Maker Pro
Maker Pro

re problems with Ohmeda-Biox 3100 pulse oximeter


W. Curtiss Priest

Jan 1, 1970
W. Curtiss Priest, Director, CITS
Center for Information, Technology & Society
466 Pleasant St., Melrose, MA 02176
Voice: 781-662-4044 [email protected]

Ohmeda Biox 3700 Pulse Oximeter Technical Note
June 9, 2006
Revised February 10, 2007

Failure Modes:

I. Probes fail
II. "Low Signal" problems
III.Calibration trim pot goes out of range
IV. Power/Standby off/on failure
V. Lead-acid battery life

Keyworks: repairing oximeter, fixing oximeter, oximeter probes,
oximeter batteries


One reason "hospital quality" medical instruments are expensive
is because such instruments must meet quality standards regarding
safety, and be FDA approved for use. The sale/purchase of medical
equipment requires a physician's authorization.

In recent years there has been a gray zone about various
medically-related devices. As a result, there are lower
cost pulse rate devices and oximeter devices. These devices
typically lack the "alarm features" of a pulse oximeter intended
for medical use and lack, say, abilities to record data internally
or, via RS-232 to an external computer.

In general, repairs to hospital equipment should only be made
by qualified facilities. This ensures the integrity of the
device. However, such devices are now regularly sold on the
used market, both on eBay and various Internet-based sellers.

Also, in general, no hospital or treatment center would
purchase such second-hand devices. Should a patient be
harmed via the failure of such a device, the care center would
likely be found negligent.

So. The information provided here is intended only for repairs
for the 3700 only for non critical care or use. In our belief
that perspective patients benefit from gaining knowledge about
their bodies, we would rather see a valuable pulse oximeter
repaired and used, rather than discarded.

Thus, this article contains no medical advice and the details
are strictly informational. We believe the information to be
reasonably accurate, but we do not assume any risks that
the misuse of this information may entail. In summary, do
not either purchase an unknown instrument nor modify it if
the intended use might be injurious or life threatening. We
describe what we have done which is appropriate for our
environment and situation.


It is typical for two or three problems with an electronic
device applies to the majority of the units built and sold.

When it was actually cost effective to repair electronics,
the repair service depends on this because while the
experienced repair service knows these faults and can often
fix a unit in a few minutes. This compensates for a unit
with an unusual failure which might occupy a full day of
diagnosing, and which, after that, might just be returned
to the customer as "not repairable."

I have a DIY approach to repairs. As I do not do repairs
for a living, I typically hold on to a piece of equipment
for many years, and obtain the service manuals. So, if
there is a failure and the "groups" have not discussed it
(especially, see, and
I must spend extra time to sleuth the problem, and, then as
the groups have given me useful leads on some problems, I
publish my results, a technical note, as a gift.

Further, some people do crossword puzzles, but I like
the "natural" puzzles of fixing something. When I had
spare time (I wonder where that has gone to), I might
get something at a flea market, to add to a museum collection,
and repair the item even if I might never use it. The
challenge of the hunt and the "solve" is what I enjoy.

My nonprofit Center conducts sleep apnea research. We plan
to offer a DIY "self-test" kit for people, concerned about the cost
of a $2000 sleep study, wanting to know if it is wise to get such
a study. So, we have monitored patients in our research and one
valuable instrument has been the pulse oximeter. We received ours
as a gift, in 1992.

The overall quality of the device has been fine, considering
that at $3000, sales are limited to the thousands. And while
today there are other units that sell for $100-200, the Model
3700 stores an 8 hour night of data in memory, and can be asked
to dump the data to a PC via its RS-232 interface. As one seldom
finds one of these instruments with its manual, for those
interested the output "MODE" is 1200,7,O,1. To retrieve
the internal data:

1. Hold the SaO2 Trend 20/60 on Power ON
2. Let the machine self test
3. Hold the SaO2 Trend 20/60 for 3 seconds
4. Press SaO2 a second time to start the output

The data are the readings for every 12 seconds. To exit during
output, press the SaO2 key again. The output looks like:

SaO2=XXX PR=XXX YY (YY -- is an error code)

An alternative to internal recording is to record directly
from the serial port while in use. We have a B&W laptop PC
that we dedicate for this. We use the shareware DOS Procomm
program, recording RS-232 in in ASCII to a named file. In
the morning we write the file to a 1.44 floppy.

To chart SaO2 on 8 1/2 x 11 paper in landscape mode, we post-process
the output file with a program we wrote. We then ask "Cricket Graph" to
chart the data. And, we send the chart to the Laser printer.

As there are too many data points for 11" paper, our program
takes 9 data points and pulls out the lowest SaO2 reading during
that period. As oxygen lows are a critical number, this proceedure
provides easily observed charts. Also, as REM periods critically
affect sleep, we prefer recording to the PC as the patient may
sleep longer than 8 hours. Should anyone wish a copy of our
program (written in C), please write the author. And, as
I believe Cricket Graph is "oldware," you may inquire about that too.

Difficulty of repair:

As we have seen five failure modes, and some related failures,
this is difficult to describe.

For someone with little electronics construction skills, the
information about the probe may help the user know whether to
purchase another.

For battery life and substitution this is fairly easy. It
does require some carving of plastic to accomodate the
different "form factor" of a single 6 VDC battery.

For, the off/on repair, purchasing the relay and replacing it
is of moderate skill. For repairing the existing one, this
is of greater difficulty, and requires steady hands.

For hunting down a thermal problem related to the calibration,
this is difficult. And, while we truly believe that if "one
does it, they all do it," we are uncertain about whether TI-
made LM358's, with a manufacturering date of around 1992 (the
date number "92," say, is above the chip number all develop
a thermal failure. As NTE does not supply a substitute, finding
another requires a Google search for providers of out-of-date
ICs. If anyone finds the same fault, kindly e-mail [email protected].

I. Pulse Oximeter Probes

In the second month, under warranty, a probe failed. It was
replaced, so I have no information about the failure mode.

Recently the unit grew fussy, indicated low signal from the
probe, and sometimes thought the probe was off the patient
when it was still on the patient.

A new probe is $250. A refurbished probe is $175. A
refurbished probe, with exchange, is $125.

Five years ago, the unit was a little fussy. Finding
that a little more pressure helped, we fitted the probe
with an O-ring located 3/4" from the pivot point. We
lightly notched the edges of the probe to hold the ring
(1/8" thick, 7/8" I.D.). We also used some silver colored
hot melt glue to assure the ring didn't wander.

This year the probe was fussier. In observing the
quality of the signal it is important to know that the
unit processes many readings from the probe and produces
a delayed output about the signal quality. To understand
why, it is useful to know how the probe senses both oxygen
desaturation and pulse rate.

I am indebted to various group discussions for an insight
about this. I had thought, perhaps, the pulse rate was
taken using, say, a "pressure sensitive" rubber. However,
both readings are derived from just 2 LEDs on one side of
the probe and one phototransistor on the other side.
One LED lases red; the second lases infra-red.

In an era of digital meters, there is still a very good
purpose for a handheld analog meter. My treasure is an
NCR Model 310-C. The probes of the meter have a nice
bite to them; they are better pointed than standard probes.

Now, in testing LEDs and phototransistors it is important
to know that the "resistance" on an analog meter with
a semiconductor is not simply a measure of resistance, but
the flow of current in a circuit which applies a known
voltage. These meters apply 1.5 VDC in the X1, X10, X100,
X1K, position and 9 VDC (or 15 VDC) in the X10K position. On
the NCR, the 15 VDC is applied in the X1K position.

This voltage is important because a simple "PN" silcon
junction does not conduct until at least .6 VDC is applied,
as that is the "avalanche" voltage for a typical silicon
diode. And, the actual observed resistance also depends
on the resistance of the meter coil. So, when you "test"
diodes such as LEDs it is good to use the same analog meter
and, when in doubt, grab a known good device for comparison.

So what I see with my NCR on X1 (with a 20,000 ohms per
volt coil) is somewhat relative. However, do understand that if
I grab a random red laser diode, I get exactly the same
"resistance" as I do with the "red" side of the LED pair
in the oximeter probe.

Like all diodes, we expect no conduction when the leads
are placed in one polarity, and we expect conduction based
on the above (and the character of the LED junction) in
the other polarity. So:

1. Red LED (mounted to the left on a tiny circuit board
in the probe):

NCR @ x1 ohms
4.5 K
NCR @ x10
22 K

2. Any Red LED in my parts drawer

NCR @ x1 ohms
4.7 K
NCR @ x10 ohms
22 K

3. The IR LED (mounted to the right on a tiny circuit
board in the probe):

NCR @ x1 ohms
500 ohms
NCR @ x10 ohms
3500 ohms

4. An IR LED in my parts drawer:
NCR @ x1 ohms
500 ohms
NCR @ x10 ohms
3300 ohms

5. The Photo-transistor
NCR @ x 1 ohms
150 ohms
NCR @ x10 ohms
900 ohms

6. Any typical NP junction for a silicon transistor
NCR @ x 1 ohms
100 ohms
NCR @ x 10 ohms
700 ohms

These measurements can be taken at two places. If you carefully
pry up the rubber on the inside of the probe with the LED --
note -- the rubber has a steel shield that is 3/32" wide, that
travels the perimeter of the rubber rectangle -- keep the
shield with the rubber, you might see a lot of green. This
green is cement used to tack things down. You can wash the
area with alcohol and a Q-tip. You will see the tiny circuit
board, and you will see 3 gold pads. Wires Green, Red, and
Orange will connect to those pads. So, when the gold pads
are cleaned, you can place the meter probes there. You can
also do this test from the probe plug -- but -- you are
now testing two things: the LED/sensor and the wires. As a common
failure mode for a probe worn during sleep is a broken
wire (wiggling, etc.), consider testing the wires separately
from testing the components.

The remaining component is the phototransistor which has
a flat cable polymer connector (DuPont trade name Parlux),
with 3 leads to the other side. One lead is not used and
appears to be directly connected to one of the other leads.
As the Black and uncovered wire are openly soldered to this
cable, it is easy to measure. With my NCR, I get 60 ohms
with leads in one polarity, and no conduction (again X1)
in the other polarity.

In summary, both the LEDs and the photo-transistor are
"semiconductors." As a rule of thumb, if one observes
conduction with meter probes in one polarity, and no
(or little) conduction in the other direction, and when
these differences are pronounced, the semiconductor is
typically good and working. If all 3 components in the
oximeter probe show good semiconductor behavior, these
components are unlikely to be a problem. If the probe
misbehaves, look for proper pressure against the finger,
cloudiness on the 2 plastic windows facing the components,
and connections/wiring. (For a few minutes you are welcome
carefully operate the probe, at the least, with the rubber piece
over the LEDs, removed. As the label on the probe reminds
us, the other side is put towards the palm. So without
the rubber cover/small O-ring/window, one's fingernail
is pressed directly against the 2 LEDs. And while we
expect the light levels to be higher because a somewhat
cloudy window is now removed, the unit automatically
recalibrates for the change in intensity.

Note: The copper traces deposited onto polymer are
difficult to connect to, with wire. So, one possible
failure point is the connection from the polymer trace
to the wire. And, the trace side of the cable is down, so
the trace is not exposed. To test such a situation, take
a new Exacto knife blade and, with magnifying goggles, scrape
the top, but to the side, the trace area. Understand, the
polymer is tough, it will be hard to penetrate. However, the
deposited copper trace is fragile and thin. You need to scrape
enough to suddenly see bright copper, and scrape no more. An
alternative is to take a shirt pin and punch a hole to the
copper -- but -- unlike using a shirt pin into insulation on
a wire, the copper will be pierced, and the connection will
be unsure. Another possibility is to disconnect both
black and bare, and to rotate the parlux so that the copper
plated side is exposed. Then we expect only a very thin
protective cover on the trace, more easily removed.

And, should you find the connection from the polymer trace
to the black or shield wires to be open, you need to very
cautiously take a very fine strand of insulated wire, around
# 32 guage, and with a very low power, very pointed soldering
pen, get some solder to take on the exposed trace (you may
use an acid (HCL) flux provided you apply it with a tooth pick
and wash the excess when done with water). Once you see, even
a tiny patch of solder has taken, you can tack the 32 guage
wire on, make a "service loop" (to remove strain), and then
solder that back at the black or bare wire. As the
photo-transistor side is a "high impedance" circuit, do cover
and seal any intrusions. A quick dab with a colored nail
polish (aka lacquer), will seal the area. And the lacquer
will bond to the polymer as the lacquer solvent (acetone and
toluene) will etch plastic. Allow the lacquer 1/2 hour to dry.

I would never suspect a break in the polymer trace, itself.
However, if you expose both traces, and don't see resistance to
the photo-transistor, you must assume the connection from
the traces on the other side of the probe has failed. I
have looked at that side. As I understand, there are two
styles of rubber probes -- soft and hard -- and the rubber
cushion on my photo-transistor side is hollow. I pealed
away some of the cushion, noted that one needs a true rubber
cement to reattach the bottom edges of the cushion (not contact

There is an alternative to soldering. This is also employed
in repair of the trim pot (below). For the repair of auto
rear windshield defoggers, NAPA, for example, sells the
Permatex product called "Quick Grid Rear Window Defogger
Repair Kit." It is Permatex Item# 765-1460. In a small
bottle is .05 fl. oz. (1.4 ml) of a repair compound. This
compound contains metallic copper in a quick drying cement.
As defogger traces cannot be soldered to, this compound will
reliably bridge breaks in such traces. (when traces under
high current are thin, they become excessively hot, and
this is a "run-away" condition -- i.e., the trace becomes
thinner, it becomes hotter, until it open circuits)

The same compound can be used where no flexibility is
required. I.e., if you have a break in the probe and can
(using, say, cement or hot melt glue) assure that both
sides of the break will not wiggle, this compound can be
applied for a lasting connection. Remember, both sides of
"the bridge" must be fully conducting. I.e., if this is
to repair a trace with any coating on it, the coating must
be carefully scraped away, say, with an X-Acto X611 blade.
And, always use magnifying goggles. Harbor Freight (and
others) sell a magnificent "Magnifer head strap w/lights"
for as little as $5. There are multiple magnifiers, allowing
for magnification from 1.8 x's to 4.8 x's. While the
two lights on each side of the band sound useful, they
are totally useless and I suggest unscrewing them and throwing
them away. They will "never" point light where you need it.

Instead, place a 120 VAC, 50 watt, "Par 20" miniature spot
in a gooseneck lamp. The light is particularly well focussed,
compared with, say, a miniature flood light or a standard bulb.


When done inspecting on the LED side I noticed a small copper
shield over the outside of the incoming cable. However, I
did not find that that copper nor the steel perimetry was
ever connected to a probe wire. I presume this shielding
is done to thwart static electricity. As I was unsure of the
connection of the left and right sides of this, I carefully
soldered a small piece of copper solder braid (used for wicking
solder) to assure conduction to this aspect. As the rubber
is still connected (mostly) and I thought leaving it that way,
convenient, I used a solder clip on the steel to make sure heat from
soldering didn't melt the rubber. (such clips have a claw
area and a spring to hold them shut) Should one bother? Hey!
Some engineer decided this should be here. Let it be. (That
this depends on the pressure of 2 metal ends onto the copper
cover -- heck -- I didn't design such a lame way of connecting
them. If the original design was good, then joining the two
halves is better.)

As the rubber cover (notice, it includes a small O-ring to
assure finger pressure is not applied directly to the 2
exposed LEDs), with the steel surrounds, presses back into
place, the only thing I see to do is to prevent intrusion
of oils and salt from the finger, and I do that with a light
coating of vaseline petroleum jelly at the edges where the
steel/rubber edges meet the probe arm. If we, instead, glue it,
we don't have easy access for next time.

Now, all this, and I find my 3 components are healthy. But,
the probe is no longer a mystery. As for how it measures
oxygen and blood pressure --

1. Blood changes color with its degree of oxygen
saturation. The haemoglobin cells which
transport oxygen will change the amount of
light received by LEDs to the receiving sensor

2. We presume the use of two LEDs, and some sophisticated
circuitry, allows the oximeter to calibrate to
the actual percentage of desaturation. Exactly
how is probably a guarded secret by instrument

3. Meanwhile, the pulse rate causes dimensional changes
in the finger which are translated into periodic
changes in the change in light. So we presume
that there are circuits that look for this periodic
change, and convert those changes into the display
for pulse rate

Now, it is not necessary to open the probe if all three
components are fine and if the wires are fine. We can
take measurements from the probe connector.

This is a nine pin, gold pin connector, with a flat spot,
and one labeled pin (8).

The typical numbering, looking into the male end -- the
end from the probe, would be as follows:

1 9 7
2 8 6
3 4 5

Official numbering may differ.

The connections are:

7 - green - red LED
5 - red -- LED common (red and IR)
9 - orange -- IR LED
1 - black -- photo-transistor
6 - bare -- photo-transistor

2 - one end - 51K ohm resistor in the probe plug
4 - other end - 51K ohm resistor in the probe plug

So, 2 & 4 "officially" tells the unit if there is a probe
plugged in.

II. Low Signal Problems

So far, as our probe appears fault-free, the only remedy
has been the addition of some added pressure via the O-ring.

As rubber stiffens as it ages, this compensates for 15
years of exposure to skin oils. (Note, skin oils are the
primary cause for vinyls and other plastics to age.)

However, our unit got quite irratic. Instead of a steady
heart beat, there was the sound of 2 or 3 beats. The
unit would either say "check the probe site" or might
say "probe off patient."

We noted a "daughter board" (inside the unit) between the
probe connector and one of the 3 circuit boards. Such
boards are inserted when a design is touchy and a redesign
is "added" to the unit. Now, all connectors from the front
panel to the probe were gold plated. However, even gold
contacts can become dirty. So, we decided to contact clean:

1. the probe connector to the panel connector
2. the probe connector inside to the daughter board
3. the daughter board connector to the circuit board
4. the panel connector J3 to the top board

We used Caig's Deoxit D5 spray. (If we are dealing with
non-gold contacts in harsh environments we use Caig's
Deoxit to clean contacts and then Caig's Preservit to
prevent further decay.) We spray the cleaner onto the
female receptacles, we then slide the cleaner onto the
male pins and shuffle the connector half a dozen times to
assure cleaner transfer and mechanical cleaning. (Would
the Radio Shack "contact cleaner" spray work as well?

The results from this were quite good. Typically
the signal strength was barely hovering at the first
level, and we would get a fault within 30 seconds. Now.
The signal strength gradually headed towards the top, and
provided the probe site was kept still, will mostly
stay at the highest level.

This was unexpected. We saw gold contacts on all except
for J3, and we, say in PCs, have gold to gold contacts
on boards that last for years. But, we will not argue
with success. (In retrospect, we note that, 1.) connectors
to the boards are difficult to get fully seated, and those
with "clamps" are not always easily seated where the clamp
comes over the connector. As for probe signal stability,
we note that the most likely cause for poor signal is when
the probe is attached to an unsteady hand. So, in making
tests, be extremely careful to clip the probe to a hand
that you do not move, and, be sure to squeeze the probe
as you place it, as it comes to rest better.)

III. Calibration / trim pot goes out of range

This was a problem we encountered in the first few years
of the device's use. One presses power-on. The unit
goes through a lengthy checkout process.

However, there is exactly one hole underneath the unit
for one adjustment. This is usually a situation where
a design engineer finds no way to electronically compensate
for a problem. As this trim pot is in an area with analog
JFET amplifiers, we know that some analog comparator
needs adjustment due to changes in components over time.

The need for this is announced on the unit's LCD display
during startup. It indicates that the adjustment is
out of range, and asks that it be set to +/- .1

When this happened, the value displayed was about +1.2
Further, the trim pot was not able to bring the value
within range. We examined the local circuit and found:

a TL082CP (U21) NTE 858M
IC-Linear, Dual Low Noise JFET Input OP Amp

We suspected it, but, adding an IC socket, and substituting
did not change the measurement. See Appendix A for how
we remove ICs and replace with sockets, for "through the
board" ICs.

In this area of the board we find:

a 10K resistor
(next to R26, in line with screw on pot)

and the trim pot:

trim pot 1K (R25)

We noted that by reducing the 10K resistor a bit, the
trim pot brought the calibration in line.

After another two years, the calibration drifted further,
and, we recognized it had a thermal component to it. And,
the drift would sometimes be to +2 or +3 and based on that
we could tell that there is a rough correlation between
this number and how far off the SAO2 reading is on the

The ICs that are most important are:

U14 -- an LF398 Monolithic Sample and Hold Circuit
U15 -- an LM358 low power, dual op-amp
U16 -- another LF398 Monolithic Sample and Hold Circuit
U21 -- the TL082CP IC-Linear, Dual Low Noise JFET Input OP Amp

Details about U14-U16 are given in the 1982 National Semiconductor
Corporation Linear Databook. U21 is by Texas Instruments and
is also shown in the NTE databook and web site (NTE 858M).

A typical design that employs two "hold circuits" is for
a slowly varying signal that has a high and and a low over,
say, seconds. Then, the difference is taken by a comparator
(here, U15) and that signal is sent to U21.

What is critical to consider is that the change in light
level at the probe's phototransistor is delayed by five-
ten seconds, either in this circuit, or in the microprocessor.
This means that the calibration number displayed is the
result of measurements in this area, a number of seconds
ago. And, in that the problem appeared thermal, we now
need to selectively heat and cool parts of the board with
these ICs, knowing that the sensitive area will be where
the heat gun or arctic freeze changed the component
temperature "a while ago." The board did respond to this
inspection, and we finally reduced the problem to U14-U16.

As these are ceramic packaged ICs we can put some heat
sink compound on each, and use the soldering pencil to
heat them and carefully drizzle arctic freeze on them
to cool them. The IC that was thermal was U15. Indeed,
if we cooled it enough, the op amp on leads 1, 2 and 3
would go from drifting to full failure, and the calibration
setting would jump to 5.4 We presume 5.4 is the upper

Fortunately we had a spare LM358, we pulled the
defective IC, we socketed the location, and inserted the
other LM358. Suddenly the calibration was rock solid,
and, we even had to remove the 10K resistor (above) to
get calibration back to +/- .1

Now, in constantly readjusting the trim pot (via the small
hole near the center of the oximeter's underside) we managed
to hurt it. And, this is a remarkably small, 10 turn
potentiometer, so we had no spare. Above we mention one way to
repair without soldering. This could be used here. In the case of
this trim pot, the screw goes to a worm gear inside. That worm gear
moves a slider on a carbon film layer deposited on a ceramic square,
about 1/4" by 1/4." The center lead (the varying resistance
lead) had become disconnected from a silver film that
connects to the slider. Carefully using very pointed meter
leads, we found the break to occur just where the silver
film goes around the edge of the ceramic, to connect to the
tinned copper lead that solders to the circuit board.

It is fairly easy to expose this area, normally encased in
the black plastic box that contains the pot. You take a
1/16" chisle bit in a battery Dremel, on low speed, and you
create an opening to the side of the brass screw, in the
center, and on the side where the center lead comes up.
Do this with magnification -- it is difficult to put back
pieces of worm gear, or slider, or ceramic.

We first coated the carbon film/slider with a contact cleaner,
and then carefully washed the area where the silver film came
to the edge of the ceramic. We carefully washing this
location with alcohol on a toothpick, and then put a small
drop of the copper-based cement. After drying 1/2 hour, we
found continuity to be excellent. As this area points downward,
we simply left the "access door" open. In any further adjusting
of the pot, be careful not to let the jeweler screwdriver
slide over and break this repair.

IV. Power/Standby off/on failure

Last year the unit simply refused to power on. We traced
the problem to the failure for the front panel membrane
switch to actuate a "power relay" located on the top board.

There are two such relays. Off/on is the relay closer
to the front. Carefully removing the top of the relay
with a rotary cutoff wheel (parts extend all the way to
the thin top cover, so, place the wheel in the same
plane as the top cover), we found that a toothpick inserted,
pressing a white plastic contact holder allowed the
unit to power on. So, for a while, it was simply used
that way.

Then we diagnosed the problem in the last few days. Again,
with no circuit diagram, we followed the circuit from the
off/on relay (K2) to the front membrane switch. This is
actually quite difficult to do, as there are many safety-
related IC gates, and following the true path of the signal
was often thwarted. However, we now know:

1. The membrane power switch on the front panel
connects to the center board of the stack of three
circuit boards via a fifty pin ribbon connector

2. Off/on is triggered by a voltage on pin 3 of this
connector where the voltage is regularly 8.2 VDC
and goes to zero when the membrane switch is pushed

3. Pin 5 is also part of the power on circuit as well
as ground provided via the three wire cables from
the front cover to the top, power board

4. The off/on signal trace wanders around the middle board
and then is connected to pin 12 of J2 (the shorter
ribbon cable that connects the middle board to the
top board)

5. That signal goes to pin 2 of U18 (a CD4093, quad
NAND gate). Capacitor C35 connects here, providing
a "de-bounce"

6. The output of this NAND, pin 3, goes to pins 5 and
6 of another NAND in the same IC. I.e., this
NAND is simply used as an inverter

7. The output on pin 4 goes to the "Set" (pin 8) on
the second flip/flop contained in U9 (CD4013B)

8. The output, Q2 on pin 13, goes into the Clk (pin 3)
of the first flip/flop contained in U9

9. The output of that flip/flop, Q1 on pin 1 goes
to the base of transistor Q1

10. The emitter of the transistor goes to ground, and
the collector goes to the "low side" of the coil
in relay K1

11. The "high side" of the coil in K1 is provided 9 VDC
via relay K2. K2's purpose is to sense that the
unit is properly connected via the power cord to
120 VAC

The fault was "mechanical." These relays K1 and K2 are
Omron part number G6A-234P-ST-US-DC9 And these relays are
still widely sold by Omron. This particular design, the
"G6A-234P" uses a remarkable white plastic slide piece that
contains a magnet and 2 pole pieces. I.e., the magnetic
flux of the coil is applied by a rectangular bar that
comes out of the coil and sits inside the 2 pole pieces
of the moving magnet. This design provides a very high
force, and the berillium copper moving contacts can
achieve up to 3 amps of current on the switched leads
of this relay.

The input power of this relay is 200 mW or 22.2 ma.
at 9 VDC. The coil resistance is 405 ohms, and there
is a "snubber" diode across the coil to prevent inductive
surges from hurting the Q1 transistor. "ST" stands for
"standard sensitivity" and the "DC9" identifies this as
a 9 volt coil.

The manufacture date of the relay, 1992, corresponds to
the manufacture date of our oximeter. Now. When Omron
terminated the varnished coil wire (an astounding 48
gauge wire, think tiny), they used an unreliable technique
of wrapping the end of each coil wire to the square edged
lead wire, with just 2 turns. I.e., the wrap must be
vigorous enough for the square edge of the lead wire to
break through the varnish and then create a "cold weld" from
the copper wire to the lead.

One of these connections failed. As the relay is a production
part, it can be purchased through various distributors. Note:
only a few of the distributors stock the 9 VDC version as the
5 VDC version is much more popular.

However, it takes less time to repair this relay than to
get another, remove the old relay, and install a new one.

As we had already carefully removed the top of the relay,
we only had to cut "access windows" to where the magnetic
wire is wrapped onto the leads. Clipping the lead on
the large electrolytic next to the relay and using a razor
blade to separate the silicone rubber holding the cap. to
the board, it was now easy to take a small carbide burr
and cut two vertical slits on the sides of the plastic
case of the relay -- finishing the cut using an Xacto
blade -- being extremely careful not to nick the coil.

With magnification, put a tiny amount of paint/varnish
remover onto the two turns of coil. Do not drip the
remover/solvent onto the coil! Now, take some flux,
either rosin or NoKorode paste or even an acid-based Dunton's
NoKorode, and, again with a toothpick, dab these turns.
Now, take a very pointed solder pencil, and solder the
leads to the pin. We do this on both sides as we don't
know which side has failed. Now, spray the area with
a light solvent (e.g., rosin flux cleaner) to remove anything

Note: we inspected an G6A-234P manufactured in 2006.
As we suspected, this fault must have been common, as
the newer device has six turns, and, is properly soldered
to the lead pins.

So, we consider all such relays manufactured the "old way"
to be truly "accidents waiting to happen."

Otherwise, in examining this "top board" we did look at the 3
+9 VDC regulators -- TL497AN's in a column, to the
right (facing machine) -- each next to a toroid

As these ICs are not listed in the NTE replacement series,
we provide the pinout:

1 Compensation Input
2 Inhibit
3 Freq. Control
4 Substrate
5 Ground
6 Cathode of interior diode
7 Anode of interior diode
8 Emitter out
9 No Connection
10 Collector out
11 Base (connects to the base of the output transistor)
The output transistor is called the "Power Switch."
The maximum "switch current" is 750 mA
12 Base Drive (connects to output of interior Oscillator,
and is the Base of the output transistor, before an
internal resistor
13 Current Limit Sense
14 Vcc -- 9 VDC

This is an early "switching voltage regulator." Curiously
I find it in the 1st Ed. Texas Instruments "The Linear Control
Circuits Data Book for Design Engineers (1976)," but
by the 2nd Ed. (1980) it no longer appears. This explains
the absence of an NTE replacement. On page 238 of the
1st Ed. (yes, we have one) we are shown 4 sample circuits. The
switcher could be used for "positive" step up, "positive" step down,
"positive" to "negative," and as a positive regulator
with "buffered output" (i.e., an external transistor to
handle the regulator's current output). The "M" version
operates over the 2nd highest temperature range -55 C to 125 C.
The "C" is -65 C to 150. No letter, the operating range is
-25 C to 85 C.

As the circuits are non-obvious, and the actual output
can be taken off from the diode or the emitter, the common
element is the voltage into pin 1 (the comparator, aka,
the "error amp." In all cases that voltage is compared to
an internal fixed reference of 1.2 VDC.

So for the regulators, back to front, pin 1 reads:

#1 -- - 3.28
#2 -- -13.85
#3 -- + 1.21

and these voltages are typically the output between
two series resistors, where one resistor runs to the
actual output voltage, and the other resistor runs to ground.

So, following the 3 "high legs" of the dividers, we
find the actual supply voltages:

#1 -- - 5.0 (R36)
#2 -- -15.15 (R40)
#3 -- +15.24 (R42)

We measure other supply voltages:

Pin 14 - LM339 - +5 VDC
Pin 14 - these regulators - 9 VDC
Caps C1/C2 - +14.50 VDC

Those familiar with instrumentation know that many
circuits were designed to run with +15 V and -15 V.
So outputs #2 and #3 are expected. Further, some
circuits require a negative voltage with respect to
ground, and -5 (#1) is a commonly use value, for example,
it is still included in PC computer power supplies.

Of course almost all TTL ICs and many CMOS ICs run
from +5 VDC. The +9 VDC was probably chosen as
a source for these 3 switching supplies.

As the switching voltage outputs are all relatively
low, we do not expect failures in the switching
transformers as we see when switchers are used to
create higher voltages, such as in CRTs. If
the designer used proper "inductive surge" protection,
and 15 years of life suggests he did, we might only
expect failures if something shorts their outputs,
but only if the "current limit sense" does not
properly "crowbar" the output, protecting the IC.

So, if one of these devices is not at the proper voltage,
the first step would be to remove the load by cutting
a board trace, disconnecting it. If the voltage then
measures OK, then find the short. Also while the
Internet is an excellent source for obsolete ICs one
could just as easily purchase a +15/-15 supply module
for around $10 and simply take the offending supply
out of the circuit. And, of course, any small supply
does 5 VDC, and if isolated, one is free to turn that
around and make it a -5 VDC supply.

However, do note that these supplies should be "instrument
quality." If the supply is inserted just where the switching
supply provides their voltages, then the further filtering
for those voltages would remove unwanted ripple. (Do
not use a cheap AC adapter as their voltages are unregulated
and they would have a 60 Hz component to the output which
the filter circuits are not designed for.)

The capacitors to pin 3 of TL497 can range from 5 to 1000 pF
which produces from 385 down to 10 kHz switching

V. Lead-acid batteries

Because the unit is built for hospital use and high
reliability, the unit can run on internal batteries.
And as small lead-acid batteries are constructed using,
not plates, but screens, these deteriorate over time
whether the unit is used, or not.

The unit requires 4 x 1.5 VDC cells rated for 2.5 AH
(ampere hours). When we first replaced these 4 cells
we used originals, called "D" cells, from
The total with shipping was around $30.

However, as high powered hand lamps have become available in high
volume, and these use a single sealed battered with 4 cells,
and at a post Christmas Target sale we bought several at $5
each, it was much cheaper to buy one of these. And, the
rating is 4 AH. The plus and minus battery clips were
compatible. We used a piece of galvanized steel wire to
go from one screw-down to another and used a rotary cutoff
wheel to remove some plastic in the cover. Total retrofit
time, about 15 minutes.

In closing, I notice one 3700 (with no probe) was just sold
on eBay for $111. I see that online resellers want
$500-$1000 for a good, working instrument. As we
encourage people with sleep problems to have access to
useful instruments, we hope this (detailed) note is useful.
And, the extra security of a unit built for hospital use
does provide a level of safety greater than consumer devices.


Chief Engineer

Appendix A:

Finally, for removing ICs without damaging traces or the
IC, I have this down (for pins through the board).

1. One makes a rectangular loop out of #12
guage copper wire (now a tool one keeps)

2. One flattens one side, inside and out so
that flattened sides go up against
the 2 x 7 IC leads on the solder side

3. One makes sure that the leads are straight --
i.e., the oximeter boards has all 14 leads
bent over -- individually take a soldering
pencil and straighten them

4. Now, take an IC puller adapted with a screw and
nut to tighten the puller to the IC.

5. Hang a 3 pound vise connected to the puller, where
the puller is pointed towards the ground (or
better inside a drawer)

6. Place excess solder so that all 7 x 2 leads
are connected by solder

7. Take the copper wire rectangle (using its
original insulation on a leg, as a handle)
and work the flat sides against the
14 leads

8. Within 4-6 seconds, the IC and vise drop into
the drawer with a resounding thunk

9. Now desolder each hole, and wack the board
onto your thigh, to remove solder

10. Finally, clear the holes for a socket by drilling
with a # 69 number drill

Total time, about 10 minutes to socket an IC.