Teardown of a Remstar DS550P

So I managed to get a hold of a Remstar DS550P (Remstar Auto C-Flex). Basically got it for free because it was contaminated with cigarette smoke and was making a high pitched motor noise.

I also opened up my AVAPS C-Series just for fun to take some photos and compare / contrast. Didn't go in too deep with the AVAPS as it is now my primary machine (though I do have one more as a backup that i'll probably sell)

Picture Album: http://imgur.com/a/vPnyz#1

For comparison of their design process change through the years, the Remstar is of Rev00 and was produced in 2010-04 and the AVAPS is of Rev01 and produced in 2013-07
Remstar PCB is Rev9 (Codename Q-series) and the AVAPS board is Rev4 (Codename 'Yoda')

They both utilize the same microcontoller: The 32-bit ARM core ST-micro produced STR911FA (http://pdf.datasheetarchive.com/indexer ... 243287.pdf)
- Crystal was 4Mhz, but the chip has it's own PLL to bring that up to whatever the CPU clock is.
- I may at some point trace the circuit out to determine what signal conditioning they are applying to reconstruct the phase for the sensorless BLDC driver.

Both use Sensirion SDP703 mass flow sensors. Earlier Respironics (such as BiPAP ST-D) used Honeywell AWM2100's or AWM9000 series mass flow sensors.

So some other cool stuff i noticed:
- The Blower has 5 wires coming out of it: U, V, W drive phases and 2 wires for a thermistor embedded in the stator for overtemperature protection (took me a while and lots of destruction to figure out what they went to)
- The above means that they are using a sensorless drive algorithm to detect the mechanical phase of the rotor (saves cost)
- The older BIPAP's used a small voice-coil actuated pressure relief valve to cycle between IPAP and EPAP. It's absense on the AVAPS machine suggests the control loops on the modern machines are fast enough and they have a quick enough decay mechanism that this is no longer necessary (obviously the part cost savings would be massive).
- The above also means that the two machines are very similar except for the PCB (AVAPS has a very loud alarm mounted in the case, not on the PCB though).
- The power stage drive waveforms are chopped at 16khz regardless of whether the pressure is set to 4cm, 20cm, output closed off or output open. Some more investigation is required because I would expect both frequency and magnitude to change.
- They are using 3 FDD8424H to build the 3-phase bridge. These are matched N and P channel power mosfets in one little TO package.
- They are using FAN3268 to drive the mosfet gates. Must of had lots of room in the BOM bottom line.
- Butt loads of via stitching on the outer edge of the PCB. Probably for EMI compliance. At least a 4 layer pcb with 2 ground planes.
- Couldn't find a JTAG / test point header on the Remstar, but there are a couple headers on the AVAPS. There's an unpopulated through hole footprint on the Remstar though.

The AVAPS and the Remstar are actually pretty similar when you get down to it. Most of the components on the PCB are the same, though some have been re-arranged. Blower and assembly seems to be the same. The major differences are the inclusion of a chassis mounted alarm on the AVAPS that is far louder than the PCB mount alarm on the Remstar. AVAPS has Backlit LCD, extra alarm button. The AVAPS also has a small EDL capacitor next to the RTCC coin cell to presumably act as an alarm backup power source (Lots of ventilators have separate internal alarm batteries, so they scream at you if power fails completely). They don't call it the System One platform for nothing.

Overall top notch construction on this thing. Mechanically put together very well, Designed for assembly and designed for manufacturing. Bodgery absent, as would be expected on a Rev9 board. Seems like no expenses were spared on the PCB. Though there was some flux residue in places.

So yeah this thing was utterly disgusting because of the buildup of cigarette smoke (Unit had 5300 hours on it). There's a picture of a paper towel I wiped a bit of the inside of the blower assembly with that shows just how much buildup of gunk there was on only maybe 4 or 5 square centimeters. Absolutely nasty. Smoking is bad for you kids, don't do it. Also it ruins perfectly good equipment. Also that stuff is in your lungs.

Probably forgot some stuff, i'll be looking in more detail at the drives and analog stuff later though. I know i could have gone into more detail in some areas. I'll keep an eye out for an S9 or (god permit) an S10 to rip apart sometime.

EDIT: Also forgot to mention the neat little humidity sensor, I think it's a Sensirion SHT11. It's positioned just above the mass flow sensor on the PCB (it's very small)

the 550 is an auto a-flex. when you run as straight cpap it has c-flex. as far as i know a-flex is for apap and c-flex is for cpap.

Very neat info, thanks.

I believe the backlight was added somewhere in the middle of the production for the 50 series PRS1 machines, along with some other changes, so some or all of the updates to the design may be from production date, not AVAPS vs. APAP.

The PRS1 used to have one set of hardware for the DS150 and DS250 (Plus), and another set of hardware for DS450 through 750. The difference between models was in the software and labeling. I don't know about AVAPS, ASV, or BiPAP ST models, or 60 series.

Interesting stuff, good work. This is only for whose who are very familiar with the insides of electronics:

Check these out these links for ResMed models, which are a lot harder to get back together:

viewtopic/t74477/How-to-open-Resmed-S9- ... epair.html

viewtopic/t96624/Resmed-S9-Elite-Teardo ... tures.html

There are more PR write-ups on teardowns, just type into your internet browser search engine, ie. Google, Duckduckgo etc. site:CPAPtalk.com PR teardown or site:CPAPtalk.com Phillips 550 repair... or something like that, you get the picture. For more ResMed try site:CPAPtalk.com ResMed teardown or site:CPAPtalk.com S9 repair into your internet browser search engine, ie. Google, Duckduckgo etc.

CPAPtalk's search feature isn't very useful for this sort of thing. As I recall Palerider likes to have some fun poking around inside of these beasts.

Hey aytikvjo,
Is you circuit board still functional? Was you getting error codes that would make the machine shutdown while in use? You should try to get the microcontroller to give access to its memory so you can download the machine code of the microcontroller.

Has anybody ever downloaded the memory of the microconroller and looked at the machine code? There is a port on the other side of the circuit board where SD card is held. There is a device that is used to access that port. You can see it in this link:
viewtopic/t101692/viewtopic.php?f=1&t=9 ... 42#p872812

This device is used to program and repair the machine from a computer/laptop using software from phillips.

Sir NoddinOff wrote:Interesting stuff, good work. This is only for whose who are very familiar with the insides of electronics:

Check these out these links for ResMed models, which are a lot harder to get back together:

pfft, the resmeds are a *little* trickier to get apart, but once you know where the clips inside are, and the access slots to release them, they're easy to get apart without any disfigurement, and they go back together very easily.

Sir NoddinOff wrote: As I recall Palerider likes to have some fun poking around inside of these beasts.

true. takes me about 4 minutes to reduce a s9 to it's component parts, and about 3 minutes to put it back together a long handled t10 and a thin flat blade screwdriver (for the inner clips) are all that's needed. aytikvjo has seen those pics, though.

borgready wrote:Hey aytikvjo,
Is you circuit board still functional? Was you getting error codes that would make the machine shutdown while in use? You should try to get the microcontroller to give access to its memory so you can download the machine code of the microcontroller.

Has anybody ever downloaded the memory of the microconroller and looked at the machine code? There is a port on the other side of the circuit board where SD card is held. There is a device that is used to access that port.

This device is used to program and repair the machine from a computer/laptop using software from phillips.

Board is still fully functional. No error codes or fault operating conditions that I know of.

As far as dumping the memory of the MCU, most times the engineers will implement security to protect the flash of the chip and or disable debugging interfaces. Some more investigation would be required. I'll have to trace out some of the possible connections. I have some other STM32 projects first.

As far as the serial interface goes, this is normally used for the modem or the service center link. It's going to be managed by the software running on the machine, and thus will have only a few permissible operations. We know that it allows you to do things like modify the RTCC, update firmware, and communicates with the modem, but something like a firmware dump would probably not be in it's operational parameters. To do that you'd access the MCU itself through it's normal debugging interface.

If i had something (like the modem) that communicates over the serial bus, I could stick my logic analyzer on it and try to figure out what it's doing, but i don't. Alternatively I could guess and just start sending crap down the pipe and see what comes back up