AI cells sucks. Newer buy them again!

Drmike

fully agree - but I think it goes more towards gross negligence than simply appalling

Cells are the most important part of our units yet they are the most unreliable.


George was right.

__________________
Get a girlfriend you sad twat - a Rebreather is an unfaithful mistress - dont blind yourself to her faults just because she goes down on you

jeff h

I've had similar problems with AI. S/n 70586708 kicked my butt last night while trying to calibrate an Accublend unit. I really appreciate Paul posting those batch numbers, I wish AI would have. The sensor would go from between 9.9 -10.3 mV in air to 0.0 or somewhere in between when any torsion was applied to the cable (R17D type cell). There have been other problems with other sensors too. After checking about 30 of these it makes me not want to dive my rebreather.

So after all of those problems I started looking at other cell designs. For rebreathers I really have to ask the following questions:

1) Why do we use, actually design units for 10 mV cells with output tolerances of +/- 25% (2.5 mV)? There are 25 mV cells (R10Ds) out there that are +/- 2 mV or 8%. It seems to me the 25 mV cell would be the way to go. Also, if we know what the lower tolerance is then we also know what the lower calibration slope is so the unit should be able to kick out any bad cell before the dive (of course there are other problems too, but this would help).

2) I've concluded mono connecters are garbage. The molex connecters may be better but still have the connection at the cell which is rigid mount. I'm guessing that wire leads which are soldered to the cell board are a better design because it eliminates any torsion that would be applied to the snap type connector.

3) Have there been this many quality issues with other style sensors? Do the Mk15/16 divers find the same problems with R10D's? I suppose the R22DHO has similar specs (25 mV +/- 2 mv)

Looking forward to any replies on the subject.

-Jeff

jradomski

the usually differnece (when size is the same and its desigend for the same application) is just the internal reference resistor.. using a lower "voltage" allows the designer to use a higher gain when using op amps.. the higher the gain the more stable op amps are in general.. You need to design it so that the multiplied output fits within a certain range.. most low voltage a/ds use a reference voltage between 1.5 and 2.5 volts.. if you design your system to be linear to say a 3.0 po2, that means you have .83v (asumoing a 2.5v ref) per 1.0po2 available, using a modest gain of 16, would put a po2 of 1.0 at around 51 mv.. this cell would be about 10.8 mv in air... to use a cell with say 20mv nominal you would have to cut the gain by half which in general reduces op amp stability and increases gain variation over the i/o range.. if they had a 1mv or even .1mv, rebreather cell I would jump on it for a design!.. You could then use instrumentation op amps with gains of 500 or even 1000.. now the differnce bewteen 500 and 50x isnt that significant as opposed to sat 14 or 15 and a nominal of 16...


the hardest stabaility point for an op amp as a buffer would be unity gain (a gain of 1).. You can get opamps that are stable at lower gains at the cost of power consumption..

he mk15 needed higher voltage becuase the were basically an analog controller with a fairly high source voltage.. back then the precision of op amps was orders of magnitude less so you tried to cover it usng large voltage swings..

__________________
Joe Radomski
CCR Trimix Instructor Trainer
ANDI Instructor Trainer Director #10

All posts are personal opinions and DO NOT reflect any affiliated agency unless specifically stated.

maninthesea

Glad to see this post.
I have put 2 AI cells in my trimix analizer in 2 months & was starting to think the analizer was killing the cells. Now I know better.

jeff h

Joe,

Thanks for the response. Without a background in electronics I assumed that part would be easy (it's just a voltmeter right). Without much effort I did find a low power consumption 12-bit-adc that can measure up to 5v analog input (has a 4.096v internal reference?). Here's the link http://datasheets.maxim-ic.com/en/ds/MAX187-MAX189.pdf. I'm sure my electronics guy would argue that this could be done with a 10-bit circuit, etc, etc.

So if the hardware/electronics weren't the limiting factor would a 25 mV sensor that was +/- 2mV be more reliable/robust than the 10 mV sensors?

-Jeff

jradomski

for low power systems, the idea is to use the lowest voltage that you can to save power.. There are plenty of of amps that can read 5v or even 10 volts.. the problem is that these are not usually designed with saving the most power and, then would require generally require extra circuitry to interface to lower voltage cpus.. The most power efficient designs would use a/d built into the cpus so that they can be using no power while idle...

also if you look at the data sheet the max the analog input can be is the voltage reference.. so if you use the 4.0 v rref then thats the limit... also take a look at the power numbers.. the voltage reference alone takes 4ma at 5v... figure another 2ma for the chip itself.. this now gives you 1 channel... now you need to either set up an analog multplexing system or in the case of devices like this seperate a/d converters.. so now you in the ranhe of 18ma just to sample the o2 sensors... not exactly low power! These numbers are pretty typical...

You can design a sysetem for any range.. for our needs we want the lowest power possible with the best accuracy...

__________________
Joe Radomski
CCR Trimix Instructor Trainer
ANDI Instructor Trainer Director #10

All posts are personal opinions and DO NOT reflect any affiliated agency unless specifically stated.

WD8CDH

Op-amp stability is not a problem at all with either high or low output cells. Most op-amps are unity gain stable and those that aren't are usually stable by a gain of 3.

Interfacing to a 10 mV cell is trivial and to a 25 mV cell is even easier.

The main limiting factor for current consumption is not the electronics rather it is the solenoid, the back light for the display and the HUD LEDs. It's easy to design the electronics to draw less than 5% of the current needed for the solenoid, the back light for the display and the HUD LEDs.

I think the use of 10 mV cells has something to do with the cells themself, not anything with electronics. Possably cost and availability were the origional deciding factors. It could be just that the cells were "in the junk box" of the designer in the early days of eCCR developement. And then everybody else just used similar cells.

__________________
Ron Schroeder
Former Commercial Diver, now Lazy Diver

cool_hardware52

I can't comment specifically on the original choice of cells, but I do agree that simple proximity can have a huge, and often unanticipated effect on design, or the evolution of the design.

Often these almost "happenstance" design decisions live on and become a "standard".

These defacto standards commonly become the topics of long wordy debates, with defenders and detractors both providing long tortured justifications to support their respective positions, when in reality greatest factor was that the designer "already had a couple."

It is useful to occasionally review one's design criteria to see if any of the assumptions the design is based upon are really "cast in stone"

Tobin

__________________
www.deepseasupply.com

WD8CDH

If I were to design rebreather electronics again, I would design around the most RELIABLE and STABLE cell that I could find. I wouldn't worry about either nominal output or uncalibrated accuracy. The electronics could be easily designed to handle either high or low output cells.

My "day job" is designing instrumentation. One of the things that I do use is O2 sensors. Of the last 600 that I have used, only 5 have failed and that was a drift to zero volts output in a 5 to 7 day period after at least 5 months use. Unfortunatly, these cells (GS Yuasa KE-25) won't handle pressure over 22 psi so they can't be used reliably in rebreathers but they are great for measuring breathing mixes.

__________________
Ron Schroeder
Former Commercial Diver, now Lazy Diver

jradomski

most of the really stable (and repeatable) cells have 2 properties

a) slow response time full scale (relative to what we want in a Rebreather)
B) are not deigned for hyperbaric applications..

The cells from cititech are a perfect example.. they last a long time and the voltage output is very consistent over the reasonable periods of time..

__________________
Joe Radomski
CCR Trimix Instructor Trainer
ANDI Instructor Trainer Director #10

All posts are personal opinions and DO NOT reflect any affiliated agency unless specifically stated.