Here I give my personal thoughts and recommendations on
test equipment and other electronic tools I use either at
work or at home.
Accuracy and Precision
Accuracy and Precision are two words bandied around when there
is talk of test equipment. Often the two will be used incorrectly, either at
random or the wrong way round. Wikipedia has a nice little
of the two words. A brief summary:
- How close to reality a reading is.
- How repeatable the reading is.
Avometer 2001 Multimeter
AVO.... now there's a name. This rare little beauty was the first DMM I used in a
my professional career as a design engineer at
Thorn EMI Electronics. And I only
just recently found out that this was in fact made by Thorn EMI Instruments Ltd
Thorn EMI owned company.
|Function ||Range ||Accuracy|
|AC Volts ||200mV - 1000V ||±1%|
|AC Current||200μA - 10A ||±1.5%|
|DC Volts ||200mV - 1000V ||±0.25%|
|DC Current||200μA - 10A ||±0.75%|
|Resistance||200Ω - 20MΩ|| |
|Digits ||3½ ("1.999")|| ||
As you can see its a bit of an odd design: a large case with a small LCD
at the bottom and the three 4mm sockets at the top. There are two slide switches: one for
mode (off, ac, dc/diode, resistance/continuity), and one for range. I can attest that it is a solidly built unit,
and being through-hole construction it is a very nice instrument to service. And in practice
the LCD is very clear and responsive, far better than those cheap DMMs you can get for peanuts these days.
Fluke 8060A Multimeter
As well as a basic multimeter it is good to have a second meter to both give
you a second display and to provide more advanced features that you do not
always need for general bench and field work. The 8060A is a classic DMM,
with excellent true RMS AC reading to well above the audio range. And you
can usually pick them up for very little on eBay these days.
|AC Volts ||200mV - 750V|
|AC Current||200μA - 2A|
|DC Volts ||200mV - 1000V|
|DC Current||200μA - 2A|
|Resistance||200Ω - 3GΩ|
|Frequency ||20Hz - 200kHz (auto-ranging)|
|Digits ||4½ ("1.9999")|
Discontinued. Replaced by model 289.
Original list price: $499.
This meter is great for audio work. The dB ranges are normalised to
600Ω so can
read dBu directly.
And in Relative mode you can
measure dB directly, both in DC and AC modes. The frequency counter
is useful for spot checks (although I'd recommend a proper frequency
counter for anything more involved).
Hewlett Packard HP 3478A Bench Multimeter
Hand-held DMMs a great for general measurements both at the bench and when out
and about, or even just head-down inside a mixer. However, once in a while you
need higher accuracy than a hand-held, or perhaps a long-term experiment needs
monitoring that would otherwise exhaust the batteries in a hand-held.
There are many good examples of bench DMMs on the used market these days.
The obvious contenders are there: HP/Agilent and Fluke. From the HP stable comes
the venerable 3478A 5½ digit DMM.
While it only does AC/DC volts, AC/DC current, and 2- and 4-wire
resistance (it doesn't do continuity, diode test, frequency, temperature, capacitance, transistor hfe, and so on),
it makes measurements to a high precision and, when calibrated, to a very high
The worst-case DC volts accuracy is ±(0.044% + 41) on the 30mV range
after 1 year and within ±5°C of calibration temperature.
More typically, if it says "1.00000V" on the 3V range then the worst-case inaccuracy
is ±(0.019% + 2), or ±210μV.
Another nice feature of high-end DMMs is the input resistance. The HP3478A (and
its later incarnations) have very high (>10GΩ) input resistances on the
low voltage ranges. Useful for work on high-impedance or otherwise sensitive
And of course it has a GPIB connector on the back for hooking up to a computer.
Tektronix TDS340A Oscilloscope
The old saying goes: "There are only three names to look for in oscilloscopes:
Tektronix, Tektronix, Tektronix". Personally I'd modify that to "Tektronix, Tektronix, LeCroy" but
only because I spent the first few years of my professional life with a LeCroy 9420.
(I've never really got on with HP/Agilent scopes -- I guess its a LoveIt-HateIt thing)
For home use these days I've upgraded from an old Tek 468 to a newer digital TDS340A.
Same 100MHz bandwidth, but it has some clever triggering modes, can do FFTs for a bit of rough spectral
analysis (it is not a spectrum analyser!), and I can dump plots to a printer
or save them to disk for later documentation or analysis.
Documentation is easy to find on the net. Here are some files I've found and saved here for posterity:
A Word on Probes
The probe and the scope form a complete tip-to-trace test system.
There is no point spending good money on a
decent scope only to spend pennies on rubbish probes. So after hunting around the net for a
make and model of scope probe that would befit this fine scope (other than Tektronix probes) I've bought a couple of brand new
Probemaster 4905-2RA 250MHz probes.
While they may not meet the highest specs of Tektronix probes they are
perfectly good enough for my home use.
Why 250MHz probes? Why not 100MHz to match the scope? To understand why
you need to understand what it means to say a scope is "100MHz". In simple
terms, a scope's bandwidth is the -3dB point of its vertical channels. So a
100MHz scope measuring a 100MHz sinewave will display it with an amplitude
of 0.707 times the actual amplitude.
Now, that does not include the probe. The probe also has a -3dB point,
so the complete system of scope plus probe will have a -3dB point lower than
the individual components. For example, a 100MHz scope with a 100MHz probe
will have a -3dB point of around 60MHz. So it is worth splashing out that little
bit extra on
higher bandwidth probes so you don't limit the bandwidth of the
tip-to-trace system so much.
One nice thing about the TDS340A is that it has a Centronics printer port
and it talks DeskJet, LaserJet and Epson protocols. I recently picked up an
old DeskJet 500 printer for next to nothing (thanks Ben!),
and with a fresh ink cartridge
I can hit the HARDCOPY button and get a printout while I'm
working at the bench. Certainly easier than scribbling down some
measurements on a scrap of paper.
I scanned in a couple of scope plots during the
development of my dualfo.
And having a hard paper copy is great for making annotations or adding other notes.
Hewlett Packard HP 5334A Universal Counter
While a scope can be used to make frequency or period measurements of a cycle or two,
a proper standalone counter makes the job much easier and provides more precise
results and more ways to measure the signals.
It also frees up the scope for probing around the circuit looking
at waveforms while the counter keeps an eye on frequency.
HP (now Agilent) produce some really good counters. This one I picked up
on eBay a while ago:
One important feature of this counter over cheaper/simpler models is that
it uses the reciprocal technique. Simple counters measure frequency by
counting the number of cycles of the input signal in a given window of time.
With simple decimal scaling you get the result in Hz, kHz, MHz, etc (my old
Racal Dana 9902A did this). Unfortunately this means that measuring low
frequencies with any degree of precision can take some time: to measure a
1Hz signal to three significant digits requires a gate time of at least 1000
seconds. Yes, I know, I could measure the period and then calculate the
frequency from that with a pocket calculator. But dammit that's what
test equipment is for!!!
The reciprocal technique takes the opposite approach: the input signal
determines how long the gate is open to count pulses from a stable reference
oscillator (*). A CPU then does the maths to work out frequency. All it needs
is one complete cycle of the input to measure the frequency to a very high
degree of precision. In the case of the 5334A, to nine significant digits.
Taking the earlier example, with the reciprocal technique it will take a little over
one second to measure a 1Hz waveform. Very useful when developing
oscillators for example.
(* - for greater accuracy use an ovened oscillator, or a rubidium oscillator, or
a GPS-locked oscillator, depending on how much you're prepared to pay).
As the saying goes: "Power corrupts. Absolute power... is kinda neat."
Bench power supplies are so important in the electronics lab, and yet so often they
are hidden away, never discussed; the family cousins that everyone acknowledges
exist but doesn't mention in polite company.
I've designed, built, and used many power supplies over the years.
Expensive ones. Cheap ones.
High voltage ones and high current ones (sometimes both at the same time).
Small ones and heavy ones. Simple ones and complex ones. Most of the time though
you can get by with a basic dual bench power supply. Which I guess is why
they're so popular.
Hewlett Packard HP 6236B
Probably one of the nicest little bench PSUs for op-amp development. Two
outputs provide tracking bipolar rails up to ±20V limited to 500mA.
Plenty for op-amp circuits.
The third output provides a higher current 0-6V up to 2A, which can be used for
a digital rail or perhaps a control voltage.
Two meters provide voltage and current monitoring, switchable between the
All in all it is a nicely packaged little bench PSU:
These single (and their dual cousins) bench supplies are a common sight
in UK R&D labs. They're everywhere! In schools, universities, and labs,
they have provided stable supplies for many years.
The basic single output model (pictured below) can supply up to 30V at 2A,
and can operate
in constant-voltage with current limit, or constant-current modes. Very useful
for bringing up new circuits with the current limit turned right down!
Hewlett Packard HP 35660A Dynamic Signal Analyzer
Oscilloscopes are great for showing you what is happening to a signal in the
time domain. But often times you want to know what is happening in the
frequency domain. This is what spectrum analysers provide. They can
measure cutoff frequencies, resonant frequencies, and filter responses amongst
Spectrum analysers fall into two general types: swept filter and FFT. The
former is the classic analogue spectrum analyser, sweeping a filter over the
required range, then measuring the signal level coming out of the filter. Great
for radio frequencies, but due to filter response times not so good for audio
The first spectrum analyser I owned was a venerable old HP 141T, with
a range of plugins, including one that covered the audio range. That was a beast
indeed! One problem with swept filter spectrum analysers is that as you
reduce the video width to get more detail in the spectrum, because of the
delays introduced by the high-Q filters, sweep times could be measured in tens
of seconds. Tedious, especially for audio work where you might want a
video bandwidth of a few Hertz.
With the advent of cheap computing power came the FFT-based spectrum
analysers (or dynamic signal analysers (DSA) as they tend to be called).
sample the signal as-is, without any filtering, then apply the power of FFT maths
to extract and display the spectral content. This provides data much faster as
you only need to wait as long as one cycle of the smallest frequency you are
interested in. For example to resolve to 1Hz requires a 1 second measurement
period. More typically, to cover the audio range with a 25kHz span would take
about 15ms per acquisition, or 64 updates per second, i.e., realtime.
Note: this is theory. Actual machines may add processing time, or require
the sample record to be full before processing, etc.
I used an HP 3561A in my first job building high power passive audio filters,
hand-winding coils and building fan-cooled capacitor banks, tuning them with
the aide of the 3561A and its internal noise source to get the right cut-off
frequency and slope.
Nowadays I use an HP 35660A DSA at home for audio development. It has
the advantage of two input channels, so together with its built-in signal source
it can also be used as a network
analyser to show both amplitude and phase response of audio filter networks.
Hewlett Packard HP 8903B Audio Analyzer
Developing and testing audio electronics (for example,
often requires the measurement of signal noise and distortion to a degree that cannot
be easily achieved with oscilloscopes or multimeters.
Measuring these characteristics is best done with dedicated test equipment,
such as the industry-standard
Audio Precision series.
However they are rather expensive, so the hobbyist has three options:
- older-generation audio analyzers,
- home-brew apparatus
- PC soundcard and software
The PC soundcard option can be the cheapest option. I say "can be" since to
get worthwhile results you need a very good soundcard, ideally external to the PC rather than built-in to minimise the noise floor, and decent software that can analyse the
data properly (e.g., TrueRTA).
To measure distortion products it is also a good idea to
build or buy
an analogue notch filter so that you can remove the fundamental test tone and get the most
from the soundcard's dynamic range. You also need a low-distortion signal
generator to minimise any masking effects that a noisy generator would introduce
(a PC soundcard might be good enough).
And then you need to set it up in a way that minimises noise-pickup from the surrounding environment.
Building your own THD Analyser
can be very rewarding if you enjoy the challenge and have the abilities to build it
to the required high standard. It most certainly is not the cheapest option, especially
if you include your time. But, when finished, you end up with equipment that
you can tailor to your specific needs.
Finally, by far the quickest overall option is to buy a used audio analyser from a
elsewhere. Names to look out for include
Audio Precision, HP, Boonton, and Tektronix.
The 8903B is one such machine, incorporating a low-distortion sinewave
oscillator, a tunable notch filter, a frequency counter, and a sensitive RMS AC voltmeter. Each
of them on their own are very useful, but having them together in one instrument
with GPIB control makes for a versatile test setup. The photo below shows my
8903B driving itself at about 1kHz. The displayed distortion+noise is -87.43dB,
or about 0.0043%, with a 30kHz low-pass filter.
Hewlett Packard HP 3311A Function Generator
The 3311A is a strange little curiosity. I believe it was an attempt by HP to produce
a low-cost signal generator to compete at the lower end of the price spectrum,
probably aimed at schools and colleges and basic function generator duties at the bench. Which is a fine idea, but on closer
inspection you can clearly see that HP, at the time, just didn't understand
how to make something cheap: high-quality Allan Bradley pots, gold-plated
double-sided PCB, a sturdy diecast clam-shell case making servicing easy,
and high-quality panel furniture (knobs, binding posts, etc).
It sports a decent spec as well, covering the range 0.1Hz to 1MHz,
three waveforms (sine, triangle, square),
a separate TTL-compatible pulse output,
fully-adjustable amplitude and offset,
and even has connections on the back to allow modulation of the internal VCO.
Hewlett Packard HP 3325B Synthesizer/Function Generator
A decent lab needs a decent signal generator. While the 8903B has a low-distortion
sine oscillator suitable for audio measurements it is not very precise in frequency and only
covers the range 20Hz to 100kHz.
The 3311A has a wider range (0.1Hz to 1MHz), has a choice of waveforms, and output
amplitude and offset control, but it is only suitable for rough measurements as the
frequency is liable to drift with time, and the dial setting is not very precise.
So this is where a machine like the 3325B comes it. It is a synthesized function
generator, giving 11-digit frequency precision, referenced to a quartz crystal, covers the range 1μHz to 21MHz sine (less for
square, triangle and ramp, or up to 60MHz for TTL-compatible pulses), has a separate modulation oscillator for AM or PM duties,
both linear and log sweep modes, GPIB, and so on.
There are options to fit an ovened crystal for better stability (like mine
has), or use an external reference clock for greatest precision,
such as a GPS-disciplined 10MHz oscillator.
Hewlett Packard HP 8165A Programmable Signal Source
The 8165A sits between the 3325B and 3311A. It is certainly superior to the basic 3311A, although one could argue that the 3U case,
the noisy fan,
and the weight of the 8165A make it a tiresome beast compared to the lightweight
3311A. And when you need a simple source to tickle a circuit, the 3311A is right
there in moments, rather than faffing around with the 8165A's parameter punching.
At the other end of the spectrum, the 3325B is visibly superior when it comes to frequency
control, with seven more digits of precision than the 8165A, over a wider range
(down to μHz), and the option (as in mine) of a built-in ovened crystal
for greater stability. The 3325B also has
superior sweeping facilities over the full frequency range, as well as a built-in modulation signal generator, and can provide a TTL clock up to 60MHz.
And yet... I find the 8165A sits in a sweet-spot all of its own. In many ways
it is inferior to the 3325B, and yet in several ways it is actually superior!!!.
Its sine, triangle and square output waveforms cover the full frequency range of 1mHz to 50MHz (the 3325B restricts the square to 11MHz and the triangle and ramps to 11kHz). With Option 002 the 8165A can also do logarithmic frequency sweeps, and while the sweep profiles are greatly restricted compared to the 3325B, what it does have is very usable for electronic music instrument
development. And where the 3325B has a built-in modulation generator (a capable signal
generator in its own right) the 8165A can generate counted bursts -- very useful for
testing the response of VU meter drivers for example.
In many ways, unless you need the higher precision or extensive frequency sweeping
capabilities of the 3325B, then the 8165A is very close to the ideal bench signal
generator for the general electronics lab. And the really crazy thing is how little
they go for on ebay!