Insulin Pump clocks

Now that DST has come again (sigh), I have a question about the digital clocks in your insulin pumps…

How accurate is your clock?

I have a MM 630g pump. I set it against my cell phone, as that’s generally the most accurate time piece anyone usually has access to. About once a quarter, I’ll adjust it, because it’s off by 3 to 5 minutes. I never really thought about this before… I just made the correction every time we did DST and usually 1 or 2 more times kind of at random throughout the year.

Is that a weird amount of clock drift? Should we expect pumps to have clocks that drift this much? (My previous MM was a 5xx series, and it was no better at keeping accurate time.)

This is not something that worries me or that I think is a technical problem. That kind of drift shouldn’t have any real impact on my insulin therapy. Or at least, not any amount that isn’t lost in the general “noise” of imperfect biological systems.

This is more just curiosity.

I think our Tandem was off by about 4 minutes.

I would consider this normal. [modern] Computers are [mostly] correct on the time only because they periodic sync in the background from reliable timesources.

I don’t bother with corrections of a handful of minutes on our various devices and just figure I will get them twice a year during the time change.

Ditto. OmniPod, just a few minutes off every clock change cycle. Not enough to matter. I only fix it when it is clock change time or for big races.

1 Like

That’s a very interesting question, CaM2. I stumbled upon some info on that recently, although I wasn’t looking for it. If you are interested in the source, I will find it and post.

So, my Dexcom continuous monitor does not have a’real time operating system,’ meaning it is not built to correspond with hard time deadlines in real life and you might expect some drift. My sensor might have a 15 min lag when it outputs my BG to me, just from the reality of how the hardware and my body works, so that might not be super surprising. A 15 min lag on a sensor (in general) is kinda big.

I expect that pumps have a ‘harder’ real time operating system that is written specifically for the device. However, because of the proprietary nature of those devices, I might not stumble upon any confirmation of this on internet documentation. I am totally not informed on if anyone has fully explored your pump on the internet. However, previous models have been pretty fully explored. If anyone finds documentation of this, please, please, please send it my way.

I found info somewhere here:
Note: These are people that you might want to talk to if you have complaints about the accuracy of your sensor. They are currently working on verifying the accuracy of manual BG machines, but are probably working their way up to sensors. Interesting stuff.

1 Like

Glad you asked this because I had the same thought when I was doing mine. It was off by ~4-5 minutes, which surprised me b/c I’d never noticed before. In the digital age I’ve gotten used to assuming electronic devices are all on network time, so it surprises me to find my pump is drifting that much, but of course it isn’t networked. Still, that much discrepancy seems like a lot even for a non-connected device.

I hadn’t even thought about it, but electronics can have clock drift from environmental factors as well as design weaknesses…

Again, I don’t expect pump manufacturers to invest in super-accurate clocks inside a pump. Correcting the time 2 - 4 times a year doesn’t bother me. It’s just interesting to observe.

From anecdotal observation / memory, the difference is more than PCs used to drift, back before Windows would self-correct via the internet. I used to use various applications to sync servers to internet time servers, but Windows now basically does all that for you. The drift would be a minute or two a year.

So PC clocks then seemed to be more accurate than pump clocks today. But those clocks had a more stable power supply – main power vs. batteries that begin to drain down before they’re replaced. And they aren’t subjected to the temperature changes that come with going from indoors to outside during winter and summer. Plus the microcontroller / onboard clock in a pump is constrained by space in ways that a PC’s clock doesn’t have to be.

So that all kind of makes sense.

That seems to be entirely in line with my expectations. Electronic devices as a rule are not great at keeping time.

The most accurate time piece most people have access to would be their GPS device.

Mobile phones (as @CaM2 mentioned) are close enough for anything of practical purposes and more convenient to walk around the house with. However, take the phone off the mobile network and your time accuracy goes out the window. Remove your GPS device from satellite communication and that time accuracy goes out the window.

I think they are all about the same. There would always be a small battery inside the computer to keep the clock running. Sometimes you can see it (button battery) and sometimes it is buried somewhere. Consumer electronic devices just do not have the need for high time precision particularly when anything that can be networked can easily be running a background/scheduled time correct process. Most times, the time correction is designed to alter the clock by less than 1 second at a time so you would never notice the time jump.

I fully expected them to have hard ‘real-time’ operating systems. I was shocked.

@Tim35, you don’t think the pump is more time centered than the sensor???
I was kinda imagining it as the most real-time system I own.

No. Really - what difference does it make if the pump or cgm system gains or loses a second per day?

1 Like

I always just sort of assumed the sensor itself has no clock at all. It just measures data on demand, and that the pump requests data from the sensor on some sort of timer. So the sensor has no “sense” of time at all. Just data.

But I have no idea if that’s how it works at all.

I don’t think it matters a bit to treatment.

I just raised the post from curiosity. And because I’m an IT guy who finds electronics and the software running on those electronics interesting.

Based on some things I read from “the internet” (lol), the Dexcom sensor has nothing. The Dexcom transmitter has a clock that gets reset to zero when the session starts (or maybe when the transmitter is used the first time?) It then sends data in terms of how many seconds offset from zero.

The actual time (as we are used to seeing time) is only on the Dexcom Receiver (or App or Pump or whatever). The Receiver gets how many seconds since session start (or transmitter start?) and computes the time as we would know it.

But I could totally have this wrong.

1 Like

My Animas pump was ‘off’ by ZERO minutes… And considering all the battery changes in between, I think that’s not bad – I do not recall having to change the clock on it except when we change to/from DST.

The time on the Dexcom receiver matters for reports. I think if it gets off by too much, the uploader complains about that…

A second a day doesn’t make much difference. But, I guess because of the urgency of deliver meds within a certain time frame. I mean, that stuff has to work. So, maybe that’s the ‘priority’ of the clock and operations. I would still call it a ‘real time’ system, I guess, because it has to respond to user commands in real time. I dunno. Definitions get fuzzy.

There is definitely an internal clock in the sensor receiver. My pump is a system on chip, which is pretty sophisticated, and might have multiple clocks. Uncertain. This is Dex data. The ‘Display’ Time is what the actual time is on your clock on the wall. The Meter InternalTIme is the system clock. All these clocks have to be synchronized when your pulling together the two datasets.


1 Like

What device is that data dump from? Your pump?

Nope, thats my Dex. It outputs as either .xml or .csv

By “Dex” - do you mean the Dexcom Receiver?

Yep. I’m still working on getting the pump data. That’s the ‘bad apple.’ As I think about this, when you calibrate your Dex, you are relying on an ability of the device to perform according to ‘real time’ events because the data ages and becomes less useful the longer it takes to calibrate. Calibration, at least, is a priority event that needs to happen as close to ‘real time’ as possible.