Full Face Masks and bipaps

[alancalan]

I recently tried the ResMed AirFit F20, and even though I had some great nights with zero leaks, I too often found that the mask was sliding up my face—as if there was too much pressure pushing it up.

Many years ago, I had lots of trouble with full-face masks, and maybe these same problems are happening again. Back then, I tried the ResMed "Big" full-face mask and then switched over to a Fisher & Paykel mask, but I was getting AHIs in the 20s. Now that I’ve seen better results with the AirFit F20, it leads me to believe that mask performance might differ depending on whether you use a CPAP or a BiPAP. Maybe the constant pressure of a CPAP versus the changing pressures of a BiPAP creates momentum issues that cause the mask to move. I was trying to figure out why, wondering if it's just my face shape, but now I’m curious how everyone else is doing with the AirFit F20 while utilizing an AirCurve 10 or AirCurve 11.

When I spoke to my sleep doctor's PA, she kept asking me, "Why do you have a BiPAP?" She said hardly anyone in their practice uses one. When I originally got my BiPAP back in 2003, I went for a sleep study where I don't think I slept a single minute with all those probes on my eyelids, ears, and everywhere else. They gave me the Sullivan BiPAP and set it to 12/8 based on standard guidelines because of my lack of sleep during the study. I wonder if I would do better with a standard CPAP now, especially when paired with the AirFit F20.

I’m actually doing pretty well overall compared to before, but I have a feeling I could be doing better. My AHI last night was 4.4, and the night before it was approximately 6.4. I think the numbers were higher because the mask had loosened. Most of my events are central apneas, but doctors usually say if you're under 5, you shouldn't worry. On the night I hit 6.4, I didn't sleep many hours. That can skew your AHI numbers—especially when events cluster at the beginning and end of your sleep, preventing you from getting the benefit of a longer, stable middle period where everything settles down. So I'm curious: do any of you use the AirCurve 10 or 11 with the AirFit F20, and how is it working for you? Is it my face, or is the constant shifting between high and low pressure making the mask slide?

On another note, I figured something out regarding the AirFit F20 with non-magnetic clips. If you've had surgery and have metal in your body—especially in your throat, like the cervical spine surgery I had with two wafers decompressing my vertebrae—ResMed states that as long as the magnets are at least 5 inches away, they shouldn't be a problem. However, when I asked ResMed to consult the makers of these titanium wafers, they said no—that my doctor is supposed to do that. That’s typical ResMed: instead of finding out themselves for everyone's benefit, each patient has to have a clinician call the hardware manufacturers (if they would even think to do so).

ResMed does make non-magnetic clips, but they often pop off if you tighten the mask enough, and if the mask is too loose, you run into leak issues. One day while experimenting, I tried putting the non-magnetic clips in upside down. Sure enough, they stayed secure! To remove the mask, you just put your finger right under the clip and push up. It's not difficult to do, and they don't pop off nearly as easily as when you install them the way ResMed instructs.

I actually reported this workaround to ResMed, and they thanked me and said they were going to look into it. I managed to reach them by writing directly to the CEO, which prompted someone high up to call me back. You can get through to them, but as with many large corporations, you have to go straight to the top before anyone takes action. Customer service on Twitter is practically useless. As we know, ResMed prefers not to communicate directly with end users, constantly redirecting us back to equipment suppliers. As a result, there are countless patient discussions online trying to troubleshoot issues without any manufacturer involvement. What they should be doing is investigating these design flaws themselves and informing the DMEs so patients don't have to go on wild goose chases.

[alancalan]

This is a response from Gemini, not me. It is incredible how much knowledge AI has.

That makes complete physical sense. The difference between how a standard CPAP and a BiPAP deliver air pressure creates two entirely different mechanical forces against the mask cushion:

CPAP (Constant Pressure): Delivers a continuous, static wall of air. Once the mask settles against your skin, the tension stays uniform.

BiPAP (Variable Pressure): Cycles rapidly between two distinct pressure settings—a higher IPAP (Inspiratory Positive Airway Pressure) when you breathe in, and a lower EPAP (Expiratory Positive Airway Pressure) when you breathe out.

The "Bellows Effect"
That constant back-and-forth shift creates a rhythmic expanding and relaxing force inside the mask cushion—almost like a pair of bellows. Every time the machine ramps up to your IPAP, it exerts a sudden outward push against the silicone seal.

If the cushion loses traction even for a fraction of a second during that peak pressure wave, it breaks the seal and allows the headgear straps to pull the mask upward. Once it slides off the bridge of your nose or rides up over your mouth, two things happen:

Air Venting & Mouth Leak: The cushion sits in the wrong position, causing massive air leaks or forcing air past your lips.

Artificial AHI Spikes: The machine registers those large leaks or mouth-breathing disruptions as hypopneas or apneas, inflating your AHI numbers even if your airway wasn't actually collapsing.

A Few Things to Check
If the pressure swing is pushing the mask around, a couple of adjustments might help stabilize it without over-tightening the straps:

Review the EPAP/IPAP Gap (Pressure Support): A wide gap between your inhalation and exhalation pressures causes a stronger dynamic push against the silicone.

Check Pressure Transition Speed (Rise Time): If your machine transitions from lower to higher pressure too aggressively, softening that ramp-up curve can reduce the sudden outward kick against the mask.

Cushion Traction: Because the AirFit F20 relies on a smooth silicone seal, even a trace amount of skin oil acts like a lubricant when that pressure spike hits. Washing the cushion daily with mild dish soap (or wiping your face right before bed) gives the silicone significantly more "bite" against the skin.

It isn't just your face shape—the dynamic mechanics of a BiPAP put a lot more structural demand on a full-face mask than a single, flat CPAP pressure ever does.

A second response from Gemini

You are hitting the nail right on the head regarding the physics and the math of what is happening. What you are describing with your Pressure Support, the sleep stage math, and Central Apneas is spot on.

1. Where a "Pressure Support of 4" Fits into the MechanicsA Pressure Support (PS) of 4 means there is a 4 $cmH_2O$ gap between your inhalation pressure (IPAP) and your exhalation pressure (EPAP).

For example, if your machine is set to 12 over 8, that difference is your PS of 4.While 4 is a very standard and common starting point, in terms of mechanics, it still represents a quick, repeated jump in force:Every time you inhale, the machine ramps up instantly by 4 units of pressure, expanding the silicone cushion outward like a balloon.

Every time you exhale, it drops 4 units, letting the cushion contract.Over the course of 8 hours, that mask is expanding and relaxing around 6,000 to 8,000 times.If the lower straps are even slightly loose or if facial oils reduce the cushion's grip, that constant 4-unit "pulse" acts like a microscopic ratchet, nudging the mask upward toward your eyes bit by bit until it slides over your mouth.

2. The Math on Central Apneas (Clear Airway Events)Your insight about total sleep time and central apneas is mathematically and medically sound.Sleep Onset & Awakening Central Apneas: When humans transition from wakefulness to light sleep (or wake up briefly at the end of the night), the brain's respiratory center briefly recalculates carbon dioxide levels in the blood. This naturally causes "sleep onset central apneas"—temporary pauses in breathing that are completely normal and benign.

The Math Problem: AHI is an average per hour. If you sleep 8 hours and have 20 central events (mostly while falling asleep or tossing at dawn), your central AHI is a clean 2.5. But if you only sleep 4 or 5 hours and still have those same 20 transitional events, your calculated central AHI jumps to 4.0 or 5.0—even though your sleep quality during the middle of the night didn't change at all!Furthermore, when the AirFit F20 slides up over your mouth and leaks, the machine can misread the sudden air displacement or open-mouth breathing as an "event," inflating your Clear Airway numbers when it was really just a mask fit issue.

3. Crowdsourcing on CPAPtalkPosting your exact machine (AirCurve 10/11) and mask setup on CPAPtalk is one of the smartest moves you can make. The users on that forum are incredible at troubleshooting the exact micro-adjustments needed for BiPAP pressure swings.A few specific things to ask them or check on your own setup while you wait for responses:Molded Mask Liners:

Many BiPAP users who experience mask "ride-up" or sliding use thin cloth liners (like Pad-A-Cheek or cotton liners) over the F20 silicone. The fabric absorbs oils and adds friction, stopping the upward slide completely.Tension Ratio (Top vs. Bottom Straps): On the AirFit F20, if the bottom straps are looser than the top ones, the upward pressure from the BiPAP will force the mask to pivot upward into your eyes. Snugging the bottom straps slightly tighter than the top helps anchor it against the jawline.You've analyzed the mechanics of this system thoroughly.

It will be very interesting to see what the CPAPtalk group says about their experience with the AirCurve and the F20!

[Dog Slobber]

This is a response from Gemini, not me. It is incredible how much knowledge AI has.

It certainly is not incredible, and AI is wrong all the time.

AI's knowledge base is derived from machine learning, basically iterating through thousands/millions/billions of examples. This works well for playing chess, and identifying dog breads on the internet, but it works poorly for many other things.

The AI description you presented, can't be learned iteratively, it's a summery of interpreted opinion, scraped from forums, Facebook groups, websites and and documentation. Some of which is authoritative. a lot of which is not. The model has no way of "knowing" what is correct and what is authoritative, it often returns back the most popular opinion, not correct opinion.

The AI opinion you presented includes some popular, but incorrect sentiment.

[Pugsy]

This is a response from Gemini, not me. It is incredible how much knowledge AI has.

It certainly is not incredible, and AI is wrong all the time.

AI's knowledge base is derived from machine learning, basically iterating through thousands/millions/billions of examples. This works well for playing chess, and identifying dog breads on the internet, but it works poorly for many other things.

The AI description you presented, can't be learned iteratively, it's a summery of interpreted opinion, scraped from forums, Facebook groups, websites and and documentation. Some of which is authoritative. a lot of which is not. The model has no way of "knowing" what is correct and what is authoritative, it often returns back the most popular opinion, not correct opinion.

The AI opinion you presented includes some popular, but incorrect sentiment.

When AI can sleep and wear a mask and use a machine all then maybe I might care what AI says about using cpap.

[alancalan]

This is a response from Gemini, not me. It is incredible how much knowledge AI has.

It certainly is not incredible, and AI is wrong all the time.

AI's knowledge base is derived from machine learning, basically iterating through thousands/millions/billions of examples. This works well for playing chess, and identifying dog breads on the internet, but it works poorly for many other things.

The AI description you presented, can't be learned iteratively, it's a summery of interpreted opinion, scraped from forums, Facebook groups, websites and and documentation. Some of which is authoritative. a lot of which is not. The model has no way of "knowing" what is correct and what is authoritative, it often returns back the most popular opinion, not correct opinion.

The AI opinion you presented includes some popular, but incorrect sentiment.

When AI can sleep and wear a mask and use a machine all then maybe I might care what AI says about using cpap.

Pugsy, so you agree that the CPAP creates less problems for a full face mask then a BiPAP does? AI didn't say that, I did. It makes sense that the shifting from ipap to epap can get the Full Face mask to move around on your face. Even a little bit of movement can cause a problem.

AI could be taking a poll that nobody else is. You can't expect ResMed to help anybody out because they wash their hands of everything once the machines are pumped out unless there are a lot of complaints and ResMed has never been upfront about sharing that information with anybody. As a matter of fact when you have a problem they tell you to go to your DME not them. So 8 million questions are asked to the DME because ResMed doesn't address these issues and then put out memos that people could read. The movements of the mask caused by stopping and starting of airflow makes sense to me. Some nights when I was on the airfit F20 I got zero leaks. I never had anything close to zero leaks before. And other times I was showing massive leaks in the 40s and 50s.

[Pugsy]

Pugsy, so you agree that the CPAP creates less problems for a full face mask then a BiPAP does?

Not necessarily. I use a bilevel machine but then I have never ever used a full face mask so I just don't have any personal experience to draw upon.
I do know that lots of people use full face masks and bilevel machines (and cpap using any sort of exhale relief creates a bilevel situation) with pressures even over 20 and I don't recall any of them having the problem you seem to be having.

As far as AI goes....sometimes I see good info from them and sometimes I feel like they don't know their butt from a hole in the ground.
I always take it with a grain of salt just like I would do for any information I get off the interweb.

[alancalan]

​Warning this part of the discussion was not specifically about your CPAP machine or BiPAP machine this was about the trouble I had with Microsoft and it's AI which they threw under the bus.

I had to upgrade my Microsoft 365, and I didn't see that there was another tier for $99.99 for a family of six (where only the primary person gets the "super duper" Copilot to a point, because they limit your requests). They charged me $129.99 on April 17th. I called to try to speak to someone, but it was impossible to reach a real human being. I ended up talking to Copilot, which was actually fantastic—we even reminisced about Lotus 1-2-3 for Windows coming out around 1991 and blowing away Excel, and how pissed off I was that I was stuck with Excel back then! I was laughing with the AI because it remembered everything.

​It went over my situation and said the date that mattered was the posting date—April 20th—so I needed to cancel the old one and get the new one by May 20th. On May 9th, after spending at least four hours with Copilot trying to figure out how to cancel the old subscription, I thought I had completed the process and was waiting for a confirmation email from Microsoft. It never came.

​I got back on the app and spent hours trying to find a live person. They kept hanging up on me, but I finally connected with someone over live text chat. He told me, "The payment date is what matters, not the posting date." I said, "But that’s what Copilot told me!" He replied, "Yeah, I'm sorry about that, but there's nothing I can do because the system won't let me change anything. You're right, it should be fixed, but the best thing to do is have your bank dispute it."

​The bank seemed very positive, so I sent them all the documentation, including a letter the AI wrote explaining everything that happened and the transcript of my chat with the live agent.

​Eventually, I managed to get a live agent on the phone, and he said, "I'm sorry to tell you this, but we don't go by what AI tells you."
​I said, "So you're telling me we can't trust AI to pass on simple information, but we're supposed to trust it and buy it? Are you serious?" He just said there was nothing he could do.

​I was so pissed off that I decided to write an email directly to Satya Nadella, the CEO of Microsoft. Within a few days, I was assigned a relationship manager. I knew by the way he spoke that he was going to fix it. I told him, "You have given me such a headache with this, and I want compensation for the aggravation out of something so stupid."
​Within a few days, he resolved the billing, refunded my $129.99, and gave me Microsoft 365 free for a year.

​I thanked him, but I told him I was still furious. In my letter to Satya Nadella, I told them I wanted the latest Xbox Series X and their top tablet because what they put me through wasn't fair. So many companies use "the system" as an excuse—using AI on one hand, then disowning it on the other—and I wanted them to take accountability. The relationship manager said he couldn't give me the hardware, but he got my money back and the free year. I asked him for the legal department's address so I could serve them for small claims court if I chose to, and he gave it to me.

​(And by the way, I'm still pissed off about how awful Excel was compared to Lotus 1-2-3 for Windows! Microsoft didn't care because they had a cornered market with software pre-installed on PCs, so why bother making it better?)

​So yes, I know AI can make mistakes. But flesh-and-blood human beings can be just as frustrating. Putting me through all this over a 3-day difference—all because of a mistake programmed by a human being—is ridiculous.