Supplying High-Energy Physics
Detector Systems
F. Szoncsó
Institut für Hochenergiephysik der
Österreichischen Akademie der Wissenschaften
Vienna, Austria
Abstract
The paper gives hints on how to properly supply power to high energy physics detectors.
It later focusses on some electromagnetic compatibility issues because it is widely
accepted that the power supply system is a crucial parameter in the system design
of detector electronics.
In some cases the power supply system remains the only variable in a given setup which
again stresses its importance in particular during the design phase of an experiment.
-
Introduction
In this paper the term "power supply" refers to a power converter from ac-mains to
a stabilized dc-source for electronic equipment.
The specifications for a power supply include many parameters besides the obvious
matching between power demand and power availability. In order to be able to choose
a power supply, or, in certain cases, to build it, a careful evaluation of some
important side-issues should be done as briefly outlined below:
a) Mechanics & cooling
Adapting the power supply to the mechanical layout and the environmental conditions
of an experiment is a difficult task of its own. Specifications of crate mechanics
like CAMAC, FastBus, VME, Euro may include very accurate power system specifications
which can be used for other purposes in some cases. It will be shown later which power
density you can expect from the various converter techniques. Cooling power can then
be calculated easily on the basis of efficiency figures.
b) Input power rail
The power requirements always being high the mains distribution system also merits
some attention when it has to carry power for noisy applications (e.g. thyristor
converters) and low noise front ends at the same time. Large numbers of switching supplies on single power rails are particularly difficult to handle.
c) Protection
Safety aspects play an increasing role in system design, in particular when talking
about power supplies for underground detectors. Most power supplies have some sort of temperature triggered shutdown for their own
protection. However, additional circuitry will be needed in order to protect connected
electronics. A digital shut-down terminal at the power supply would be a useful asset.
d) Implementation
Serviceability and reliability are to be kept in mind since most supply systems are
installed in restricted access areas. The distance between the power supply and the electronics it has to supply is a parameter
that will heavily interfere with system design of densely packed detector electronics. Voltage drop on the interconnection to the load may exceed the voltage range of a power supply. Interconnection done with more cross-section will interfere with space requirements inside the detector area.
e) Monitoring
Monitoring as well as voltage control should, whenever possible, be incorporated into
the power supply. Many power supplies offer remote sensing but only poor monitoring
facilities such as built in monitoring terminals showing the actual conditions at
the point of sensing, i.e. at the load. Additional monitoring leads brought to special
monitoring ADC's are considered obsolete for many reasons.
f) Remote sensing
In order to maintain a certain voltage far at a certain distance from the the power
supply with changing load the power supply senses the voltage at the load and adjusts
its output voltage for drop compensation. However, this feature is subject to changes in load regulation, interference and system outlay.
The sensing lines should be routed together with the DC-power lines. They should not be used for monitoring purposes.
g) Electromagnetic compatibility
Neither should the supply system close any unwanted loops nor should it introduce
additional noise into the detector area. The marginal signal levels at the input of preamplifiers, sometimes amounting only to the
order of magnitude of the amplifiers' inherent noise level, require an absolutely
"clean" power supply. Supply voltage feedthrough of many thousands of channels of
electronics cannot always be kept at a negligible level economically which imposes additional
requirements concerning the spectral purity and long term stability of the power
source.
f) Filtering
Parallel connection of many loads on a single DC-power rail requires decoupling elements
that avoid voltage surges to proceed to connected loads in the neighbourhood. DC-power
rails fed by distant power supplies, a common picture in high energy physics, will need distributed elements to ensure low rail impedances seen by the loads.
2) Power requirements as seen by the users
a) Preamplifiers
The most important asset of the supply system for a given set of preamplifiers is
its capability of providing a low impedance DC-power source with the least possible
coupling to other systems of any kind. Capacitive coupling brings transients and
high-frequency noise, magnetic coupling causes humming and electromagnetic coupling may provoke
high-frequency susceptibility. Monitoring of the output voltages should be done without
creating coupling paths.
Table 1 Specifications of a typical preamplifier power supply
output voltage 6V to 15V (typical)
output current(residual/during test-pulse) 3A cont/ 7A 2% duty cycle
ripple 1mVpp, tr 0.5ms
output spectral purity @ nom. load level "K" VDE 0875 [4.875]
insulation impedance to mains |Z| 10k up to 1MHz
MTBF 105 h
A few terms need further explanation because they do not normally play a major role
in power supply design:
"Least possible coupling
"
Assuming an insulated power supply (i.e. the electronics is grounded at the detector
or elsewhere) all magnetic and electromagnetic fields in the experimental area will
cause a certain spectrum of voltages on the power lines. In order to avoid residual
currents that flow into the preamplifiers' ground system the coupling impedance to the
mains and, of course, to possible monitoring facilities, has to be as small as possible.
Most of this impedance depends on the winding-to-winding capacitance of the mains
transformer of the power supply. Values in the order of a few hundred pF are a good assumption. In case of monitoring systems with dedicated ADC's connected
to the output of the power supply an additional galvanic impedance is introduced.
If such a system cannot be avoided it is recommended to use an ADC with at least 1M input impedance and 10k series resistors for all leads departing from the load (fig. 1).
Fig 1 Monitoring and sense connections
Monitoring without creating additional coupling paths
Closing unwanted loops will cause unwanted currents on all the supply lines and from
there to other systems. Monitoring lines connected directly to the load, i.e. the
preamplifier, introduce additional impedance and hence additional currents via the
ground system.
In principle the voltage monitoring can be left out at all. Any power supply will
"know" about voltages and currents at the load provided the sense lines are connected
as shown in fig.1. However, the supply is not able to tell you the exact voltage
at the load unless it is equipped to do so. Tapping the sense lines where they enter the
power supply will put an "antenna" to a very sensitive point of the control loop
and hence should be avoided.
In most cases it will be sufficient to monitor the output current of the power supply
because only sufficient output voltage will cause the residual output current.
Some power supplies output a digital "output voltage(s) present" signal which provides
sufficient information together with the current values [6].
Applications that are sensitive to power supply voltage drift definitely require a
voltage monitoring but such sensitivity in general suggests bad electronics design.
Abusive applications of deregulated power supplies, e.g. to generate threshold voltages
via divider chains, cannot be discussed in this context.
Electromagnetic coupling
For frequencies above 100kHz the power lines will, as all other lines, act as antennas
for electromagnetic fields originating inside and outside the detector area. In particular
broadband amplifiers for phototubes (like, e.g., in [3]) will show a certain susceptibility to radio frequency stray currents. Shielding will only alter the path
of the currents.
Fig 2 Radio frequency current paths
When grounding detector electronics a set of loops are formed by the inherent connection
between electronics and power supply. Electromagnetic coupling cannot be avoided
by a change of layout of the electronics. Its effects, however, can be cured either by inserting a high impedance spot into lines carrying radio frequency stray currents
or made ineffectice by an electronics layout that is capable of absorbing them. The
well-known toroids may serve for this purpose, another way of decreasing RF-currents
is the RF-filter in its various forms. A spectrum-analyzer should be used in order
to determine spectrum and amplitude of the interference. The information gathered
will allow the proper hardware choice.
It should be noted, however, that radio frequency stray currents are not very likely
to be the main source of interference. The electromagnetic field strengths brought
in from outside a detector area are generally small compared to the electromagnetic
noise generated by digital links via flat cables or by badly screened switching power
supplies.
b) Analogue Readout Electronics
Shapers, line receivers, ADC's are densely packed in modern physics experiments. Their
power requirements are much less critical and the setup will involve only few restrictions. The market offers an abundant variety of setups including the power supply.
Spectral purity specifications of the power supply will depend on the bandwith of
the electronics and, of course, on its specifications for supply voltage feedthrough.
It can be assumed that a 40dB rejection of supply ripple can be handled when referring
to the signal level on the cables between preamplifier and readout electronics at a
dynamic range not exceeding 104.
Table 2 Typical parameters of analogue readout systems
minimum level on signal lines 100 V
signal dynamic range 104
power supply ripple 5mVpp
supply output spectral purity@nom. load level "K" VDE 0875 [4.875]
voltage feedthrough of electronics -40dB 50 V
gain dependence on supply voltage 2%/V
Evaluating the numbers leads to some elementary conclusions about power supplies at
this stage:
a) Switching power supplies (ripple 50-200mVpp) cannot be used at low signal levels and high dynamic range.
b) Supply voltage feedthrough may be a major issue when designing readout electronics
and, of course, must be specified and tested both in terms of amplitude and frequency
dependence.
g) The power supply remains an integral part of the setup also at this stage. Electronics
designers should bear in mind the relevant specifications of the future power supply.
Setups with a dynamic range exceeding 104 are difficult to handle which also applies to the power supply. Tight margins on
line and load regulations are followed by a close look at long term stability and
temperature dependence of the output voltage.
c) Digital readout electronics
Less sensitivity to noise of any kind and operation of exclusively digital circuits
permit the use of power supplies with higher ripple. Supply voltage feedthrough is
not supposed to trigger false logic state changeovers. Specifications of linear supplies
exceed by more than an order of magnitude the requirements. Today's state of the art
is switching power supplies with ripple voltages up to 200mVpp.
Most of the digital experimental electronics comes in standard crates equipped with
standard power supplies. In case of custom-designed digital electronics power leads
should be kept as short as possible and crates should be shielded.
Fast digital electronics tends to very high power density and will also be sensitive
to the supply voltage level at the integrated circuits. It should be made sure that
none of the digital circuits' specifications gets touched by transient drops on bus
bars and printed circuit boards. Sensing the busbar voltage(s) should be done at several
well separated points.
Power supplies for digital circuits have to deliver current surges when large groups
of digital circuits decide to switch into a different logical state at the same time. Busbar impedance [11], line impedance, power supply impedance
all depend on frequency.
Power supply output impedance over frequency is specified for good power supplies
but it is very difficult to keep a low impedance across busbars [11], connectors and printed circuit boards. On-board busbar impedance can be pushed
down to a few ohms. Decoupling from the busbar is done with LC-filters and parallel
decoupling capacitors of different kinds in order to cover the frequency range required by the electronics.
2) Power Supply Techniques
a) Series regulated (linear) power supplies
Power supplies with series dissipation regulators are widely used for general purpose
and low noise applications. Low output ripple as well as good spectral purity, in
particular for frequencies exceeding 100kHz, make series regulated supplies compatible
with almost any type of electronics.
A linear power supply is essentially a current limited power amplifier running at
fixed output voltage. The voltage difference between the power source (rectifier
output) and the output voltage is dropped on the pass transistor that has to dissipate the product drop voltage by output current. Ripple
and spectral purity depend on the precision of the employed regulator and on the
dynamic parameters of the power stage. Output impedances can be brought to extremely
small values (few m ) up into the MHz-range. The bloc diagram is shown in fig. 3.
Fig 3 Bloc diagram of series regulated "linear" supply
A well-hidden disadvantage is the strong coupling to the mains by the transformer's
winding-to-winding capacitance. Most supplies offer a third winding, the electrostatic
shield, to overcome this problem.
The connection to this winding is brought to a special output terminal (ES, fig.3).
A separate ground connection will be needed in order to make proper use of this terminal.
The coupling will move from the mains to the ground system.
The connection to the load via wires and busbars increases the impedance of the power
source seen by the load. A coarse regulated supply far away from the load will offer
the same result as a precision supply. In addition, supplies located at some distance from the load usually are accessible and do not have to be of a high reliability
precision type.
Table 3 Parameters of a typical
series regulated "linear" power supply
output voltages and currents any
ripple 0.5 - 2mV RMS
load regulation 0.02 - 0.1 %
line regulation 0.02 - 0.05 %
input voltage range 10 %
power density 0.03 W / cm3
short circuit current mostly low foldback current
short circuit protection current and/or power limit
efficiency 40%<h<55 % @ full load
Linear power supplies are in general equipped with a crowbar circuit which bears this
odd name due to its rather crude function of shorting the power supply output in
case of overvoltage. Releasing the output may be performed by switching the power
supply off and back on again.
Heat dissipation of some 4-7 W per Ampère output current and, due to efforts in putting
limits to the dissipation, a rather restricted input voltage range are the most obvious
drawbacks of linear supplies. Weight and volume are in the order of five times higher compared to a switching power supply. In addition the hold-up time (=time interval
during which load can be properly supplied without mains voltage present) falls far
short of other types of supplies.
b) Magnetically stabilized supplies
Today's computer-designed magnetically stabilized supplies perform almost like linear
regulators. In spite of its clear merits this type of power supply is almost unknown
in physics research. The major disadvantage is a given load dependence of the output
voltage. Setting precise voltage values is not possible but, in most cases, also unnessecary.
For crates with fixed configurations the voltage variation remains within DV/V 1% including line voltage variations of +20/-10%. A performance breakdown is
given in table 4.
Table 4 Parameters of a commercially available
magnetically stabilized power supply
output voltages and currents 12V 30A 100% duty cycle
ripple 75mVpp @ 30A
load regulation 900mV (25% - 100% load)
line regulation 65mV (line volt.176-242 V)
mains transient suppression >60dB
mains transient protection few kJ
short circuit current 0.8 times nom. load current
short circuit protection thermal breaker
efficiency h = 72 % @ 30A
Supplies of this type are used for the analogue readout of the UA1-hadron-calorimeter
[2], [5]. The rather high ripple of 75mVpp at full load is an extremely soft one. Its spectrum does not extend above 300Hz because
of the heavy filtering at the output. Operational amplifiers will properly suppress
ripple of this kind. No difference in electronics performance compared to the use
of a linear supply can be observed.
The schematic diagram is shown in fig. 4. The transformer is a three-leg transformer with a saturable core. The amplitude
of the magnetic flux density in the leg with the secondary winding is held constant
by saturation and is given a characteristic shape by adding reactive current with
an additional winding connected to a capacitor. For this application a soft trapezoïdal waveform
will guarantee best output stability. This waveform is also best suited for rectifiers.
Excess ripple is compensated by a differential smoothing reactor coil.
Fig 4 Magnetically stabilised power supply 12V
Fig 5 Load dependence of output voltage for a
magnetically stabilised power supply (24V@6A)
A typical voltage dependence versus load is shown in fig. 5. It should be noted, therefore, that setups where frequent changes of the load are
likely to occur will not favor this type of supply.
A high efficiency magnetic stabilizer will lose part of its regulation capabilities
if the load current drops below 25% of the nominal load (fig. 5).
The magnetic stabilizer should be considered an interesting alternative for low noise
applications that have to operate on highly contaminated mains outlets. Its inherent
transient suppression (around 60dB) will swallow most of the incoming noise. Radio
frequency coupling to the mains is extremely small because primary and secondary windings
of the power transformer are put on different legs of the yoke.
Some computer manufacturers use magnetic stabilizers even in front of the usual set
of switching power supplies [9] both for protection, mains transient suppression
and easier switchover to different line voltages.
Magnetic stabilizers appear to be attractively priced at about 1$/W at the level of
a few hundred watts. Applications requiring high reliability und high transient suppression
include cable-TV, high power operational amplifiers and computer supplies.
c) Switching power supplies
Fast high efficiency power switches paved the way for switching power supplies operating
at ultrasonic frequencies. Depending on the configuration of the power converter
a certain amount of energy per unit time is transferred into a smoothing/conversion
circuit by means of semiconductor switches. The duty cycle of this transfer is adjusted
by a regulator according to the actual output voltage after the smoothing components.
A very general bloc diagram is shown in fig. 6.
Table 5 Parameters of a typical
switching power supply
output voltages and currents any
ripple 20mV to 250mVpp @ full load
load regulation 0.1 - 1 %
line regulation 0.05 - 0.1 %
input voltage range 20 %
power density 0.14 W / cm3
short circuit current nominal load current
short circuit protection current or power limit
efficiency 70%<h<90 % @ full load
Fig 6 Possible bloc diagram of switching power supply
Depending on the configuration of the power supply a number of inherent properties
can be derived but this paper does not focus on power supply descriptions. Detailed
information can be found in [1].
Although the efficiency of switching power supplies can be found in the 70% - 90%
range depending on configuration and specifications a number of effects restricts
the merits of low heat dissipation and small volume.
Weight reduction is achieved by operating the converter at ultrasonic frequencies
thus allowing for very small transformers and smoothing reactors. However, fast switches
operating at some rather low fundamental frequency will exhibit a frequency spectrum
extending far into the MHz-range. Noise can be measured even at 200MHz when testing
very small supplies equipped with VMOS.
Radio frequency noise can be contained inside the power supply box but it is a question
of price. Radio frequency components, compatibility measurements, compliance with
safety rules will push the cost of switching power supplies. The market being extremely competitive only tight regulations will force supply makers to keep noise levels
below certain limits. Limits for a wide range of applications can be found in [4].
Power supplies should be clearly stated to comply with these standards. Output ripple
values have to be interpreted cautiously. Parameters listed in table 5 always state the same noise level but the values show large differences which is
due to the fact that the output ripple of a switching power supply shows a very typical
behaviour.
Table 5 Possibilities of listing switching power supply ripple
Without grading
Output ripple @ full load (peak-to-peak) 150mVpp
Output ripple @ full load (RMS) 2.4mV RMS
Output ripple @ full load (peak-to-peak,
bandwidth restricted to 20MHz) 60mVpp
Output ripple @ full load (with additional filter,
drop on filter unregulated) 10mVpp
Input ripple @ full load not spec.
Parameters according to IEC or VDE rules:
Input and output ripple (or spectrum) specified according to grade
The action of the fast switches result in switching transients with high amplitude
during a few 100ns, whereas charge and discharge of the output smoothing capacitor
will show as an asymmetric triangular waveshape of low amplitude with the switching
frequency as its fundamental.
d) Exotic power supplies and configurations
Many forms of nonconventional power supplies or configurations made their way into
physics research. The system in [6] (described in amendment 1) was the answer to rather particular boundary conditions. No dramatic simplification
is to be expected except for very special system parameters such as coupling (e.g.,
battery operated amplifiers) or cabling (e.g., power feed via the signal lines themselves). If harmonics generation on the mains lines has to be avoided there could be
a reason for special power supplies such as switching power supplies with sinusoïdal
mains current. Weight and volume usually are not considered a problem except in very
special cases [6].
Power feed to preamplifiers via signal cables
Feeding power via the signal lines might reduce front-end cabling considerably. It
is not worked out, though, whether this is likely to cost less. The basic outline
of such a system is shown in fig. 7:
Fig 7 Feeding power to the front end via signal cables
For AC-coupled readouts (fig. 7) splitters take over the function of the power lines. The filter chokes actually
are very small but somewhat expensive. The resonant frequency of the splitter has
to be well outside the spectrum used for the signal. Voltage stability during the
outgoing pulse at the preamplifier end can be guaranteed by the stored energy inside the filter
capacitors. The preamplifier should be capable of running on more or less constant
current.
DC-coupled systems (fig. 8) will inevitably need specially designed electronics operating at a residual constant
current on the signal lines. The signal line becomes an inherent part of the electronics.
Getting rid of interference on the signal lines involves high quality design work taking into account all possible interference paths.
Fig 8 DC-coupled power feed and signal connection
The accuracy of the DC-offset, hence also the signal amplitude depends on the parameters
of three constant current sources and one constant voltage source in addition to
the actual signal. Of course some of these elements can be found in single chips,
and they can all be compensated against each other. However, such an approach is not easy
to handle. Only very good reasons (e.g. space restrictions inside a detector, loop
cancellation, definite need for DC-coupling) might force the designer to live without
a DC power rail inside the detector. The constant voltage source U1 (fig. 8) is set to:
U1 = (Ri ampl + Ri driver + RL1 + RL2) * (Is + Ip res)
with Ri ampl load resistance due to amplifier power
Ri driver load resistance of line driver
RL1, RL2 line resistance
Is supply current
Ip res pulse current offset
It essentially compensates the voltage drops of the residual current through the amplifier.
This can be done by a small control loop configurated as a voltage-to-current converter
in order to keep the offset at a certain value independent of other parameters.
Switching power supplies with sinusoïdal mains current
Mains connection of transformerless rectifiers (such as switching power supply input
circuits) in single phase configuration considerably distorts the mains waveform
because of the typical shape of the mains current and its slightly leading phase
presenting a capacitive load (fig. 9). In addition a switching power supply will, due to its alsmost constant efficiency
over a wide input voltage range, present a load with a negative dynamic resistance.
Fig 9 Mains voltage and currents of a switching power supply
Low mains impedances provoke harmonic currents, high mains impedance lets these currents
develop into harmonic voltages distorting the sinusoïdal waveform mainly at phase
angles just before 90o. Oscillations due to the negative dynamic input resistance are very difficult to
come by in large setups. Several blowouts of entire sections of the supply system
of UA1's underground electronics building triggered a series of urgent protective
measures against the resulting surges.
Medium-size physics experiments may need in excess of 100kW of power for readout and
computers, whereas large experiments such as UA1 need as much as 1MW to run the electronics.
The distortion of the mains waveform may have very bad effects, and it also means losing the voltage level to a certain extent.
In order to properly maintain sinusoïdal waveforms for all users of an experiment
a fast power factor compensator can be installed. It should be noted, however, that
the dynamic compensation of harmonic currents results in magnetic stray fields in
the kHz-range around the power distribution network.
Three phase rectifiers create much less harmonics but they do not enjoy widespread
acceptance at power levels below 1kW because of increased cabling complexity. For
fastbus-electronics requiring several kW of crate power three phase power supplies
are state of the art.
A much better way is to avoid at all the development of harmonics by drawing sinusoïdal
mains currents at the fundamental frequency of the mains.
Recent research on switching power supplies [12] concentrates on being compatible
with low noise levels, medium to high mains impedances, parallel connection and sinusoïdal
mains waveforms. The latter requires the implementation of an additional control
circuit that approximates the primary current waveform to a sinewave by storing the
energy of a mains half cycle. The supplies will draw an approximated sinusoïdal current
(fig. 9) in phase with the voltage but the overall power supply efficiency drops by 10%.
3) Layout of Power Systems
A considerable part of the electronics budget is spent on the power supply but only
little effort gets invested when choosing it. Hints given here also apply to the
proper use of already existing supplies from previous setups.
a) Grouping
Most detector systems are grouped geometrically and by the signal cabling. Any inherent
grouping should never be broken by the power supply. The groups of channels supplied
by one power supply should be held as small as possible.
Large groups
of amplifiers running on rather powerful supplies are a poor approach although the
solution looks extremely attractive and simple in the first place (fig. 10).
Fig 10 Large groups of preamplifiers on single power rail
Preamplifiers that are fed via long cables and power rails have to be designed to
operate under such conditions. Care has to be taken to avoid channel intercoupling
via the power rails and hence oscillations. In most setups the interconnection of
power ground and detector ground runs via the amplifier boards from the signal cable connector
to the power connector. There is no more flexibility once the system has been assembled.
Inherent grounding via the power rail forces the designer to use symmetric readout
cables connected to differential receivers with a common mode rejection ratio specified over the bandwidth it has to reject.
Termination of the signal cables cannot be done via the receiver ground system because
there certainly is a difference in ground potential between detector ground and receiver
ground. A potential difference of this kind can only be rejected if the output impedances of the preamplifier's positive and negative going line drivers remain exactly
equal which is never the case.
Small groups
connected to small power supplies located, if possible, inside the detector avoid
the delicate system handling associated with remote sensing and high DC-power. Power
rail impedance and ripple can easily be held at low values by the power supplies
themselves. In case of failure only small groups of amplifiers will go off. Fusing, switching
and monitoring will have to be done remotely which is convenient during running.
Natural groups are formed by all the channels on one receiver or all the channels
on one signal cable. A separate power supply for all natural groups is desirable
but expensive.
Local power rails for somewhat larger groups will not be a problem provided the departing
signal cables run along the same path to the same final destination.
b) Mains contamination
Switching power supplies will contaminate the mains lines in two ways. Nonsinusoïdal
currents distort the mains voltage as explained before. In addition the switching
frequency and its harmonics will also come through. Connecting several such supplies
to the same mains network creates random contamination because the switching frequencies
are randomly distributed in the 20-100 kHz range.
Compatibility regulations can be found in [4.IEC], [4.875], [4.871].
c) Power rating
The rating of a power supply is the most delicate point. In addition to the obvious
parameters defining power output specifications must include ripple, spectral purity,
environmental details, electromechanics, monitoring capability. Although good power
supplies are capable of delivering slightly more than the rated power at a 100% duty cycle it is always advisable to have some spare power available. For applications
drawing large power at small duty cycles special power supplies are needed.
Custom-built electronics will always have rather vague specifications concerning its
power supply. Extrapolations of power measurements on single channels should be avoided.
Power measurements have to be done using the final setup with all loads present including simulated transferred data.
d) Compatibility issues
Large experiments (should) have engineering committees that are supposed to deal with
electromagnetic compatibility which generally covers all specifications concerning
noise generation and noise susceptibility. A safety margin of non-interference should
be worked out for a particular experimental area. In physics research it is of the
utmost importance to calculate and cross-check interference levels because many circuits
will not work properly under experimental conditions although their performance on
the bench was satisfactory.
In particular the use of switching power supplies including their physical location
and connection should be investigated.
Parts of a detector might permit the use of switching power supplies. However, switching
power supplies should always be physically separated from any low-noise equipment.
They should not at all be used inside a detector area unless very good reasons dictate such an arrangement. Their implementation may require decoupling networks such
as indicated in [11] or a separate ground connection.
Series regulated supplies will come into use at all levels of noise sensitive equipment.
The dynamic properties of the power source are very good at the terminals of a linear
power supply. However, connecting leads will introduce impedance and hence severely degrade the capability of the power supply to swallow surges. Sense lines will
help keeping the voltage value but they cannot guarantee the same dynamic properties
at the load.
If power supplies are to be located inside a detector area they should be fed with
clean, if possible also prestabilized mains. Magnetic stabilizers with sine wave
output offer all desireable properties for protecting noise sensitive equipment against
power line noise and transients. This will be particularly important for setups with only
one mains distribution transformer feeding both low-noise and normal equipment at
the same time.
Series and parallel connection of power supplies should, wherever possible, be avoided.
Most power supplies cannot at all be connected in parallel properly. For series connection
protective measures against short circuit conditions (reverse voltage) have to be taken.
Parallel connection of power supply primaries creates a network with leading power
factor (compare with fig. 9). Switching operations on large groups of power supplies allow the stored reactive
energy in the mains network after the main power switch to decay in a damped oscillation
with its peak amplitude determined by the L/C-ratio of the network. Only the stray inductance of the cabling will contribute to the generally very small inductance
which means that powerful voltage spikes will occur during switching. Transient protection
rated for the total stored energy has to be installed.
e) Remote sensing
Sensing the power rail voltage(s) is done via a set of independent sense lines connected
to the voltage sense input of the power supply. Regardless of the length of the power
cables a constant (static) voltage will be put across the load. Dynamic load, however, will not see the same impedance as if connected directly to the power supply.
Power rail bypass capacitors (separate capacitors for the frequency ranges to be
covered) offer some spare energy during current surges.
Of course the total voltage drop, i.e. on the positive and the negative line, is not
to exceed the voltage range of the power supply. Very long (more than 15m) power
lines equipped with remote sensing will cause excessive drop and difficult voltage
regulation. No bypass capacitors should be put to the power supply end of the line. Routing
of the sense should be done together with the power lines.
4) Conclusion
A summary on the often mystic side of DC-power supplies for electronics in high energy
physics has been given. New ideas for minimizing cabling and volume for high-density
detectors are briefly outlined. Power supply principles are summarized, a typical
performance breakdown will allow useful comparisons in order to better understand the
way how to properly design a DC-power system. The following summary allows for a
quick crosscheck on any such setup.
Crucial parameters of a DC-power supply system and its load:
- DC voltages and currents
- add 25% to nominal load
- stability
- avoid gain dependence of electronics on supply voltage
- power supply ripple
- test electronics against supply ripple
- distance
between supply and load - keep at minimum
- power cable resistance and impedance
- keep as low as possible
- monitoring and remote control
- avoid direct DC-sensing
- interference levels generated
by supply - high for switchers
- ambient interference level
- check with spectrum analyzer
- grounding
- avoid double grounding at supply and electronics
- reliability
- mean time between failures, see specifications
- environment
- dissipation at supply needs adequate ventilation
- mechanics
- keep to supplies manufactured for crate standard
- safety
- start up, short circuit, transient suppression - check data
A careful study of the entity of power supply parameters will exhibit a very small
margin of viability.
Amendment 1:
Exotic power system for adhoc-solution of interference problem
Severe interference problems on broadband-amplifiers supplied by remote power supplies
feeding a DC-power rail 70m away inside a particular detector area [3] [10] pointed at a replacement of the power supply system as the only practicable solution
within a rather restricted time interval. It also suggested that the amplifiers have
never been tested against the interference levels present a detector. In addition to the very high frequency interference created by the passing-by
of charged particles close to the beam tube of the accelerator the magnetic stray field of the main detector coil (I=10kA, Iripple 100A@600Hz) caused by the magnet power supply ripple induced a powerful standing voltage on all lines attached to the amplifiers.
Having exhausted quickly all possibilities of interference cancellation a new power
supply system was conceived. It grouped the electronics "naturally", taking into
account the modularity of the cabling and the detector geometry. It also ensured
as short as possible power supply cabling and, above all, avoided low impedance connections to other systems.
Mains separation was accomplished by a power converter delivering a 172Hz-sinewave
for a set of miniature series regulated power supplies located inside the detector.
Using a frequency way above 50Hz enabled the use of superflat miniature transformers
inside the detector where the available volume was a given boundary condition [6]. Spectrum
analyzer measurements permitted a clear distinction between the interference inside
and outside the power supply system which was very important when the final cable
routing was tested for minimum interference levels.
The new configuration was found to have an interference level reduced by almost two
orders of magnitude. It ensured proper use of the full precision of all channels
throughout the lifetime of the experiment. However, it would have been much cheaper
to conceive a noise-immune system from the very beginning.
Amendment 2
EMI-levels G, N, K from 0.15 to 300 MHz (VDE 0875)
Please note that fig.11 shows noise power@120kHz bandwidth whereas fig.12 is calibrated
in noise voltage@9kHz bandwidth to be measured with a calibrated receiver. Ref. [4]
and [7] show how to determine interference levels. Electronics connected to, e.g.,
power supplies of a certain interference level can easily be crosschecked against the
levels in figs. 11,12 by means of artificially generated noise levels, preferably
inside an anechoic chamber.
Amendment 2:
EMI-levels G, N, K from 0.15 to 300 MHz (VDE 0875)
Fig 11 Noise power versus frequency according to VDE-levels
(measured at 120kHz bandwidth)
Fig 12 Noise voltage versus frequency according to VDE-levels
(measured at 9kHz bandwidth)
References
[1] F. Zach
Leistungselektronik
Springer, Wien 1979
[2] W. Majerotto, H. D. Wahl
Ein Meilenstein auf dem Weg zur einheitlichen Feldtheorie
Die Entdeckung des W- und Z-Bosons
E & M 101 Nr.8 (1984), 365-380
[3] J. Colas, J. Lacotte
A Low Noise, Large Dynamic Range Pulse Amplifier
NIM 176 (1980), 283 - 286
[4.IEC] IEC Publ.Nr.161
[4.871] VDE 0871/6.79, DIN 57871
Funk-Entstörung von Hochfrequenzgeräten für industrielle,
wissenschaftliche, medizinische (ISM) und ähnliche Zwecke
Radio frequency interference avoidance of high-frequency apparatus
for industrial, scientific, medical and similar applications
[4.875] VDE 0875/6.77, DIN 57875
VDE-Bestimmung über die Funkstörung
von elektrischen Betriebsmitteln und Anlagen
VDE-specifications about radio frequency interference
by electrical apparatus and installations
equivalent to 76/889/EWG+76/890/EWG and
CISPR 1, 2, 3, 14, 15, 16
[5] Z. Kulka, F. Szoncsó
Hadron Calorimeter Front End Electronics
for the upgraded UA1-detector
Nucl. Instr. and Meth. A 291 (1990) 587-594
[6] F. Szoncsó
The Power Supply System for the Gondolas
CERN internal report UA1/TN 84-92
[7] T. Birchmeier, E. Bolliger, F. Ineichen (Timonta AG)
Electro-Magnetic-Compatibility
Brochure available at Timonta AG, CH-6850 Mendrisio
[8] J. E. Foster
Electromagnetic Compatibility in Spacecraft
and Space Instruments
RAL-84-035 (Rutherford Appleton Laboratory, Chilton UK)
[9] The Case of an Input Transformer in a System of
Transformerless Supplies
Proceedings of Powercon 9, Power Concepts, Inc. 1982
[10] F. Szoncsó
Entstörung der Meßelektronik des Experimentes UA1
am CERN Proton-Antiproton Collider
Thesis at the Technical University of Vienna, Vienna, 1985
(not published)
[11] H. W. Ott
Noise Reduction Techniques in Electronic System
John Wiley & Sons, New York, 1976
[12] Keith Billings
Switchmode Power Supply Handbook
McGraw-Hill New York
[13] Donald. R. J. White
Handbook series on Electromagnetic Interference and Compatibility
DWC Inc., Germantown MD 20767, USA (1973, 3rd ed.1981)
[14] M. Ivanovici, J.-J. Morf
"Compatibilité Électromagnétique"
Presses polytechniques romandes, Lausanne (1983)
Écoles Polytechniques Fédérales de Lausanne et de Zurich
Fritz Szoncso, CERN ECP,Friedrich.Szoncso@cern.ch,October 1996