A junior engineer investigates noise problems with a cardiac imager and
learns the value of writing everything down and not
forgetting that the smallest fix may save a
lot of work
My first job out of college was with the Ultrasound Division of a
major medical manufacturer. At the time, this company was the leader in
Medical Ultrasound, and, as a tribute to its management, the company
these days no longer even has an ultrasound division.
I was hired into the development group for the Cardiac Imager, which
was about to be the first real-time dynamic ultrasound scanner. The
company’s flagship system was the ‘Lamnigraphic’ unit, which depended
on a probe mounted on an arm with encoders to generate a still image of
a slice of the body.
At this point in the mid -70s, that was all ultrasound did. The
Cardiac Imager mounted a probe (actually a crystal) on a oscillating
transport to give a constantly refreshed sector of the body.
The probe was actually a crystal which was used as both a send and a
receive transducer. The crystal was hit with a high voltage pulse to
make it ‘ring,’ then it was monitored as it picked up the returning
echoes. It was accepted knowledge at the time that the faster the
leading edge of the initial pulse, or ‘main bang, the higher the
output and the better the image. The standard pulser was a circuit
based around a 600 Volt SCR with a rise time in milliseconds. So when
nanosecond-reacting SCRs appeared in the late 70s, the cardiac
development group jumped on it and created a new, faster pulser.
The Cardiac Unit was released to rave reviews and as a reward the
development engineers were moved into sustaining engineering without
anyone bothering to ask. I did get a small development project working
on a chart recorder interface to the phonocardiographic channel. This
was an electronic version of the stethoscope and was featured on the
As it happened the unit I was working on had the new pulser in it.
Yes, it had the nice, faster edge, but the1khz repetition frequency was
all over the phone channel. Since that wasn’t really my task, I
disconnected the pulser. However, any good engineer will always try to
solve problems, even if they aren’t his. After a couple of weeks, I
dove into the noise problem, and, as it was my first foray into this
area, learned a lot. The main thing I did learn that applied to every
noise problem I’ve ever had to solve was this: Write Everything Down!
Because of the high voltage, the pulser was housed in a solid
aluminum box with only DC power & frequency drive in, with an SMA
output. A coax cable with an SMA connector ran from the box to the
front panel where the transducer cable plugged in. The circuit was
designed such that removing the load, unplugging the SMA, and the
pulser ceased to work. So it was exactly ‘divide and conquer’ but
more like ‘probe and pray.’
From the start it was obvious that pulser noise was everywhere.
However, the proximity of the card cage to the pulser nor the cable
routing seemed to matter. This seemed to be a grounding problem.
The original grounding system was designed for the lazy risetime of
the old pulser, not much care was taken to separate chassis ground from
signal ground. Redesigning the grounding system would involve a
redesign of the main power supply and two motherboards, just as a
start. So I turned my attention on how the pulser noise was getting
into the phono channel.
Considering how this was essentially an audio channel in a digital
and high voltage system, it was probably not good engineering practice
that the phone channel was just another card shoved into the card rack.
It did, however, have a differential amp on the front end and on the
face of it, the phono pickup ran differentially from the front panel to
the card. That is, until I examined the backplane and discovered that
the negative input of the differential amp was tied solidly to ground.
So I’d come full circle back to the place where grounding was the
problem. After several days of trying to redress cabling and reassign
ground points—and a good number of pages in my notebook—nothing seemed
to budge the noise.
The first breakthrough came when I cut the ground connection to the
negative input. This necessitated hooking the shield of the coax
directly to that input--not the ideal solution. This made a small
decrease in the noise, so I now knew what I was dealing with. After
scouring the engineering department, I found some shielded twisted pair
and replaced the coax. This contributed a further small reduction.
Probing the inputs of the differential amp, I still noticed that the
noise on the negative input was larger than on the positive input, so I
still had not made the noise common mode.
My main problem now was firmly attached to the front panel. The phono
sensor jack, I noticed did not really use the two pins, but rather one
pin and chassis ground. Did I mention that the system grounding was
messed up? Then the sensor itself used the same method, shielded
twisted pair wire where they used one conductor and the shield. At
least they were consistent.
At that moment, I was looking at a mechanical mod to the front panel
and a consultation with the company that provided our phono sensor—and
a nice history of back-incompatibility.
So I tried something stupid. Here’s the reasoning. I had a low
frequency audio channel. Phonocardiology really looks at sounds less
than 100hz, with most heart problems topping out at 800hz. This was a
little too close to the 1khz noise, unless one realizes that although
the noise was at that rate, the individual pulses were sharp and high
frequency. Then I knew I had an imbalance between the two inputs of the
differential amp. Simple! Just put a small high pass filter to route
some of the noise from the negative input to the positive input and
make it common mode.
It was stupid enough to make it work. With just a single resistor
and capacitor, the noise was reduced to the level where it could just
barely be heard with the gain all the way up. Also, I now had a nice
fat journal of what didn’t work—and what worked. I then returned to my
project. Being a junior engineer and steadfastly ignorant of office
politics, I had no idea that what I had just done for my own amusement
was important to anybody else. The ‘L’ unit still used the old pulser
based on the premise that ‘if it ain’t broke, don’t fix it.’
A few months later, a move took place a couple of layers in the
hierarchy above me, where someone started asking the question as to why
we were building two different pulsers for two very similar units. The
‘L’ group shrugged, did about 15 minutes of testing and wrote the ECO.
The new, faster pulser was now being installed in all ‘L’ units and
nobody thought to look, or rather, listen to the phono channel.
Until an ‘L’ unit arrived at a prestigious New York hospital where
the doctors found that all they could hear was a 1khz whine. It seemed
like the whole management structure was running in circles screaming,
“Oh Shoot! Oh Shoot! Oh Shoot!”
I did not realize what the problem was until I accidentally walked
into a panic meeting the Director of Engineering was holding in the
hallway. After listening for 5 minutes, I told them I had fixed that
problem months ago. There was dead silence. Being a junior engineer, I
did not take advantage of the opportunity to remind management why they
should involve some engineers in their decisions. However, I did
demonstrate the fix. A complicated fix it was, too! A resistor and a
cap were added to the phono board and the input coax was replaced with
shielded, twisted pair cable.
The fact that a junior engineer had fixed the problem before it
existed was mind-blowing, especially to the Director of Engineering. He
wanted to know who would ‘vouch for‘ this solution. I was a bit
surprised when one of the veteran sustaining engineers ‘stood up’ and
said he’d take responsibility for it.
And the rest was just documentation.
Of course it was not the ideal solution, but it was good enough for
a peripheral function. It was so good that the actual grounding
problems were never addressed.
Author Steve Tomporowski graduated from Northeastern University with a BSEE and the University of New Haven with an MSEE. After working for three different medical manufacturers, he's currently Sr Electrical Design Engineer at Kaman Precision Products.