General > General Technical Chat
What did we learn from the "open source ventilator" mess.......
tggzzz:
One perspective, from comp.arch.embedded
On 08/07/20 04:36, Rick C wrote:
> On Tuesday, July 7, 2020 at 10:02:45 PM UTC-4, Paul Rubin wrote:
>> Rick C <gnuarm.deletethisbit@gmail.com> writes:
>>> Wow! Designing in a group is strange. I'm helping with the design of a
>>> ventilator and it is a unique experience.
>>
>> Sounds like fun, but why a ventilator? Is the ventilator shortage still
>> happening and are there more urgent shortages? I thought there were
>> already a bunch of ventilator projects.
>>
>> What is the software going to be like, and is it being taken care of? This
>> sounds like a good opportunity to write some Ada code.
>
> Yeah, that was one of my first observations that there are tons of ventilator
> projects. They nearly all are based on squeezing an Ambu-bag to push air
> into the patient. The details of not harming the patient are tricky.
> Ventilating a patient while conscious is harder and very few are doing that.
> They seem to be having problems with the plumbing at the moment, getting all
> the right valves in the right places to do the right things. At the moment
> there is a lot of hardware right at the patient's mouth. I'm working to
> understand what is needed and what isn't.
>
> There's a valve called a Ruben valve which allows the patient to be connected
> to the inspiratory limb (branch from the ventilator) or to the expiratory
> limb (the path for exhaling) without a leak to the outside via a straight
> through path. It's not entirely clear just how the details of this valve
> work. I have concerns that the valve can be stuck in the inhale position by
> the pressure in the line. It doesn't seem to have a way to leak out once
> pressurized.
>
> I'm not sure the software is too hard at the higher levels, but there are
> interface procedures that have to convert differential pressure to flow rate
> and monitor other marginal sensors. The O2 sensor is only ~18% accurate.
> The guy running the project doesn't seem to understand the difference between
> accuracy, resolution and noise when we were talking about increasing the
> resolution of a reading by averaging many samples. Not that this was a big
> deal. The real issue is I'm not sure improving the resolution of the
> measurement will give him what he wants. The airflow measurement is very
> sensitive at the low end. I'm wondering if there is an issue with the flow
> being more or less laminar at the low end so that it is not going to be an
> accurate reading. The interesting part is they verified the home brew flow
> rate sensor by using a commercial flow rate sensor that uses the same
> technique.
>
> Whatever. I'm trying to help without being a drag. I think sometimes my
> questions are not what they want to hear. Maybe I'm too pessimistic.
>
> The software is C I believe. I don't think anyone in this crowd is going to
> use Ada.
Smokey:
I'm a little surprised how casually some of you are suggesting Medtronic should have just "ramped up", especially given that I know some of you actually have manufacturing experience. I would expect the market for ventilators was pretty stable and small volume so why would they have that extra capacity in place already? I can think of a few companies that failed because they had an unexpected bump in demand and botched the "ramp up".
Another big question for me would be how many of the components used in the Medtronic ventilator were outsourced to China or India or one of the other countries that was effectively shut down at the beginning of the pandemic. Given modern parts sourcing and volume manufacturing strategies, I'm going to guess a significant number of parts (both custom and off the shelf) were not made locally and out of their direct control. So now you can't get your overseas factories to make higher number of parts you need and you can't get the parts shipped quickly anyway since the bulk transport was all messed up. At that point your only option would be to go through the pain and suffering of finding and qualifying a new local vendor at a time when the businesses are shut down and people are afraid they will die if they leave their house. The medical device certifications probably prohibit you from arbitrarily swapping vendors anyway, which could be waved in the emergency but that's one more hoop that needs jumping. Good luck with all that.
/speculation/
In my opinion its plausible that Medtronic realized that what the world needed in that moment was essentially what it looks like the JPL team made, which was a high quality but stripped down machine built from parts that were actually available at that moment. To facilitate this, they released their full proved specification which as I've said elsewhere is essentially the highly valuable roadmap and up front investment to quickly making a functional device for anyone that's capable. They got that out first because that's the most important. Having the BOMs or the mechanical assemblies may be helpful from a reference standpoint, but if you are making things from scratch based on what you can get now they aren't necessary so they release those later for whatever reason. Since the expectation was that the result would be a new custom device anyway, they protected their specific machine design elements with a restrictive license long term.
Now you may be asking why Medtronic didn't just make the new stripped down thing themselves, which is a legit question too. It takes a certain type of engineer and organization to function well in the skunkworks type environment. Unless that's what you do, you will probably fail if you try to sprint through a design. Developing things super quickly is a different skill and process, and I'm going to guess a big medical device company with multi-year design cycles is not set up that way.
Of course, their evil open-source hating lawyers may have been scheming about how to trap unsuspecting do good engineers with subtle license clauses, but I'm not sure I'd go there first when there are other possibilities. With the spec being the investment, you don't just give that away hoping to get some publicity and eventual license royalties and especially not from a mega-corp. Maybe they were just actually trying to do good, and they still got shit on.
EEVblog:
--- Quote from: Smokey on July 08, 2020, 11:51:15 am ---I'm a little surprised how casually some of you are suggesting Medtronic should have just "ramped up"
--- End quote ---
I'm saying that if the government put in a huge order Medtronics (and others) could have ramped up production, and done so vastly easier and quicker and to certification than anyone dicking around with the source files they released.
EEVblog:
--- Quote from: Smokey on July 08, 2020, 11:51:15 am ---/speculation/
In my opinion its plausible that Medtronic realized that what the world needed in that moment was essentially what it looks like the JPL team made, which was a high quality but stripped down machine built from parts that were actually available at that moment. To facilitate this, they released their full proved specification which as I've said elsewhere is essentially the highly valuable roadmap and up front investment to quickly making a functional device for anyone that's capable. They got that out first because that's the most important. Having the BOMs or the mechanical assemblies may be helpful from a reference standpoint, but if you are making things from scratch based on what you can get now they aren't necessary so they release those later for whatever reason. Since the expectation was that the result would be a new custom device anyway, they protected their specific machine design elements with a restrictive license long term.
Now you may be asking why Medtronic didn't just make the new stripped down thing themselves, which is a legit question too. It takes a certain type of engineer and organization to function well in the skunkworks type environment. Unless that's what you do, you will probably fail if you try to sprint through a design. Developing things super quickly is a different skill and process, and I'm going to guess a big medical device company with multi-year design cycles is not set up that way.
Of course, their evil open-source hating lawyers may have been scheming about how to trap unsuspecting do good engineers with subtle license clauses, but I'm not sure I'd go there first when there are other possibilities. With the spec being the investment, you don't just give that away hoping to get some publicity and eventual license royalties and especially not from a mega-corp. Maybe they were just actually trying to do good, and they still got shit on.
--- End quote ---
Jones's Razor: Never attribute to genuine helpful intention that which can be adequately explained by a publicity opportunity.
donotdespisethesnake:
--- Quote from: EEVblog on July 09, 2020, 07:40:03 am ---Jones's Razor: Never attribute to genuine helpful intention that which can be adequately explained by a publicity opportunity.
--- End quote ---
:-+ Good one, probably right in this case. I expect someone in the boardroom though it a good idea and made the order, after being filtered through the rest of the company (enginering, business, legal) they publish the design for the most obsolete product they have which has the least impact on their sales, and with the most restrictive license possible (CYA).
But rather than rehashing old arguments about motivations, did any of these open designs actually get deployed in real hospitals?
I guess the real lesson is that there rarely is a quick techno-fix for this type of problem, despite how much people would like one. What works is monitoring and tracking disease vectors before they become human-human transmissible, and contact tracking when they do. All of which is boring but well-known by the WHO, who try to encourage countries to do it, despite lack of funding or other political agendas getting in the way.
Unfortunately, the interaction with humans and wild animals is generally increasing, as humans continue to expand into wild habitats. Nature has a way of dealing with "too many humans" (like any other excess population). Expect more pandemics. Build all the ventilators you want, it won't really help much.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version