage dips on your measurement?
What happens if you have a 100nF capacitor across the zener?
Or is it the output scaling amplifier?

with best regards

Andreas
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: LM399 based 10 V reference
« Reply #601 on: November 09, 2017, 09:42:46 pm »
Andreas,

beside amplified popcorn noise I again do have a temperature influence on the 6,95V --> 10V amplifier. This is why I will replace the 82k 25ppm/K resistor by a 82k2 5ppm/K one, that is in parallel to a 10k 2.5ppm/K in the amplifier. Hope to get the resistor by the middle of next week.

The original 5ppm/K smd resistors where fine, after they where preaged incircuit (with crystal heater on) and after crystal heater turned of. However, I wanted to see what difference I can achieve by through hole components, so I replaced them by AE 2.5ppm/K resistors (10k and 20k).

-branadic-
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: LM399 based 10 V reference
« Reply #602 on: November 16, 2017, 04:16:15 pm »
Vishay SFERNICE Foil resistor 82k2 received today, so I stopped the current measurement (updated graphs in former comment) and replaced the 82k metal film resistor. Here are the first measurement points captured. I will update the graphs from time to time, so we get an idea of what has changed.

-branadic-
« Last Edit: December 10, 2017, 12:36:44 am by branadic »
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: LM399 based 10 V reference
« Reply #603 on: November 16, 2017, 07:32:55 pm »
beside amplified popcorn noise I again do have a temperature influence

Hello branadic,

I did not mean the 0.5ppm popcorn noise but the larger (short) spikes of around 4 ppm especially at the beginning of the measurement.

with best regards

Andreas

 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: LM399 based 10 V reference
« Reply #604 on: November 16, 2017, 08:12:22 pm »
Quote
I did not mean the 0.5ppm popcorn noise but the larger (short) spikes of around 4 ppm especially at the beginning of the measurement.

Got you, but I don't know where they are coming from. Maybe some external emf coming via power line into the lab power supply and thus into reference voltage? There is a lot of test equipment, test benches but also a photo table top with background lights inside the lab, that is used irregular and that I can't control. Mains line is also messy by the maschines running inside our building. So I wouldn't pay to much attention on that spikes as the reference is not running on batteries.

-branadic-
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: LM399 based 10 V reference
« Reply #605 on: November 24, 2017, 07:04:16 pm »
Updated the graphs above. After the weekend I will stop measurements on this LM399 for a while and will turn over to the 4xLM399 of Philipp and see how stable this unit is. I'm quite happy with the performance by now. Even though the temperature changes haven't been very big it's obvious that the most temperature sensitive part of the reference is the gain amplifier.

Compare reference output from post 619

and amplified output from post 627


I think I will perform some investigation on this monitoring the difference between reference output and amplified output soon, to make this sensitivity somewhat more visible.

-branadic-
« Last Edit: December 03, 2017, 08:01:31 pm by branadic »
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14072
  • Country: de
Re: LM399 based 10 V reference
« Reply #606 on: November 24, 2017, 08:24:46 pm »
The amplified reference does not look that bad. The sharp peaks going down should be popcorn noise from the reference itself. It somewhat depends on how stable the temperature was how good the performance of the 7 to 10 V amplifier is. Unless the temperature was really stable the performance looks good.
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: LM399 based 10 V reference
« Reply #607 on: November 24, 2017, 09:20:15 pm »
Hello branadic,

was the measurement on different time / different measurement instruments?
It looks a bit confusing to me for a direct comparison. (85K vs. 170K samples).

with best regards

Andreas


 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: LM399 based 10 V reference
« Reply #608 on: November 24, 2017, 09:43:48 pm »
Hello Andreas,

reference output was from post 619 and amplified output from post 627, two different measurements as I do only have one 3458A and can't measure both outputs at the same time. Thus I will perform a measurement with the difference between reference output and ampflied output.

By the way: 85.000 samples = 94,44 hours as measurement rate is 0,25 Hz (100NPLC with Auto Zero on)

-branadic-
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: LM399 based 10 V reference
« Reply #609 on: November 29, 2017, 09:26:34 pm »
With the latest data given, some dependancy can be plotted:

1. voltage vs. temperature
2. voltage vs. humidity
3. voltage vs. dew point temperature (includes temperature and humidity)
4. voltage vs. ambient pressure

-branadic-
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: LM399 based 10 V reference
« Reply #610 on: December 03, 2017, 08:26:38 pm »
I added the dependancy plots for the raw reference output https://www.eevblog.com/forum/metrology/lm399-based-10-v-reference/msg1320837/#msg1320837

While the raw reference voltage is almost insensitive to changes in temperature, humidity and pressure, the amplified 10V output indicates a correlation. I'm currently not sure if the resistors haven't yet stabilized in their new environment or if it's the LT1001 which is a CDIP type. But as they are all within a Hammond aluminium die cast and covered with batting I can't find a simple and thus logical explanation. Any ideas on this?

-branadic-
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14072
  • Country: de
Re: LM399 based 10 V reference
« Reply #611 on: December 03, 2017, 08:54:26 pm »
The voltage versus temperature slope is around 3 µV/K. This would be ratehr high for an LT1001 after amplification by a factor of about 1.5.  So it is either is rather poor OP or it's more like the resistors. The resistor dependent part is about 3 V or the 10 V.  So it would be something like a 1ppm/K TC matching for the resistors.
 

Offline Insatman

  • Frequent Contributor
  • **
  • Posts: 278
  • Country: ph
Re: LM399 based 10 V reference - First 650 hours of data
« Reply #612 on: December 12, 2017, 09:45:03 am »
I built this 10volt LM399 based reference on perfboard inside a cast aluminum enclosure.  The LM399 is insulated inside a cube of Styrofoam about 1" square.  The +18v supply is from a decent quality wallwart unit (from digikey).  Attached is voltage data taken approx every 24hours on an HP3457a meter.   The average value for "overnight" is logged.  The typical standard deviation is less than 7.0E-6 during an overnight run.   Also attached is as-built schematic.   All parts are thru-hole.   The resistors marked LTC are low temp co.  Typically 15-25pm.  The pot used for adjusting the voltage was the best I had rated at 100ppm.  The measurement cable used is shielded twisted pair with gold pomona plugs on both ends. 

Question here is this.  Is the upward slope of this LM399 typical for >500 hours of operation?  I doubt the meter is drifting but it is possible I suppose. 
Retired Pulsed Power Engineer/Physicist...now I just dabble in electronics
 

Offline TiN

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: LM399 based 10 V reference
« Reply #613 on: December 12, 2017, 12:38:09 pm »
Maybe it's your R7 trimmer resistor affecting the stability? I'd use smaller value resistor just for finetune so effect from trimpot would be minimal, with further replacement to  proper fixed  resistance network.
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 
The following users thanked this post: Insatman

Online David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: LM399 based 10 V reference
« Reply #614 on: December 12, 2017, 02:24:40 pm »
While the raw reference voltage is almost insensitive to changes in temperature, humidity and pressure, the amplified 10V output indicates a correlation. I'm currently not sure if the resistors haven't yet stabilized in their new environment or if it's the LT1001 which is a CDIP type. But as they are all within a Hammond aluminium die cast and covered with batting I can't find a simple and thus logical explanation. Any ideas on this?

Thermocouple effects?

Did you use the offset voltage adjustment terminals of the LT1001?  They will produce an offset voltage drift of 1uV/C for every 300uV of offset away from zero.
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: LM399 based 10 V reference
« Reply #615 on: December 12, 2017, 03:10:16 pm »
Quote
Thermocouple effects?

Did you use the offset voltage adjustment terminals of the LT1001?  They will produce an offset voltage drift of 1uV/C for every 300uV of offset away from zero.

Well, the LT1001 is in a precision socket, but everything inside the Hammond die cast is filled with batting. I don't use the offset trim pins, they are left open.

I've started measuring the difference between reference voltage and amplified voltage, thus the amp itself. Here are current diagrams. The FFT indicates 1/f-noise as well as the popcorn (burst) noise of the reference. Temperature influence on the gain stage is also clearly visible.  :--

-branadic-
« Last Edit: December 15, 2017, 07:48:22 pm by branadic »
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: LM399 based 10 V reference
« Reply #616 on: December 12, 2017, 06:40:12 pm »
Quote
Thermocouple effects?

Did you use the offset voltage adjustment terminals of the LT1001?  They will produce an offset voltage drift of 1uV/C for every 300uV of offset away from zero.

Well, the LT1001 is in a precision socket, but everything inside the Hammond die cast is filled with batting. I don't use the offset trim pins, they are left open.

The increase in offset voltage drift actually applies to any offset so trimming to zero offset reduces the offset voltage drift to zero as well but given the already low trimmed offset voltage of the LT1001, I doubt this is the problem.

In the past I have built a test circuit to grade amplifiers for offset voltage and offset voltage drift but that should not be necessary here with the LT1001.

Quote
I've started measuring the difference between reference voltage and amplified voltage, thus the amp itself. Here are current diagrams. The FFT indicates 1/f-noise as well as the popcorn (burst) noise of the reference. Temperature influence on the gain stage is also clearly visible.

I do not think the FFT is showing anything useful about the amplifier.  Maybe operate the circuit with the reference output disconnected and the amplifier input grounded to get just the performance of the amplifier?
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14072
  • Country: de
Re: LM399 based 10 V reference
« Reply #617 on: December 12, 2017, 08:12:32 pm »
One would expect the noise to be mainly from the LM399 reference. Compared to OPs the noise of the LM399 is really high (100 nV/Sqrt(Hz) and around 10 Hz 1/f noise cross over) Only for the higher frequencies (e.g. > 100 Hz) where filtering is practical the reference noise can be reduces.

On the other hand drift and TC would be most likely to a large part due the two main resistors that set the gain. Here it depends on the resistor types. The TC can still be quite a bit better than the nominal TC of the resistors, as the second oder TC component can often compensate and only about 1/3 of the 10 V is from the amplified part. Drift of the OP would be in the 1 µV/K order of magnitude and thus 0.1 ppm /K range for the output.

The FFT and other data only show the very low frequency noise part. This 1/f and maybe thermal variations.
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: LM399 based 10 V reference - First 650 hours of data
« Reply #618 on: December 12, 2017, 08:54:04 pm »
Question here is this.  Is the upward slope of this LM399 typical for >500 hours of operation?  I doubt the meter is drifting but it is possible I suppose.

Hello,
it really depends. See examples here:

https://www.eevblog.com/forum/metrology/lm399-based-10-v-reference/msg796829/#msg796829

So you should wait for at least 5000 hours or 200 days before you can really judge a LM399.
But it could also be the ageing of your resistors. 25ppm/K is not really precision.
And your trimming scheme could also be better (see LM399 data sheet).

with best regards

Andreas
 
The following users thanked this post: Insatman

Offline Insatman

  • Frequent Contributor
  • **
  • Posts: 278
  • Country: ph
Re: LM399 based 10 V reference
« Reply #619 on: December 13, 2017, 02:06:12 am »
While the raw reference voltage is almost insensitive to changes in temperature, humidity and pressure, the amplified 10V output indicates a correlation. I'm currently not sure if the resistors haven't yet stabilized in their new environment or if it's the LT1001 which is a CDIP type. But as they are all within a Hammond aluminium die cast and covered with batting I can't find a simple and thus logical explanation. Any ideas on this?

Thermocouple effects?

Did you use the offset voltage adjustment terminals of the LT1001?  They will produce an offset voltage drift of 1uV/C for every 300uV of offset away from zero.

I did not use the offset terminals in my build.   After looking at branadic's data, I should probably see if the amplifier is the source of the drift.
Retired Pulsed Power Engineer/Physicist...now I just dabble in electronics
 

Offline Insatman

  • Frequent Contributor
  • **
  • Posts: 278
  • Country: ph
Re: LM399 based 10 V reference - First 650 hours of data
« Reply #620 on: December 13, 2017, 07:48:45 am »
Question here is this.  Is the upward slope of this LM399 typical for >500 hours of operation?  I doubt the meter is drifting but it is possible I suppose.

Hello,
it really depends. See examples here:

https://www.eevblog.com/forum/metrology/lm399-based-10-v-reference/msg796829/#msg796829

So you should wait for at least 5000 hours or 200 days before you can really judge a LM399.
But it could also be the ageing of your resistors. 25ppm/K is not really precision.
And your trimming scheme could also be better (see LM399 data sheet).

with best regards

Andreas

I have no problem waiting but I thought the drift would settle down somewhat after about 200 hours like most other references.  I've also started monitoring the LM399 output directly, so I can isolate if the drift is the LM399 or op-amp.   I will take a month or so of data and see what comes out.

I measured the value of the 5k pot today and ordered a 15ppm resistor to replace it with.  Being in Philippines, I'm pretty much stuck with Digikey for good parts as they deliver here quickly.  Otherwise Ebay is my only other alternative but parts take 1-2 months to make through the postal system here.

 

Retired Pulsed Power Engineer/Physicist...now I just dabble in electronics
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: LM399 based 10 V reference
« Reply #621 on: December 13, 2017, 12:39:40 pm »
Did you use the offset voltage adjustment terminals of the LT1001?  They will produce an offset voltage drift of 1uV/C for every 300uV of offset away from zero.

I did not use the offset terminals in my build.   After looking at branadic's data, I should probably see if the amplifier is the source of the drift.

In case it was not clear, the reason I brought the input offset terminals up is that using them to trim the input offset voltage to zero also trims the input offset voltage drift to zero.  Where people sometimes get into trouble is using the input offset terminals to trim offset from other areas of the circuit resulting in very high input offset voltage drift.

 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: LM399 based 10 V reference
« Reply #622 on: December 13, 2017, 07:47:44 pm »
Quote
Did you use the offset voltage adjustment terminals of the LT1001?  They will produce an offset voltage drift of 1uV/C for every 300uV of offset away from zero.

From where do you get those numbers? I do have LT1001ACJ8 with the following numbers for input offset voltage behavior given in the datasheet:

Input Offset Voltage: typ. 20µV, max. 60µV
Long Term Input Offset Voltage Stability: typ. 0.2µV/month, max. 1.0µV/month
Average Offset Voltage Drift: typ. 0.2µV/°C, max. 0.6µV°C

-branadic-
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14072
  • Country: de
Re: LM399 based 10 V reference
« Reply #623 on: December 13, 2017, 08:02:41 pm »
The ratio of 300 µV of offset producing 1 µV/K of drift is a good approximation, but not 100% accurate.

The main reason for offsets in BJT based amplifiers is usually a mismatch in current or effective size of a transistor pair. This results in an offset voltage proportional to absolute temperature (This is one way to make a PAT source)  and thus the approximate ratio.

With most BJT based OPs adjusting the offset to zero will also reduce the drift to a low value. However there can be small extra contributions to the drift. The relation between offset and drift is usually better at higher offset.

The min / max values in the datasheet are usually what is tested - so there ratio may not reflect the physics behind it.
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: LM399 based 10 V reference
« Reply #624 on: December 13, 2017, 08:25:09 pm »

but I thought the drift would settle down somewhat after about 200 hours like most other references. 

Hello,

200 hours would be a typical time to stabilize for a plastic housing to environment humidity after soldering.
I guess the drift during this phase is much larger than the initial drift of a LM399.

With best regards

Andreas
 
The following users thanked this post: splin


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf