Author Topic: Building your own voltage reference - the JVR  (Read 127213 times)

0 Members and 1 Guest are viewing this topic.

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #50 on: October 15, 2015, 09:22:04 pm »
Well, my two prototypes (5V with U440 and 10V with J211) are running nicely and manage to stay inside +/-10ppm range at a somewhat varying room temperature, without any obvious drift. I am waiting for some high stability wire-wound and bulk foil resistors to be delivered so I can start building a "superstable" version, where all influences except that only of a JFET itself would be below 1ppm/C. The aim is to get a stability better than +/-10ppm in 20-30C temperature range (or better than 2ppm/C) without temperature stabilisation and then check the long-term drift for 1000h with a temperature log.

Cheers

Alex

P.S. - here is a 30min run on the bench of a somewhat "fine tuned" version on J211 device. Vertical scale 1ppm/div, Keithley 2015 works with a 10x averaging filter at NPLC10 to reduce the noise.

« Last Edit: October 15, 2015, 10:30:37 pm by Alex Nikitin »
 

Online Marco

  • Super Contributor
  • ***
  • Posts: 6694
  • Country: nl
Re: Building your own voltage reference - the JVR
« Reply #51 on: October 16, 2015, 05:38:02 am »
Might it be possible to get a low chosen voltage (say 2.5V) at near zero tempco by using an opamp based "constant" current sink in parallel with the source resistor?
« Last Edit: October 16, 2015, 06:33:08 am by Marco »
 

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #52 on: October 16, 2015, 06:47:49 am »
Overnight (from 12am till 6am, in open air), 1ppm/div:



Cheers

Alex
« Last Edit: October 16, 2015, 07:49:10 am by Alex Nikitin »
 

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #53 on: October 16, 2015, 07:16:31 am »
Might it be possible to get a low chosen voltage (say 2.5V) at near zero tempco by using an opamp based "constant" current sink in parallel with the source resistor?

It is not difficult to get lower voltages - just by choosing a different JFET. The sample of J211 I'm using right now has the reference voltage around 2.6V and J112 I've tried earlier - even lower, about 1.7V. I suspect however that it could be better to have this voltage as high as possible for a low noise and better stability.

Cheers

Alex
« Last Edit: October 16, 2015, 07:50:50 am by Alex Nikitin »
 

Online Marco

  • Super Contributor
  • ***
  • Posts: 6694
  • Country: nl
Re: Building your own voltage reference - the JVR
« Reply #54 on: October 16, 2015, 08:11:07 am »
The point wasn't to get a lower voltage, the point was to get a given voltage. It would be nice to have the elegance of the first circuit (where variations from everything but the JFET on output voltage are highly attenuated) but at a specific voltage. If you have 2.5V you can make 5v out of it with a switched capacitor circuit.
 

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #55 on: October 16, 2015, 10:13:53 am »
The point wasn't to get a lower voltage, the point was to get a given voltage. It would be nice to have the elegance of the first circuit (where variations from everything but the JFET on output voltage are highly attenuated) but at a specific voltage. If you have 2.5V you can make 5v out of it with a switched capacitor circuit.

Well, in that case an additional current source should also have a very high stability (means at least some top quality resistors) and would complicate things. An additional resistor gives you the required voltage above the reference, nothing more is required.

Cheers

Alex
 

Online Marco

  • Super Contributor
  • ***
  • Posts: 6694
  • Country: nl
Re: Building your own voltage reference - the JVR
« Reply #56 on: October 16, 2015, 11:07:33 am »
I'd have to the math to be sure, but as far as I can see drift in the current source has an attenuated effect due to transconductance just like variation in the source resistor (but NOT the second resistor you are adding in your current circuits).

PS. doing even a small bit of math in my head showed me I was wrong, I was overestimating transconductance by many orders of magnitude.
« Last Edit: October 16, 2015, 11:31:55 am by Marco »
 

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #57 on: October 16, 2015, 11:13:39 am »
I'd have to the math to be sure, but as far as I can see drift in the current source has an attenuated effect due to transconductance just like variation in the source resistor (but NOT the second resistor you are adding in your current circuits).

Yes, however there is only one reference voltage point and you can not move it by adding a current, you only can adjust the voltage to that point by an external current if the source resistor has not quite correct value. It is easier IMHO, just to get the resistor right in the first place (or to use a trimmer to get the last 100ppm).

Cheers

Alex
 

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #58 on: October 16, 2015, 07:22:06 pm »
OK, Friday night - a torture night  >:D . As the J211 version runs rather well and I am still waiting for the bits to build a controlled temperature chamber, I've decided to torture the JVR on record, first with a freezer spray for 10sec (about -50C or 70 degrees drop), then with a soldering iron for 20 seconds flat on the plastic case of JFET (I would estimate it as at least 70 degrees rise). Unlike the U440 in a metal case, some hysteresis is clearly visible, about +20ppm after the cold session and about -30ppm after the hot one. Here I have to note that the freezer session was not the first and on the first session I didn't freeze the FET so deeply, however the hysteresis was more - about 50ppm. Clearly for a stable reference a metal case is much better. (A note - the "wave" on the second graph after 400 samples is just me adjusting the voltage back to 10.00000 Volts).

Otherwise it is clear from the graphs that in both cases the voltage drops about 0.2-0.25% or 2000-2500ppm for 70-75 degrees change, so at 23C the reference sits nicely on the very top of the curve and has the lowest possible tempco. With a metal case and the temperature stabilised around, say, 45-50C (and the tempco adjusted for zero at that temperature), the stability should be very good.

Cheers

Alex

Each record is about 6min long:




« Last Edit: October 16, 2015, 07:51:57 pm by Alex Nikitin »
 

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #59 on: October 17, 2015, 02:03:10 pm »
A quick estimate - the temperature coefficient for this version (10V with J211 and INA133) is about 5ppm/C with few degrees change near 23C and up to 30-40ppm/C at temperature extremes.

Cheers

Alex
 

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #60 on: October 17, 2015, 05:21:30 pm »
To compare my results with a quality band-gap reference, I've decided to repeat my torture tests on LT1019-2.5 in DIL8 I had in the lab (with a stated 5ppm/C typical and 20ppm/C maximum tempco). This chip did show however a lot of hysteresis - about +600ppm after a freeze and -400ppm after a soldering iron application, leaving a total permanent shift about +200ppm. Ouch!  :o

Cheers

Alex
« Last Edit: October 17, 2015, 06:09:50 pm by Alex Nikitin »
 

Offline Messtechniker

  • Frequent Contributor
  • **
  • Posts: 774
  • Country: de
  • Old analog audio hand - No voodoo.
Re: Building your own voltage reference - the JVR
« Reply #61 on: October 18, 2015, 11:16:12 am »
Since I am new to this game please bear with me.

Simply as a first test, I built your circuit without the amplifier and with the components at hand. FET was BF 246 C and 2N 3819 new old stock resting for some 40 years in the box. Resistors were all standard metal film ¼ W, 1%, 50 ppm. Stability of these FETs was pretty awful +/- 300 to 500 ppm over 1 hour. So I had a look at the power supply. This was a LM 317 set to 24 V (to power the amplifier at some later date). Load was 10 mA approx. (dropping resistor and LED) and fed with 30 V from a benchtop power supply. Surprisingly drift was around +/- 5 ppm at room temperature.

Then to investigate further I changed the output to around 10 V. Feed voltage now being 20 V. What I then got is for you to see.

At the beginning it’s the test rig without me in the room. The droop before the peak is me entering the room whereby the “air blast” seemed to cool the LM 317 a little. The big peak is due to the soldering iron placed at the LM 317 for a short while. Then left the LM 317 to settle with some hysteresis being apparent. 25 ppm approx.

So I’ll probably follow this route and put the thing in a metal box and heat it to some temperature. Since the TO-220 casing is flat, I plan to attach flush one BD 139 (TO 126) as the temperature sensor and two further BD 139s on the outsides as heaters bolted together as a single compact pack. Hope this makes for a fair voltage reference, and this with standard components only.

Yours Messtechniker.

P.S: This is great fun. :)
Agilent 34465A, Siglent SDG 2042X, Hameg HMO1022, R&S HMC 8043, Peaktech 2025A, Voltcraft VC 940, M-Audio Audiophile 192, R&S Psophometer UPGR, 3 Transistor Testers, DL4JAL Transistor Curve Tracer, UT622E LCR meter
 

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #62 on: October 20, 2015, 03:13:08 pm »
P.S: This is great fun. :)

Oh, yes!

I've just bought the Fluke 731B and a lot of precision wire-wound resistors. I would like at least match the stability on the Fluke (and, as a side point, to be able to calibrate in-house the DC parameters of my two Keithley 2015).

Cheers

Alex
 
The following users thanked this post: Roman oh

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #63 on: October 21, 2015, 06:20:29 pm »
While I am waiting for the Fluke and resistors, here is a 21 hours run of the same 10V version using J211. The Keithley 2015 runs with 40 averaging, vertical scale is 1ppm/div. I've noted the room temperature at some time points. It looks like the TC around 23C is ~2ppm/C  8) .

Cheers

Alex

 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: Building your own voltage reference - the JVR
« Reply #64 on: October 22, 2015, 04:42:05 am »
I've noted the room temperature at some time points. It looks like the TC around 23C is ~2ppm/C  8) .

The TC of the Keithley (according to datasheet) or of the JVR?

With best regards

Andreas
 

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #65 on: October 22, 2015, 07:31:13 am »
I've noted the room temperature at some time points. It looks like the TC around 23C is ~2ppm/C  8) .

The TC of the Keithley (according to datasheet) or of the JVR?

With best regards

Andreas

Hi Andreas,

Your guess is as good as mine - the Keithley TC, noise and drift all included! The only possible clue is that the "fast" changes are more likely to be that of the JVR itself due to the air movements. As you may see there are variations in the evening when I was moving around the lab and in the morning hour when I was preparing to go to work, while at night and during the day there was very little air movement and the changes are more gradual and uniform. I hope that the Fluke 731B will provide me with a more stable reference than the 2015s (I'm running two in parallel though recording only one - and the difference between two 2015 varies from 4 to 6ppm). I plan to record the Fluke against 2015s after a good warm-up and then record the difference between the Fluke and the JVR. I also plan to build the "ultimate" 10V JVR using U440 and a Vishay Foil resistor divider for the 5 to 10V step up.

Cheers

Alex
« Last Edit: October 22, 2015, 07:41:16 am by Alex Nikitin »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: Building your own voltage reference - the JVR
« Reply #66 on: October 22, 2015, 09:26:23 pm »

 the "fast" changes are more likely to be that of the JVR itself due to the air movements.


Hello Alex,

air movements?
or changes by parasytic capacitance?
have you checked what happens when your hand moves over the cirquit or touch the power supply and voltage output lines?

My "zero quiescent current voltage regulators" were very sensitive in this regard.
 (several volts output change without blocking capacitors).
I had to use the blocking capacitors to keep the noise away from the gate diode.

With best regards

Andreas
 

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #67 on: October 22, 2015, 09:40:53 pm »

 the "fast" changes are more likely to be that of the JVR itself due to the air movements.


Hello Alex,

air movements?
or changes by parasytic capacitance?
have you checked what happens when your hand moves over the cirquit or touch the power supply and voltage output lines?

My "zero quiescent current voltage regulators" were very sensitive in this regard.
 (several volts output change without blocking capacitors).
I had to use the blocking capacitors to keep the noise away from the gate diode.

With best regards

Andreas

Hi Andreas,

No, these "fast" changes are actually quite slow - minutes. And the reference is not sensitive to the external interference or capacitive coupling. I can touch it, move it, shake it and, as long as I don't transfer any heat from my hand to the circuit, it stays rock solid. I am now building a 0.1Hz-10Hz filter-amplifier, with the gain 10000 and self-noise around 100-150nV p-p in the pass-band, so I can accurately observe and measure the LF noise.

Cheers

Alex
 

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #68 on: October 27, 2015, 08:09:45 pm »
Well, the Fluke 731B arrived yesterday (no battery though, so I've disconnected R30 to avoid the indicator overload and added noise). It was last calibrated by Fluke in 2009 and agrees with both of my Keithley 2015s within 10ppm on the 10V output. I've run it overnight and it did show about -1ppm/C TC  - see the graph below, taken with NPLC10 and averaging over 100 samples. On the previous night I did run my 10V JVR with J211 using the same settings on the 2015. Below both graphs using the same scale. Again, I've noted the room temperature at some points. Next step would be to put the JVR in a suitable enclosure as the "faster" variations are most likely due to the air movements in the room.

Cheers

Alex






« Last Edit: October 27, 2015, 08:15:54 pm by Alex Nikitin »
 

Offline Messtechniker

  • Frequent Contributor
  • **
  • Posts: 774
  • Country: de
  • Old analog audio hand - No voodoo.
Re: Building your own voltage reference - the JVR
« Reply #69 on: October 28, 2015, 08:08:56 am »
Well, here now the next development step of my poor man’s 10 V reference based on the venerable LM 317 with its internal bandgap reference. The LM 317 is now heated by two BD 139 trannies on each side with a third BD 139 serving as a temperature sensor sandwiched in between making a single compact block bolted together with a M3 bolt. The arrangement is now insensitive to air movements although not yet contained in a metal box.

The next development steps will be:
1.)   Replace the metal M3 bolt and nut of the block with a plastic M3 bolt and nut.
2.)   Put the LM 317 block with the heating circuit on an new and clean hole dot matrix epoxy pcb. (instead of the currently used hard paper hole dot matrix pcb. material and avoiding flux residues) in a small metal box.
3.)   Provide separate power supplies for the LM 317 block (+15 V) and the heater (+15 V, -15 V and - 5 V)
4.)   Do some extended burning-in and subsequent ppm measurements. Watch this space.

By the way, my Keysight 34465A DMM set to the 10 V range and with its input shorted produces an output fluctuating at about 0.2 ppm (as measured over 1 hour). So my measurement limit, i.e the contribution from the 34465A in the enclosed graph will be about 10 times that at 2 ppm re. 10 V.

My aim is to produce a voltage reference circuit specc’d for around +/- 10 ppm (20 ppm total) perhaps, using cheap off-the-shelf components only.

What do you think?

Yours – Messtechniker

P.S. Still having fun with this. :)
Agilent 34465A, Siglent SDG 2042X, Hameg HMO1022, R&S HMC 8043, Peaktech 2025A, Voltcraft VC 940, M-Audio Audiophile 192, R&S Psophometer UPGR, 3 Transistor Testers, DL4JAL Transistor Curve Tracer, UT622E LCR meter
 

Offline BravoV

  • Super Contributor
  • ***
  • Posts: 7547
  • Country: 00
  • +++ ATH1
Re: Building your own voltage reference - the JVR
« Reply #70 on: October 28, 2015, 12:01:11 pm »
Messtechniker, schematic and photos please, tia.

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #71 on: October 28, 2015, 12:05:31 pm »

What do you think?

Yours – Messtechniker

P.S. Still having fun with this. :)

Hi,

Where are the voltage setting resistors placed? You need to keep these at a stable temperature too if you are looking at <20ppm stability.

Cheers

Alex
 

Offline Messtechniker

  • Frequent Contributor
  • **
  • Posts: 774
  • Country: de
  • Old analog audio hand - No voodoo.
Re: Building your own voltage reference - the JVR
« Reply #72 on: October 28, 2015, 01:25:02 pm »
Messtechniker, schematic and photos please, tia.

Circuit is still at a very early stage (see remark below). Will provide it as soon as I (hopefully) get improved results.

Where are the voltage setting resistors placed? You need to keep these at a stable temperature too if you are looking at <20ppm stability.

The voltage setting resistors are currently placed close to the heated LM 317 but horizontally on the pcb. Based on your remark I'll put them upright and against each side of the LM 317 in thermal contact. Will look messy, but will probably be more effective. Thanks for the hint.

Yours - Messtechniker
Agilent 34465A, Siglent SDG 2042X, Hameg HMO1022, R&S HMC 8043, Peaktech 2025A, Voltcraft VC 940, M-Audio Audiophile 192, R&S Psophometer UPGR, 3 Transistor Testers, DL4JAL Transistor Curve Tracer, UT622E LCR meter
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: Building your own voltage reference - the JVR
« Reply #73 on: October 28, 2015, 09:12:19 pm »
Well, here now the next development step of my poor man’s 10 V reference based on the venerable LM 317

My aim is to produce a voltage reference circuit specc’d for around +/- 10 ppm (20 ppm total) perhaps, using cheap off-the-shelf components only.

What do you think?


I think you could get better results when using a LM723.
the LM317 has alone a "typical" noise of 30 ppm RMS = 2mVpp.

with some clever design you could use the current limiting transistor
as chip temperature sensing diode for the oven.

with best regards

Andreas
 

Online Alex NikitinTopic starter

  • Super Contributor
  • ***
  • Posts: 1150
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Building your own voltage reference - the JVR
« Reply #74 on: October 28, 2015, 11:25:31 pm »
Two more graphs, taken at NPLC1, no filtering, for 30 minutes. The room temperature was about 24.5C, the JVR is still in open air on the bench. I was hoping to compare the noise, however the noise in the graphs is completely dominated by the noise of Keithley 2015  :( .

Cheers

Alex



 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf