ADR1001 - Ovenized Voltage Reference System

[magic]

But I fear for accurate readings of temperature a pull up resistor will not be sufficient.

IIRC the change in forward voltage from 1°C difference is equivalent to 7% change in bias current.

[iMo]

What is interesting - the LTspice model outputs temperature in degreesC (at least it did - see the TSET sims in this thread).

PS: a 50-100uA current source may help..

[Andreas]

PS: a 50-100uA current source may help..

Hello,

shure, but since they are connected to reference ground: how much will the additional current across the bond wires/PCB traces affect the output voltage?
Is it ~ 100 mOhm * 0.1 mA = 10 uV or more?

The remote sensing ICs normally have at least 2 different current levels to increase accuracy to 0.5 .. 1 deg C without calibrating the sensor.

with best regards

[iMo]

The current source should be stable and then the additional voltage drop on the traces will be constant too (my above sim shows it).. With a simple resistor there in sensor's collector the drop will vary based on the temperature..

[Kleinstein]

The 2 current levels would only be needed to get an absolute value. Alternatively one could use only the temerature sensor with a cold chip to get a starting point. Just looking at changes in the temperature, e.g. to compensate a TC in unheated mode could use a constant current. Chances are one could use less, like 10 µA and still have good enough a temperature signal.

[magic]

As far as I see there is no GND force/sense, all reference ground current flows through one fat trace to the two REF_GND pins.

Thermal diode current also joins this path at one point so extra voltage drop is inevitable, but not sure if you need much current there? Wouldn't 1MΩ to REF6P6 do the job?

Also, note that this trace already conducts a substantial reference transistor current, which is not stable with temperature (proportional to zener voltage). A resistor from TCHIP to REF6P6 would simply increase this current by a small fraction.

[iMo]

Simplest variant ala magic..

[kerwenard]

https://www.analog.com/en/resources/app-notes/an-82f.html

[Non-English text removed by Moderator]

[jaromir]

One cold winter day I received a message informing me that a package is underway and I should expect it in no time. Indeed, a guard in blue, red, orange and yellow striped uniform appeared and handed me an urgent package containing a powered ADR1001 reference, built by Imo, a valued member of this community.
Since I had only a few days before he returned to pick up this valuable cargo, I started a quick evaluation of this reference.

1, At first I went for noise measurement. My setup consisted of DUT, connected to LFLNA-80 preamplifier, followed by HP34401A triggered by Rigol DG822 function generator to obtain a 25Hz sample rate.
Three measurements were performed, with various DUTs - a 50R termination resistor (logged as 25SPS_n-50r.csv), then ADR1000 containing ADRMU reference from Marco Reps (logged as 25SPS_adrmu.csv) and finally, reference from Imo (25SPS_im.csv filename).
Python script was used to generate a NSD chart attached below (see 1_NSD_result.png). It is obvious that ADR1001 in this circuit (utilizing its internal scaling resistor and amplifier) does exhibit higher noise than ADR1000 with scaling resistors and an amplifier, especially at lower frequencies.

2, Then I connected the reference to Datron1281 DMM and logged voltage overnight. I measured ADR1000 based reference under similar conditions. Measured values were visualized in a chart, and for both logged charts I added a 20 point long moving average to point out the low frequency noise somehow better, see 2_6h_log.png
This measurement confirmed previous findings - ADR1001 in this circuit seems more temperamental than the ADR1000 reference.

3, Last test I performed was a rough temperature coefficient test. I left the reference enclosure thermally insulated and the loss heat from its PSU did the rest. Over the course of a few hours internal temperature slowly rose by 6 degrees Celsius from healthy 37,5 to over 43 degrees. HP34401 logged the voltage from an LM35 sensor near the ADR1001 IC, built into the reference circuit by Imo. Logging this data and the reference voltage provided me with an indication of the reference temperature coefficient. See the picture 3_tempco.png
The result is relatively noisy and it's difficult to put a good trend line over the data, but it seems like the sensitivity is a bit below 1uV per degree Celsius in this narrow temperature range (though probably wide enough for lab use).

At this point I had to return the reference box to its author. Results of the processed data are below, as well as source data in case anyone wants to play along.

[aronake]

Anyone have any kicad or other pcb design software files to share from some ADR1001 project?

[iMo]

I would only add to Jaromir's info above (Thanks!) - that myADR1001#1 reference was switched off from July 2024 to mid December 2024 (aprox 5.5 months), and switched on 2 weeks prior to the delivery to Jaromir's Labs.

My 34401A showed 9.999.890 at 23C couple of days back, and now both are switched off again.

Also I am missing the 100nF capacitor in the feedback of the ADR1001's internal output buffer, as it is indicated in the DS for 10V out (well, I built it 18 months before we got the first official DS handy  )..

Hopefully we will see more projects with that reference here soon.. 

[iMo]

FYI - you may enjoy new pcb renderings of TiN's ADR1001/REF80 compatible boards

https://xdevs.com/article/ref_ex/

[Andreas]

Hello,

I joined the club.
To gain some experiences I choosed the Evaluation Board first before I do a PCB.

First Test: Startup duration of the EVM until 1 uV stability against 34401A with 100 NPLC.
Result ~300 cycles * 4 sec ~ 1200 sec = 20 minutes warmup time.

with best regards

Andreas

[Andreas]

Next test: tilting. again with supply from USB port.
Is there any weak point due to position of heater and temperature sensor on the chip?

Result:
stange behaviour with orientation "4" = USB port on top as shown in pictuer IMG3475w.
And also the step after this orientation. (totally 2 ppm change).

The question is: is this really thermal (which I doubt due to the large "spike" at begin of orientation "4") or is it EMI-related.
I fear I have to repeat the test with battery supply.

with best regards

Andreas

[Andreas]

First 1/f noise measurement (0.1-10 Hz) in my cookies box covered with a cloth.

Battery supplied with 14V regulator LT1763 at the input of the board internal LT3045 (11V) regulator.

result of 19 measurements with 100 seconds duration each at 5V output.

Peak-Peak
average     0.892 uVpp
std dev      0.047 uVpp

calculated RMS
average    116 nVRMS
std dev     2.07 nVRMS

typical example (near average) attached

so I slightly measured above the data sheet value of 0.7 uVpp @15V (0.9 uVpp @ 11V)

The question is: is noise dependant on supply voltage (like on LT6655)?

with best regards

Andreas

[Kleinstein]

It is possible that the noise depends on the supply voltage. The temperature regulation should see more internal gain with a higher supply voltage and this can lead to better thermal regulation.
With the voltage regulator up front a different voltage can also change the temperature profile with more or less heat at the regulator. The point here is likely more the temperature gradients and not so much the absolute temperature.
The noise shown has quite some peaks. So the details of the higher frequency roll off could make a difference. Not all 0.1 - 10 Hz filters for noise measurements are exactly the same. So take the numbers with a pinch of salt.

[Andreas]

Repeated noise test with 14V at output of 11V Regulator (since I have no 15V regulator at hand).

Peak-Peak
average   0.861 uVpp
std dev    0.046 uVpp

calculated RMS
average  115 nVRMS
std dev   2.71 nvRMS

so no significant change in noise due to supply change from 11V to 14V for the ADR1001.

Supply current 14V at input   of LT3045 (11V)  29.2 mA @ room temp
Supply current 14V at output of LT3045 (14V)  21.2 mA @ room temp
So factor 1.38 current reduction for factor 1.27 voltage increase.

with best regards

Andreas

[iMo]

@Andreas: fyi - here are my 0.1 to 10Hz measurements at my 10V config and 14.3V powering.
https://www.eevblog.com/forum/metrology/adr1001-ovenized-voltage-reference-system/msg5154198/#msg5154198
Basically 2x yours..
As Jaromir wrote above, the ADR1001 is more "temperamental" than the ADR1000, it could be the on-chip resistors contribute to that noise.
Could be interesting to measure the noise at the 6.6V output (with the first internal divider disconnected) and compare that with 5V or 10V output, trying to identify how the on-chip resistors contribute.
Btw., what is your chip revision?

[branadic]

Behind the scenes TiN reported increased noise in the 5 V and 10 V configuration, while the bare zener noise is mostly on par with ADR1000, which indicates either the resistors or the amp contribte to the noise.

-branadic-

[iMo]

A pity TiN is reporting behind the scenes, and not in the section he founded.. 

Behind the scenes TiN reported ..

[Andreas]

Hello iMo,

the datecode is 2408 as you can see on the photo above:
https://www.eevblog.com/forum/metrology/adr1001-ovenized-voltage-reference-system/msg5894856/#msg5894856

I am not familiar with chip revision coding.

With noise the interesting fact is that the 5V output is specced with higher noise 0.7uVpp as the 6.6 V output (0.6 uVpp).
Normally the noise should scale with output voltage.
I guess that the output OP-Amp contributes to noise.

with best regards

Andreas

[dietert1]

Behind the scenes TiN reported increased noise in the 5 V and 10 V configuration, while the bare zener noise is mostly on par with ADR1000, which indicates either the resistors or the amp contribte to the noise.

-branadic-

Isn't that the expected relation? I think for all zener references the 10 V output will have higher noise. Don't know whether the internal opamps of the ADR1001 are of zero-drift type.

Regards, Dieter

[iMo]

Behind the scenes TiN reported increased noise in the 5 V and 10 V configuration, while the bare zener noise is mostly on par with ADR1000, which indicates either the resistors or the amp contribte to the noise.

-branadic-

Isn't that the expected relation? I think for all zener references the 10 V output will have higher noise. Don't know whether the internal opamps of the ADR1001 are of zero-drift type.

Regards, Dieter

But it [resistors/opamp] may contribute "much more", than it would be expected. The amplifier in the ADR1001 is not a zero-drift one.
There are 2 dividers - 6.6V to 5V (for 5Vout), and afterwards aprox 10k/10k for 10Vout..

You may try to measure the noise:
1. at 6.6V output
2. at first divider's 5V output
3. at 5V buffered by the opamp
4. at 6.6V buffered by the opamp
5. at 13.2V amplified by the opamp
6. at 10V amplified by the opamp
7. at the buffer/amplifier output (with its input loaded with an external resistor)
8. perhaps the noise of all 4 individual resistors in both dividers..
9. all above at various temperatures (40-120C).

[branadic]

Behind the scenes TiN reported increased noise in the 5 V and 10 V configuration, while the bare zener noise is mostly on par with ADR1000, which indicates either the resistors or the amp contribte to the noise.

-branadic-

Isn't that the expected relation? I think for all zener references the 10 V output will have higher noise. Don't know whether the internal opamps of the ADR1001 are of zero-drift type.

Regards, Dieter

Replace the word "increased" by the word "excess" to get what I meant? The noise didn't scale with amplification, but was way larger than expected.

-branadic-

[Andreas]

T.C. at 5V output of my Eval Board.

14V battery supply at Input of 11V LT3045 regulator.
0.1K/minute slope at heat spreader.
temperature measured between both VRef out 4mm sockets.

result (linear interpolation)
-94 ppb/K
no visible early drift within successive ramps.

(compare to ADR1399 measurements in LS8 package)
https://www.eevblog.com/forum/metrology/adr1399-reference/msg4879958/#msg4879958

with best regards

Andreas