Skew flow bug may be due to SD card speed...

DreamDiver wrote:
That said - it looks like the data is written to the card for two parameters: flow and pressure at mask 25 times per second. Sounds to me like a slower card could hamper that.

I don't think so. It appears from my testing that the 25Hz data is only written to the SD card once a minute. Also, if you look at the size of the BRP file compared to the number of minutes' worth of data being written, and then compare that to the rated speed capacity of the SD card, it's not that onerous. The machine has almost a minute to complete writing to the BRP file before it's time to write to it again.

As I said above, I think the EVE (event) file is a more likely candidate for data problems because it appears that data is written to it as and when events happen.

DreamDiver wrote: That said - it looks like the data is written to the card for two parameters: flow and pressure at mask 25 times per second. Sounds to me like a slower card could hamper that.

Morning DD and alterego...

That edf software looks great. That also means that we have the ability to look at the data from the S9 without having ReScan 3.10 and ask questions about skewing if its showing there. ResMed cannot say that we shouldn't be looking through their proprietary software. I'm going to DL the tools and play with it later today.

DD: On your graph illustration... is that 1 second intervals across or... ??? Also, are you still getting skewing of data right now ???

BTW: My skew this AM for the Mask Event #19 and Day #7 is... 11 seconds (another 2 seconds)... but no missing Flow Data this date.

It also leads me to some further thought about skewing by either SD cards or memory: Our skewing is progressive by either Days or Mask Events and since it is progressive and each of the sessions is completed before another begins then... If is is the SD speed, then there must be some interaction or overwriting by the S9 to cause it to be progressive... and it has to overwrite the Hi Rez Flow data somehow... unless its all the other data that skews and the Flow data is correct each day. I don't think that the Hi Rez Flow data is written more than once.

If the overwriting doesn't take place... then we have to look at memory or cache memory as likely.

Nord

alterego61 wrote:...It appears from my testing that the 25Hz data is only written to the SD card once a minute. Also, if you look at the size of the BRP file compared to the number of minutes' worth of data being written, and then compare that to the rated speed capacity of the SD card, it's not that onerous. The machine has almost a minute to complete writing to the BRP file before it's time to write to it again.

As I said above, I think the EVE (event) file is a more likely candidate for data problems because it appears that data is written to it as and when events happen.

The once-per-minute write would coincide with the fact that all data begins and ends on the same second. The EVE.edf evidence would also coincide with the fact that some events are written after the flow recording has ended - sometimes half a minute past the flow recording end.

Nord wrote:...DD: On your graph illustration... is that 1 second intervals across or... ??? Also, are you still getting skewing of data right now ???

...It also leads me to some further thought about skewing by either SD cards or memory: Our skewing is progressive by either Days or Mask Events and since it is progressive and each of the sessions is completed before another begins then... If is is the SD speed, then there must be some interaction or overwriting by the S9 to cause it to be progressive... and it has to overwrite the Hi Rez Flow data somehow... unless its all the other data that skews and the Flow data is correct each day. I don't think that the Hi Rez Flow data is written more than once.

If the overwriting doesn't take place... then we have to look at memory or cache memory as likely.

That particular image a couple posts above is a five minute window - 300.0000 seconds - bottom right in the image.

I started using the 4GB class 4 SDHC card as of last night. Class 4 should be sufficiently fast enough to overcome any write or access problems. Given the above information, it begins to sound less like SD card problems, and more like something to do with the S9's memory or caching. I'm going to continue using this card until I see skew. If others are doing similarly, and the process is repeatable, and we see continuing and increasing skew, this avenue of inquiry will be closed. If we don't see skew, perhaps we'll have found an answer.

This will be my standard practice:

• No locking of the card. This seems useless since the data is a non-proprietary format.
• No powering down unless I have a good reason. I can't think of a reason to power down as it doesn't seem to matter to my machine with regards to skew.
• No three-second power-button power-downs. Again, I cannot think of a good reason to do this. It doesn't seem to make any difference.
• Hit the power button to start a mask event.
• Hit the power button to stop a mask event.
• Wait a couple minutes before removing the card - bathroom break, breakfast, whatever.
• Remove the card.
• Stick it in the SD card reader.
• Stick the card reader in my usb slot.
• Open ResScan once per day in the am.
• Wait for the quick start.
• Download the data.
• Discard duplicate data.
• Read detailed graphs.
• Save data.
• Close ResScan 3.10.
• Undock USB card reader.
• Remove card.
• Put card back in the S9.
• Repeat next morning.
• Occasionally, maybe every three to four days, I will back up the data on the card in zip format.

alterego61 wrote: I don't think so. It appears from my testing that the 25Hz data is only written to the SD card once a minute. Also, if you look at the size of the BRP file compared to the number of minutes' worth of data being written, and then compare that to the rated speed capacity of the SD card, it's not that onerous. The machine has almost a minute to complete writing to the BRP file before it's time to write to it again.

As I said above, I think the EVE (event) file is a more likely candidate for data problems because it appears that data is written to it as and when events happen.

Hi alterego...

It seems unlikely that speed of the card is the problem if it is only writing once per minute. I thought before based on some info in the Clinician Manual that the S9 writes to memory... then writes once to the SD card. That might account for some skewing if that were the case.

There was some information out there that memory on the S9 held 365 days Summary Data, 30 days Detailed Data and 1 day Hi Rez data. That seems to be an error on our part. On the S8 you could plug the card into the Flow generator and "claim" all the data once per day. We have not yet tested the S9 to see if there is memory to hold the Hi Rez data... if there is NO card in the machine does it collect the data for us for later. I believe so far, everyone is leaving their card in the machine at all times.

I have limited testing in this area... I knew that the S9 will not write to the card a second time to try to re-collect data. It may be that it does not collect Hi Rez or Detailed data if the card is not there. I will test it after 12 noon today.

The other issue then, is when the skew happens. I believe for me, and others, there is very, very little skewing of data after the S9 starts to write to the card. It basically seems there is a delay of writing that gets progressively worse. The length of writing to Detailed and Hi Rez data seems to co-ordinate properly and write for the correct amount of time... just Detailed Data being delayed by some 3, 5, 7, 8, 9, 11 seconds etc.

I thought that might be due to overwriting the data if there was some interchange between files for Detailed, Summary and Hi Rez. But if the case is, that Summary data stored on the S9 is strictly collected separately by the S9 and never overwritten... then it could only happen if the S9 overwrites data on the SD card for Summary and Detailed. Can we eliminate that possibility through the edf software ???

Just some random thoughts...

Nord

Nord wrote:... if there is NO card in the machine does it collect the data for us for later. I believe so far, everyone is leaving their card in the machine at all times.

I have limited testing in this area... I knew that the S9 will not write to the card a second time to try to re-collect data. It may be that it does not collect Hi Rez or Detailed data if the card is not there. ...

No detailed or high-res data are saved. I've already done that by mistake by forgetting to leave the card in one night. I think others have found similarly. The card must be in the machine to save more than compliance data.

DreamDiver wrote:[
I started using the 4GB class 4 SDHC card as of last night. Class 4 should be sufficiently fast enough to overcome any write or access problems. Given the above information, it begins to sound less like SD card problems, and more like something to do with the S9's memory or caching. I'm going to continue using this card until I see skew. If others are doing similarly, and the process is repeatable, and we see continuing and increasing skew, this avenue of inquiry will be closed. If we don't see skew, perhaps we'll have found an answer.

This will be my standard practice:

• No locking of the card. This seems useless since the data is a non-proprietary format.
• No powering down unless I have a good reason. I can't think of a reason to power down as it doesn't seem to matter to my machine with regards to skew.
• No three-second power-button power-downs. Again, I cannot think of a good reason to do this. It doesn't seem to make any difference.
• Hit the power button to start a mask event.
• Hit the power button to stop a mask event.
• Wait a couple minutes before removing the card - bathroom break, breakfast, whatever.
• Remove the card.
• Stick it in the SD card reader.
• Stick the card reader in my usb slot.
• Open ResScan once per day in the am.
• Wait for the quick start.
• Download the data.
• Discard duplicate data.
• Read detailed graphs.
• Save data.
• Close ResScan 3.10.
• Undock USB card reader.
• Remove card.
• Put card back in the S9.
• Repeat next morning.
• Occasionally, maybe every three to four days, I will back up the data on the card in zip format.

Hi DD...

I presume you will be using the same card without removing any data, until the end of the test.

To give us another picture:
I will follow exactly the same practice for SD... except:

I will use a ScanDisk Class 2 2 Gig card freshly formatted
I will have multiple mask events each day (session)
I will only remove card once in AM to DL data to ResScan
Until skewing begins or 10 days whichever is first.

That should give us an indication of the multiple events but only one DL per day...

Nord

Nord wrote:Hi DD...

I presume you will be using the same card without removing any data, until the end of the test.

To give us another picture:
I will follow exactly the same practice for SD... except:

I will use a ScanDisk Class 2 2 Gig card freshly formatted
I will have multiple mask events each day (session)
I will only remove card once in AM to DL data to ResScan
Until skewing begins or 10 days whichever is first.

That should give us an indication of the multiple events but only one DL per day...

Nord

Same card. However, I have a sleep lab on Sunday, so that data will be absent. Depending on whether and/or when I get a different machine, my experimenting on this machine may be discontinued. I feel I should say this beforehand. So I'm not likely buy any more new SD cards any time soon.

Hi DD... looking after your CA's... good luck on your SS...

TEST #1 to test for standard memory in S9 and writing times to SD card:

Today at 11:20 AM ResMed clock time.
No card in S9.
Placed the mask ON event.
Turned the S9 on.
Faked 2 Apnea Events for 20 seconds each.
Turned off S9 at 11:31 AM.
Placed card in S9 at 11:35 AM.
Removed card from S9 at 12:06 PM (shortly after a new 24 hr session had begun)
Reviewed data on ResScan....

Results...
No Detailed data was written to the SD card for that 10 minutes.
No Hi Rez data was written to SD card for that 10 minutes.
Summary statistics were written to the card for the 10 minute event time and recording the 2 Apneas (the only 2 of that session)

Conclusions...
There is no memory storage for Detailed of Hi Rez data during any session.
The SD card must remain in the S9 for recording Detailed and Hi Rez data.
Detailed and Hi Rez data writes directly to the card.
S9 writes data to the card as it is collected and there is no final write (at 12 noon) as a final write to summarize data for that date.
Summary statistics are written to memory in the S9 and is stored there unless cleared by some other method and recorded to the SD card when re-inserted.
There is NO standard memory that is used by Detailed or Hi-Rez data.
Unknown if S9 uses cache memory for writing Detailed or Hi Rez data.(not part of this test)
There is no overwrite of files for Detailed or Hi Rez data.
There may be an overwrite of files to correct Summary data for each session.

Move on to TEST #2

Power Down and Power Up S9 to remove any stored cache memory data.
Placed SD card ScanDisk Class 2 2 Gig freshly formatted in S9 at 12:15 PM this date... further testing to eliminate variables with multiple mask events.

Nord

alterego61 wrote:As I said above, I think the EVE (event) file is a more likely candidate for data problems because it appears that data is written to it as and when events happen.

Since most of my problem is RERAs, I have few events with 1 my highest S9 AHI. I still see skew. I am presently working on a 'interesting' memory issue at work so outside of work I am boycotting detailed analysis of anything. I would be happy to look for patterns in skew and events in the data in ResScan if someone wants to tell me what to look for. I don't archive the edf files.

Hey, all!

I'm expecting my new S9 to arrive any day now.

I was thinking that starting with a virgin machine, are there some tests that would make sense to do that could help? Or, assuming I simply start using it, is there a protocol I could follow that could provide you all with another possibly useful testcase?

I unfortunately don't have a lot of time to spend experimenting, but I'd be happy to at least try to provide some additional data points ...

ken

kennethryan wrote:Hey, all!

I'm expecting my new S9 to arrive any day now.

I was thinking that starting with a virgin machine, are there some tests that would make sense to do that could help? Or, assuming I simply start using it, is there a protocol I could follow that could provide you all with another possibly useful testcase?

I unfortunately don't have a lot of time to spend experimenting, but I'd be happy to at least try to provide some additional data points ...

ken

Thanks Kenneth...

As soon as you get your machine and look it over... can you come back to this thread and we can go over some stuff.
We should be able to solve this issue with all the good help we got now.

Thanks
Nord

fiberfan wrote:

alterego61 wrote:As I said above, I think the EVE (event) file is a more likely candidate for data problems because it appears that data is written to it as and when events happen.

Since most of my problem is RERAs, I have few events with 1 my highest S9 AHI. I still see skew. I am presently working on a 'interesting' memory issue at work so outside of work I am boycotting detailed analysis of anything. I would be happy to look for patterns in skew and events in the data in ResScan if someone wants to tell me what to look for. I don't archive the edf files.

Hi fiberfan...

Most of us have been looking at skew in ResScan by looking at the Flow graphs in the 10 or 30 second range for when Flow begins and ends and when the others begin and end. Usually off a clean SD card the difference is 2 seconds and when the skew happens... it starts at 3 sec, then 5, 7 etc.

Just have a look through the threads for Skew and you'll find the archival evidence.

How long have you had the S9 ???
How many mask sleeps do you have during each session (24 hrs)
How often do you DL to ResScan
Have you been following any set pattern
When did the Skew start for you and if its still there... what is the skew that you have

Since you have few AI's it would be a good comparison for the amount of skew that is happening versus an S9 with many AI's...

Thanks
Nord

Nord wrote: Thanks Kenneth...

As soon as you get your machine and look it over... can you come back to this thread and we can go over some stuff.
We should be able to solve this issue with all the good help we got now.

OK will do.

I have to say my event log and recording for the next few weeks is likely to be ... interesting. I am getting hammered by
allergies, and I have a very powerful cough. Last night I crashed my S8 at least three times I'm aware of...

Nord wrote:Just have a look through the threads for Skew and you'll find the archival evidence.

I have been following and occasionally contributing to the skew threads.

Nord wrote: How long have you had the S9 ???
Since March 17

Nord wrote: How many mask sleeps do you have during each session (24 hrs)

It varies. On a bad RLS night I might have 4 or 5, some nights I have 1.

Nord wrote: How often do you DL to ResScan

every day

Nord wrote: Have you been following any set pattern

The last week I have unplugged the machine when I get up - more to save power than the possible impact on the skew problem. The first 2 weeks I left the machine on all the time and turn it off with the start/stop button for a few days before unplugging each day.

Nord wrote: When did the Skew start for you and if its still there... what is the skew that you have

March
18 - 3 second
19 - 5 second
20 - 8 second
22 - 13 second
23 - 13 second
24 - 29 second
27 - 5 second
28 - 54 second
29 - 70 second
30 - no detail data - not sure what happened here - I expect it was operator error
31 - 3 second

April
7 - 3 second
10 - 3 second
11 - 5 second
12 - 7 second
13 - 26 second

I think I formatted the card on April 2, 4 and 5.

fiberfan wrote:

March
18 - 3 second
19 - 5 second
20 - 8 second
22 - 13 second
23 - 13 second
24 - 29 second
27 - 5 second
28 - 54 second
29 - 70 second
30 - no detail data - not sure what happened here - I expect it was operator error
31 - 3 second

April
7 - 3 second
10 - 3 second
11 - 5 second
12 - 7 second
13 - 26 second

I think I formatted the card on April 2, 4 and 5.[/quote]

Hi FF

Did you keep copies of these SD cards ??? Did you copy the contents into a folder ???

Thanks
Nord