Transcript furnace_improvements_lne_inm_11_06

Improvements to Furnaces and Controllers
Euromet 732 Progress
LNE-INM
23-24 Nov 06
D HEAD, M de Podesta, J Pearce
NPL
Talk
• Electrical Noise Reduction
– Pick up from thermometer in furnace
– Pick up from standard resistor in bath
– Optimal Averaging
• Furnace Uniformity
– Three zone performance
– Cascade control
• Thermal Modelling
– ‘Swiss’ Roll
Electrical Noise
in furnace
• AC furnaces – Thyristor controller Mains voltage
– Cheap but noisy
• DC furnaces – variable voltage
– Expensive but quieter (except magnetic field near PS)
• Reduced Voltage AC Thyristor heaters
– Cheaper
– Crude external max power limitation
– Quieter
– But…
• Need to shield temperature controller (mild steel)
• Additional Transformer cost
• Extra Power leads – safety
Furnace Power
1. Mains Thyristor
Thyristor
PID Controller
Control Box
Furnace Power
2. Reduced Voltage Thyristor
PID Controller
Thyristor
Control Box
Furnace Power
3. Reduced Voltage Thyristor
PID Controller
Thyristor
Control Box
Electrical Noise
Standard Resistor
Baths & UPS
• Noise pick up via Standard Resistor
– In oil bath systems eg Grant, Huber and Hart (wtpc bath) etc.
– Reduced by peltier systems? Untested at NPL
– Reduced by keeping Standard resistor further from heater
– Noise depends on resistor value/design?
• UPS may increase F18 NOISE!
– unless phase locked
• Noise Traces (F18 in manual mode– no “Balancing Noise”)
– Standard resistor bath AC, Variac®, DC
– AC, cycle time, Variac, DC
Noise
Noise reduced
when Standard
Resistor Bath
switched off
Standard
Resistor Bath
switched off
DC furnace
reduces midtrace
DC Power
Supply
Noise Traces
AC, DC
& Hc
DC furnace quieter
than 230V thyristor
Standard
Resistor Bath
switched off
DC Power
Supply
Noise Traces
Heating
Cycle Time
Reduced
firing
rate
reduces
noise
Cycle Time
3s
Cycle Time
30 s
Cycle Time
10 s
Noise Traces Time
ASL F18
Manual mode
improves
resolution
Reduced noise
from reduced
thyristor voltage
Manual
Mode
DC
34 V
50 V
D/A
Mode
Allan Variance
Allan deviation
• PRT measurement systems pick up noise
• Average measurements – but over what period
• Calculate Allan variance/Allan deviation
– Average of increasing number of data points
– Calculate variance for each of these numbers
– Plot variance against number
Allan Deviation
Assessing stability
Standard Deviation:
 Characterises variability of a set of results
 One Number
Allan Deviation:
 Characterises variability of a set of results on different time scales
 Several numbers
Signal
V
V
Time
tt
t
The Allen Deviation of the
Residuals
Residuals (Bridge ratio)
-7
10
Variations with this time scale
are commonly caused by air
conditioning.
Still need to decide where drift
arises
–electronics
–thermal furnace “noise”
–real physics of ingot
•Allan variance gives a tool to
separate these effects
Allen Deviation (Bridge Ratio)
Averaging reduces the noise up
to 1000 s (16 minutes).
Window Size (hours)
0.01
0.1
1
10
Standard Deviation
-8
10
-9
10
10
4
100
1000
10
Window Size (seconds)
10
5
Moving Average
Based upon 16 and 100
Points
TIN Sn-A Freeze (24-25 August 2005)
furnace C5 Set pt 228.0 variac 22%
0.4763961
0.1 mK
0.476396
Bridge ratio
16 points about 5
minutes.
100 points about
28 minutes
0.1 mK
0.4763959
0.4763958
0.4763957
1
3
5
7
9
11
13
15
Time elapsed (hours)
17
19
21
23
25
End zones and Uniformity
Long freezes require
• “Temporal stability”
• “Spatial uniformity” (or controlled gradient )
– Permits meaningful investigate hydrostatic head
Poor “Spatial uniformity” leads to
• Apparent early run off (Ancsin)
• “Dough-nuts” (Frozen material dropping into melt)
• Cracked cells (Melt expansion?)
To achieve “Spatial uniformity” use
• heat pipe (pressure controlled?)
• 3 (or more) zone furnace
Three zone furnaces
Normal
Furnace Temperature D
T/K
Control sensor in centre of
each zone
• offset to one side of work
tube
• Zone centres very close
in temperature
– but zone junctions
hotter !
– ‘Two hump camel’
profile
Standard 3 zone
Single zone
Z / cm
T/K
Adjust end zone temperature set point
Three zone furnaces
Reduced End Zone T
Z / cm
T/K
Adjust end zone temperature set point
To improve uniformity
• Reduce end zone
setting by T
– T changes with
furnace T
ΔT=f(T)
Z / cm
tandard 3 zone
ingle
zonezone
Three
furnaces
Vertical Furnace
Standard vertical furnace
(more complicated distribution)
Z / cm
• Vertical use adds other
complications
Z / cm
• Breaks axial symmetry of
temperature distribution
• IGNORE!
Heat
rising
zone temperature set point
T/K
ΔT=f(T)
Thermocouple tip
Furnace Temperature D
T/K
Three zone furnaces
Postulate!
• Move end zone control
istribution
thermocouple(s) closer to
centre zone
• Make end zone shorter
• Should give a smoother
axial profile (2B
T/K
continued…)
Adjust
thermocouple position to test concept
Adjust end zone thermocouple position
Z / cm
Adjust end zone length (shorter)
Z / cm
Three zone furnaces
Radial
• Radial variations in T vary with
– Immersion/load
– Furnace design
– New design under test
– May need to sacrifice some axial uniformity?
Controllers
• Proprietary eg Eurotherm® or own manufacture
• Newer ones have sophisticated PID, with several settings for different
temperatures eg 2604
• Control thermocouple position
– Near ingot or near heater?
• Control block T or furnace T;
– Slow versus fast CONTROL response
• CASCADE control uses two thermocouples
– one near ingot in block
– one near heater
– heater set point is calculated by response of thermometer near ingot
– 2 B Continued !
Cascade Control Diagram
Thermal Modelling
• Can we use thermal modelling to guide us in furnace,
block or cell design?
• Consider block – ideally want good vertical uniformity
– avoid transmitting any furnace non uniformity into
block
– vertical or circumferential
» Concept of “Swiss Roll”® block – ie extruded spiral
‘Swiss Roll’
uniformity device
• Can be made out of any
flexible materials
• Must not touch at top or
bottom
• Must be radially insulated
• Will work at any temperature
– suitable for all fixed points
• Example is around the Sn
point
• “Independent” of thermal
conductivity – it is a
geometric effect.
–
model calculations: J V Pearce
F900 and F900 future;
Standard Resistor Baths
• NPL finds F900 superior to F18
• 2nd F900 on order BUT with improvements (due 2007)
• Standard resistor baths
– Expensive (?12k€) versus cheap (?5k€)
– Some cheap baths have better laminar flow
• but dip cooler introduces water into oil
– Some more expensive baths have cooling on bottom
• can cause vertical gradient – especially if standard
resistor is sitting on the bottom
• More noise close to stirrer/ heater
• Larger cooler Larger heater  more noise