Author Topic: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x  (Read 234891 times)

0 Members and 2 Guests are viewing this topic.

Offline technix

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #125 on: November 07, 2016, 05:31:12 am »
I have talked with factory that originally made those 2DW233's, Shanghai No. 17 Radio Factory, and I can order them for you fresh from the factory if you want to.
 
The following users thanked this post: evb149

Offline zlymexTopic starter

  • Frequent Contributor
  • **
  • Posts: 556
  • Country: cn
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #126 on: November 07, 2016, 05:44:35 am »

If I recall correctly I think I saw that the noise of a zener is related to its area and current density which makes sense if a larger area is like putting more devices in parallel.  If that is the case then perhaps part of the lower noise of such devices could simply be that they are larger area devices?

That is not the case. If we parallel four identical devices and run the same current, the noise would be the same.
Parallel four devices and the noise is halved, this only happens if the total current is quadrupled. but this is equivalent to running the larger current on a single device.
The advantage of large area is the ability to run at larger current.
"2DW232 has much better noise than LTZ1000", this is compared at the same zener current.
 
The following users thanked this post: evb149

Offline 2N3055

  • Super Contributor
  • ***
  • Posts: 6407
  • Country: hr
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #127 on: November 07, 2016, 07:36:16 am »
You can consider that large crystal is equivalent of many small crystals connected in parallel. All stochastic noise related events combine same as they would from dozens of separate parallel connected diodes..
Of course,  if  you run it as same current density as with separate diodes...
 
The following users thanked this post: evb149

Offline technix

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #128 on: November 08, 2016, 04:07:46 am »
Thank you very much, technix!

What information did you receive about prices for various quantity ranges?

I would be interested to order some, maybe others here will also be so.

I have talked with factory that originally made those 2DW233's, Shanghai No. 17 Radio Factory, and I can order them for you fresh from the factory if you want to.

They gave me a solid 75 US cents each regardless of amount, and for order more than 50 (a whole pack) they will pay for the shipping from their factory to my home. When I send it on the shipping cost would be $5 for less than 2kg.
 

Offline technix

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #129 on: November 08, 2016, 04:33:26 am »
Here is the test circuit I am going to use when selecting the 2DW233 for you.

ADC is the built-in 10-bit unit inside ATmega328P, while DAC is TLC5615 (8-bit) referenced from ADR03B. The microcontroller will perform a search to find the best approximate of the zero TC point in the range of 0mA to about 50mA.
« Last Edit: November 08, 2016, 05:59:11 am by technix »
 

Offline zlymexTopic starter

  • Frequent Contributor
  • **
  • Posts: 556
  • Country: cn
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #130 on: November 08, 2016, 11:33:32 am »
Here is the test circuit I am going to use when selecting the 2DW233 for you.

ADC is the built-in 10-bit unit inside ATmega328P, while DAC is TLC5615 (8-bit) referenced from ADR03B. The microcontroller will perform a search to find the best approximate of the zero TC point in the range of 0mA to about 50mA.
2DW232 is very sensitive to the supply current, therefore I suspect the short term stabllity of the current mirror unless they are paired and thermally bonded.

Theoritically, to test the zero T.C. points(there are many for one device), there is the need both for varying the current and the temperature.
Practically, a step current can be used, then monitor the voltage variation direction to determin whether the current is Ok.
 

Offline technix

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #131 on: November 08, 2016, 12:40:45 pm »
Here is the test circuit I am going to use when selecting the 2DW233 for you.

ADC is the built-in 10-bit unit inside ATmega328P, while DAC is TLC5615 (8-bit) referenced from ADR03B. The microcontroller will perform a search to find the best approximate of the zero TC point in the range of 0mA to about 50mA.
2DW232 is very sensitive to the supply current, therefore I suspect the short term stabllity of the current mirror unless they are paired and thermally bonded.

Theoritically, to test the zero T.C. points(there are many for one device), there is the need both for varying the current and the temperature.
Practically, a step current can be used, then monitor the voltage variation direction to determin whether the current is Ok.

In practise I am using the BCV62 for the current mirror - the two PNPs are built on the same piece of silicon and are factory matched. I have already breadboarded the setup and tests on my stock of 1N4728 and 1N4734 are showing the correct behavior. I can scan through 0 to 25mA (I ended up using a 199.7 ohm resistor instead of 100 ohm one as the current sense resistor) at about 0.1mA step and measure the voltage with this automated jig.

I will implement a faster search algorithm that can find the optimal current on the MCU and a component insert sense feature, so when the parts arrive I can bulk test them very rapidly.
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14016
  • Country: de
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #132 on: November 08, 2016, 01:02:29 pm »
To measure the voltage change one should use a way more capable ADC / DMM or maybe measure only difference towards a fixed voltage close by. For a 10 ppm/K TC and 10 K Temperature step, one has only bout a .6 mV change. So something like a 14 Bit resolution would be the minimum - better more.

Changing the current can also change the internal temperature - though at least the temperature increase should be about proportional to current and reproducible.
 

Offline technix

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #133 on: November 08, 2016, 02:29:52 pm »
To measure the voltage change one should use a way more capable ADC / DMM or maybe measure only difference towards a fixed voltage close by. For a 10 ppm/K TC and 10 K Temperature step, one has only bout a .6 mV change. So something like a 14 Bit resolution would be the minimum - better more.

Changing the current can also change the internal temperature - though at least the temperature increase should be about proportional to current and reproducible.

I was searching for the Iz with maximum Vz. Maybe I do need a better ADC though, but otherwise my rig is fully automated and can detect the insert of DUT and can search for that Iz point.
 

Offline Cerebus

  • Super Contributor
  • ***
  • Posts: 10576
  • Country: gb
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #134 on: November 08, 2016, 04:36:12 pm »
To measure the voltage change one should use a way more capable ADC / DMM or maybe measure only difference towards a fixed voltage close by. For a 10 ppm/K TC and 10 K Temperature step, one has only bout a .6 mV change. So something like a 14 Bit resolution would be the minimum - better more.

Changing the current can also change the internal temperature - though at least the temperature increase should be about proportional to current and reproducible.

I was searching for the Iz with maximum Vz. Maybe I do need a better ADC though, but otherwise my rig is fully automated and can detect the insert of DUT and can search for that Iz point.

If you measure differentially against a stable voltage source that's close to the expected zener voltage and then add a bit of gain you'll get much better sensitivity. I'm concerned that you've looking for effects that are on a very close order to the LSB step of the ATMega's ADC. You could derive your stable voltage from the same reference you use for the ADC. You might even gain a tiny advantage from this giving you a ratiometric measurement vs your voltage reference.

I don't exactly know what the expected spread of voltage with individual part variations and temperature is, but if we say for argument sake it is +/-300mV and your original measurement range was 0-10V that would give a resolution improvement of 17 times for the cost of an op amp and a few resistors. That's an LSB of 585 uV versus your original LSB of 9.76 mV. I think you need this and at least a couple more bits on the ADC. The latter would give you a ~125 uV LSB.
Anybody got a syringe I can use to squeeze the magic smoke back into this?
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14016
  • Country: de
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #135 on: November 08, 2016, 04:59:34 pm »
One is not looking for the current with maximum Vz. The Voltage should go up with current over the full range - unless the temperature rise is to much. This point is not what one is looking for. The interesting point is the current at which the voltage is not temperature dependent in linear approximation for the right temperature range, e.g. with 30 C environment or a higher temperature like 50-60C when used with temperature stabilization.

So one would really need to resolve small changes (like 10 µV) in the voltage. So the test rig would be a little more complicated. More like using a commercial DMM (or high resolution ADC) to measure the voltage. The µC internal ADC is more like good enough to read the temperature from the other diode.
If you really want to go with the µC internal (10 Bit) ADC one might be able to only measure the difference to a second reference and use the ADC only for a maybe +-50-100 mV.

So the procedure could be:
0) at RT measure the forward voltage of the second diode, to calibrate it as a thermometer.
1) set test current for a first try
2) Adjust the second ref. voltage so that the difference to the DUT is small (e.g. < 50  mV)
3) Put a relatively high current (e.g. 10-20 mA) through the second diode in the chip, to cause some heat up
4) Change the current to the second diode to a small one in forward direction to use it for temperature measurement
5) do a fast measurement of voltage difference and temperature during cooling (e.g. 5-30 seconds). Temperature measurement is not that critical - decay curve should be similar for the devices.
It more to get the average temperature and an estimate for the order of magnitude.
6) calculate / estimate TC for the given test current and temperature range
7) adjust the current to get lower TC   (e.g. interpolate / extrapolate form older points and maybe typical curve).
8) Repeat if needed (TC to large).

For just a few parts one could do this also by hand - reading the drop on a DMM.
If you really need to automate, I would look for a better external ADC (e.g. 16-24 Bits).
 

Offline VintageNut

  • Frequent Contributor
  • **
  • Posts: 534
  • Country: 00
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #136 on: November 08, 2016, 05:49:04 pm »
Or just use a SMU......
working instruments :Keithley 260,261,2750,7708, 2000 (calibrated), 2015, 236, 237, 238, 147, 220,  Rigol DG1032  PAR Model 128 Lock-In amplifier, Fluke 332A, Gen Res 4107 KVD, 4107D KVD, Fluke 731B X2 (calibrated), Fluke 5450A (calibrated)
 

Offline technix

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #137 on: November 08, 2016, 06:20:39 pm »
So how do you guys think about this idea:

On the board I have an ADR03B that is used by the constant current generator. I can wire up the resistive divider so the divided buffered output from 2DW233 is also close to the output voltage of ADR03B. Then I can use the built-in differential PGA of the ATmega to measure the ratiometric difference of 2DW233 and ADR03B with a relatively high gain.

Rinse and repeat in a fridge or on top of a boiling kettle.
 

Offline Alex Nikitin

  • Super Contributor
  • ***
  • Posts: 1148
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #138 on: November 08, 2016, 06:38:40 pm »
So how do you guys think about this idea:

On the board I have an ADR03B that is used by the constant current generator. I can wire up the resistive divider so the divided buffered output from 2DW233 is also close to the output voltage of ADR03B. Then I can use the built-in differential PGA of the ATmega to measure the ratiometric difference of 2DW233 and ADR03B with a relatively high gain.

Rinse and repeat in a fridge or on top of a boiling kettle.

For low noise high stability applications mostly the performance in a reasonably narrow temperature range (say, 20C-50C) is important. I would use a temperature cycling device made out of a small Peltier cooler and a local heater for the zener, so the temperature swing can be arranged by just switching the heater on/off (as Peltier modules do not like switching polarity quickly) and the optimum current will be found for the smallest voltage change from cold to warm and back.

Cheers

Alex
 

Offline technix

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #139 on: November 08, 2016, 06:47:30 pm »
I don't know what kind of power element should be used to heat it up. Scanning the entire current range takes about 2.5 seconds using my fully automated design but things can be cooled down quite a bit by then. If I can digitally control the temperature even only with 8 bit resolution I can still characterize those diodes fairly well and can fully automate it (so you just plug in the component and wait until a full report is spit out from the AVR.)
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14016
  • Country: de
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #140 on: November 08, 2016, 07:01:33 pm »
The amplifier inside the AVRs is not really good, but in principle it could work with the difference.
With a gain of 200 the signal range would be somewhere in the 12 mV range and resolution at about 10 µV. With a divider to about half to third - this is 20-30 µV resolution for the reference.

The reference (ADR03) might need some noise filtering. Due to noise (e.g. amplifier) it may also take some averaging to get good enough data.
The divider needs individual adjustment (but still really stable) for each diode to bring it in the range. It might limit the current range used in one go.

The interesting temperatures would be more like normal room temperature or a little up, maybe up to 50 C environment. No real need for data below room temperature. Getting data on TC versus current at near room temperature (only self heating, maybe increased from the second diode) is already an important part. The optimum current should not be that much temperature dependent.

If you want a real characterization you would need a better ADC, to get the full resolution and thus also get direct data, so no adjustable divider. This is something a SMU is made for - if you have one.
 

Offline technix

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #141 on: November 08, 2016, 07:18:17 pm »
The amplifier inside the AVRs is not really good, but in principle it could work with the difference.
With a gain of 200 the signal range would be somewhere in the 12 mV range and resolution at about 10 µV. With a divider to about half to third - this is 20-30 µV resolution for the reference.

The reference (ADR03) might need some noise filtering. Due to noise (e.g. amplifier) it may also take some averaging to get good enough data.
The divider needs individual adjustment (but still really stable) for each diode to bring it in the range. It might limit the current range used in one go.

The interesting temperatures would be more like normal room temperature or a little up, maybe up to 50 C environment. No real need for data below room temperature. Getting data on TC versus current at near room temperature (only self heating, maybe increased from the second diode) is already an important part. The optimum current should not be that much temperature dependent.

If you want a real characterization you would need a better ADC, to get the full resolution and thus also get direct data, so no adjustable divider. This is something a SMU is made for - if you have one.
I don't have a SMU but I do have MCP3911, a dual channel 24-bit ADC with internal 1.7V reference and differential inputs. Maybe I can use that?

And still how do I control the temperature of the diodes? If I heat up the chip using one of the two diodes, will PWM work?
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14016
  • Country: de
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #142 on: November 08, 2016, 08:02:39 pm »
Using the MCP3911 could work to build a higher resolution measurement system. Only the low voltage range (like +-600 mV) might be a little inconvenient and need some care against thermal EMF. So it is possible, but not very easy and more like a separate project first.

The resolution is good enough to eliminate the subtraction. And stability should be good enough to measure Vz versus current at different temperatures instead of changes in Vz on fast modulated temperature. Though you still have the option to measure fast.  Later you can than sort the data towards  voltage versus temperature for the different current settings.

For heating the chip PWM on the second diode could work. The downside is that you can only measure well when the heating current is off, as they share a common pin. So one would have a heating phase and one for measuring the zener ref. and the other diode as a temperature sensor with a much lower current (e.g 100 µA) in the other direction. This could work for more moderate temperatures (e.g. 10-40 K temperature rise). The internal heater is fast, but limited power. PWM on the second diode could also introduce thermal gradients and thus mechanical stress, so this would not be the most accurate way.

For higher temperature or very accurate data an external heater would be needed. The external heater might be controlled by PWM. One might not need to have a control loop, but could do the measurements during slow heat-up and cool-down. The second diode could be used as a temperature sensor.
 

Offline VintageNut

  • Frequent Contributor
  • **
  • Posts: 534
  • Country: 00
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #143 on: November 11, 2016, 02:43:24 am »
I mounted one of the 2DW232 parts on to a banana plug, plugged it into a KE2461 forcing 5mA. I am monitoring voltage with a DMM7510.

The voltage is 6.02xxxx and wanders about 600uV p-p, 166 uV standard deviation.

I traded some PMs with chuckb. He tried forcing current into a 2DW23X part with a calibrator and saw the voltage wandering as well.

My opinion is that the instruments that can force current are not ppm stable and cannot be relied upon to maintain stable enough current for a zener reference device.

Any other opinions?
working instruments :Keithley 260,261,2750,7708, 2000 (calibrated), 2015, 236, 237, 238, 147, 220,  Rigol DG1032  PAR Model 128 Lock-In amplifier, Fluke 332A, Gen Res 4107 KVD, 4107D KVD, Fluke 731B X2 (calibrated), Fluke 5450A (calibrated)
 

Offline Vgkid

  • Super Contributor
  • ***
  • Posts: 2710
  • Country: us
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #144 on: November 11, 2016, 03:38:53 am »
What about making a precision current source? You have precision resistors, right.
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline technix

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #145 on: November 11, 2016, 05:11:39 am »
I mounted one of the 2DW232 parts on to a banana plug, plugged it into a KE2461 forcing 5mA. I am monitoring voltage with a DMM7510.

The voltage is 6.02xxxx and wanders about 600uV p-p, 166 uV standard deviation.

I traded some PMs with chuckb. He tried forcing current into a 2DW23X part with a calibrator and saw the voltage wandering as well.

My opinion is that the instruments that can force current are not ppm stable and cannot be relied upon to maintain stable enough current for a zener reference device.

Any other opinions?

You need a more stable source to compare it against, for example LM399.
 

Offline VintageNut

  • Frequent Contributor
  • **
  • Posts: 534
  • Country: 00
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #146 on: November 11, 2016, 09:38:22 am »
I mounted one of the 2DW232 parts on to a banana plug, plugged it into a KE2461 forcing 5mA. I am monitoring voltage with a DMM7510.

The voltage is 6.02xxxx and wanders about 600uV p-p, 166 uV standard deviation.

I traded some PMs with chuckb. He tried forcing current into a 2DW23X part with a calibrator and saw the voltage wandering as well.

My opinion is that the instruments that can force current are not ppm stable and cannot be relied upon to maintain stable enough current for a zener reference device.

Any other opinions?

You need a more stable source to compare it against, for example LM399.

Have you watched Dave's teardown of the DMM7510? the DMM7510 has an ovenized LTFLU as its reference. This same ovenized LTFLU is what provides the Fluke 5700A DC volts reference .

The DMM7510 can measure the 3ppm daily drift of my Fluke 731B (two of them) in my lab/office.The 731B drifts 1 ppm for every 2 or 3 degrees F.

The LM399 is what is inside my calibrated KE2000 and it is a very stable and it agrees with my DMM7510. The DMM7510 has 10X resolution measuring the daily drift of the 731B compared to the KE2000.
working instruments :Keithley 260,261,2750,7708, 2000 (calibrated), 2015, 236, 237, 238, 147, 220,  Rigol DG1032  PAR Model 128 Lock-In amplifier, Fluke 332A, Gen Res 4107 KVD, 4107D KVD, Fluke 731B X2 (calibrated), Fluke 5450A (calibrated)
 

Offline VintageNut

  • Frequent Contributor
  • **
  • Posts: 534
  • Country: 00
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #147 on: November 11, 2016, 09:41:04 am »
What about making a precision current source? You have precision resistors, right.

I do have some precision resistors. They would probably require tempco tweaking to make a better current source than what is in the KE2461. Its a good idea. That will have to go on the long list of future projects.
working instruments :Keithley 260,261,2750,7708, 2000 (calibrated), 2015, 236, 237, 238, 147, 220,  Rigol DG1032  PAR Model 128 Lock-In amplifier, Fluke 332A, Gen Res 4107 KVD, 4107D KVD, Fluke 731B X2 (calibrated), Fluke 5450A (calibrated)
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14016
  • Country: de
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #148 on: November 11, 2016, 10:20:53 am »
If would also suspect the current source, though this should not be that bad. Also EMI could be a problem, when using an external unit.
There is also a chance that there is quite some temperature effect, as the TC is low only for a suitable current and this current can be vary quite a lot from device to device. Having one with 6.0x V suggest this could be one of the more extreme ones. So if possible I would also monitor the temperature via the second diode.

The normal test circuit would be similar to the one right at the beginning of the thread.
 

Offline VintageNut

  • Frequent Contributor
  • **
  • Posts: 534
  • Country: 00
Re: Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x
« Reply #149 on: November 11, 2016, 11:29:56 am »
If would also suspect the current source, though this should not be that bad. Also EMI could be a problem, when using an external unit.
There is also a chance that there is quite some temperature effect, as the TC is low only for a suitable current and this current can be vary quite a lot from device to device. Having one with 6.0x V suggest this could be one of the more extreme ones. So if possible I would also monitor the temperature via the second diode.

The normal test circuit would be similar to the one right at the beginning of the thread.

I think that the SMU is not the largest contributor to the 100+ uV wander of the zener.

I just changed the forced current to 10mA and the average wander is now 6 uV instead of 100+ uV at 5mA.

It looks like an automated measurement will be required to characterize all 10 pcs of the 2DW232 that I will be using.

The 2DW232 mounted on the banana plug is wrapped in 3 tissues tied with a twist tie.

Another thing to mention here is that it takes 30+ minutes for the 2DW232 to stabilize. In open air, it never stabilizes.
working instruments :Keithley 260,261,2750,7708, 2000 (calibrated), 2015, 236, 237, 238, 147, 220,  Rigol DG1032  PAR Model 128 Lock-In amplifier, Fluke 332A, Gen Res 4107 KVD, 4107D KVD, Fluke 731B X2 (calibrated), Fluke 5450A (calibrated)
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf