We continue in this set of
tutorials for the audio test software suite, Visual Analyser. So
far, we've had:
In this article, we look at
calibrating the system. If you haven't gone through the previous tutorials,
I'd suggest doing that first.
Without calibration, VA can measure levels in the only
language it understands, proportion of full scale. After
calibration, VA can translate that to real-world quantities like Volts.
Much more useful! Indeed, given the computing power available in
even the most modest computer, really useful!
Before calibration, I did some tests. The test
bench computer sound-card is a plain vanilla SB5.1, so certainly nothing
special. I wanted to know at what input levels it runs into
clipping, to make sure I remained well below that level. I fed
both channels from my test bench (hardware) oscillator, and monitored
the level with a Rigol DS1053 Digital Storage Oscilloscope and on
speakers. Input clipping of the soundcard occurred at 5.92V
peak-to-peak. Interestingly, it was a pretty hard form of
clipping, visible on VA, but also on the external DSO and audible in the
speakers. So it was actually killing the oscillator. I
imagine it's some form of input clamping to protect the soundcard.
I'd done that test with the sound card's input gain set
low - I wanted to detect hardware clipping, not running out of bit
range. So next question was when do I run out of range? I
increased the input gain to max, and found that the maximum input signal
was now 2.9V peak to peak, or 1.02V RMS. With VA's Zoom channels
set to 1 (minimum zoom) the wave takes up most of the screen. Any
increase in level shows as clipping on the screen. The two
headroom indicators (bargraphs) in the Scope input channels showed
levels to be 0.6dB down from full scale. I'd definitely used up my
But reducing the soundcard input level doesn't really
help that situation. VA clips on Zoom 1 at top and bottom of
screen, no matter what level that really represents at the input.
And setting the input gain lower will probably have the
undesirable effect of bringing up the noise floor - no real issue to the
Scope function, but probably significant to the FFT (spectrum analyser)
and Noise & Distortion measurements.
So, what to conclude? Clearly 1V RMS is the max I
can input if I leave the input gain at Max (100%). There's a lot
to be said for doing that - you are in no doubt when you are at max,
while setting anywhere else is not so reliable. So I'm going to
calibrate for 100% input gain and see what trouble I get into.
What levels can we expect anyway?
Standards, as someone once quipped, are so useful;
that's why we have so many of them! And that definitely applies to
audio. For example, professional audio levels are usually +4dBu or
+6dBu, depending on where you are, and domestic ones -10dBv. Since
our sound card is of the domestic variety, that implies levels peaking
to around 0.316V RMS, leaving us 10dB headroom but still a theoretical
86dB above the noise (assuming 16 bit operation). Should be safe.
Preparing to calibrate
We're going to need a sine wave of convenient and known
level, somewhere well within the level and frequency range of the sound
card. I used a combination of my hardware oscillator and my Rigol
DSO, as it has True RMS, Peak and Peak-to-peak measurements where my
multimeter doesn't. I do have a home-made Noise and Distortion
analyser/millivoltmeter, but it has an analogue meter, so I can't rely
on that. I think everyone is going to be a bit different on this
one - it will be a case of looking at your options and picking the one
that you have most confidence in. Some calibration source options:
Use VA's signal generator, monitored by your best
multimeter. Make sure to pick a frequency within the
multimeter's range! Your local mains frequency seems like a
Ditto, but using a mains transformer feeding a
Build the calibrator from the Nuova Elettronica
magazine article. It appears intended to generate a reliable
The probe calibrator on your hardware oscilloscope?
You might be able to think up other ideas - let me
(*Interesting aside. You'll often see cautions
about the poor frequency response of your multimeter. I think this
might be old-fashioned advice - most multimeters I have checked have a
pretty wide and flat response. But you probably can't go wrong at
Have VA running in A and B mode
Input the known signal mentioned above either to
both channels, or to the left channel at first
Tick the Values box on each channel so you can enjoy
watching the calibration kick in. At the moment, all the
amplitudes will be shown as %fs.
Open VA's Settings, and choose the Calibrate tab
Select your input source from the drop-down list
Use the vertical fader(s) to set the level at which
the soundcard is to be used (I set 100%)
Choose the Unit you want to calibrate (I chose Volts
Enter the level of the calibration signal being
applied to the inputs (I chose 1.00V as read on the DSO.)
Press Start Measure Signal (L). A few seconds
elapses and a value for Detected Levels (% full scale) appears.
Tick Apply calibration left channel
Now check the Values columns - you'll find the
amplitude values are now reading in volts!
Repeat for the right channel
SAVE! If you don't save, you will loose this
calibration when you close down the program. Press the Save
button near bottom left of the Calibrate window. It will
prompt you for a name. I chose Volts RMS 100%, but I
don't really know what you should consider as the most important
Now I pressed Save Config (top left of window) as
well. My guess is that this tells VA that you want it left
with that particular calibration chosen (assuming you have a range
of saved calibrations). Note that I could also have chosen to
save as a specific config. I guess time and experience will
suggest why you might do that.
Windows 7 - VA 2012 Bug?
The process above worked fine with my test-bench
machine, using an SB5.1 card running under XP. But I cannot so far
get it to work on my office/lab machine, using its built-in Realtek
high-definition soundcard and running under Windows 7. The machine
appears to calibrate and save, but once relaunched the calibrations are
clearly not reapplied properly. However VA version 2011 seems to
calibrate fine on this machine. This might suggest a bug in the
2012 Beta version. (Alfredo?)
I can get around this in the meantime by recalibrating
any time I need to depend on absolute values. Alternatively,
calibrating at a precise 1V level appears to work.
Test your calibration
In theory, you should now be able to close down A,
relaunch it, restart the generator (if you were using it) and the
analyser, and the calibrations should still be in place. Go
Let me know if that doesn't work for you, especially if
you can find a solution!
Enough for now
Because there appear to be some issues with the
calibration function at least with some machines, I might leave it there
until we get resolution of those. There are a few remaining
buttons and options, but they don't appear essential to the operation.
If you delve a little deeper and want to tell us what you found, do get
It will now be tempting to look into the Voltmeter
option. Next episode!