LEAKS and their measurement.

[dataq1]

One of the biggest issues that users have to contend with is leaks. 
But what do we know about how leaks are measured?
I'm making an assumption there are sensors (pressure and flow) are in the body of the cpap machine.
The pressure sensor , after adjusting for pressure loss through the tubing, can monitor the mask pressure.

But how does the flow sensor "measure" the loss of air volume when there is a leak?

(My interest here is to understand how direct a measurement (of leakage) actually is. Seems like a lot of so-called measurements are actually inferences drawn from other measures. For example oxygen saturation inferred by measurement of color absorption, or sleep cycles inferred from "tone" of capillaries. )

[BoxcarPete]

It's a direct reading from a single sensor on the machine's air intake (ResMed Airsense10). The raw data is filtered through an algorithm that attempts to find the "zero point" of flow into and out of the user's lungs inferred from the feedback of flow (and maybe also references pressure for this?) during inhale and exhale cycles. Anything greater than the expected value for the mask used (based on a simple FFM/nasal/pillows setting) is reported as a leak.

All told, it does an admirable job of estimating this from a limited amount of raw data, but it's pretty easily fooled. While awake, I have drawn in a long slow breath that was reported as a leak and a CA followed by a mysteriously large exhale.

[dataq1]

Engineer here as well (it's a cross we just have to bear!)

If I follow your reasoning, the leak calculation* is influenced by what the device infers the users tidal volume is (because the device doesn't directly measure tidal volume either).  The tidal volume is calculated by way of pressure differentials assuming constant supply of airflow. 

* I tend to say measurement value when there is a determined by an actual measurement; and I tend to say "calculated value" when the value is determined by an algorithm (aka "blackbox").

[BoxcarPete]

Tidal volume is (I believe) calculated from the integration of deviation in real-time flow rates from the moving average. The moving average is where you see the red dotted line in OSCAR, which is straightened out for us in the report to make it easier to tell what's going on. When the real-time flow is up from that average, the machine attempts to determine if the flow is into the lungs of the user, or out the side of the mask (or in my case, more than likely right out my open mouth). When real-time flow is below that average, it gets assigned to exhale flow.

Tidal volume is the area between these curves and the baseline. An attempt appears to be made to keep the area above and below equal, which we can see on the charts of people with palatal prolapse. There is a normal inspiration curve, but flow is locked off to the mask and goes out the mouth due to the one-way valve action of their palate. This flow rate should technically be zero, on the red line, but the machine doesn't really know that, so assigns a portion of the mask venting as exhale flow to keep a balanced tidal volume.

Just some observations I've made and the explanations of what I think is behind them. If you want the real scoop, try calling ResMed to see if they'll feed you anything.

[dataq1]

Is it your understanding that the blower motor speed is variable? Such that the blower tries to maintain the command pressure?

[BoxcarPete]

Yes, blower motors in ResMed machines are controlled to maintain pressure, not flow. Whatever flow results is a measured value, not a control value according to my understanding. I think Respironics machines use flow as a control parameter in addition to pressure, and I'm not sure about any other manufacturers.