Service Manuals, User Guides, Schematic Diagrams or docs for : Agilent Calibration of Precision Step Attenuators - White Paper 5991-1226EN c20140515 [14]
<< Back | HomeMost service manuals and schematics are PDF files, so You will need Adobre Acrobat Reader to view : Acrobat Download Some of the files are DjVu format. Readers and resources available here : DjVu Resources
For the compressed files, most common are zip and rar. Please, extract files with Your favorite compression software ( WinZip, WinRAR ... ) before viewing. If a document has multiple parts, You should download all, before extracting.
Good luck. Repair on Your own risk. Make sure You know what You are doing.
Image preview - the first page of the document
![Calibration of Precision Step Attenuators - White Paper 5991-1226EN c20140515 [14]](/pics/5c4d6d0026430.gif)
>> Download Calibration of Precision Step Attenuators - White Paper 5991-1226EN c20140515 [14] documenatation <<
Text preview - extract from the document
Keysight Technologies
Calibration of Precision
Step Attenuators
White Paper
Abstract
This paper describes an automated parallel IF substitution system for precision
attenuator calibration which has been in use for over 15 years and presents results
of tests made on some very accurate attenuators.
The calibration system was originally developed to meet the calibration needs of
a step attenuator in a new synthesizer. Years of data have shown this attenuator
to have phenomenal accuracy. The system has been continually refined and is
now used for high-accuracy calibration of several other types of step attenuators.
The data presented establishes confidence in the accuracy of the system and the
attenuators.
Presenter and Author
Bill Bruce
Hewlett-Packard Company
1998 NCSL Workshop & Symposium
Introduction
During 1978, there was a need to verify the accuracy of an excellent
attenuator in a new synthesizer product. The author attempted to calibrate
one attenuator using the existing manual method and found it to be
an exasperating process. The natural consequence was to develop an
automated calibration system. The system has been in use since that time,
and with some refinements over the years, it has been used to calibrate
a large number of very accurate step attenuators. This paper describes
the system and some of the considerations necessary for achieving high
accuracy. One particular data set has become more or less standardized
and many attenuators have been measured using this data set. Graphs
and histograms of some of the data from approximately 90 attenuators are
presented. The data demonstrates the capability of the system and also
indicates that a well-designed attenuator can exhibit very small errors.
Description and Basic Operation of the System
The method chosen for the automatic calibration system is parallel IF
substitution[1]. This method has the capability to achieve excellent accuracy and
fairly wide dynamic range. A programmable ratio transformer is used as the
system standard. The system capability covers frequencies between approxi-
mately 0.3 MHz and 80 MHz, and attenuation between 0 dB and 100 dB.
With some re-connection to a non-mixer configuration, 1 kHz is also a usable
frequency.
Figure 1 shows a block diagram of the system. Its operation can be compared
to a superheterodyne radio receiver which has a 1 kHz IF and has the capability
to precisely measure changes in input signal level. A doubly balanced mixer
is fed by two signals, one from a local oscillator at a fixed level, and one from
an RF source, which is phase locked to the local oscillator. Figure 2 shows a
photograph of the system.
Coax choke
Synthesizer
local oscillator
Coax choke
Phase
lock Mixer Lowpass Coax Matching HP 3581C
filter switch transformer detector
Synthesizer DUT Pad
RF source
DVM
Attenuator
AC source w/ Ratio Isolation
remote sense transformer transformer
HP-IB to instruments
RF section
Audio (1 kHz) section
Digital section PC controller Printer
Figure 1. Block diagram of system
3
Figure 2. Photograph of system
The signal from the RF source is passed through the attenuator under test
before it goes into the mixer RF input. Appropriate padding in the signal path is
used to minimize uncertainties due to impedance mismatch. The mixer output
should be terminated in its characteristic impedance for all frequencies exiting
the mixer. Also, some low-pass filtering is done on the mixer output to prevent
LO feedthrough and unwanted mixer products from causing abnormal response
by the detector. These two requirements are filled by a simple "constant Z" two-
stage low-pass filter, with the first stage fc at 32 MHz and the second stage at
32 kHz. Figure 3 shows the filter details.
2.5 H 250 H
R = 50 R = 50 R load= 50
Zin = 50 1000 pF 0.1 F
R+ 1 R + SL
SC
Zin = (For each section)
R + 1 + R + SL
SC
R = 50
1 R 2R
It can be shown that if Zin = R, then + = .
RC L L
L
The result, R 2 = will result in constant input Z for all frequencies.
C
Figure 3. "Constant Z" low-pass filter design
4
The detector input is routed through a switch that allows selection of either
mixer IF output or output from the ratio transformer. The ratio transformer is
driven by a stable AC voltage source at approximately 1 kHz, with remote sens-
ing at the ratio transformer input. The load regulation of this AC source must
be adequate to negate the effects of varying reflected load as the ratio of the
transformer is changed. The burden on the output of the ratio transformer is a
100 k to 50 resistive divider.
The general operation of the system is described by the following sequence.
First, with the device under test (DUT) set at its reference position, the IF level
is read by the detector. The reference setting is normally zero, but other settings
can be accommodated. Next, the detector input is switched to the output of the
ratio transformer and the ratio is adjusted to obtain the same detector reading,
within a defined window. Then, with the DUT set at the first step to be tested
(10 dB for example) the matching process is repeated and a second ratio is
obtained. The attenuation is then 20 log (ratio2/ratio1).
Note that the above sequence produces error data for the attenuator under test
with the errors at each setting referenced to its zero (or other specified) setting.
The actual value of its insertion loss when set at zero is not measured. If this
were required, a reference would have to be set using a lossless "through"
connection, then the test device connected and measured. The requirements for
this system were to measure errors relative to the zero setting.
Additional System Considerations
The parallel IF system uses a programmable ratio transformer as its standard,
operating at a frequency of approximately 1 kHz. The ratio transformer is
the backbone standard of the system, in that all measurements are derived
from settings of the ratio transformer. It must be periodically checked against
a traceable standard. The practical range over which a six-decade ratio
transformer may be used to match IF levels is determined by the desired
resolution and the nominal ratio at which it is being used. For a resolution
of 0.001 dB, a range of approximately 40 dB can be covered. In order for the
system to calibrate DUT attenuator settings above 40 dB, it is necessary to set
a second reference level using a step at or near 40 dB which has already been
measured. This is accomplished by increasing the input signal to the DUT by
40 dB, which results in a detector level very close to that used for the first 0 dB
reference. Note that for DUT settings of 40 dB to 80 dB, the detector and mixer
are operating over the same range that was used to test 0 dB to 40 dB. The
error of the second reference setting is then added to the measured errors of
subsequent steps to reference them to the zero setting.
5
An additional 20 dB of DUT attenuation can be measured if the lowest detector
signal is sufficiently above its noise floor to provide an acceptably small error
contribution from the noise. With the system parameters used, this is possible,
allowing a total range of 100 dB, with some degradation of uncertainty near
100 dB. To cover the 80 dB to 100 dB range, the AC calibrator driving the ratio
transformer is switched down one range (20 dB). Normally, the uncertainty
requirements for large values of attenuation are less stringent than for low values
of attenuation and fortunately this fits well with the system's performance.
After all the steps have been tested using a reference, the reference level is
measured again to determine if it has changed. Repeatability of the reference
is part of the uncertainty budget, and it must be within acceptable limits. There
are several possible causes for changes in the reference level, involving both
the device-under-test and the system. Some of these are drift of the RF source,
non-repeatability of switch contacts, either in the DUT or the system, change in
connector resistance, and self-heating effects. The problem of non-repeatability
of contact resistance is always present at some level in step attenuators. Figure 4
relates changes in contact resistance to change in attenuation. Once the system
has been well characterized, the reference change is an excellent indicator of
switch contact problems in the DUT.
In a perfect 50 system, the output will be exactly half the generator voltage.
50 W
Vo1 50
Vo 1 =Vg
50 + 50
Vg 50
Now if a small resistance, R, is introduced in series with the generator, what is
the relationship between its value and the change of the output (in dB)?
50 R
Vo2 50
Vo 2 = Vg
50 + 50 + R
Vg 50 Vo 1 100 + R
=
Vo 2 100
100 + R
dB = 20 log
100
dB Results
100 + R
10 20 = For .001 dB, R = .0115
100
For .01 dB, R = .115
dB
20 For .1 dB, R = 1.16
R = 100 10 -1
Figure 4. Analysis of effect of contact resistance
6
Switching of the detector between the IF and ratio transformer was initially
accomplished by an expensive high-quality SMA coaxial relay with excellent
isolation. It was found that although it had great isolation, the repeatability of
the contact resistance of this device left something to be desired. Data was
obtained on relay contact performance using a precision DMM to make four-
terminal resistance measurements at its connectors. Since the IF frequency of
1 kHz does not really require 50 ohm geometry to be preserved, an inexpensive
power relay with large contact area mounted in a shielded box was found to
provide superior performance. Contact resistance repeatability was evaluated
and the calculated switching uncertainty in a 50 ohm system was predicted
to be well under .001 dB. Isolation was also found to be sufficient. Without
sufficient isolation, either the IF or ratio transformer output signal appears at
reduced level at the detector input when measuring the other and can cause
problems.
The use of a 1 kHz IF frequency requires the use of high-quality synthesizers for
the system. Unwanted sideband energy close to the carrier frequency caused
by phase noise of the synthesizers must be at a low enough level to not cause
adverse effects.
In order to minimize ground loop currents in the system, two coaxial chokes[2]
are used. Each choke consists of about 16 turns of .141" semi-rigid coax wound
on a toroidal form. The choke in the RF path between the DUT and the mixer has
a powdered iron core and the choke in the IF path ahead of the detector has a
high-permeability tape wound core. Another measure to protect against ground
loop currents is a double-shielded isolation transformer that is inserted between
the ratio transformer output and the IF-Ratio switch.
Ground loop currents in the outer conductor of the coaxial path between source
and detector can cause errors in measured attenuation, particularly at low
frequencies and large attenuation settings. Reference [3], page 218 has a
discussion and analysis of this so-called "classical attenuator" problem. The same
problem has also been observed as occurring within some step attenuators,
probably due to some sort of internal grounding problem.
The detector used (HP 3581C) has automatic frequency control, which centers
its IF passband on the incoming signal. This allows the IF and ratio transformer
output frequencies to differ slightly without affecting detector performance. The
detector has a selectable IF bandwidth, which is normally set at 30 Hz. Also,
an audio transformer is used to improve power transfer to the high-impedance
detector input from the 50 ohm system. Detector linearity has no effect on
system accuracy because it is only required to measure two nearly identical
levels, the IF and ratio transformer outputs. The system must be tailored so
that the lowest level measured by the detector is sufficiently high such that
its input noise (which RMS adds to the signal) is sufficiently below the input
signal. The detector DC output is read by a DMM which sends its readings to
the control computer via the HP-IB interface. The HP 3581C does not have a
remote-controllable input attenuator, so some operator involvement is needed
when level changes are made. A programmable 50 ohm step attenuator ahead
of the matching transformer could be used to eliminate operator involvement,
but it would decrease signal-to-noise ratio because the detector would always
be set at maximum gain.
7
The main requirements placed on the detector are that it be repeatable (short-
term stability), have sufficient resolution and sufficiently low noise, in addition
to the AFC capability mentioned above. A possible way to improve the system
would be to phase lock the AC source and synthesizers together and use some
sort of lock-in amplifier as a detector.
Mixer non-linearity will contribute directly to system errors, and therefore the
linearity of the mixer must be checked over the range where it will be used.
Fortunately, this is possible using the ratio transformer and a step attenuator
which only needs to exhibit excellent repeatability. The low-frequency limit of the
system in IF mode is dictated by the ferrite core transformers inside the mixer.
Various step sizes can be used for linearity verification. The apparent attenu-
ation of a given step is measured at many different input levels, spaced over
the mixer dynamic range of interest. Differences in the apparent attenuation
which exceed the attenuator's repeatability are attributed to mixer non-linearity.
Again, the lowest level used must be sufficiently above the detector's noise. The
high-end limitation of usable mixer dynamic range is due to compression. The
maximum dynamic range over which the mixer is used in this system is slightly
over 60 dB, and its linearity error is approximately .002 dB or less. The software
must insure that the mixer is always used in the range over which it has been
characterized. The uncertainty analysis assumes that the mixer linearity follows
a pattern, which doubles its uncertainty for the second reference.
When each IF and ratio transformer output level is measured by the detector,
at least 20 measurements are made and simple statistical methods are used to
help assure data integrity.
Summary of Uncertainty Analysis
The following table summarizes the various contributions to the uncertainty
budget for the system (not including mismatch effects). Each number represents
approximately 2 standard deviations.
Table 1. Summary of uncertainty analysis.
Range 0-40 dB 40-80 dB 80-90 dB 90-100 dB
dB dB dB dB
Ratio transformer output negl. negl. .002 .002
Ratio-IF-match .0014 .002 .0028 .0028
Mixer linearity .002 .004 .008 .008
Reference drift .001 .001 .001 .001
Noise .0006 .0006 .005 .03
RSS sum .0027 .0046 .0100 .031
Analyzing the uncertainty of the system has not been straightforward because
it is very difficult to separate and quantify each source of uncertainty. The
system design approach and careful assessment of the system's various error
mechanisms has resulted in an analysis of system uncertainty which seems to
agree very well with practice.
8
Uncertainty Caused by Impedance Mismatch
After it has been calibrated on the system, the return loss of both ports of each
DUT attenuator must be evaluated and an estimate made of the uncertainty in
attenuation caused by impedance mismatch. Sometimes this uncertainty will be
greater than that of the system. In this case, the uncertainty of the calibration
must be increased to include the effects of mismatch. We have been assuming
worst-case phase conditions for calculating mismatch uncertainty, although a
less conservative approach seems to be acceptable. A practical and verifiable
level of return loss for the input and output ports of most step attenuators
seems to be somewhere around 40 dB. This is equivalent to a VSWR of 1.02 to
1. Modern network analyzers are easily able to verify this return loss (actually
their capability is quite a bit better than this). With 40 dB return losses at both
ports of the system and both ports of the device under test, and with the device
under test set at zero, the worst case uncertainty due to mismatch is predicted
to be about ◦ Jabse Service Manual Search 2024 ◦ Jabse Pravopis ◦ onTap.bg ◦ Other service manual resources online : Fixya ◦ eServiceinfo