Ultra Precision Reference LTZ1000

[TiN]

Let's see some ppm's. 
Looking forward, making me desperate to find time to build few of my own...

[Andreas]

Let's see some ppm's. 
Looking forward, making me desperate to find time to build few of my own...

Keep calm,
before we can get into ppm´s I need to verify if all is running smooth.

LTZ1000A#5 (with all LTC2057)

First picture:
CHA (blue) 14V regulated voltage LTC1763
CHB (red)  current (as voltage over R1 = 120R)
CHC (green) voltage on base of heater transistor (J6 pin 12)
CHD (yellow) zener voltage (U1 pin 3)

Its amazing: only 200ms full heater current
Steady state reached after around 500 ms.

But whats that: the zener current (red) decreases with rising temperature.
(starting with 511mV down to 435mV).
Now its getting clear why HP uses a 111R resistor with the higher temperature setpoint.
Its just to compensate the reduced base emitter voltage at higher temperatures.

2nd picture
CHC (green) voltage on buffered output (J5 pin 1)
Obviously the chopper (LTC2057) has some startup delay.
I can live with that.

3rd picture
Start delay from "ON" signal
CHC (green) on signal (J5 pin 1)
around 17ms until the zener current rises.

4th picture like 1st with pwm signal on "On" signal

5th picture like 2nd with pwm signal on "on" signal
so sometimes the startup delay of the buffered output
adds some extra voltage on top.
Thats a bit annoying.

With best regards

Andreas

[Andreas]

Hello,

and the same set of pictures for LTZ1000A#3
with LT1013A for the LTZ1000A and only the buffered output with LTC2057

1st + 2nd picture: obviously the heater makes a rest during startup when the zener current overshoots.
3rd picture: buffered output with startup delay of chopper.
4rth picture: around 16ms delay from on signal to zener current rising.
5th picture: heater reaction to pwm on on signal
6th picture: buffered output with pwm: in this case there is always annoying overshoot on the buffered output voltage.

with best regards

Andreas

[chuckb]

Attached are some measurement I gathered from a new LTZ1000 and an aged LTZ1000A. The chips were in an environmental chamber from -50 deg C to +150 deg. I have not seen the data altogether in one spot like this so I decided to post it. Testing details are in the last attachment.

The heater resistance and the Zener resistance both seem to have a 4000 ppm / deg temp co. So maybe the resistance is all from the aluminum traces on the die. The kelvin voltage sensing of the Zener voltage by the base of Q1 (at the center of the die) may remove the resistance of one of the Zener traces. The next step is to activate part of the chip to understand that.

With a passive chip the Zener voltage changes at high bias current because of chip heating. I will need to activate the heater to control chip temperature while checking voltage at higher currents (20ma).

Suggestions for other testing are welcome.

[ltz2000]

Attached are some measurement I gathered from a new LTZ1000 and an aged LTZ1000A.

Great contribution!

Suggestions for other testing are welcome.

Zener + Q1 from -50 to +150 and back. I know, not easy to do, requires a stable current source and another LTZ1000 to compare with...

[Andreas]

Hello chuckb,

thanks also from my side.
For me the zener voltage in a built up cirquit at around 50 deg C would be interesting.
And the Q1 base voltage with the standard cirquit. (120 Ohms / 70K)

I have just recorded some operation points on my 2 built up references.

What I am concerned about is that there is a large difference in Q1 base voltage
between your measured value at 100uA and my totally built up cirquit.

The Q2 base values at 50 deg C are about the same 530 vs 550 mV @ 100uA.
But the Q1 base values differ totally.
Around 440mV in the complete cirquit on my side. (resulting in 3.7mA zener current).
And around 580 mV on your measured value with base + collector shorted.
(but this should give around equal values).
So did you test the VBE voltage already in a complete cirquit?

To the measurements:
most of them are done with a 4.5 Digit multimeter which is around -0.13% off (shows too small numbers).
The values show that all regulation loops work as intended.
At a direct short cirquit at the buffered output the zener voltage changes by around -1mV. (ground line impedance).
Total current with short cirquit is 50 mA so short cirquit current at output is around 30mA (which corresponds to the typical LTC2057 datasheet value).

With best regards

Andreas

[Galaxyrise]

Attached are some measurement I gathered from a new LTZ1000 and an aged LTZ1000A.

Thanks! nice to have someone else's (way better) data to compare against!  Since my previous post on measuring Q1 temp co, I tried again with a different approach and a lot more logging.  The number that fit my new data best was -2.22mV/C, but I know there's still error--I need a proper thermal chamber.  Your data, at 100uA from 25C to 0C gives -2.2mV/C; matches my new number very closely!

[chuckb]

What I am concerned about is that there is a large difference in Q1 base voltage
between your measured value at 100uA and my totally built up cirquit.

The Q2 base values at 50 deg C are about the same 530 vs 550 mV @ 100uA.
But the Q1 base values differ totally.
Around 440mV in the complete cirquit on my side. (resulting in 3.7mA zener current).
And around 580 mV on your measured value with base + collector shorted.
(but this should give around equal values).
So did you test the VBE voltage already in a complete cirquit?

When TiN check the LTZ1000 in his K2002 he had 60mv difference between the Vbe of the heater transistor and the current transistor.
https://www.eevblog.com/forum/testgear/keithley-2002-8-5-digit-dmm-review-and-teardown/msg482280/#msg482280

I just checked 8 operational LTZ1000As running on a burn in pcb. The circuit is from the LTZ data sheet but the resistors are all cheap.
One had 30mV difference between the Vbe of the heater and the current control transistors. The current control voltage (Q1) is lower than the temperature sensor (Q2). This is just the opposite of what I measured in an environmental chamber when I just checked one component at a time. I will have to triple check my wiring and run it active.

Two LTZs on the burn in pcb had 60mV difference.
Three had 70 mV difference.
One had 100mV.
One had 110 mV.

So your difference of  90mV is in the middle of my distribution and similar to what Tin measured (60mV) on the K2002. 
Sorry my data caused you a concern but we will learn either I messed up or there is a very interesting thing going on inside the chip when it's active.
I'm leaving on a 10 day vacation in a few days so I may not get to much resolved in the next few weeks.

[chuckb]

Well the setup wiring is ok so that leaves a strange operation of the chip.

The Vbe of Q1 is 20 mv greater than Q2 when they are all check by themselves. In a full, active circuit the Vbe of Q1 is 30-110mV below Q2. That's something to look into.

Have a great new year!

[Andreas]

Well the setup wiring is ok so that leaves a strange operation of the chip.

The Vbe of Q1 is 20 mv greater than Q2 when they are all check by themselves. In a full, active circuit the Vbe of Q1 is 30-110mV below Q2. That's something to look into.

Have a great new year!

Hello chuckb,

thanks for checking.
and also thanks for the link to TiNs measurement values.
He has also 440 mV on the current regulator inputs with slightly higher temperature setpoint.
Now I am confident that my cirquit is working as it should. (even with only 3.7 mA).

Together with the statement of Ken
it would explain that the Vbe on Q1 is higher than on Q2 with base + collector connected.

 The two transistors in the LTZ are *not* matched-- Q1's geometry is significantly smaller than Q2, and so much less current will be going through Q2 than the Zener when the operation point reaches equilibrium. 

So a possible explanation would be that the hfe of Q1 is significantly higher than that of Q2 for the active cirquit.
And the large stray of 50 mV from your samples in active cirquit could indicate that the hfe is very different between samples.

I think I should get a socket for LTZ1000A to check this. (too late for the current 2 references).

I wish you also a great new year!

Andreas

[TiN]

Happy volts everyone!
I'll be back home 3rd Jan, so I can do some more measurements, just let me know what exactly.

My plan is to build one module with socket and wirewound resistors to test my chips (three LTZ1000CHs from 1990 and one LTZ1000A from 3458A A9 board (with popcorn noise). I can also compare with my three older modules with VPG resistors.

[Andreas]

Hello,

the first ppm measurements. (TiN will have to wait until after his holydays) 

The resistor sensitivity measurements are done by simply paralleling a 1% resistor to the existing 0.01% resistors to get at least around 10ppm Uz change.
The values are corresponding to the results of Lymex/Janaf with except of R3 which seems to have a larger stray and is somewhat higher in my case.
So I will have to keep a eye on this resistor.

As I got some requests, I will post more, but give me a couple of days.

Thanks for posting the Lymex data. If we multiply my data by 100, for 100ppm resistor changes we get:

-	Datasheet	Lymex	Janaf
R1	1	0.14	-0.14
R2	0.3	0.4	-0.4
R3	0.2	0.03	-0.07
R4/R5	1.0	0.95	+1.2

A couple of conclusions:
- It seems the R1 and R3 values are not as critical as given in the datasheet. Or both Lymex and I are wrong.
- Put your money on R4/R5 and R2 while R1 and R3 are less critical.

Thanks to Janaf I have selected the resistors in a manner that the more uncritical 70K resistors have the higher T.C.
And the R4/R5 resistors are paired with a maximum T.C. difference of 0.2ppm/K.

As a side effect I get the T.C. measurement of the unheated zener which is 53 or 56 ppm/K in my case.

with best regards

Andreas

[Kleinstein]

The different size of the transistors does not mean a different amplification factor. It just means the smaller one needs less (emitter-) current to get the same U_BE voltage or at the (nearly) same current as in the normal circuit the smaller transistor has the higher U_BE.

The direction of change with the resitors changed should be the same in all cases.

[Andreas]

The different size of the transistors does not mean a different amplification factor. It just means the smaller one needs less (emitter-) current to get the same U_BE voltage

Hello,

That is exactly that what I wanted to say.
Smaller area gives larger Vbe voltage in the diode connected case.
(collector + base connected)

The different behaviour in the datasheet cirquit might be a matter of hfe.

with best regards

Andreas

[Andreas]

Hello,

before mounting all the thermal isolation stuff and the inner shield I am doing preliminary 1/f noise tests.
I use a 4th order bandpass 0.1Hz-10Hz with 10000 fold amplification and a analog scope for this.

LTZ#3:
evaluation of 22 measurements 0.2uV/Div 1s/Div gives:

Average   1.097272727
Stddev     0.122146775
min           0.92
max           1.4  (outlier? all others below 1.26uVpp)

so this corresponds to the datasheet.
A typical sample with 1.08uVpp (img2244) attached.

LTZ#5

first also measured with 0.2uV/Div
but found that there is massive popcorn noise which exeeds the screen.

Average   1.325263158
Stddev     0.228297323
min          1
max         1.68  (= maximum screen heighth)

so did 2nd measurement with 0.5uV/Div

Average   1.206521739
Stddev     0.263052461
min          1
max         1.9

Pictures attached also (2268 + 2276 are with 0.2uV/Div / 2290 + 2295 with 0.5uV/Div)
one sample typical without popcorn noise the other with popcorn noise.

So now I have a problem:

Where does the popcorn noise come from.
Is it the LTZ
Or is it the LTC2057 perhaps together with the removed EMI capacitors?
This is the first LTZ1000 where I see this.
And it is the first time that I use no LT1013.

Any opinions or own measurement results for the LTZ?

With best regards

Andreas

[TiN]

What is your amplifier setup? I'd like to join for this too. Also even for preliminary test i'd recommend having everything in cookie metal can to avoid airflows. Those can also cause big jumps.

[Kleinstein]

The modified circuit with all the extra caps is more sensitive to OP noise, at the higher frequencies. In principle this is good, as at higher frequencies the OPs noise should be lower that that of the reference itself.

However the low frequency part (e.g. < 10 Hz) is still controlled by the transistor from the LTZ1000, so even if the OP is producing such noise, the LTZ would suppress the noise quite a lot (up to about 200 fold) - this is also why the noise (and drift) of the OP is not critical in the original circuit.

My suggestion would be to try a slightly higher current (e.g. test with a resistor in parallel to the 120 Ohms) - the zener diode might work better at a different current.

[3roomlab]

@andreas, i have same "pop corn" problem before, and i didnt quite know the way to solve it.

i had so much of it when i did "wrong" stuff to my K2015. the 2 things in my case that gave the problem is SOT23 BJT, due to my poor soldering work (mechanical problem), the leads are stressed. if i tap on the SOT23 lightly with a insulator stick, it can trigger/reproduce the noise. the 2nd is temperature problem, when ever anything warm up over approx 41C, readings/log also show popping more often. it appears (or at least to me) when i tap around on parts i have soldered before, if it aid in the noise, it showed that i either did it wrong or the part is not good post soldering/warm-up, and that become my method of "checking". all the parts that show popping early or after warming, i threw away . maybe the problem in your case could be similar ? maybe some leads is accidently "supplied" some stress and need to resolder?

[Andreas]

Hello,

What is your amplifier setup? I'd like to join for this too. Also even for preliminary test i'd recommend having everything in cookie metal can to avoid airflows. Those can also cause big jumps.

Sorry, did not mention that. Have a look into my cookie box:
Without this you will have only 50Hz line frequency hum.
The noise amplifier in the background has its own aluminium case within the cookie box.
And also a thermal shielding of the pre-amplifier.
The LTZ carries its preliminary thermal shielding.
(also on the solder side of the PCB).
The setup actually shows heater noise voltage measurement.

Attached is the schematics of the filter amplifier.
The input electrolytics are selected for < 20uA nA leakage current.

Caution you cannot connect a unbuffered LTZ reference to the low impedant input,
which is necessary to get a low current noise of the input stage.
(this will degrade the LTZ/introduce a large hysteresis because the heater setpoint is set to infinite temperature).

So you have to remove the jumper when attaching the voltage reference.
After 1-2 minutes when the input capacitors are charged you have to install the jumper for the measurement.

You need a oscilloscope with 2mV/div (or at least 5mV/div) sensitivity with DC coupling if you want to measure a LTZ.
Otherwise you will need a further amplification stage.

My suggestion would be to try a slightly higher current (e.g. test with a resistor in parallel to the 120 Ohms) - the zener diode might work better at a different current.

I will try this (I have a UPF 1K resistor with 2 ppm/K)  but this would prove that the LTZ itself has a problem.
I fear that the popcorn noise is gone at the current working point and will re-appear at another. (eg over temperature).
So in this case I would opt to replace the LTZ.

it appears (or at least to me) when i tap around on parts i have soldered before, if it aid in the noise, it showed that i either did it wrong or the part is not good post soldering/warm-up, and that become my method of "checking". all the parts that show popping early or after warming, i threw away . maybe the problem in your case could be similar ?

Also a idea at least worth a try.
But it might be difficult to tap on parts in a cookie box.
I will look how I can manage that.

In the mean time I tried to measure the noise on the heater voltage (J6 pin 12 = base of heater transistor).
AC-coupling by a 10uF foil capacitor to the 1Meg input of the scope.
10 kHz bandwith limit on the oscilloscope.

LTZ3 has larger heater noise
than LTZ5 (the one with popcorn noise at the output).
But most of the noise is from the oscilloscope.

With best regards

Andreas

Edit: changed uA to nA

[Kleinstein]

The base of the temperature controller is mainly reproducing the noise of the reference, there should be only minor traces coming from the temperature noise. The better place to look at temperature noise would be the output of the OP.

Just for noise testing even a more normal resistor would be OK to test a different current. Its allways the possiblity to have a certain current or temperature  the zener diode does not like. So it might be interesting to do a slow temperature ramp (e.g. 1 degree), while doing a longer noise measurement.

One could also do slight changes in current or temperature by changing R2 and R3 - these are the less critical ones. So this wold likely be the point where one could couple in the ramp. The collector of the current sensing transistor is changing at about 400 mV/K with an impedance of some 60-70 K. So something like a 1 M resistor to a 0-10 V ramp would result in a 1.6 K ramp.

[chuckb]

Andreas
I like your low noise amplifier design. Capacitive coupling of all the stages should make the overall design easier to operate.

Unstable capacitor leakage currents can look like popcorn noise in the Zener. In AN124 Jim Williams took special care with the coupling capacitor to reduce capacitor noise. "Selected commercial grade aluminum electrolytics can approach the required DC leakage although their aperiodic noise bursts (mechanism not understood; reader comments invited) are a concern." Page 7 of the app note, Note #1.

I have had better luck using 35V (or higher voltage) Aluminum electrolytics for the input coupling capacitors for a low noise 7-10V application. The leakage current really drops off when you have a larger margin between rated voltage and working voltage. About 5 years ago I checked several brands of 1000-2000 ufd 35V caps and there was one version of either Panasonic or Nichicon that had lower, more stable, leakage current than the rest. I am on vacation so it will be a bit before I can supply that partnumber.

[Andreas]

@chuckb

according to branadic standard 85 degree types are better than low ESR 105 degree types.
I simply choosed among that what I had in drawer (25V types) after 2 days formation time with 10V.
The 2200uF is a ELNA and the 1000uF is a YAGEO. Among 10 capacitors you will usually find 2-3 which are usable.
The rest of the 1000uF capacitors you can use instead of the 470uF bypass capacitors.
So you are right: having a good voltage margin will be a advantage.

And there is a trick: If not used I put a 9V Block on the input of the amplifier to keep the capacitors charged.
Otherwise I would have to wait 2 days before each measurement until the leakage current
drops again below the 20nA limit. (and creates strange noise patterns).

Soldering heat is also a problem. It easily changes the leakage currents by magnitudes of order.
But the discussion of the 1/f amplifier would be worth a own thread.

With best regards

Andreas

[branadic]

I've measured a batch of several capacitors and found, that the 85°C types from Yageo deliver fair leackage currents <5nA after 24h forming in a combination of 2200µF/35V and 1000µF/25V that were finally used for the noise amplifier. 105°C types had decades worse leakage.
Wet tantal are hard to get and Sanyo Oscon deliver only small voltage ratings.
There was a document available, that delivers some explanation, but it's in german. (see attachement)

[Andreas]

The base of the temperature controller is mainly reproducing the noise of the reference, there should be only minor traces coming from the temperature noise. The better place to look at temperature noise would be the output of the OP.

Hello,

Do you really think that I am doing noise measurements at the base of the LTZ internal temperature sensing transistor?
I thought it is clear when I write J6 pin 12 as measurement point that it is the heater power transistor (BC639).
But obviously this is not the case.

@Branadic:
I do not think that OS-Con have low leakage current.
They are optimized for low ESR and high ripple current in low size.
I have some old stock 150uV/16V that I will charge over night.

With best regards

Andreas

Edit: measured values of leakage current as voltage drop over 100K resistor after 20 hours charging.
OSCON 150uF/16V (through hole, pink shrink tube, new old stock)
#1   1.6mV = 16nA
#2 11.7mV = 117nA
#3   0.7mV = 7nA

For 3200uF one would have to parallel 21 pcs of those to get the capacity.

[Galaxyrise]

Well the setup wiring is ok so that leaves a strange operation of the chip.

I'm looking at a puzzling behavior that may be related.  Started a separate thread for it here rather than lengthen this thread with back and forth over my circuit.  (but if it's interesting, I'll reproduce the conclusion here.)  Short version is that I'm seeing an effect on Pin4 from Pin6 even when I don't expect there to be one.