Open Ventilator Project
Half the world's population lives on an income of less than $300 usd/month. Who will make ventilators for them?
Consider Egypt, 100m people living on a thin strip of land by the Nile, the Philippines, rural Ukraine, India. A yearly salary may not cover a Chinese ventilator.
Please see Johnny Lee’s ventilator project (the video gives an overview): https://github.com/jcl5m1/ventilator). That’s a $50 ventilator. It's a working starting point.
This document is a proposal for progressive further development, component by component. The goal is to create a more flexible and robust design that could be easily produced in quantity, license-free, by factories that would otherwise probably be producing consumer electronics.
While $50 is the current target, the other aim of this initiative is to gather people with know-how in one place to discuss and share ventilator construction knowledge. The ventilator crisis in developed countries has already begun, and ICU ventilators have complex supply chains and the tooling that is not set up for high-rate production.
Why a centrifugal compressor?
A ventilator with a microcontroller, airflow and pressure sensors, and a digitally controlled blower, can potentially perform any ventilation ‘mode’, it’s just a question of code. A rough analogy of the lungs for EE:. a capacitor in series with a resistor. By modulating the voltage across them (well, pressure), you can induce any signal you like. (well, not taking into account the patient’s own breathing efforts etc.)
This design has potentially few parts. Tooling up means machining a few molds. Email gerbers to a PCB house. Get a few automatic screwdriver machines etc.
“In Italy there are currently two issues. One is the number of mechanical assisted ventilation. Meaning very severe cases, intubated, machine replicates the breathing pattern. Second issue is that there are plenty of people that need ventilation, needs to be separated by the general population and their respiratory assistance needs seems to be sufficient with something like Johnny suggested. It almost looks like there are two different needs.” (Pertusa)
A fully functioning and safe ventilator for adults, designed and built from a 'grass-roots' remote working team, is a tall order. A simple CPAP or BiPAP, non-invasive system is more feasible. However, we will spec out the hardware so as not to close any doors at this point.
A blower-type design allows easy regulation of pressures, but volume can only be regulated indirectly by monitoring airflow with a sensor (integrate the flow.. and the error, to get volume). A piston or bellows (positive displacement) type ventilator allows volume to be regulated directly, but regulation of pressure is indirect.
This is not the only way to make a ventilator, there are many viable approaches to create a low-cost ventilator. This one was selected as a promising approach.
See also:
https://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/20080012540.pdf
Proposed Organisation of Design Work
Interested people are welcome to participate!
Github is used for hosting design files: https://github.com/OpenVentOrg
Trello is used for organizing task: https://trello.com/invite/b/UvdAbU5e/c90c6832751c6f0f11998d9e4aaf2b82/open-ventilator-overview
Discussion takes place here on slack: Open-ventilator.slack.com
Join link is here: https://join.slack.com/t/open-ventilator/shared_invite/zt-cwh6py1x-C5tO4JewMUf2tRN7rwWB~g
Please, use private messages and discussion threads inside channels when appropriate, to keep from cluttering up the communication too much.
Various groups are forming with their own slack channels. A directory of people working in each group may be compiled shortly, allowing them to be contacted with questions, by voice call or email etc.
It is highly encouraged to invite specialists to help with anything that could benefit from their knowledge. There are lots of professionals eager to help on helpfulengineering.slack.com and elsewhere. The most important thing for this project is speed, everything needs to happen quickly.
Medical staff and ventilator technicians would ideally be observing all design work and making suggestions.
The information below is out of date, please update with current information!
The follow groups exist, or will shortly:
Electrical & Electronics Engineering Group (‘electronics_engineering’ slack channel)
E.E. Documentation
E.E. Github
E.E. Trello Board
Design Group - Blower (‘design-blower’ slack channel)
Challenging, Critical Task!
There is an initial design for a print cooling blower (by Mark Rehorst) that replicates a CPAP blower and is modeled to be printable with FDM using a repurposed harddrive motor: https://drmrehorst.blogspot.com/search?q=blower.
Multiple variations on this preliminary design were modelled in CAD (by Glen Lovsin) to make use of off the shelf RC BLCD motors such as the common 2205 drone motor. These printable designs can be found at: https://github.com/necro-nemesis/open-ventilator-centrifugal-pump. The master branch contains the latest design undergoing bench testing. Within the development branch of the repo there are several alternate designs providing variations on impellers and cases. Readily printable .stl and CAD files .stp and 123dx are included. Some of these designs have been tested fully (latest in master branch), others are at the design phase with some undergoing initial tests (dev branch). The most up to date information with regards to blower design is available on the slack blower-design channel.
Construction and testing of blowers has been conducted by Henry Palonen to test various parameters of the 3d printed pump design. Testing is largely focused on pump volume and pressure but in addition simultaneously recording various other parameters in order to be able to offer data in order to generate algorithms which are able to modulate pump speed in order to achieve the necessary output parameters. The test bench arrangement and collected is posted to:
https://github.com/randomev/openventilator-test-bench
Using this as a starting point, the group will develop an improved centrifugal compressor design using principles of turbomachinery and software tools like ANSYS. If faced with limitations of FDM during prototyping (material strength at high RPMs, or dimensional accuracy, for example), the designers should not feel constrained to move to CNC milling, SLA printing etc.
Tiny brushless motors that can put out 500W are available for $5. (banggood.com/search/brushless-motor.html) They come from ${Unknown factory in China} and might last 5 mins or 5 years. The motor control board can run a 'test mode' at high output, for 12 hrs or so. Blowers can be used in pairs in ventilators. Motor currents and speeds can be monitored, and faulty blower units shut down, with the other blower increasing output to compensate. Some of the motors are '2-4s' (2-4 3.7V lipo cells), so up to 15V. Others, '2-6s', so up to 22V. For sure there are better quality motors available, I'm just not sure where, and for what price. (David)
Airplane motors can be operated well below rated specs for voltage and current and they will run cool and last until the bearings give out. (Mike)
“Ideally turbomachinery designers would magically appear. There are 20 people when I search for 'Turbomachinery' on Upwork.. people could be found by other means too. I think Mark's design is probably not far off where we need to be though, in terms of dimensions, and definitely a good starting point.” (David)
Euan French 9:17 AM (helpfulengineering.slack.com)
Another call for any projects that might need help with CFD! Currently assisting with the pneumatic ventilator project, but happy to help where I can.
#project-turbofan-ventilator (helpfulengineering.slack.com)
Some discussion occurring here: https://github.com/jcl5m1/ventilator/issues/8
Blower specifications:
40cm3 at 240L/min (from comment on helpfulengineering.slack.com)
Additionally: Deliver at least 400ml of air/air 02 mix in no more than 1.5 seconds. The ability to change the rate at which air is pushed into the patient is desirable but not essential. (https://www.britishchambers.org.uk/media/get/Specification%20For%20RMVS%20Challenge.pdf)
Check out the specifications here too: https://www.aarc.org/wp-content/uploads/2014/11/19802-001-F-LTV-1200-and-1150-Ops-Manual.pdf
Procurement Group - Blower (‘procurement-blower’ slack channel)
Critical Task!
It's also possible to buy assembled blowers for CPAP machines. For example: https://www.alibaba.com/product-detail/24-volt-dc-motor-fan-middle_60663035973.html?spm=a2700.wholesale.deiletai6.10.275a3bafIvRGvX This model I believe has sufficient output for a ventilator. Someone mentioned 240L/min and 40cm h2o in the other slack group as requirements. There are almost certainly other factories producing similar products. (David)
This group would try to contact manufacturers of suitable blowers. Search on Alibaba, search Baidu in Mandarin, look for manufacturers in US/Europe etc. On establishing contact, asking about their current production capability, and request samples be sent out for evaluation. For Chinese manufacturers, best if that person speaks Chinese.
If there's a suitable existing turbine from a reliable supplier, manufacturing can be simplified.
Design Group - Motor control (‘design-motor-control’ slack channel)
This group will develop a motor control board. Johnny’s design uses a hobby ESC. The firmware on these might do something unexpected or inappropriate (i.e. shut down when a certain temperature is reached). A better motor control board can provide diagnostic information and test/stress modes.
Some starting points:
STSPIN32F0 Advanced BLDC controller with embedded STM32 MCU evaluation board
https://www.st.com/en/evaluation-tools/steval-spin3201.html . STSPIN32F0 chip is $1.90 in q1000 (from LCSC), and has plenty of processing power for advanced motor control techniques. Motor control SDK and PC tools available.
Small tidy board based around the ATMEGA328. These simpler 8-bit MCUs don’t have the processing throughput for ‘FOC’ control, but their simpler architecture and shorter pipeline have some advantages (less latency).
http://www.electronoobs.com/eng_arduino_tut91.php
VESC - open source custom ESC based around STM32F4, DRV8302 MOSFET driver, ChibiOS. Extremely flexible and with nice diagnostics and control PC application. Possibly too many moving parts for high-reliability design, and generally overkill.
http://vedder.se/2015/01/vesc-open-source-esc/
Other ESC projects:
https://github.com/bitdump/BLHeli
See also:
https://www.embedded.com/designing-electronic-speed-controllers-for-drones/
Design Group - Power supply (#battery-and-power)
Plan of Record:
- Powered by 120/240 50hz/60hz mains power
- Purchased AC->DC power supply
- Backup power with batteries
- Batteries will be 12V automotive lead-acid
- 2X batteries for redundancy and to enable hot-swap
- Modular battery electronics; identical boards, one for each battery
- Charging circuit
- Boost converter to match system bus voltage requirements
- Priority selection circuit to avoid discharging both batteries at the same time
- Basic monitoring and fault detection
TBD:
- Power consumption on mains and on battery
- System bus voltage
- Both batteries external, or one internal:
- Argument for one internal, one external:
- With both batteries external, there's more of a risk of someone accidently disconnecting both.
- If one battery was internal, and one was external, we could set it up so that the internal battery was always secondary. So essentially it would only be discharging if the primary circuit breaks, or during a hot swap.
- As soon as you complete a hot swap, it would switch back to the external battery.
- It would still only charge the internal battery when mains was connected.
- With two external batteries, there are more failure modes from incorrect connections
- Argument for both external:
- Continuous hot-swaps could be achieved by alternating between batteries
- If a battery was internal, we would have to design the enclosure around a specific battery size
- With one battery internal, we would also require the end user to open the ventilator to install the battery.
The group should provide (design or source) a (12-24V) power supply, for a single blower.
Regarding power supply.. they are easily available from China (banggood.com/AC110V220V-to-DC12V-10A-120W-Switching-Power-Supply-1989842mm-p-1457072.html?rmmds=search&cur_warehouse=CN). These are probably not fantastic.
Laptop chargers around 90W can be good quality and easy to find.. the old round-barrel thinkpad chargers come to mind. The connectors are easy to get and they put out 20V. (David)
I have some experience with cheap stuff from China. Don't buy those crappy switching power supplies. They are very unreliable. MeanWell in Taiwan makes quality supplies that are still pretty cheap, and they can usually be sourced in the US and other countries because they distribute everywhere. (Mark)
ATX computer power supplies are extremely plentiful in the e-waste stream and are basically everywhere.
They would be the most readily available power supplies, can provide 12V high amperage and 5v and 3.3v for logic etc. (Stephen Sheehan)
Design Group - Control board hardware
Produce a design for a board that will interface with the blower(s) motor control board and sensors, that will operate the overall logic of the ventilator. The board should have interfaces for sensors, motor control board(s), and a data/control interface, that can communicate diagnostics information, control messages and other data to a desktop application or other UI device.
MCU yet to be determined. STM32 processors are ‘vanilla ice cream’, unless some special capabilities are needed. Snippets of circuits and sensor driver code can be lifted from drone control boards (betaflight, pixhawk, ardupilot).
The user interface/controls for the ventilator have not been considered. Check the ‘Drager portable ventilator demos’ in ‘Useful Resources’ for some ideas.
A header/footprint on the PCB could allow an optional ESP32 board to be attached that could relay the data/control messages over wired ethernet, wifi, or bluetooth.
Control board firmware group
Initially, this group will research the requirements for different ventilator modes, start to plan the firmware architecture, and prepare some useful code. When hardware boards are available, they will implement and test firmware.
Philip Kamenarsky 10:14 AM (helpfulengineering.slack.com)
for anybody who's interested in software correctness: i've submitted a new channel request, #discussion-high-assurance-software
Andrew Tergis 17 hours ago (helpfulengineering.slack.com)
There are other tools typically used in safe embedded applications: No dynamic memory allocation at all, use of a watchdog timer, systems designed to be non-reentrant (single-threaded).
I'd also offer up the Unity testing framework, i've found it to be nicely lightweight: http://www.throwtheswitch.org/unity
Lastly, smart system design: ensure that the MCU cannot be stopped in a state that will be hazardous. Put in fuses and circuit breakers to avoid hazardous current draw, ensure actuators are disabled by default or automatically disabled if not serviced by the MCU in a certain amount of time. This is all hardware EXTERNAL to the MCU.
Pedro 11 hours ago (helpful engineering.slack.com)
Hi, I've been working for 8 years designing electronics and software for safety critical railway applications.
Design Group - Sensors
Challenging, Critical Task!
To provide accurate volumetric ventilation modes, there is a need to precisely measure the airflow.
There are different designs for sensors that could achieve this (https://en.wikipedia.org/wiki/Flow_measurement). Such a sensor is important to measure the total volume delivered to the patient, and also possibly the volume of oxygen that is supplied to the ventilator by the pressure regulator/flow meter.
"450ml +/- 10ml per breath" - I have doubts whether an airflow sensor, based on a differential pressure sensor, can meet this level of precision for volume. It may need a different approach (metering 'wheel'?), but this can be an additional sensor that plugs into the control board. Would be very helpful to speak to a ventilator technician here. (David)
There would need to be a pressure regulator/flow meter for the oxygen supply. I think existing devices could be used, seperate from the ventilator. A sensor that measures oxygen flow and reports it to the control board could be helpful. (David)
In addition, the control board needs to know the output pressure of the ventilator.
Some ventilators measure the pressure near the mouthpiece.
“I made a CPAP manometer using MPXV7002, same part from Ardupilot”
https://github.com/jcl5m1/ventilator/issues/22#issue-583945054 (Bin S.)
“IMPXV7002 has desired measuring range and sensitivity for ventilators” (Bin S.)
$9 from Digikey for quantity 1000, about 40 CNY from sellers on 1688.com (David)
A couple of interesting links:
https://www.sensirion.com/en/flow-sensors/sfm3xxx-sensor-platform/
https://hackaday.io/project/170446-helpful-engineeringopen-source-mass-airflow-meter
Discussion of a low cost flow sensor is currently ongoing here:
See also #project-flow-sensor on helpfulengineering.slack.com
Design Group - Qt Application
Create an application for Diagnostics/UI that will receive data from the control board (motor speed, motor current, output pressure, output airflow etc)., visualise it, and allow control over control board parameters and modes. This application will be useful for developing control algorithms for different ventilation modes etc.
Consider the UI of commercial ventilators for ideas for how to present useful data and graphs.
I will consider a design for a protocol that will be used over a serial link.. to communicate between a motor controller, ventilator control board, possibly sensors, and PC. Mavproxy/Modem AT commands/Wiimote data format/something. It could probably be quickly implemented on Johnny's Arduino design, and communicate with a Qt app on a PC to provide control and visualisation of data. Qt isn't the only framework, but, it's a good fit here, imo. (David)
Also, any Qt programmers in the house? I'll go looking for one. [...] I spotted six in helpfulengineering.slack.com . (David)
I am not fully into QT, but I am fully into helping in this. Is there a channel in the group we can talk? (Joao)
I have Qt Application development experience and have started working on this with another group. Definitely interested in collaborating. Any channels to communicate about this? Open repositories to join? (Julian)
On Materials and Manufacturing (from Slack discussion)
My only concern with 3d printed parts is FFF/FDM prints having crevices for bacteria to get stuck in which can't easily be cleaned. There are solutions to this such as coating, and it might be fine to use them uncoated for the short term. The 3d printed air valves worry me for the same reason. (Alex)
The PETG etc. can also withstand normal disinfectants such as alcohol. (Bin S.)
Resin prints should also be relatively safe from the bacterial issue, though they need to be fully cured. (Alex)
Aluminium mold can be made quickly, then you can make thousands of parts. Factories in Shenzhen are running, I think. (David)
PCBs are also quick to make, and inexpensive. Once manufacturing files are available (gerbers), a PCB house in China can source the parts, make the PCBs, and assemble them, in four days even. JLCPCB does it in 24hrs, but only if they stock all components in house. If you keep in mind their catalogue when designing, that's not too difficult. (David)
Thoughts and Criticism
“It does it a bit.. strange.. to design a machine from scratch, when there are many manufacturers who know exactly what needs to be done. They are in the best position to design a low-cost machine that could be rapidly manufactured. ?” (David)
Useful Resources
There is a set of 5 videos on Youtube by the aptly named Medi-Cram which outline the basics of ITU mechanical ventilation.
NXP Ventilator/Respirator Hardware and Software Design Specification (excellent info!)
https://www.nxp.com/docs/en/application-note/DRM127.pdf
Drager portable ventilator demos:
https://www.youtube.com/watch?v=eCBdXQDXyj4
https://www.youtube.com/watch?v=UT40WUKmuCA
https://www.youtube.com/watch?v=AVev0DYYWQ4
LTV ® 1200, 1150 Ventilator, and MR Conditional LTV ® 1200 System Operator’s Manual (Modern Centrifugal Compressor ventilator)
https://www.aarc.org/wp-content/uploads/2014/11/19802-001-F-LTV-1200-and-1150-Ops-Manual.pdf
(especially check p244, ‘specifications)
COVID-19 Cheatsheet
https://app.slack.com/client/TUTSYURT3/CV96HC6DN/thread/CV96HC6DN-1584433853.499600
George Birchenough: Hi Guys. I am an aerospace engineer Uk based with a background in product design including ventilators. My masters thesis at Imperial College was a portable, low-cost CPAP device. […] I would love to join the project wherever possible. See attached my project documentation.
Indicative Specification for a Rapidly Manufactured Ventilation System (RMVS)
“We are looking for an existing, proven technology that can be rapidly adapted to be built in the UK. The winning technology will be adapted for manufacture and use in the UK through UCL’s Institute for Healthcare Engineering”
EU Request for ventilators
Code Life Ventilator Challenge
“We need YOU to design a simple, maintainable, easy-to-manufacture ventilator to provide life support to COVID patients anywhere in the world.”
https://www.agorize.com/en/challenges/code-life-challenge/pages/guidelines?lang=en
Page 27 paragraph 3.2 outlines ventilator parameters for Covid-19 patients and some ventilator filter requirements, page 48 outlines nursing procedures for ventilator maintenance and care
https://video-intl.alicdn.com/Handbook%20of%20COVID-19%20Prevention%20and%20Treatment.pdf
VENTILATION, VENTILATORS and HUMIDFICATION - USyd Lecture notes
http://www.anaesthesia.med.usyd.edu.au/resources/lectures/ventilation_clt/ventilation.html
Some Curated Comments from helpfulengineering.slack.com
KosenHitachi 11:21 PM
Folks, quick word, i'll try to source them later.
I've been talking to my dad who is an emergency doctor in France (basically the first in line).
Please be careful, I'm not speaking about what is going in the ICU resuscitation service.
He was telling me that he only needs to be able to control 3 things on the machine (when on the frontline):
%O2 concentration
Tidal volume (Vt) ( computed as follow : weight*alpha ml/Kg )
Respiratory rate (Rr) (in cycles/minutes) -> refer to https://helpfulengineering.slack.com/archives/CV96HC6DN/p1584402876402200?thread_ts=1584397302.392900&cid=CV96HC6DN
Regarding the rest :
PEEP (residual lung pressure) should stay below <20 mm of mercury
The device they are using is pretty simple (OSIRIS 2), it does not control CO2 concentration.
They can get this info thanks to an oximeter..
https://urgences-serveur.fr/fiche-technique-smur-osiris-2,981.html
(maintenance et safety check guide) https://sofia.medicalistes.fr/spip/IMG/pdf/OSIRIS_2.pdf
Eric Griffin 18 hours ago
The real value in these would be if a standardized proven open source design could be manufactured using standard tools and processes that most medium sized assembly facilities around the world already have in place, using COTS components that are commercially available everywhere. That way, consistent, identical units could be mass produced in a decentralized manner using local resources.
Those designs already exist and have been cast aside as too practical, too corporate, or too non-prusa related. Because unless it looks like it was made by homeless people from dumpster found bits then its not open source.
Paul Perera 2:25 AM
Re- Viral aerosols: One adaptation to think about is that the exhaled gas vents into the room. This will be full of virus and is a really major concern among clinicians now, especially with gown and mask shortages. HEPA filter attachment on the exit port perhaps? If any experts in that area might want to look into this, it might save time later on. I am thinking of medical viral HEPA filters but also cleaner motor HEPA filters (Roomba or Dyson for example) and domestic air purifier machine HEPA filters (Philips, Honeywell). This may become a really big deal.
Paul Perera 2:24 AM
Interesting input from John Dingley [johndingley98@gmail.com]
Hi Paul,
The OxyLog 3000plus is their latest current model. If you Google it, you will see it has a much more detailed large screen, internal battery etc and also has a non-straightforward breathing circuit emanating from it. This is to make it fully compliant with all the alarms and safety functions that current legislation requires and of course the temptation to succumb to "mission creep" where they keep adding extra features to each successive model.
The OxyLog 1000 however does still have some alarms, some monitoring electronics and useful functions when compared to the very original ultra-basic oxylog (image attached) which was entirely pneumatic with hardly any alarms and no electronics. I am not sure if the 1000 is still manufactured by Drager as they now have their newer models. I will find out from their website.
It is a very good compromise choice to re-engineer even so. It does have a special breathing circuit that you have to use with it so this would have to be included in the design process.
It is also in use in hospitals to transport ventilated patients around or between hospitals.
Matt Aldridge
Hi all, I'm an anaesthetist and ICU doctor from the UK. I love all the enthusiasm but wanted to share this across several groups as I think we should be focusing on the reason I joined this group in the first place which was to create plans for a simple mass producible ventilator.
Basically what we need for covid are plans for mass producible invasive ventilators that can supply high concentration 02 from hospital supplies with high levels of peep (up to 20cmh20). If these can be at least partially 3d printed then all the better.
Hospitals will take whatever they can get once it gets bad and in the UK alone we are going to need 1000s of new ventilators. If it works then there will be a place for it. Yes staffing will be an issue, but we are expecting to adapt and stretch existing personnel. We can't do this without the right hardware however. 02 supply in many hospitals also far outstrips the available ventilator supply at present.
Ideally any device would be simple, robust and mechanical (minimal electronics). Covid patients are not difficult to ventilate so any ventilator types/modes would work but they do need peep and high 02 as oxygenation is the problem. We don't need anything near as sophisticated as current ICU vents for the majority of patients.
Otherwise vents could be pressure or flow limited with time or volume cycling. Normal vent settings are rates of 10-25bpm, inspiratory: expiratory ratio of 1:2, volumes of 200-600ml and max inspiratory pressures of 30-40cmh20, with peep up to 20cmh20. Supply pipeline pressures are 440kpa for 02 and air, with tubing needing to fit 15/22mm connectors.