Author Topic: T.C. measurements on precision resistors  (Read 271667 times)

0 Members and 4 Guests are viewing this topic.

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
T.C. measurements on precision resistors
« on: June 15, 2014, 04:19:35 pm »
Hello volt-nuts,

in this thread I will describe temperature coefficient (T.C.) measurements on precision resistors like Z201, UPW50, 8E16 ... i.e. with ageing <50 ppm/year and T.C. <= 5ppm/deg.

Motivation:

I got the idea from Mickle in the LM399 thread with his statistical divider. The idea is to select from several individual resistors a matching pair with nearly same tempco for either the temperature setpoint of a LTC1000 reference (12K5 + 1K) or a output voltage divider 10V/7V for a output buffer of a precision reference voltage. My temperature range for this is the "extended room temperature" range which is from 18 to 33 degrees C at minimum or with some reserve around 10-40 deg C.

There are many different informations regarding T.C. on the web especially for the Z201 metal foil (Z-foil) resistors. While the manufacturer gives a typical spec of 0.2ppm/deg or even  0.05ppm/deg for a limited 0-60 deg C range other "typical" measurement values from users of Z-foil resistors (VHP202Z) are more in the range from -0.3 up to -1ppm / deg for a  around 10K resistor which is not very far from wire wound resistors with 3-5ppm/deg.
https://www.eevblog.com/forum/reviews/precision-resistor-standard/msg420396/#msg420396
https://www.eevblog.com/forum/projects/precision-dc-current-source-diy/msg382583/#msg382583

So I wanted to make my own picture and do some comparisons at least with Z201 and UPW50

Also from other articles I got the impression that the metal foil resistors have from principle more hysteresis than good wire wound resistors. The metal foil is bonded do a ceramic substrate and the bond may create some hysteresis which is not available on wire wound resistors. So it is interesting for me how large the hysteresis actual is for the 10-40 degrees range. Dr. Frank reports 5ppm hysteresis for a 25 deg to 125 deg excursion for a hermetically 10K VHP202Z which is the hermetically tight upgrade of Z201. Others report soldering drift (hysteresis?) of even 50 ppm.
https://www.eevblog.com/forum/reviews/precision-resistor-standard/msg420729/#msg420729
https://www.eevblog.com/forum/projects/vishay-bulk-foil-drift-after-soldering/msg445297/#msg445297

Further questions:

Another question is wether the resistor value of a Z201 resistor has a influence on T.C.
Is there a sweet spot resistor value which really is able to be below 0.1ppm/deg for selected resistors?
So perhaps it is better to use for e.g. all values created out of 2K or 5K resistors with zero T.C. instead of using a 1K with large positive T.C. and a 10K with large negative T.C.

Humidity sensitivity:

On the other side a volt-nut friend of mine has reported a humidity sensitivity between 5-50ppm for 30% RH change of 8E16 resistors which are similar to the UPW50 wire wound resistors. I hope I will get further information on this topic from him when he is back from his trip.
My problem is that I do not have any hermetically resistors up to now so that I cannot measure the RH sensitivity with my equipment. But 50ppm is really serious compared to 1ppm/K over some 10K temperature change.

If you have any measurement data on these topics: Feel free to contribute.
The minimum numbers that I need is the type of resistor, the resistor value and the measured values preferably as temperature cycle from around 25 down to 10 up to 40 and back to 25 degrees C so that a hysteresis can be easily shown.

For my part I will need some days to post the first results and update the thread. So don't be too impatient.

With best regards

Andreas


 
The following users thanked this post: quarks, Inverted18650

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #1 on: June 15, 2014, 04:20:00 pm »
first pre tests:

The first test that I did is simply a voltage divider built of a 1K Z201 resistor and a 1K UPW50 resistor.
The resistors are soldered to 1.5m long lines (twin 0.14qmm) so that they can easily put into a temperature controlled environment.
Of cause the soldering is done at the end of the Resistor pins using aligator clips of a "3rd hand" to keep the resistor elements cool during soldering.

Because of the line length which consumes around 0.5mV (100ppm) for each line I need a 4 wire connection for each resistor.
The current path is fed from the 5V voltage reference of a 24 bit ADC.
From one to the other resistor. And finally to the star ground of the voltage reference.
The voltage path measures the voltages by a MAX4052A multiplexer which is attached to the ADC.
So I have a quasi differential measurement over the resistors.
Since the 0.5mV on the copper lines are not constant over temperature they are not usable for T.C. calculation.

The 24 bit ADC is similar to branadics hardware:
https://www.eevblog.com/forum/projects/oshw-24bit-adc-measurement-system-for-voltage-references/
The main differences are: my FT232R and the pre regulators for the reference voltage are outside of the ADC board. So the design of branadic is more compact. I also use different voltage references. But since I do all measurements in a ratiometrical manner the stability of the voltage reference has a minor effect on the T.C. measurement results in my setup. My highest quality ADCs with AD586LQ and LT1236AILS8-5 are needed for my daily ageing measurements.

First measurement was simply taking the Z201 between my fingers to heat up from room temperature (around 25 deg C) to around 35 deg C.
This gave a around +9 uV voltage increase measured over the 1K Z201 resistor (ADC_CH1 - ADC_CH2). (minute 5 to 9 in the measurement).
On the other side a warming up of the 1K UPW50 gave a -27uV decrease over the Z201.
So both measured resistors have a positive TC and the UPW50 is only about a factor of 3 worse than the Z201.



« Last Edit: June 15, 2014, 05:59:25 pm by Andreas »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #2 on: June 15, 2014, 04:20:26 pm »
measurement setup:

the whole setup is built without using pricey instruments.

IMG1625w shows an overview:
The (blue) temperature chamber on the floor is a 40 EUR (price may vary) 12V/230V (car) cool box.

Within the box there is a aluminium sheet (~1mm thick) as heat spreader for the heater foils.
(IMG1627w = bottom view).
http://www.reichelt.de/Heizfolien/THF-77110/3//index.html?ACTION=3&GROUPID=5159&ARTICLE=108462&SEARCH=THF-77110&SHOW=1&OFFSET=30&

In the middle between the foils a small NTC 33K (red legs + black line) glued to the heat spreader.
http://www.reichelt.de/Heissleiter-Varistoren/NTC-0-2-33K/3//index.html?ACTION=3&GROUPID=3114&ARTICLE=13561&SEARCH=NTC-0%2C2%2033K&SHOW=1&OFFSET=30&

Edit: the fan + plastic cap is now obsolete and is no longer used (see below)
On the other side (IMG1635w) there is a low noise fan (low power) under a plastic cap (CD-spindle against direct cold draft + dripping water from the cooler).
http://www.reichelt.de/Luefter/FAN-ML-6015-12-S/3//index.html?ACTION=3&GROUPID=6215&ARTICLE=110414&SEARCH=FAN-ML%206015-12%20S&SHOW=1&OFFSET=30&

The resistor under test is mounted under the air current of the fan together with 2 additional NTCs to sense the temperature of the resistor rather than the aluminium sheet. The fan shall keep the NTCs and the resistor at the same temperature.
Unfortunately there is no possibility to place the sensor within the resistor. So when ramping up + down the temperature there will be always small differences between actual and meaured temperature.

On the desk in foreground from left to right:
ADC3 (grey box) measuring the NTC (27K pull up) of the heat spreader. The black cable is a USB/RS232 converter connected to the laptop.
ADC18 (grey box) with a MAX4052A multiplexer attached measuring the resistors with 4 channels (4 wire pseudo differential).
ADC14 (grey box) with a MAX4052A multiplexer with 27K multiplexed pull up measuring the 4 NTCs of the 2 resistors under test.
All 24 bit ADCs (LTC2400) are isolated with photocouplers on RS232 line and battery supplied to keep mains noise away.

The measurement setup uses only ratiometric measurement. So all reference voltages cancel out.
On the resistors under test the offset of the ADC is also canceled out by the pseudo differential measurement.
The noise of the ADC (around 10uVpp for a single measurement) is reduced by averaging over 1 minute to around 1uVpp.
The LTC2400 does a self adjustment of offset and full range at every conversion so it is very stable over temperature compared to other 24 bit converters. With the LTC2400 a temperature controlled environment is not necessary.

On the right: Instrument to measure the battery state before and after measurement.

Middle left:
2 Power stages (VNP5N07 + 1K series at the gate + 100K pull down) for the heater foils.
http://de.rs-online.com/web/p/transistoren-mosfet/3133080/
The heater foils are connected in series to a 17-18V unregulated DC supply (not visible)
 (transformer 2*12V/30VA with middle tap, MBR20100CT schottky, 4700uF/35V Cap)
The power stages are connected to the RTS-lines of the USB/RS232 converters and can be.
The 3rd USB/RS232 converter is connected to a 230V SCR relay (not visible) for the cooler.

Background left:
Cardbox with temperature controlled (27.5 deg C) reference resistor between 2 heat spreaders with heater foil
together with 2 NTCs. See also IMG1631w.

The cooler is switched on for negative slopes below 32 deg C and switched off at 15 deg C for rising slopes.
The temperature control is done by the heaters.
The temperature range is around 10 - 45 degrees. (10 degrees only below 25 degrees room temperature).

The laptop is generating temperature setpoints calculating the temperature control loops + sampling all measurement data as average values over 1 minute.
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Enhancements to the measurement setup.

1)
the most essential tip came from Emmanuel:
https://www.eevblog.com/forum/projects/t-c-measurements-on-precision-resistors/msg600033/#msg600033

Previously I had the body from the resistor within a metal block and contacted the wires outside the block.
But this gives unequal temperatures between the 2 wires (thermoelectric) and additional to the body.
(different hysteresis curve when placing resistor face down or face up).
I tried a 12V fan for temperature equalizing but with little positive effect.

The most heat transfer from / to the resistor is through the wires.
So it is most essential to keep the wires at the same temperature.
My approach from 14.02.2015 on is with aluminum sheets + silicone foil (see pictures).
https://www.eevblog.com/forum/projects/t-c-measurements-on-precision-resistors/msg615638/#msg615638
https://www.eevblog.com/forum/projects/t-c-measurements-on-precision-resistors/msg637341/#msg637341
The whole setup is now in a shoe box (card box) within the car cooler to keep air drafts away.

2)
AC-multiplexer
See also:
https://www.eevblog.com/forum/projects/t-c-measurements-on-precision-resistors/msg693460/#msg693460

This was a approach to eliminate completely the thermocouple voltages.
I tried this with low ohmic resistors where it works good.
But with higher ohmic resistors 12K and 70K the filter capacitors which
I need to compensate EMI noise from the long lines had a bad effect on the measurements
since they have to be completely charged/discharged every measurement..
With the enhancement 1) the thermocouple voltages play nearly no role,
so I discarded the AC-Multiplexer.

3)
For the 70K resistors the noise level of the measurement increased dramatically.
Usually the LTC2400 has around 10uVpp noise level.
But with the 70K resistors the average level increased to about 50uVpp
with single events far above the average level.
Increasing the EMI capacitors was useless:
So I changed from unshielded wires to shielded lines and grounded
the heat spreaders of reference and DUT resistor.
With shielded lines + EMI filters the noise level is somewhere between
12-18uVpp as average level with the 70K resistors.

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------

error estimation:

resistor self heating:
--------------------------
since there are no values of thermal resistance I use the power derating curves for a worst case estimation.
self heating of the resistor is relative constant over temperature so only the whole curve is shifted.
-> has only (a small) influence on the 25 degree related values.

Z201/S102 1K self heating
---------------------------------
Z201 derating 0.3W / 50K -> 6mW/K
power 1K 6.25mW -> 1K
due to thermal grease / fan a reduction of factor 3-5 can be estimated
-> resulting 0.2 - 0.33 K error


UPW50 1K self heating
-----------------------------
UPW50 derating 0.5W / 20K -> 25mW/K
power 1K 6.25mW -> 0.25K
due to thermal grease / fan a reduction of factor 3-5 can be estimated
-> resulting 0.1 - 0.05 K error


NTC-02 33K self heating with 27K pull up @ 5V reference
------------------------------------------------------------------------
Datasheet: dissipation factor 7mW/K

max power (power match): 2.5V @ 27K = 0.23 mW
-> heating with 1:4 multiplex for NTC1-NTC4 = 8.3mK average
will be further reduced due to thermal grease.


Error due to T.C. of reference resistor Z201#1
----------------------------------------------------------
Temperature is controlled to 27.5 degrees C nominal.
There are slight variations (or ADC noise) during measurement of up to 0.05 K.
Together with the T.C. of 0.9 ppm/K this gives a
maximum error of 0.05 ppm for the whole temperature span.





to bee continued...
« Last Edit: April 25, 2018, 04:23:34 am by Andreas »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #3 on: June 15, 2014, 04:20:53 pm »
Z201 results:

Datasheet values:
typical +/-0.2ppm/K +/-0.6ppm/K max. spread from -55 .. +125 deg C
typical +/-0.05ppm/K from 0 .. 60 deg C

http://www.vishaypg.com/docs/63187/zseries.pdf

first pictures: explanation + further evaluation will follow

----------------------------------------------------------------------------------

Z201 #1 1K 0.01% datecode B0940-

to measurements of 14.06.2014 with Z201 #1 with 0.3 K / minute slope

Regression curve

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 = -5.7873783799925560E+0000
A 1 =  8.9822403793771305E-0001
A 2 = -1.0237801410870104E-0002
A 3 = -1.6531589738371225E-0004

max. deviation to regression curve:  1.6041069332224885E+0000 ppm

So measured T.C. at 25 degrees is  0.9 ppm / K. (= A1)

----------------------------------------------------------------------------------

Z201 #2 1K 0.01% datecode B0940-

measurement of 27.07.2014 with 0.12 K / minute slope

T.C. with box method: -0.27 ppm / K  (-9.75 ppm / 36 K)

Regression curve

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 = -8.21920230407688E-0002
A 1 = -2.59088412027236E-0001
A 2 = -4.62391535615482E-0003
A 3 =  1.60378673959659E-0004

max. deviation to regression curve: 1.58584984755668E+0000

So measured T.C. at 25 degrees is  -0.26 ppm / K. (= A1)

----------------------------------------------------------------------------------

Z201 #3 1K 0.01% datecode B0940-

----------------------------------------------------------------------------------

Z201 #6 1K 0.01% datecode B1305-

----------------------------------------------------------------------------------

Z201 #7 1K 0.01% datecode B1315-

« Last Edit: March 19, 2015, 06:04:09 am by Andreas »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #4 on: June 15, 2014, 04:21:22 pm »
UPW 50 results:

Datasheet values:
- typical +/-3 ppm/K  (0..85 deg C)
- max. +/-5 ppm/K (-55..125 deg C)

http://www.te.com/commerce/DocumentDelivery/DDEController?Action=showdoc&DocId=Data+Sheet%7F1773299%7FB%7Fpdf%7FEnglish%7FENG_DS_1773299_B_UPW-0312.pdf%7F1624323-6

--------------------------------------------------------------------------------------------------------
UPW50 #1 1K 0.1% Datecode 0816 resistor value @25 deg about -100 ppm against Z201#1

First results from 17.06.2014

on 17.06.2014 the 2 NTCs are fixed to the UPW50 on opposite sides (top + bottom) with tape.
The resistor is taped to the heat spreader directly above the NTC which is used for the temperature controller of the heating foils.
The Fan is installed. (But obviously I have used too much tape).

There is a relative large difference between the temperature of the 2 NTCs visible.
see: 20140617_UPW50_Tdiff_NTC2_NTC1.PNG
And the difference is also dependant on temperature slope.
The difference between +0.3K/minute and -0.3K/minute is up to 0.8K between the 2 NTCs.
Whats the real temperature of the resistor then?
It is not possible to tell wether the observed hysteresis of the resistor curve
is really from the resistor or if it is simply a temperature measurement error.
So for hysteresis I will have to improve the measurement setup thermal coupling (see discussion below).

For the T.C. measurement the thermal coupling has less influence,
since I have waiting times at the temperature extremes for temperature settling.
So the final results will not significantly differ from that what I have up to now:

Picture: 20140617_TC_UPW50_1K_1_raw_temp.PNG
For the UPW50 1K sample #1 this is after the box method:
   dev (ppm)   temp   (deg C)
min   -42.04121928   8.4112
max   32.44987977   44.5383
diff   74.49109905   36.1271
TC = 2.061917482

The measured curve is not straight linear with slightly decreasing T.C. at higher temperatures.
So the T.C. after box method is dependant on measured temperature span.

I calculated a 3rd order regression curve after normalisation to 25 deg C.
See picture 20140617_TC_UPW50_1_LMS.PNG
So the A1 coefficient is the linear part of T.C. at 25 degrees.

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg)

A 0 = -7.1365771688644140E-0001
A 1 =  2.0716957976478127E+0000
A 2 = -2.2325669247677248E-0002
A 3 =  1.1149432124690903E-0004

max. deviation to regression curve:  1.4751894259493142E+0000

So measured T.C. at 25 degrees is  2.1 ppm / K.

23.06.2014: further results from 20.06.2014:
change in setup: UPW50 mounted between 2 stacked SK09 heat sinks with thermal grease for the UPW50 + 2 NTCs.
NTC1 is between bottom heat sink and UPW50. NTC2 is above UPW50. Some silicone foil between upper and lower heat sink for better thermal contact. Ramp is 0.3 K / minute.
The difference between NTC2 and NTC1 is much smaller between rising and falling temperature.
But no significant effect on UPW50 hysteresis.


further results from 21.06.2014:
same setup as on 20.06.2014 only setpoint ramp changed to 0.12K / minute.
The "hysteresis" did not change significantly.
If there was a thermal settling time I would have expected a factor 2.5 lower hysteresis for UPW50.
But this is not visible in the measurement.

T.C. curve coefficients:

A 0 =  7.5167802876286093E-0001
A 1 =  2.0586430917904010E+0000
A 2 = -2.2608480477597587E-0002
A 3 =  1.6552949180267930E-0004

max. deviation  1.6919232571042099E+0000 ppm


further results from 22.06.2014:
same setup as on 20.06.2014+21.06.2014 only setpoint stepwise starting from 30 degrees (over night) then
3 hours 25 degrees,
2 hours 8.5 degrees (not reached),
3 hours 25 degrees,
2 hours 46 degrees,
4 hours 25 degrees.

still a remaining hysteresis of around 1.3 ppm (zoomed picture)

By the way: NTC1 below UPW50 shows some overshoot on setpoint changes. NTC2 lags somewhat. So the true temperature might be somewhere inbetween.

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------

measurements of 29.06.2014
-------------------------------------
Against 21.06.2014 2 isothermal blocks where used.
UPW50 #1 is mounted in the brass cube 40mm of branadic.
NTC1 in a hole below the UPW50.
NTC2 near the wire of the UPW50 in the 6.5 mm dia hole for the UPW50
Thermal grease is used. Fan is within the plastic cap.

the Z201 is mounted in a smaller aluminium 19.5x19.5x40mm together with NTC3+4
Also with thermal grease.

The results are calculated with linearisation correction for the NTCs.
They do not differ much from all previous results (except for NTC temperature tracking).
So why all the effort? (I hate thermal grease).

Resistance: about -100 ppm @ 25 deg C against Z201 #1

Average T.C. from box method:

Drift UPW50 #1
min -38.96856779
max 33.85651927
diff   72.82508706

temperature
min 8.778501769
max 45.22255957
diff   36.4440578

T.C. = 1.99827054

so average T.C. from box method = 2.0 ppm/K

Regression curve:

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 = -1.1959881148839021E+0000
A 1 =  2.0982250580817275E+0000
A 2 = -2.3221150504742209E-0002
A 3 = -5.7552059855277272E-0005

max. deviation to regression curve:  1.7295248354935827E+0000

So measured T.C. at 25 degrees is  again 2.1 ppm / K. (= A1)
Hysteresis not changed against previous measurements.

--------------------------------------------------------------------------------------------------------
UPW50 #2 1K 0.1% Datecode 0841 resistor value @25 deg about -580 ppm against Z201#1

Measurement of 17.08.2014: (slow ramp 0.12 deg / minute)

Average T.C. from box method: +0.59 ppm/K

Regression curve:

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 = -2.80998469524063E-0001
A 1 =  6.29693967898549E-0001
A 2 = -1.70300759394405E-0002
A 3 =  1.32133545088887E-0004

max. deviation to regression curve: 2.02996451023940E+0000 (hysteresis + noise)

So measured T.C. at 25 degrees is  +0.63 ppm/K

So UPW50#2 has only 1/3 of the T.C. of UPW#1

--------------------------------------------------------------------------------------------------------
UPW50 #3 1K 0.1% Datecode 1220 resistor value @25 deg about -100 ppm against Z201#1

Measurement of 23.08.2014: (slow ramp 0.12 deg / minute)

Average T.C. from box method: +1.85 ppm/K

Regression curve:

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 =  5.45175269191540E-0001
A 1 =  1.95948365674911E+0000
A 2 = -1.56801576131755E-0002
A 3 = -1.03623042954573E-0004

max. deviation to regression curve: 2.02918100035813E+0000  (hysteresis + noise)

So measured T.C. at 25 degrees is  +1.96 ppm/K

--------------------------------------------------------------------------------------------------------
UPW50 #4 1K 0.1% Datecode 1345  resistor value @25 deg about -50 ppm against Z201#1

Measurement of 28.08.2014: (slow ramp 0.12 deg / minute)

Average T.C. from box method: +4.76 ppm/K

Regression curve:

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 = -9.06731926559255E-0001
A 1 =  5.09683804579340E+0000
A 2 = -2.83692334469616E-0002
A 3 = -4.77182333446752E-0004

max. deviation to regression curve:  2.41485072237296E+0000  (hysteresis + noise)

So measured T.C. at 25 degrees is  +5.10 ppm/K

This candidate is also one (the first wire wound) where a ageing drift of about 1.2ppm in 3 days is visible.
Maybe due to the high T.C. value.

« Last Edit: August 28, 2014, 08:22:35 pm by Andreas »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #5 on: June 15, 2014, 04:22:27 pm »
Metal film resistors (for comparison only).

----------------------------------------------------------------------------------------------------------------

UPF50 results (seems to be relatively new)

Datasheet values:
- T.C.: +/- 5 ppm/K (20..80 deg C)

unfortunately there is no data for shelf life.

http://www.te.com/commerce/DocumentDelivery/DDEController?Action=showdoc&DocId=Data+Sheet%7F1773299-1%7FA%7Fpdf%7FEnglish%7FENG_DS_1773299-1_A_UPF-1113.pdf%7F2176163-6

---------------------------------------------------------------------

UPF50 #1 1K 0.1% Datecode CA051  resistor value @25 deg about +490 ppm against Z201#1

Measurement UPF50 #1 of 31.08.2014: (slow ramp 0.12 deg / minute)

Average T.C. from box method: +1.81 ppm/K

Regression curve:

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 = -1.11413825943835E-0001
A 1 =  2.01228648499642E+0000
A 2 = -2.18642911697272E-0002
A 3 = -2.09357556163436E-0004

max. deviation to regression curve: 9.57837118906888E-0001  (hysteresis + noise)

So measured T.C. at 25 degrees is  +2.01 ppm/K

This is the first time that I have a resistor with no visible hysteresis. (hysteresis if any is below noise level).
There was a warm / cold / warm cycle on 30.09.2014 after soldering.
---------------------
Measurement UPF50 #1 of 01.09.2014: (fast ramp 0.3 deg / minute)

Average T.C. from box method: +1.84 ppm/K

Regression curve:

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 = -3.19004540786116E-0002
A 1 =  2.02623371441288E+0000
A 2 = -2.12673046301004E-0002
A 3 = -2.45915701944138E-0004

max. deviation to regression curve: 1.55865402178475E+0000  (hysteresis + noise)

So measured T.C. at 25 degrees is +2.03 ppm/K

So with fast ramp some hysteresis shows up for the first temperature cycle (heating from min to max temperature).
The two following cycles have lesser hysteresis.
---------------------------------------
Measurement UPF50 #1 of 02.09.2014: (slow ramp 0.12 deg / minute)

Average T.C. from box method: +1.82 ppm/K

Regression curve:

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 =  3.30940300218205E-0001
A 1 =  2.00605814295572E+0000
A 2 = -2.15233506217060E-0002
A 3 = -2.15338666959163E-0004

max. deviation to regression curve: 9.80511753771691E-0001  (hysteresis + noise)

So measured T.C. at 25 degrees is +2.01 ppm/K

And again nearly no hysteresis with slow temperature ramp.

---------------------------------------------------------------------

UPF50 #2 1K 0.1% Datecode CA051  resistor value @25 deg about +310 ppm against Z201#1

Measurement UPF50 #2 of 03.09.2014: (slow ramp 0.12 deg / minute)

Average T.C. from box method: +2.17 ppm/K

Regression curve:

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 =  2.17091994721987E-0001
A 1 =  2.35167500256426E+0000
A 2 = -2.13936252691839E-0002
A 3 = -2.85364735533639E-0004

max. deviation to regression curve: 1.37799986290797E+0000  (hysteresis + noise)

So measured T.C. at 25 degrees is +2.35 ppm/K

UPF50 #2 shows some hysteresis in the cold cycle.
Against UPF50 #1 I did only a warm cycle on 02.09.2014 after soldering.

--------------------------

Measurement UPF50 #2 of 04.09.2014: (fast ramp 0.3 deg / minute)

Average T.C. from box method: +2.15 ppm/K

Regression curve:

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 =  3.98056255275583E-0001
A 1 =  2.36173851561087E+0000
A 2 = -2.16714481631818E-0002
A 3 = -2.90793789458481E-0004

max. deviation to regression curve: 1.36220327821289E+0000  (hysteresis + noise)

So measured T.C. at 25 degrees is +2.36 ppm/K
Again some (moderate) hysteresis.

-----------------------------------------

Measurement UPF50 #2 of 05.09.2014: (slow ramp 0.12 deg / minute)

Average T.C. from box method: +2.13 ppm/K

Regression curve:

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 =  5.39247317441195E-0001
A 1 =  2.33157020163661E+0000
A 2 = -2.18089184079106E-0002
A 3 = -2.32087081335189E-0004

max. deviation to regression curve: 1.15652373887686E+0000  (hysteresis + noise)

So measured T.C. at 25 degrees is +2.33 ppm/K
Again some (low) hysteresis.

----------------------------------------------------------------------------------------------------------------

RC55Y results: (is not really high precision but one of the best easily available metal film)

Datasheet values:
- T.C.: +/- 15ppm/K (20..70 deg C)

Shelf life typ: 300ppm/year
Shelf life max: 1000ppm/year

http://www.welwyn-tt.com/pdf/datasheet/rc.pdf

----------------------------------------------------------------------------------------------------------------
RC55Y #1 1K 0.1% no Datecode  resistor value @25 deg about -250 ppm against Z201#1

Measurement RC55Y #1 of 06.09.2014: (slow ramp 0.12 deg / minute)

Average T.C. from box method: -8.13

Regression curve:

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 = -3.80443706117962E+0000
A 1 = -8.88999969660920E+0000
A 2 =  5.61290695727147E-0002
A 3 =  1.10751465297150E-0003

max. deviation to regression curve: 3.57710183792049E+0000  (hysteresis + noise + ageing)

So measured T.C. at 25 degrees is -8.89 ppm/K

relative large ageing drift (4-5ppm) during measurement.

-----------------------------------------

Measurement RC55Y #1 of 07.09.2014: (slow ramp 0.3 deg / minute)

Average T.C. from box method: -8.26

Regression curve:

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 = -9.90226686827984E+0000
A 1 = -8.80002275616066E+0000
A 2 =  5.95676322397589E-0002
A 3 =  6.15950680297529E-0004

max. deviation to regression curve: 5.87999450789868E+0000  (hysteresis + noise + ageing)

So measured T.C. at 25 degrees is -8.80 ppm/K

relative large ageing drift (8-9 ppm) during measurement.

-----------------------------------------

Measurement RC55Y #1 of 08.09.2014: (slow ramp 0.12 deg / minute)

Average T.C. from box method: -8.10

Regression curve:

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 = -1.28036495163400E+0001
A 1 = -8.77294237200453E+0000
A 2 =  5.15156773448553E-0002
A 3 =  1.05699451708786E-0003

max. deviation to regression curve:  2.63539353853565E+0000  (hysteresis + noise + ageing)

So measured T.C. at 25 degrees is -8.77 ppm/K

total ageing drift over 3 days: around -14 ppm.


« Last Edit: September 08, 2014, 09:13:09 pm by Andreas »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #6 on: June 15, 2014, 04:22:52 pm »
reserve 8E16
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #7 on: June 15, 2014, 04:23:21 pm »
Vishay S102JT results  (C-alloy)

datasheet values
- typical +/-2 ppm/K +/-2.5ppm/K max. spread
- max  ( = +/- 4.5ppm/K max ????)

http://www.vishaypg.com/docs/63001/63001.pdf

-----------------------------------------------------------------------------------------------
S102JT #1 1K 0.01% date code B1312-

Measurements of 13.07.2014:

From the Z201 datasheet: "Vishay Foil Resistors’ new Z-Foil technology provides an order of magnitude reduction in the Bulk Metal Foil element’s sensitivity to temperature changes - both external and internal."

When I compare the T.C. of Z201 and S102J then I really ask me: what do they really mean?
Ok I do not test at the temperature extremes. And also the hysteresis is much larger than on every tested resistor up to now.
In my temperature range T.C. is almost zero but with a large hysteresis of +/-5.5 ppm

Deviation against Z201#1 near zero ppm.

Now its getting difficult. Up to now I could define the "box method" as min/max drift span divided by min/max temperature span.
With this definition I will get: 11.1 ppm / 37.2 K = 0.3 ppm / K.
If I calculate with the 3 cardinal points method the T.C. gets around doubled values.

Regression curve

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 = -2.11289497717753E-0001
A 1 = -1.63006512501367E-0002
A 2 = -1.19211117177579E-0002
A 3 =  2.81642336236479E-0004

maximum deviation to regression curve:  5.49062470812308E+0000

So measured T.C. at 25 degrees is  -0.016 ppm / K. (= A1)

So the main improvement seems to me that the packaging or bonding (hysteresis) has been largely improved on Z201. But T.C. at least on these samples got worse.

Measurement 14.07.2014 on S102 #1: temperature setpoint jump from 21-30 degrees.

shows that the thermal time constant is very short.
The hysteresis is issued by another (2nd) much slower time constant.

-----------------------------------------------------------------------------------------------
S102JT #2 1K 0.01% date code B1312-

Measurements of 16.07.2014:

this guy has similar hysteresis than #1 but additional positive T.C. of around 0.55ppm/K

Deviation against Z201#1 @25 deg around -65 ppm

T.C. with box method: +0.55 ppm/K

Regression curve

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 = -2.98885401103763E-0001
A 1 =  5.45546004290231E-0001
A 2 = -1.30854146011424E-0002
A 3 =  4.17914762203182E-0005

maximum deviation to regression curve:  6.25321618704226E+0000 (hysteresis +(noise))

So measured T.C. at 25 degrees is also +0.55 ppm / K (= A1)

Hysteresis on #2 (+/- 6.3 ppm) is slightly higher than on #1 (+/- 5.5ppm)

« Last Edit: July 16, 2014, 08:55:48 pm by Andreas »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #8 on: June 15, 2014, 04:23:47 pm »
UPW25 results:

Datasheet values:
- typical +/-3 ppm/K  (0..85 deg C)
- max. +/-5 ppm/K (-55..125 deg C)

http://www.te.com/commerce/DocumentDelivery/DDEController?Action=showdoc&DocId=Data+Sheet%7F1773299%7FB%7Fpdf%7FEnglish%7FENG_DS_1773299_B_UPW-0312.pdf%7F1624323-6

I got 2 UPW25 1K last week (couldnt resist since they were cheeeap at RS)
and made a isothermal block for them this weekend.
So here is the result of the first of them:


UPW25 #1 1K  0.1% date code 1244 measured on 08.07.2014

Resistance: about +630ppm @ 25 deg C against Z201 #1

T.C. with box method: -4.047

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg)

A 0 = -1.89847058482001E+0000
A 1 = -4.27204490196377E+0000
A 2 = -1.35428364585836E-0002
A 3 =  8.08056748472276E-0004

Max deviation: 2.59646041441657E+0000 (hysteresis)

I did not expect such a large deviation of the values against UPW50.
Since it is the same manufacturer.
Ok the UPW25 are much "newer" than the UPW50 from my drawer.
Temperature gradient is doubled in amount.
And this is the first resistor measurement with negative T.C.
The curve is relative "linear" against the others.
Hysteresis is slightly increased.

At 25 degrees the gradient is around -4.3K/deg

----------------------------------------------------------------------------------------------

UPW25 #2 1K  0.1% date code 1244 measured on 10.07.2014

Resistance: about +10ppm @ 25 deg C against Z201 #1

T.C. with box method: -4.117

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg)

A 0 =  2.17582422582792E-0001
A 1 = -4.29694066234232E+0000
A 2 = -1.36897923948147E-0002
A 3 =  7.03952866822879E-0004

Max deviation: 3.67844904416664E+0000 (hysteresis)

So hysteresis is again higher than UPW50 #1 and UPW25 #1

At 25 degrees the gradient is around -4.3K/deg


« Last Edit: July 12, 2014, 03:48:28 pm by Andreas »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #9 on: June 15, 2014, 04:24:06 pm »
Riedon USR2 results

Datasheet values:
- typical +/-3 ppm/K  (-55..+125 deg C)
- max. +/-5 ppm/K (-55..+125 deg C)
- upon request +/-1ppm/K (0..60 or -55..+125 deg C)

http://www.riedon.com/media/pdf/USR2-0808.pdf

I got two 1K 0.01% TCR1 versions (whatever this means)

http://www.digikey.com/product-detail/en/USR2G-1KX2/USR2G-1KX2-ND/1650022

Oh I see "country origin" GERMANY

So the resistors had to travel from Germany to Minnesota and back to get them.

---------------------------------------------------------------------------------------------

USR2 #1 1K  0.01% date code 1309 measured on 03.08.2014

Due to the "sticker" for the resistor value which makes the total thickness of the
resitor larger than that of the usual 8mm housing I had to use some force to
put the resistor into the slot of the isothermal block.
I do not know how the influence on T.C. is for the mechanical stress.
So perhaps I have to repeat the measurement.

Its also the first time that I see a large "ageing" on the measurement.

T.C. with box method: -1.2 ppm / K  (-43.2 ppm / 35.9 K)

Regression curve

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg C)

A 0 =  3.41653005510780E+0000
A 1 = -1.09885912220308E+0000
A 2 =  1.12101896460950E-0002
A 3 = -2.66093117429204E-0004

max. deviation to regression curve:  7.49589680302222E+0000

So measured T.C. at 25 degrees is  -1.1 ppm / K. (= A1)
(or somewhat higher if I did not have the ageing).

Update from 08.08.2014 for USR2 #1

I have done a measurement outside the iso thermal block on 08.08.2014 just to see wether the sticker has some influence.
There was no significant difference to the previous measurements (06.08.2014) except some higher "hysteresis" due to difference between sensor and resistor.
During ageing the T.C.  (A1 coefficient) decreased in amount from -1.1 down to -0.95 ppm/K

I have added a ageing curve.
Directly after soldering the resistor had +13 ppm against my "reference" Z201.
6 days later the value was nearly 42 ppm -> a change of 29 ppm within 6 days.

Question: is this ageing or is this humidity change of a resistor after freshly opening the polyethylene bag?

---------------------------------------------------------------------------------------------

USR2 #2 1K  0.01% date code 1309 measured on 09.08.2014 + 10.08.2014

Resistance: about  -90 ppm @ 25 deg C against Z201 #1

T.C. with box method: -1.354

T.C. Curve = A0 + A1 * t + A2 * t * t + A3 * t * t * t   with: t = (Temp - 25 deg)

A 0 = -1.41707170509333E-0001
A 1 = -1.41352317040293E+0000
A 2 =  9.45352247510392E-0003
A 3 =  4.40837707667485E-0005

Max deviation: 2.21010836378330E+0000 (hysteresis)

so T.C. at 25 degrees is -1.4 ppm/K which is somewhat higher than "TK1"

Ageing drift seems to be much smaller than on USR2 #1.
Only the first "fast hysteresis" cycle on 09.08.2014 shows about 7-8 ppm shift.
But this candidate came out of the same polyethylene bag than #1
so it had now one full week to acclimate.

---------------------------------------------------------------------------------------------
to be continued

« Last Edit: August 11, 2014, 01:08:46 pm by Andreas »
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #10 on: June 15, 2014, 04:24:26 pm »
reserve 3
 

Offline Vgkid

  • Super Contributor
  • ***
  • Posts: 2518
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #11 on: June 16, 2014, 03:17:30 am »
Looking forward to your results.
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline quarks

  • Frequent Contributor
  • **
  • Posts: 821
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #12 on: June 16, 2014, 04:30:47 am »
Will be interesting,  thanks for sharing
 

Offline jpb

  • Super Contributor
  • ***
  • Posts: 1622
  • Country: gb
Re: T.C. measurements on precision resistors
« Reply #13 on: June 16, 2014, 12:56:44 pm »
I too am looking forward to reading your results as you produce them.
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 415
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #14 on: June 16, 2014, 04:14:06 pm »
I was pointed to this thread by DiligentMinds, there are a lot methods and questions to be answered.  I will address some of them here.

Hysteresis of resistor characteristics can be caused by a number of effects and also depends on the type of resistor being used.  Hysteresis can be caused by temperature excursions, power cycling, humidity and soldering.  A correctly made precision wire wound resistor (PWW) exhibits near zero hysteresis effects from the aforementioned effects.  Large temperature excursions (- 55°C to +125°C) whether by thermal shock (5 cycles) or power, should cause very small changes in the resistor value, somewhat less than 15 PPM.  A correctly made PWW should be essentially immune to humidity without hermetic sealing (if it is, this is a result of poor design, i.e. foil resistors for example).  Soldering should cause no significant shift in value if done according to specification, no more than 10 seconds at normal soldering temperature.  Any resistor which shows significant shifts in value from any of these effects are flawed by design or its own limitations.  Soldering limits are different for foil resistors, they do exhibit a higher sensitivity to soldering temperatures due to their very small mass and inability to handle high heat for even brief periods.  Check the data sheet for the particular foil resistor to find the soldering specification.

There is no 'sweet spot' in resistors, their characteristics are determined by the materials and manufacturing processes used to make them.  Any given type of resistor has limitations, no perfect resistor can be achieved unfortunately even though everybody chases after one and tries to specify them, in practice it cannot be done, despite what Vishay's marketing BS may state or anybody else's who claim otherwise.

Personally, I am very wary of using statistics to come up with a 'typical' value for most resistors, the fact is that any given TCR specification does have a +/- limit over a given temperature range and that TCR will vary between those limits in a limited random manner.  I have not found any method of accurately predicting just how many resistors in a given batch will produce a given TCR range.  The tighter the specification, the more likely that selection has to be used to find those resistors, no matter what may be said to the contrary.  Metallurgy cannot control the parameters of a given batch of alloy that closely, I've talked with metallurgists in the industry and they all agree with that statement.  The only way to get a given batch of alloy close to zero TCR is by a hit and miss heat treating procedure which is used on all resistor alloys.  The difference here is that this hit and miss procedure can produce alloys with near zero TCR but it is costly to do and few if any alloy suppliers will agree to this procedure.  To my knowledge, it must be done in-house and that costs a lot.  The current level of heat treating by the alloy supplier does produce a fairly consistent yield of a given TCR range but the tighter the TCR specification, the lower the probable yield of a given TCR range and you are going to pay a higher price for that tighter TCR spec, this is true of foil or wire forms.

Primary resistors standards, such as the SR-104 uses this procedure to tweak the alloy down to very low TCR values and you all know how expensive an SR-104 is.  This procedure is not the only way to do it, given a really good resistor design, a very good yield of low TCR resistors can be had without the expensive in-house heat treatment.  What this means is that for any given batch of resistors, a certain percentage will yield low TCRs which must be selected out of the batch.  This does cost in labor but is cheaper than the other procedure and actually takes less time.

In the real world, a resistor should be able to operate under any given conditions within its specifications without suffering significant changes to it characteristics.  Yes, there will always be some effect produced, particularly from extreme conditions (i.e. maximum specified limits) but those effects should be minimal in a good resistor design.  With the exception of primary standards, a resistor should not have to be handled with kid glove care like a standard to maintain its specifications.  Unfortunately, foil resistors are closer to standards handling in this respect than not.

Foil resistors still, after decades of development, have a 'wavy' TCR curve, even though they have been able to bring that 'lumpy' curve down to a pretty darn good level, it is still not linear by any means.  Vishay has a habit of drawing graphs which has ridiculous ranges to compare against, for instance, a resistor with a TCR of say 0+/- 1 PPM/°C will have its TCR drawn on a graph with a 0+/- 100 PPM/°C range, making the graph of the resistor appear far flatter than it really is.  This illusion has been used by Vishay for many years to confuse customers.  If your TCR is good, then draw a graph that shows it in detail, not some intentionally obfuscated nonsense.
 
The following users thanked this post: Echo88, cozdas, Muxr

Offline Vgkid

  • Super Contributor
  • ***
  • Posts: 2518
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #15 on: June 16, 2014, 04:48:50 pm »
Glad to see your input E.G Pettis.
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 415
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #16 on: June 16, 2014, 06:47:08 pm »
Thank you, Vgkid, I've worked with resistor alloys, i.e. precision resistors since before 1975.  I've worked with some of the best people in the industry in precision wire wound resistors.  I am the co-author of the articles mentioned in the thread:

http://www.edn.com/design/analog/4427151/The-last-half-century--Wirewound-resistors-Part-one

http://www.edn.com/design/analog/4427940/The-last-half-century--Wirewound-resistors-Part-two

I am also the co-designer of the precision wire wound resistors I allude to here in comparison to other resistors.  The PWW resistors I co-designed are, in my opinion (which is backed up by extensive testing by other laboratories) are the best precision wire wound resistors.  The specifications are real world, not best case marketing hype.  I do not claim a typical TCR, that tends to be mostly statistical manipulation, they are specified by a range of TCR which is guaranteed by design.  I will not claim a TCR of 0.2 PPM/°C as typical because it isn't.  What I do claim in my resistor specifications is that the standard product TCR is 0 +/-3 PPM/°C which is easily and repeatably achievable and in any given batch of resistors, approximately 60% of the yield will be within 0 +/- 1 PPM/°C.

What I cannot predict with any certainty is how many of those resistors will be within a given range of TCR within the +/- 1 PPM/°C yield.  If a customer wants a sub-PPM TCR, it will have to be selected from the given batch, pretty much like anybody else.  The exception here is that I don't have to fiddle around with any heat treating, time consuming and expensive alloy tweaking like everybody else.  You will, of course have to pay a bit more for the selection process, but then, you are already doing that with the other suppliers and getting sub-par resistors.

You want long term stability and reliability?  My resistors were put through 50 (not 5, but 50!) thermal shock cycles by a major aerospace contractor, according to MIL-STD 202, -55°C to +125°C, most resistors changed less than 10 PPM (that's everybody else's shitting on a shelf spec) with zero failures, which no other resistor has ever achieved.  You can expect similar changes running at full rated power and slightly higher changes if you're going to run full power at 125°C.

There are other differences between a good wire wound and a good foil resistor.  Foil does have very low parasitics  compared to PWWs, however the parasitic card tends to be overplayed; except at very low values, the 'Q' of wire wound resistors is well below 1 which means that they have far less effect in a circuit than would be thought.  Yes, PWWs do have parasitics like all components, but depending on the application, they can be dealt with and in many instances, they actually have little or no effect.

Noise is another area where a good PWW excels, wire wound resistors do not produce shot noise, it is not inherent in wire but foil resistors do have shot noise and tend to have slightly higher Josephson noise levels than wire wounds, this is because the etching of the foil paths causes irregularities along the edges of the paths, this causes more noise because of the increased irregularity in the electron movements along the 'rough' edges.  It is a lot like water flowing in a stream, even if the bed of the stream was perfectly smooth, the rough edges of the stream would cause irregular eddies in the water.

Foil resistors have inherently higher drift than PWWs (well, at least mine), maybe even sitting on a shelf, even sitting in hermetically cans.  I don't make hermetic resistors, they are a waste of time and money, they are only a band-aid for a poor design.

Foil resistors are affected by humidity in their common molded packaging, hence the hermetic can to protect the foil inside.  The humidity, if it gets inside and onto internal connections or foil, directly affects the metal, with oxidation occurring which affects the characteristics.  It is also possible that the bonding agent between the foil and ceramic substrate absorbs water as well.  In a correctly designed and built PWW, humidity has exceedingly little effect on the resistor, the wire is coated with Pyre ML which is good to 220°C to 240°C, well above most resistor encapsulation ratings and does not easily absorb water (and if the resistor is under power, what little humidity may get in is evaporated by the heat) plus the wire is welded (well, at least mine are) in which case water will have little effect there as well and the weld joints are encapsulated too.  This has been verified by tests in a humidity chamber, both under power and not powered.

I could go into more details about why other PWWs don't seem to work as well as one would think but that is another lengthy story, suffice it to say that their resistor designs are flawed, hence the less than stellar performance.

Vishay does make good resistors but their predilection for buttering up the specifications is annoying at best and borders on lying at the worst.  They just need to cut out the marketing hype and publish real world specifications (which sometimes are buried in the fine print somewhere).  While Caddock may not seem to make as good as Vishay foil resistors, I find their specifications much more closer to the real world with little hype thrown in.  As a design engineer, it makes it a bit easier when you know just what you are working with instead of what statistical manipulation produces.  Sorry if I sound like I'm beating up on Vishay, in some ways I am, but it is mostly of their own making and it is ingrained in the company.

One more thing, the 'lumpy' TCR curve of foil resistors is inherent in the design, you should have seen the original TCR curves 30 or 40 years ago, it would make you seasick.  My TCR curves are linear across the entire temperature range (well, there is a slight change at the extremes, but then what wouldn't?), it is part of the design.  Whatever TCR you get from a foil, it is going to be 'lumpy' and that may give your circuits fits, it has in the past.
 
The following users thanked this post: Muxr

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #17 on: June 16, 2014, 09:23:32 pm »
Hello,

I am very surprised on the resonance to my experiment.
Sorry that I will not read+answer all posts at the moment.
I am still busy with doing measurements and generating the calculations for the results.
One measurement has around 300-800 data points each averaged out of 1 minute of the ADC values to reduce the noise of the ADC.
So one measurement lasts for up to 1 Day + several hours evaluation.
Sometimes I have to repeat the measurement if it is obvious that something went wrong ...
My wife is very tolerant, but all has its limits.

I am also not shure if it is a good idea to work with (possibly hand picked) samples which are not easily avaliable from stock of common suppliers.
How can I be shure that this would not give a wrong picture? And how can other hobbyists get those resistors?

I also still do not know if my setup will work with higher ohmic resistors or if I will have to do some changes due to the input impedance of the LTC2400.

so long

Andreas
 

Offline Vgkid

  • Super Contributor
  • ***
  • Posts: 2518
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #18 on: June 16, 2014, 10:39:12 pm »
For E.G. Pettis, and others.
What are your thoughts on the precision vintage resistors on ebay.ERC/NRC/Mepco/Ohmite.
What about the various standard resistors(you might have some insight, others can chime in)?
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 415
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #19 on: June 17, 2014, 12:29:43 am »
Hello Andreas,

To quote you:  "I am also not sure if it is a good idea to work with (possibly hand picked) samples which are not easily available from stock of common suppliers.  How can I be sure that this would not give a wrong picture? And how can other hobbyists get those resistors?

I also still do not know if my setup will work with higher ohmic resistors or if I will have to do some changes due to the input impedance of the LTC2400."


My, you do seem to be a bit skeptical.....let's see if I can put your mind at rest. 

1. All of my resistors are custom made to order, even my 'standard' grade resistors are made to order, I generally do not have very many 'stock' resistors lying around.  While some manufacturers do make a point of having some 'standardized' values in stock, I don't, my order turn around time is faster than most with the exception of stocked values on a shelf somewhere.  I do not charge a lot more for making a 12.5K resistor instead of a 10K resistor, that is just silly, the resistor price should be a function of the actual costs, a 12.5K resistor only costs a little bit more in wire and a few more seconds of labor than a 10K resistor given the same bobbin style.  You are being taken advantage of by some(?) resistor sources.  Non-stocked resistor values.....quite laughable, from where I sit, that is just an excuse to charge more money than necessary.

2. If I have the materials in stock to make a given resistor order, I do not usually put any minimum on the order, you pay the quoted price plus shipping and you've got your resistors.   Unlike many manufacturers, I will not make a customer buy some unwanted quantity or quote some sky-high price for the part(s) just because you are a small quantity annoyance.  The only exception is if I do not have the required materials in stock, particularly the wire, I am usually forced to buy a certain minimum amount which is often in the hundreds or even thousands of dollars, in which case I will either suggest another resistor type that I do have or I will simply refuse the order with an explanation why.

3. My 'standard' resistor turn around time, given materials stock, is about 5 days.  There are various options which will lengthen that out some by days, not weeks like the other guys.  Number two above is the only exception to number three.  The 'standard' part is 0 +/- 3 PPM/°C TCR, see my earlier post for more details.  If you want tighter TCR than that, they will have to be selected from the batch and depending on how tight that TCR is determines a given yield.  This of course involves more time and does carry a price point with it.  Not to put too fine a point on it, I do not select resistor TCR unless the customer specifically asks for it and agrees to the price.  If a customer is asking for free samples, they will get the 'standard' resistor, not any hand picked part, it takes too much time and effort to cull out lower TCR parts to give them away for free.  The samples are to verify the major characteristics of the parts, unless the customer is a potentially large user of 'special' parts (say NASA, etc.,) and requests a special part for qualification, then and only then would I supply 'selected' parts.  This is pretty much standard policy among resistor manufacturers.  I am sympathetic to hobbyists as I am one too, which is why I am willing to provide reasonable samples under the right circumstances to hobbyists (no I can't give you a set of resistors to build your prototype but I am willing to provide them at a competitive price).

4. I can enhance the stability (which is already better than just about anybody else's) even further if required, naturally that will take a little longer and add a bit to the price but usually hobbyists do not really need that much stability.  (I am a bit curious that hobbyists think that they need state-of-the-art components that are rather expensive, this is usually left to places like R&D labs, the military or aerospace).  I can see wanting to try and make a really ultra stable voltage reference, etc., but wanting it and needing it are two different things.

Andreas, I am capable of producing resistors over a very wide range, none of the resistors you have mentioned in your posts are unusual, I do not understand your last sentence above.  If I do send you any samples, they will be 'standard' grade, there will be no 'hand selection' of the TCR.  Since many people in these various posts and threads are talking about wanting sub-PPM TCR resistors, it may not be of any use to provide you with samples that do not meet these 'ultra' requirements.  RhoPoint and any of the other wire wound resistor manufacturers named here do not provide sub-PPM TCR resistors and if they say they do, they're not only fooling you but themselves.  They have enough trouble getting under 3 PPM/°C to 5 PPM/°C TCRs, anything much less than that is by accident.  You didn't seem to have any problem with testing their sub-standard parts so why are you questioning mine?

I have some reservations about the measurement methods being used in these threads, there are uncertainty terms, for instance, which do not seem to be observed.  Even ratio measurements have uncertainties, even short term measurements have uncertainties.  I have seen the HP 3458A referred to in several threads being used as a ratio measurement, it does have very good linearity but linearity is not accuracy, there are still uncertainties to be observed.  Additionally, the smaller the delta temperature deviation, the greater the likely error in the TCR measurement, particularly when attempting to measure small sub-PPM TCR values.  Just because the error appears to be repeatable, doesn't make it accurate.

I also have some reservations about your ADC test setup as well, from your statements, it sounds like you have to use a lot of math on those 'measurements' due to various influences such as noise, I am very leery of measurement results that use statistics to iron out garbage.  It calls into question the results accuracy.  Unless the known accuracy plus the known uncertainty is at least 5 times better than what you are trying to measure, your measurements are invalid, even if they repeat.  For example, if you are trying to measure a 0.2 PPM/°C TCR and you delta shift the part (an accurate) 30°C from ambient for an expected 6 PPM in resistance, your system accuracy plus the uncertainty must be no bigger than 1.2 PPM but that still leaves a fairly large error possibility, the 6 PPM reading could be off by +/-20%, it actually might be as much as 7.2 PPM or as little as 4.8 PPM, that is a very large error in fact when trying to measure a 0.2 PPM TCR and it just gets even worse as you try to measure even smaller TCRs.

Granted, a hobbyist does not usually have access to a high quality resistor bridge but that method coupled with an appropriate delta temperature range gives a more accurate and repeatable TCR measurement.  I have an ESI 242D (yes, they are very expensive too), one of the most accurate resistor bridges made, the only method which is more accurate than a 242D is a Direct Comparison Current bridge and those are real heavy weights in the price category.

I am not trying to throw cold water on these projects, on the contrary I think they are quite commendable and a valuable teacher but I still question a hobbyist's need for such component performance.

If you want to know more about my resistors you can ask here or I can tell you how to contact me more directly.
 
The following users thanked this post: Svgeesus

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 415
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #20 on: June 17, 2014, 01:04:29 am »
Hi VgKid,

Most of the old vintage 'precision' resistors do not meet today's more rigorous specifications and most, if not all of the vintage resistors have likely drifted out of tolerance after all these years.  Of course, the resistors are still eminently usable if you don't mind them being out of their stated tolerance.  Their other specifications are likely to still be as good as when original, but beware that TCRs can creep up or down over time with these old parts.  Alloys were not as good as today's are and were not as stable so don't expect TCRs much better than 10 PPM/°C at best.  The original Evanohm alloy was originally specified at 0 +/- 20 PPM/°C which dates back to being introduced in 1943.

The vintage power resistors are just as good as today's for the most part and while I wouldn't pay any extra for them, they have no drawbacks to using them.

For those of you who may be thinking of buying an old General Radio or early ESI resistor decade box, these also suffer the same fate, they may have drifted out of tolerance over the years and the really old ones will have higher TCRs than the newer 5 PPM/°C models today.  As long as you are not expecting these great old vintage pieces to perform like new, they deserve a place on the display shelf at the very least and may be good enough for your purposes to boot.
 

Offline quarks

  • Frequent Contributor
  • **
  • Posts: 821
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #21 on: June 17, 2014, 09:51:33 am »

Granted, a hobbyist does not usually have access to a high quality resistor bridge but that method coupled with an appropriate delta temperature range gives a more accurate and repeatable TCR measurement.  I have an ESI 242D (yes, they are very expensive too), one of the most accurate resistor bridges made, the only method which is more accurate than a 242D is a Direct Comparison Current bridge and those are real heavy weights in the price category.


If you want to know more about my resistors you can ask here or I can tell you how to contact me more directly.

Hello Edwin,

great to have you here.

About ESI 242D, when I searched for the best possible accuracy it was top of my list. But unfortunately when I tried to buy one, I ran into trouble with missing stupid European Union CE mark on the mains powered DC Generator 801. So I tried to "make" my own kind of 242 around a beautiful ESI Kelvin Ratio Bridge 240C. My first setup with a Burster high precision Decade and a DC Calibrator as Generator showed very good results. Later I tried to find RS925D but only got a RS925A. And finally looked into Fluke 8508A.

Can you suggest a setup to do a (in your opinion) propper TCR measurement with "242" and Fluke 8508A?
Because the absolute value accuracy will not be that important, can you judge if the Burster Decade (with very good <1ppm TK) will be the better choice than RS925 in the "242" setup?

If I find enough spare time I would like to check/compare different technologies (wire wound, VPG Z-Foils, and other goodies in my collection).
 
Also I would like to know more about your resistors. If you have a link, datasheets, where to buy, ... please share.

bye
quarks
« Last Edit: June 17, 2014, 10:48:37 am by quarks »
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 415
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #22 on: June 17, 2014, 02:32:47 pm »
Hello quarks,

Thank you, glad to be here, I was not aware of this forum until recently.  I usually only visited the professional forums and after being told about this one, I thought I could be of assistance.

It is difficult to find any ESI 242D (or earlier versions) here in the USA as well and prices for even used beat up units have become quite high.  A brand new 242D can still be had by special order, the price is about $30.000,00; you would be lucky to find a decent condition used one for under $10.000 plus some hefty shipping charges.  While I do see an ESI RS925 for sale now and then, it isn't very often any more and a 925D can set you back a good $4.000,00; a RS925A is still a very good decade standard, while it can't be calibrated like the 'D' version, it still has excellent accuracy.

I agree, many of the EU regulations are just plain stupid without much thought.  The original ESI 801 generator/detector is not the greatest and they tend to break down more often than one would like.  The DC power supply in the 801 is a very simple, unregulated DC supply circuit with various current limiting resistors in series with the output.  It is junk, its output varies from 600V at 6 mA on the high range (>100K ohms) to 2 volts at 2A on the 1 ohm range.  There is a variable rheostat which is used to "adjust the approximate power" going to the bridge.  Nothing complicated in the design, because it is a bridge, the voltage does not have to be regulated but should be clean of noise.  It is very important that you use the lowest POWER setting to get a stable resistance reading, too much power will cause a drift in the readings and a reading should be taken as quickly as possible.

The foregoing is standard calibration procedure, you may want to check a resistor's characteristics at something higher than minimum power levels.  The problem here is that your standard resistor (such as the RS925) will be dissipating the same power as your resistor of interest which is generally not good for the standard, even though it is rated to dissipate a few watts spread over its decades (this is spelled out in the RS925's operating manual).  The increased power will cause the standard to drift some as well, hence the reason for minimal power levels.  You can use a higher power level if it is done as quickly as possible, it will not damage the standard (within its ratings).

The 801 detector is about average in performance, you can substitute a better null detector such as a Fluke 845A or Keithley 155 or other similar null meter.  The HP 419A is also better than the ESI but I prefer the other two over it.

The Fluke 8508A is an excellent DMM, in many instances I have been able to calibrate Flukes to much better accuracy than the specifications would indicate and they are very stable instruments.  Over the mid-range of resistance, a properly calibrated Fluke could outperform an RS925A in accuracy although the lowest range may not be quite as good.  The use of a DMM with a resistance bridge setup would limit its usefulness to setting the bridge voltage (no real accuracy needed there) and as a null detector which I am not fond of using a digital instrument for that function as the digit flipping is annoying and may not give you as good as an analog null detector reading.  It can be more difficult to 'find' the best null with a digital DMM.  The ESI 240C and RS925A forms the most important parts of a Kelvin bridge and it is there that your accuracy in resistance readings resides.  The power supply is not critical but a good analog null detector is, you are 85% of the way to where you need to be with the two Kelvin bridge components you already have, in conjunction with your Fluke for the higher resistance readings, you are very well set for good accurate readings.

I am not familiar with the Burster Decade but if it has an actual TCR of 1 PPM/°C then it would be better than the 925A in that respect.  As I recall, the 925A is specified at an initial accuracy tolerance of +/-20 PPM.  The RS925D can be calibrated to as good as +/- 1 PPM except on the lowest and highest ranges where it can be as loose as +/- 3 PPM.

One other slightly less important detail, I know it is difficult to control temperatures in the typical hobby environment but at least knowing the temperature reasonably accurately and tracking it can help minimize those pesky errors that creep into measurements.  Measurements, particularly TCR, require accuracy in temperature as well as resistance to give an accurate result and as I have stated elsewhere, the lower the TCR being measured, the greater the demands on measurement accuracy become and the wider the temperature shift must be to accurately measure the TCR.

I am going to post again a bit later here about the TCR values being talked about here on several threads.

Currently, I have some restrictions on producing resistors, I am limited to the material stock on hand for the time being.  There has been substantial price increases in many of the materials necessary to manufacture these resistors.  I must wait until I can purchase additional materials which will likely cost in the thousands of dollars unfortunately.  I sell directly at this time with no distributors as yet.  Distributors add to the buyer's cost for one thing and I do not operate like most other resistor houses.  I also ran into a major problem with my bobbin supplier who apparently decided they didn't want to bother with making bobbins any more and upped the prices to the "I don't want to bother with you (or anybody else)" level, in other words, they priced the bobbins so high that they are too expensive, it causes the resistor prices to be too high to even compete.  So I am working off of my inventory and working on a solution which is going to take some time.

If I can make the resistors I would certainly quote them, otherwise I would explain why I can't.  I will post more details shortly, have some other things to attend to, thank you for asking and watch for further posts soon to answer your other questions.
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 415
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #23 on: June 17, 2014, 06:25:05 pm »
Hello gazelle,

Yes I can make dividers with matched TCRs, they are in the form of individual resistors with an option of encasing.  After some study of the LTZ1000/A, the basic circuits given in the LTZ1000?A have nominal values which can be varied to improve LTZ1000/A performance, of course these values are not shelf stock, it is also possible to improve the performance even more by modifications to the basic circuitry.  As I've mentioned elsewhere, all the resistors I make are, by definition, custom made, but I don't indulge in the pricing games other resistor houses do.  I charge only for the cost of the resistor by specification and the other related costs.  Just for your information, I do have stock on the materials required to make all of the precision resistors in the LTZ1000/A circuits including matched TCR performance for the dividers.

I suppose this is as good as any time to post the notes concerning these resistors and associated TCR requirements.  I have seen a lot of posts fussing over extreme low TCR values, sub-PPM values, this is foolish for two reasons; first is the cost of sub-PPM resistors and secondly, a voltage divider, in this LTZ1000/A application in particular, does not require sub-PPM TCR resistors to achive the needed ratio performance.  What is needed are resistors whose TCR track each other closely, the absolute TCR is of no real concern only the matching TCR.  When you are not trying to get absolute  'zero' TCRs, the resistors not only become easier to obtain but are also lower in cost as it is easier to match higher TCRs than sub-PPM TCRs.  While sub-PPM TCR matching is often called for, few circuits really need sub-PPM TCRs as many other circuit components will cause bigger errors than the resistors.

As such, I can provide matched TCRs in these voltage dividers, no matter the values, to sub-PPM matching where it really counts.  My standard TCR specification is 0 +/- 3 PPM/°C but in the case of these dividers, it is essentially irrelevant.  I have known a fair number of engineers who have made this same mistake, over specifying a parameter, such as TCR because they did not fully understand the function of the circuit.  The other non-divider resistors should have reasonably low TCRs, but again, extremely low TCR specification is unnecessary.

I believe that by tweaking the nominal values and the circuit configuration in the LTZ data sheet, even better performance can be achieved but I will leave that up to the person doing the design work.  Proper layout of the circuit is also critical, particularly for the external components, any resistors that need to track any other resistor should be mounted as close to each other as possible and shielded from air currents, even a little air movement can cause uneven deviations in temperature, a Styrofoam cup works really well and can't be beat for cost.  There are two options here, one is to put the resistors inside a small case and encapsulate them in a thermally conductive material or apply a thermally conductive material to the resistors on the circuit board.  The latter costs the least to do.  To anticipate a question, no, hermetic resistors are not really any better, they will have a longer 'tail' as we say, the response to changes in temperature are much slower, both in warming and cooling, unless the resistors are inside the same can and close together, if you check on pricing, this will cost you dearly for this option.

The best thing to do is to talk with your (potential) resistor supplier about your application and what options and costs are available.  Most often, the best solution is not to be found on a shelf or in a high priced component or exceptional specifications.  Just because the 'big boys' did it one way doesn't mean that is the only way to do it and get the same results.
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 415
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #24 on: June 17, 2014, 06:37:03 pm »

As promised, my General Resistance Specification:


 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf