Ultra Precision Reference LTZ1000

[iMo]

Yeah. I have lost  insulation material and now have to order a new one. Any recommendation on insulation please?

Aerogel
https://en.wikipedia.org/wiki/Aerogel

[dietert1]

What kind of controller are you using? Oscillation is typical for an on-off controller. A linear "PI" controller will also behave like that when using excessive gain factors. Also the P and the I gain need to be near a certain ratio for fast settling time.
A well adjusted linear controller will show small/slow changes of heater output and even smaller changes of temperature. As Andreas wrote, a good oven needs a sensor close to the heater, with no air gap in between. My LTFLU oven keeps that "fast" sensor to within +/- 0.5 milliK over 24 hours, using an Arroyo TecSource 5235 as controller. The Arroyo PDF manuals are quite instructive.

Regards, Dieter

[alex-sh]

Solved!

Attaching one NTC directly to the heater (inside the box to the wall against the source) source and putting another NTC in the middle of the box solved the problem. Thank you Andreas!

Now the heater is having an output of 210 mA rather than going from 400mA down to zero and back to 400mA. The inside temperature is now 34.6 (needs fine-tuning) within 0.1C

[Hermann W]

I wonder if this is due to the oven not being properly insulated?

I fear that the NTC for the temperature controller is too far from the heat source.

Thats why I typically use 2 temperature sensors.
One for the heater and one for the point of interest.

There are many ways to get to the goal. But that's not a classic control technique. The actual value must be measured at the point to be regulated. Otherwise the actual value does not correspond to the setpoint. The heater may be constant, but not the test object. The example shows quite clearly that the control parameters have a too high gain - as Kleinstein said.
What kind of controller is used? Temperature control works quite well with a PI controller. The integration time constant must then be chosen to be the same as the control system time constant.

[iMo]

How are your 2 thermistors wired, plz?

[dietert1]

As far as i understand, once you have a rock-solid heater temperature control loop, you can stack another control loop on top, this time using the DUTs temperature sensor.

Regards, Dieter

[Hermann W]

As far as i understand, once you have a rock-solid heater temperature control loop, you can stack another control loop on top, this time using the DUTs temperature sensor.

This is a common way when the manipulated variable (heating) requires separate control. This is not really necessary for temperature control.

[Hermann W]

Now i know what a rocksolid heater is. This is a good example. I once did that for my wood boiler. The internal regulation was for the position of the air throttle and the external regulation for the exhaust gas temperature.

[Andreas]

How are your 2 thermistors wired, plz?

In my ageing box for references one NTC is on the PCB in the middle of all voltage references.
(left in the middle of the PCB)

The other is in the middle of the top heater foil (with the red wires).
The bottom heater foil (below the PCB) is in series with the top heater foil so it needs no extra NTC.

The heater is controlled by a simple P-Controller (fast) from the heat spreader NTC.
The slow contol loop with the NTC on the PCB is a I-Controller with some anti windup limitations when the current temperature is far away from setpoint. The I-Controller output gives a Offset which is added to the setpoint for the P-Controller.

Since there is some self-heating due to many references the Offset is negative (I-Out in deg C)

here a log of the heat-up phase:

Time                                         T-PCB  T-HTR   14V      17V      24V      I-Out   P-Out (hex from 0x0000 to 0x2400 or 0..10000 dez)
08:21:01;D7   0  0 9999.9 H9; 38.21; 31.79; 13881; 16974; 24006; -4.18; P2400;   
08:22:01;D7   0 12    0.0 D7; 39.34; 39.41; 13881; 16981; 24012; -4.18; P2400; 
08:23:01;D7@  1 12    0.0 D7; 40.87; 45.15; 13874; 16974; 24033; -4.18; P2400;   
08:24:01;D7@  2 12    0.0 D7; 42.55; 49.58; 13867; 16988; 24053; -4.18; P2400;   
08:25:01;D7@  3 12    0.0 D7; 44.15; 51.91; 13860; 17021; 24593; -4.18; P2400;   
08:26:01;D7@  4 12    0.0 D7; 45.24; 51.88; 13860; 16981; 24087; -4.18; P2400;   
08:27:01;D7@  5 12    0.0 D7; 46.16; 51.91; 13847; 17028; 24600; -4.18; P2400;   
08:28:01;D7@  6 12    0.0 D7; 47.02; 51.91; 13847; 17028; 24600; -4.18; P2400;   
08:29:01;D7@  7 12    0.0 D7; 47.77; 51.88; 13840; 17028; 24094; -4.18; P2400;   
08:30:01;D7@  8 12    0.0 D7; 48.45; 51.91; 13827; 17028; 24600; -4.18; P2400;   
08:31:01;D7@  9 12    0.0 D7; 49.06; 51.79; 13827; 17028; 24600; -4.18; P0000; 
08:32:01;D7@ 10 12    0.0 D7; 49.44; 50.55; 13820; 17028; 24600; -4.18; P0000;   
08:33:01;D7@ 11 12    0.0 D7; 49.71; 49.49; 13820; 17028; 24600; -4.18; P0000; 
08:34:01;D7@ 12 12    0.0 D7; 49.85; 48.49; 13813; 17028; 24600; -4.18; P0000;   
08:35:01;A4   0  0    0.1 A4; 49.98; 47.63; 13820; 17028; 24600; -4.18; P0000;   
08:36:01;A5   0  0    0.1 A5; 50.01; 46.86; 13813; 17028; 24593; -4.18; P0000;   
08:37:01;A6   0  0    0.1 A6; 50.03; 46.12; 13820; 17035; 24593; -4.18; P0000;   
08:38:01;A7   0  0    0.1 A7; 49.96; 45.49; 13813; 17035; 24593; -4.18; P0133;
08:39:01;B4   0  0    0.1 B4; 49.94; 45.17; 13820; 17035; 24600; -4.18; P0257;   
08:40:01;B5   0  0    0.1 B5; 49.92; 45.06; 13820; 17035; 24600; -4.15; P02bf;   
08:41:01;B6   0  0    0.1 B6; 49.92; 45.01; 13806; 17035; 24593; -4.13; P0302; 
08:42:01;B7   0  0    0.1 B7; 49.87; 44.99; 13820; 17035; 24593; -4.08; P0330; 
08:43:01;C4   0  0    0.1 C4; 49.85; 45.01; 13813; 17028; 24593; -4.03; P0343;   
08:44:01;C5   0  0    0.1 C5; 49.87; 45.03; 13820; 17028; 24593; -3.99; P035c;
08:45:01;C6   0  0    0.1 C6; 49.85; 45.03; 13813; 17028; 24593; -3.94; P0383; 
08:46:01;C7   0  0    0.1 C7; 49.87; 45.10; 13813; 17028; 24600; -3.90; P0373;
08:47:01;D4   0  0    0.1 D4; 49.87; 45.10; 13813; 17035; 24600; -3.86; P039a; 
08:48:01;D5   0  0    0.1 D5; 49.87; 45.15; 13813; 17035; 24600; -3.81; P0397;
08:49:01;D6   0  0    0.1 D6; 49.87; 45.19; 13813; 17021; 24600; -3.77; P0395; 
08:50:01;D7   0  0    0.1 D7; 49.89; 45.21; 13820; 17035; 24593; -3.73; P03a7;   
08:51:01;E4   0  0    0.1 E4; 49.92; 45.26; 13806; 17028; 24593; -3.71; P039e; 
08:52:01;E5   0  0    0.1 E5; 49.92; 45.28; 13813; 17028; 24600; -3.68; P03a2; 
08:53:01;E6   0  0    0.1 E6; 49.94; 45.31; 13813; 17035; 24593; -3.66; P03a7; 
08:54:01;E7   0  0    0.1 E7; 49.96; 45.31; 13813; 17035; 24600; -3.65; P03b9; 
08:55:01;F4   0  0    0.1 F4; 49.96; 45.37; 13813; 17035; 24587; -3.63; P0387;
08:56:01;F5   0  0    0.1 F5; 50.01; 45.35; 13813; 17035; 24607; -3.64; P03a7; 
08:57:01;F6   0  0    0.1 F6; 50.01; 45.40; 13813; 17021; 24600; -3.64; P037b;   
08:58:01;F7   0  0    0.1 F7; 50.05; 45.37; 13806; 17035; 24593; -3.66; P038e; 
08:59:01;G4   0  0    0.1 G4; 50.03; 45.37; 13813; 17028; 24593; -3.67; P037e;
09:00:01;G5   0  0    0.1 G5; 50.05; 45.35; 13813; 17035; 24600; -3.68; P038a;
09:01:01;G6   0  0    0.1 G6; 50.07; 45.37; 13813; 17035; 24593; -3.71; P0366; 
09:02:01;G7   0  0    0.1 G7; 50.07; 45.35; 13813; 17035; 24593; -3.73; P0364;
09:03:01;H4   0  0    0.1 H4; 50.10; 45.33; 13813; 17028; 24600; -3.76; P0362;
09:04:01;H5   0  0    0.1 H5; 50.10; 45.31; 13813; 17028; 24600; -3.80; P0359; 
09:05:01;H6   0  0    0.1 H6; 50.07; 45.28; 13813; 17035; 24600; -3.82; P0350;   

With best regards

Andreas

[iMo]

@Andreas: has your thermostat got "digital" control (2xRth+MCU) then?
I've been looking for some 2xRth+(1-2x)opamp "linear" version I will add directly on the 10V Vref board.

[Andreas]

Hello

yes it is a PIC24FV301 with 12 Bit ADC where the NTCs and the 100 Hz PWM heater output is connected.
It is only used as dumb measurement machine, multiplexer controller and heater actuator (with overtemperature limitation).

The actual control loop is on the connected PC.
Every 10 seconds the P-controller part is calculated.
The I-controller only all 60 seconds.
The long time constants and especially the anti-windup strategy are difficult in analog domain.

with best regards

Andreas

[iMo]

But volt-nuts would rather avoid digital+pwm close to their lovely reference Therefore the idea with analog PI even it would be difficult (we do it not because it is easy..).

[Andreas]

Hello,

yes you are right, I only use it for those ugly AD586, LTC1236 and LT1027 references for pre-ageing.
But 100 Hz PWM will be canceled out by most integrating instruments with mains suppression.

The principle can also be used in a analog controller. A inner P-Controller for the heater. And the setpoint with a offset from a I-Controller (where you have carefully limit the output signal range to avoid very long settling times).

with best regards

Andreas

[alex-sh]

What kind of controller is used? Temperature control works quite well with a PI controller.

A simple P-Controller.
However, the problem is not in controller, but in placement of NTCs.
I had to place one NTC right next to the source and I failed - this gave me oscillation 

[alex-sh]

Made a stupid mistake last night while burning the midnight oil - applied +15V to the buffer output +10.000000V (Dr Frank's PCB).
Now both the BUF and BUF 10.000000V outputs are dead. VREF on the LTZ1000 is giving me 7.08823V (just like before this accident). So I guess LTC1052 was killed? 

I paid the price for staying very late 

[guenthert]

But volt-nuts would rather avoid digital+pwm close to their lovely reference Therefore the idea with analog PI even it would be difficult (we do it not because it is easy..).

Whoa, that's a slippery slope there, ending with batteries, potentiometers and galvanometers like it's 1899. 

Just because PWM is easy for the uController, there doesn't have to be a high (EMI indcucing) current PWM to the heater.  Nothing stops you from adding a cap or two before the series pass transistor (assuming a sufficiently capable transistor of course) ...

[iMo]

But volt-nuts would rather avoid digital+pwm close to their lovely reference Therefore the idea with analog PI even it would be difficult (we do it not because it is easy..).

Whoa, that's a slippery slope there, ending with batteries, potentiometers and galvanometers like it's 1899. 

Edit: It took me 2 hours to wire and simulate
Added the temperature difference between the heated case and board.
The TIP121 has to be thermally isolated from the case 

[Andreas]

Nothing stops you from adding a cap or two before the series pass transistor (assuming a sufficiently capable transistor of course) ...

Ok I forgot to mention that I use a rather slow automotive power stage (BTS432) which is optimized to EMI with turn on/off times around 100 us.

with best regards

Andreas

[iMo]

Sure, the simulation of the thermostated box for the Vref is off topic here, therefore I stop with the below picture - I've moved the thermistor (the heater one) a bit off the heater - I get a nice oscillation

PS: added a version with 2x TIP121 as the heaters with added white phonon noise

[alex-sh]

I wonder if this is due to the oven not being properly insulated?

I fear that the NTC for the temperature controller is too far from the heat source.

Thats why I typically use 2 temperature sensors.
One for the heater and one for the point of interest.

With best regards

Andreas

Hi Andreas,

I managed to obtain temperature stabilisation between 34.81C to 34.93C which is within around 0.1C.
I used two NTCs - one attached right next to the heating resistor and one is next to LTZ1000A.
I am using 2  x TE Connectivity GA10K3A1A (+/-0.1C tolerance, B value tolerance 0.5%).
Do you think this is sufficient? The enclosure is covered in one layer of insulation. I think I have to work more on insulation putting the second layer.

Thanks
Alex

[alex-sh]

I can see from time to time these spikes (marked in red in the photo). This is a new ltz1000ach ageing, but the measurements were taken (a) in the temperature oven (b) in the shielded enclosure (c) powered by a battery pack. The only thing which comes to my mind are cables. I’m using ordinary 22AGW cables rather than shielded ones. However, from the DMM I’m running banana to BNC shielded cable with hooks at the end.

What do others think? Where these spikes are coming from? What cables are you using for your boards please (links to Digi-Key would be appreciated)

[iMo]

Your DMM6500 Vref is an LM399.. Popcorn?

[hwj-d]

@alex-sh
Looks like typical popcorn noise. That's the LTZ itself. Should decrease over the months, but should be observed. My LTZ are reasonably quiet after 18 months.

[Kleinstein]

There are several possible reasons for just a single out-layer. One possibly source could be EMI (e.g. mobile phone sending a rather strong sync signal once in a while,  radar).  Popcorn noise of the DMM reference would usually look different and would not be visible near zero.

With just a relatively low impedance reference to a DMM the cables are not that critical - a poor choice could give some thermal EMF, but this is more a problem with the connectors that the cable itself.  BNC connectors may not be the best choice for DC.

What is exactly measured ? It does not look like directly reading the 7 V with the DMM. This would not make that much sense anyway as the DMM internal reference would dominate.

[dietert1]

You seem to have 1.5 uV drift in 10 minutes, which is a lot and may indicate there is something else going on besides ageing. Or is the plot in hours?

Regards, Dieter