Corometrics 506 Medical Monitor
 Teardown of a medical device from 1981

A product that I helped to design

By Dave Erickson

Schematics

qual1

Intro: HP Medical Roots

In this continuing series of buying old products that I developed on Ebay and tearing them down, here is a medical monitor from 1980. It's the Corometrics 506 Neonatal medical monitor.
I was at Hewlett Packard Medical in Waltham, my first real job after college. HP is known as one of the top technology companies in the world. In the early 60s, HP bought Sanborn Co. in Waltham, MA in order to get into the medical device business. Sanborn produced electrocardiographs and other test and measurement instruments used in hospitals. HP also bought a Defibrillator company in Corvallis OR and another medical division in Boebligen Germany.  HP Medical was one of the top medical monitoring companies. They had mini-computer based arythmia, defibrillators, were working on the first pulse-oximetry, and were working on their first phased-array medical ultrasound.  The ultrasound group spun off to a new facility in Andover, and ultimately all of Waltham moved to Andover. Then when HP decided it was a computer company and not an instrument company, it sold the medical product line to Phillips Medical.

I was hired in the summer of '76 to work on a new generation hospital central station, the HP75800 series, code named "Speakeasy". This was a multi-patient central-station system that displays multiple patient waveforms from patient monitors throughout the ICU. The project was very successful. I left after only 2 years though. I worked in the Display group. We developed the first system that could display diagnostic quality waveforms on a high resolution raster (TV type) monitor. Our group developed everything from the ECG waveform data to the display monitor. I was the video guy. I designed the video D/A, sync, and video amplifier, and worked on display monitor circuits such as the horizontal and vertical sweep circuits and the high voltage. In 2 years our group went from a hand-wired prototype of the new patented display technology to five production-ready systems including all environmental and regulatory testing. It was my first of many times doing EMC testing.

Here is the November 1980  HP Journal issue that features this system.

HP Journal Cover
I was proud of the work that our group had done. In the summer of '78, while our group had met our schedule and had production-ready hardware, the software group was still deciding which language to write the system software in. Recognizing that this project was destined to continue for a few years and that the hardware was done, I asked my boss at the time what he saw me doing for the next say, 6 months. His response: "Well Dave, I want you to make sure that the bill of materials is really correct". If I really dragged my feet, that mindless task could possibly take one day. I was looking at at least a year of little or nothing to do. I had already designed plenty of home projects (G-Jobs) and at 24 years old, needed real EE work. On top of that, the only feedback that I received in my first 2 years of employment was one negative review. Apparently being the only group that met all of our goals wasn't important.

Meanwhile I had watched another HP hardware team, the 'front-end' group, developing advanced new patient monitoring technologies that were destined to go nowhere. The managers and product planners spent nearly 3 years in closed meetings, unsuccessfully trying to define the next generation of CPU based patient monitors. Meanwhile our competing German division developed the next generation patient monitors.  I felt that the Waltham patient monitoring division was mis-managed at that time, and was shocked at how many engineering person-years were squandered. In summer1978 it was time for me to leave HP. In '78 to '79, a dozen other talented engineers also left for similar reasons. People went to Analogic to develop patient monitors for Siemens, to Corometrics and to other medical device companies. Not only did HP Medical lose a lot of key talent, but they wound up spawning several of their own major competitors.

In HP Medical's defense, they finally finished Speakeasy and were busy creating the ultrasound group, and developing the first phased-array ultrasound. This product was to become wildly successful. They closed Waltham and moved it to Andover. And got their act together.

I learned a lot about managing by watching HP and other companies mess it up. There were plenty of examples of this in the 70s and 80s. Remember Digital? Data General?

Octek

I went from HP to Octek in 1978, a small startup in Burlington, to be employee #3. The founders were John and Arthur. Arthur was another HP Medical refugee, and John was formerly a consultant at Arthur D. Little in Cambridge. Octek did consulting and product design for other companies, to support their product development: The Octek 2000, a video frame grabber for the Data General Nova. I loved the projects I worked on. Each project was typically a few months long and every one was different. I learned to write proposals, to quote and develop products quickly and efficiently. I developed about 10 products in the 3 years I was at Octek. Mostly analog and digital design with some microprocessor code. I designed floppy and hard disc testers for BASF, a Colorimeter, a data terminal for a credit card printer for Dymo, a video measurement system for cell biology, a switching power supply and others.

One other HP refugee, Patrick, went to Corometrics in Wallingford CT. They were leaders in fetal monitors. At the time they sold a portable neonatal monitor which was designed and built by another medical company, Becton Dickinson. They wanted to make improvements, and to own the design and manufacturing. In 1980, Pat called me at Octek and asked us to quote on the new product design. I called two other smart MIT EEs, Jeff and Dave who had also left HP. We spent 2 months writing a 100 page detailed technical and business proposal, and were hired to do the design. We spent the next 9 months designing the Corometrics portable 505/506 neonatal medical monitors and building and testing the first unit. We had the original Becton - Dickinson 504 design as a starting point, but the new monitor was a complete redesign of every circuit. It used the same CRT, battery, and patient connectors and that's pretty much it. 

Meanwhile, my girlfriend Alex (now wife) was a pediatric nurse who conveniently worked on the Neonatal Transport team at Childrens Hospital in Boston. She helped on the features and user interface for the new monitor. We negotiated the specs with Corometrics and began the design.
I designed:
 Jeff designed:
 Other Dave designed:
 Dave's wife Sherry designed:
 Here is an old block diagram for the 505/506 monitor.
 block

Here is the team hard at work at Octek. The other Dave, me, and our technician Wayne. I suspect Jeff was behind the camera. Check out the breadboard CRT.

team

Low Power Design

Because battery life was critical, the entire unit including the CRT had a power budget of 5 watts. We came in at 4 watts. Key technologies were:
 We used the LM346 as our go-to quad op-amp. It has 2 programming pins that allow each op-amp's power draw to be tuned for bandwidth. Most of the signals were under 1KHz, so slow, low power opamps were ideal. Where needed for speed or low bias current, we used low power BiFet opamps like the low power TL062 and the occasional LM352.

The monitor has 9 PCBs containing:
 Pull-up resistors were high value for minimum power. 4000 logic draws virtually no power, microamps at 5V and low clock rates.
The master clock frequency is 923KHz, which is immediately divided by 4 to 231KHz to drive most of the logic.

We considered using a microprocessor, but processors and memories at the time were N-MOS and so too high power. The RCA 1802 was the only CMOS processor at that time. We had hard real-time requirements for the display and signal processing, and no budget for development tools or time for firmware development, so analog and digital hardware for processing and displays.

PCB1: Isolated Front Ends

PCB1 has all the isolated front end circuitry. From left to right:


Here is PCB1, the front end board. The two small dual-row headers near the holes are for the patient connectors. ECG / Resp on the left, Pressure and Temperatures on the right.

The shiny metal object is a shield around the respiration coupling transformer. The two other transformers are for isolated power and for the multiplexed patient data.Note the neon lamps and spark gap (lower left)  for defibrillator protection. The left edge-connector provides chassis ground only. The right one is for the non-isolated power and data.

You can see the technologies we used: Bifet and LM346 low-power op-amps, 4000 CMOS logic, and CMOS switches (DG211 and 405x). We had an assortment of yellow Electrocube axial film capacitors and used them everywhere. For long time-constant capacitors, we used reliable wet-slug tantalums.

We minimized the use of trimpots and used some precision resistor values instead.

pcb1

ECG

We originally planned to use a clever 2 lead ECG. 2 leads on a tiny neonate is an advantage over 3 leads. It was originally designed at HP, and was patented by HP. I thought I had a different implementation that got around the HP patent, and built the board to do both 2 and 3 leads. But Corometrics marketing people and the IP lawyers weren't so sure, so we stayed with the standard 3 leads.

For 3 leads, you drive the 3rd lead leg electrode with the common mode signal. This reduces the common mode AC line and other noise on the main 2 leads. For 2 leads, instead of driving the patient leg with the 3rd lead, a +/- 40V amplifier drives the chassis ground through a high-value resistor to reduce common mode.
       
A challenge in designing an ECG is that the unit must detect 1mV signals, but must recover from a defibrillation very quickly.  Defib. pulses can be up to 400 joules and over 1000 volts.  The ECG must recover the display and detect QRS within 2 seconds.

In addition to two levels of voltage and current protection, differential and common mode are detected, and the high-pass filter time constants are sped up until the signal recovers.

Invasive Blood Pressure (IBP)

The IBP transducer is a sensitive strain-gauge bridge. It requires a stable voltage reference and precision differential amplifier. Instead of a pricey precision amplifier, I chose to use a chopper method instead. The bridge is driven by an AC square wave, AC amplified, and demodulated in the isolation barrier.

Respiration

28KHz Impedance pneumograpy is sensed through the ECG leads.  The circuit detects ~1 ohm thoracic impedance changes. Most respiration monitors are known for flaky operation. They are very sensitive to muscle and motion artifacts, and so is quite difficult to detect valid breaths. Jeff spent several years of his life developing a state-of-the art respiration front end.
        
The front end uses a transformer to couple a stable 28KHz sine wave into the ECG leads, and demodulates it. It has overload recovery and provides the leads-off detection.

2 Temperatures:

The temperature channels use Thermistor sensors and precision opamps: OP-07's.

Patient Isolation barrier

The non-isolated part of the isolation barrier sends 28KHz, 12V P-P AC to the power transformer. A second transformer receives the 7 signals from the isolated front ends.

A transformer cannot pass DC. So for each channel's waveform sample is converted it to 2 consecutive bipolar pulses, one positive and one negative, to send across the transformer. This converts DC signals to AC. The chop rate was 28Khz, from the isolated power. It provides the the clock for the MUX / DEMUX logic. So each sample is 1 / 28KHz = 35.7uS. 2 samples, plus and minus for each channel, and 7 channels. So the sample rate for each channel is effectively 28KHz / (2 * 7) = 2KHz.

To demodulate each channel on the non-isolated side, a single 'difference-and-hold' circuit samples the first positive pulse, and then subtracts the second, negative pulse. Then a 4051 3:8 de-mux acts as a 'sample-and-hold' for each channel.

Each side of the barrier has a free-running divide-by-14 counter. One of the time slots is used to pass an extra pulse that is detected,  and used to synchronize the the non-isolated counter to the isolated counter.  Despite the DC -> AC -> transformer -> DC conversions, the AC and DC accuracy is quite good, approaching 0.1% or 60dB. Same with channel-to-channel crosstalk.

PCB2: Respiration processing

PCB2 has the respiration processing circuitry and the non-isolated part of the isolation barrier. The 2 transistors drive the 28KHz power to the front-end power transformer.

Jeff spent 3 years at HP developing an advanced respiration system that HP never used. This was his chance to put his knowledge to good use.  Respiration and reliable apnea alarm is critical for neonates. Respiration is about a 1 ohm impedance change on a 3K ohm base impedance. It is very sensitive to lead motion, muscle artifact and other interference. Reliably detecting a breath is a very hard problem.

pcb2

PCB3: ECG and Pressure Processing

ECG Processing

Pressure Processing

pcb3

PCB4: Waveform display

Dave is a real good digital designer. The RAM requirements are two banks of 1K x 8 for the 960 8 bit samples of two waveforms. He originally considered using a new-fangled 16Kx1 DRAM for waveform memory. This would require multiple shift-registers to convert the 8 bit read and write data to serial. Instead, 2Kx8 SRAMs were available.

The master clock for the system is  a 923.52 KHz Everything in the system is derived from this clock to prevent sensitive analog circuits from getting any beat-frequency interference.

2 waveforms:
 pcb4

PCB5: Numeric display

pcb5
Below is the rear side of the Numerics board. This is the densest board of the product. In fact it is so dense that there was not room to route all the signals in the backplane area. So we added 10 wires to the back of the board (below). This board was a good candidate for a 4 layer board, and we discussed this option for the future. But 8 years later, the board was still being built with the 10 wires. Strange since this was Rev7 of this board.  Hey, if it works...
I couldn't help noticing the sparse bypass caps: only about five 0.1uF caps and a tantalum for all those logic ICs. My bad. I guess slow 4000 CMOS doesn't need much bypassing.
 pcb5r

Designing this board was a blast for me. I had worked on numerous raster video devices, and this was my first vector display. The stroke display font is contained in a small 2K x8 EPROM. each byte contains 3 bits for the X motion, 4 for the Y motion, and 1 bit for blanking. The strokes are offset binary: mid scale is no motion, + is up / right, and - was down / left. An Octek buddy worked at Adage, a local vector-based CAD system company. I visited the company, and asked him to print out their numeric font. I proceeded to redraw the digits on graph paper and hand-digitized all the strokes for each digit. Since blanking was only ON or OFF, it was important that each vector be about the same length, otherwise, long vectors would be dimmer than short ones. So I made the longest vectors out of 2 or 3 shorter vectors.

There were 16 vectors per character. "8" is the most complex digit. There are up to 16 characters across the screen.

The ADC that measures each of the 7 parameter is a simple first-order Delta-Sigma design. It has an 8:1 input multiplexer ('4051), a summing integrator, a comparator, and a D-Flop. The D flop drives a precision 1 bit DAC and the counter which is BCD. The counter is 0 to 399 and built of '4518 dual decade counters for the LS digits, plus 2 flip-flops for the MS digit.

The 1 bit DAC for the ADC is built with six  '4049 CMOS inverters in parallel, powered by the -5V reference and driven by a +5V to -5V level shifter.  Because the DAC is negative, the difference circuit is a simple summer on the input of the integrator. The summing integrator uses 0.1% resistors to achieve about 0.2% gain accuracy with no gain or offset trimmer.

The BCD values from the ADC are written into 2 tiny 16 x 4 RAMs ('40114) configured as 32x4. Write logic waits for the ADC to be complete and writes the 4 digits for each parameter into the RAM The RAM address is 3 bits for the parameter, and 2 bits for the digit. These are provided by binary counters.

To generate the characters, a Write state machine waits for the main display timing until the parameter display time. 3 counters, one for the parameter, one for the digit, and one for the segment are all held in reset. So are the stroke vector integrators. Then as the counters start, the 16 segments for each digit are read out of the EPROM. The integrators draw each digit in sequence. Between digits, the digit counters are incremented, and the integrators are reset, ready to draw the next character.

To position each character and group of characters horizontally on the display,  two analog mux'es and resistor ladders generate the X positions for the parameter and for the digit. These are scaled and summed along with the X integrator, and sent to the X deflection amp. The Numeric board also handles the waveform horizontal sweep via a simple integrator.

For the vertical (Y)  axis, the two waveforms, a  vertical offset for the numerics, and the numeric Y integrator are summed.

For the Z axis (blanking) the numeric blanking is gated with the waveform blanking and output to the Z axis amplifier.

Board 6: XYZ and Alarms

The XYZ deflection amps are on the left. They are powered from +150V and connect to the CRT via the left single row connector.

The alarm logic is on the right. It connects to the front panel board via the backplane, and to the alarm limit controls via the upper-right single row connector.

pcb7img

Board 7: CRT Power Supplies

Board 8: Power Board

Main Power supply push-pull switcher, multiple outputs

Board 9: Backplane (Mother board) and Card Cage

The backplane connectors are standard 44 pin 0.156" pitch connectors from the 60's. For the front end,  and to provide extra pins where needed, additional 12 pin connectors are used.

brd9

Magnetics: 9 potcores

The Project

The first phase of this project was to write a detailed proposal. Corometrics paid about 4 person-months of consulting effort for us to write a detailed proposal. I still have a 42 year old copy of the Phase I report. It is 120 pages and discusses the proposed design approaches for each section in detail, with trade-offs of the various approaches considered. It has the detailed block diagram (above), details of several critical circuits, and the detailed development schedule. Phase II was for three engineers for 9 months, plus a technician.  I was project and technical lead, and although I was quite good at analog and digital circuits, video audio and CRT circuits, it was my first ECG and isolation barrier. I learned ECG and IBP as I went. I knew little about Respiration, or the digital logic for the moving waveform display.  I did know that having top engineers on a team is key to success. Jeff and Dave were the best.

We developed a hand-wired prototype with wire-wrapped proto-boards. I wish I had a photo. We partitioned the design and submitted schematics to Corometrics. Their PCB designers laid out the boards per our instructions. We reviewed and approved the layouts.  Corometrics manufacturing group returned built boards which we tested and reworked as needed. We built them into a single prototype for final system testing.

The Phase II (design) project cost was about $250K and the duration was 9 months. We came in on time and under budget, and met all specs. I kept copies of all the documentation, with the intention of being available to support the transition to manufacturing. Other than one or 2 simple calls, Corometrics never called. They handled the mechanical design and the transition to Manufacturing. They built and sold many 505 and 506 monitors. Fast-forward 42 years to 2023, and there are still working units on Ebay.

This was the first big design project I was in charge of. I went on to consult on my own for about 6 months before I was hired at Datacube. 6 months later I was manager of engineering there.

But, but, but.....

I would love to dive into the design and schematics, and do a very detailed tear-down video. Despite the fact that I have copies of all the original hand-drawn schematics and Lab notebooks, these documents are owned by Corometrics (now GE). And so are the copyrights of these documents. Unfortunately, like most industry designs, these 42 year old designs are still copyrighted. A shame, since there is a lot that could be learned from this aging product.

The "CAD" tools

The other Dave had an Apple II computer at home. He used it for word processing and some circuit math calculations in BASIC. Otherwise, everything was done by hand. Hand drawn schematics, mostly D size, done on drafting tables, using logic and plastic electronic symbol templates. The electric eraser was key. HP calculators and lab notebooks for circuit calculations and timing diagrams. Karnaugh maps for logic minimization. All the PC boards were hand-taped by the excellent PCB designers at Corometrics.

No PCs, word processing, spreadsheets, schematic capture, simulations, or PC layout CAD. Most formal documents were hand typed by our secretary. Ah, the good old days.

Back to Dave's Home Page
Last updated 4/7/2024