I’m totally mystified by the Dexcom approach to transmitters. Why don’t they just ship one with every box of sensors? Sure, it would minorly increase the cost of the sensor box strapping the transmitter to it, but then they would be able to junk all the expensive transmitter handling, separate boxes and all, that they have to use every three sensor boxes anyway!
1 Like
It used to be that the transmitter was massively expensive, so they certainly wouldn’t supply extras. And they need separately boxed transmitters anyway for when they replace a failed transmitter under warranty (not common, but it does happen.) Anyway the whole issue goes away in a year or so when the G7 comes out: it’s a disposable sensor with built-in battery and transmitter.
3 Likes
We are really looking forward to the G7, especially the reduced size. Should be a lovely bit of tech with not having the transmitter hassles. I am sure Dexcom’s supply chain is excited as well. Only 1 SKU!
4 Likes
I don’t know about the earlier transmitters but the G6 cost seems to be under $10; i.e. manufacturing an extra 10,000 transmitters costs $100,000 more.
Here are the pictures of the PCB (from the FCC web site):
So that looks like one module and, most likely, one SoC, possibly the module for the bluetooth (BLE) and the SoC for the processor and memory. It’s an 8G transmitter (according to the test report). Quite possibly someone has decoded the numbers on the right hand module/SoC but I couldn’t find anything on the web.
I’m guessing maybe $5 for the SoC, $1 for the BLE module plus a couple of dollars for the battery, passive components, incremental manufacturing costs, potting compound. I admit that is pretty expensive and possibly Dexcom weren’t confident enough about reliability of the transmitter to manufacture in large runs; there were a lot of transmitter code revs.
However, as you say, they’ve worked out how to get the cost down so that they can reduce costs using nine times as many transmitters rather than just three. They most likely don’t see any profit in disrupting the G6 supply chains, it was probably bad enough when they changed the sensor box!
I’m guessing the G7 will just have a single custom SoC and the main cost increase over the G6 will be having to put it into a sterile environment for final assembly.
5 Likes
And as a pharmacy benefit, all the headaches of DME and related billing simply vanish for the users and the prescribers.
3 Likes
While the material cost may only be $10 in parts, the machines that are doing the assembly aren’t cheap and aren’t going to make millions and millions of the transmitters so that cost needs to be amortized over a relatively small number of units (compared to what the machines could be making), there is manufacturing personnel and test personnel. I would guess each transmitter costs something on the order of $30 - $40 and then you have to spread the regulatory costs across the units. With that being said, your point is valid and they could overbuild if they wanted for a relatively small cost.
2 Likes
Yes; quite so. Yet the vast majority of people who read this don’t understand how very important amortization is, so:
And if you double, triple, the number of units it has to be amortized over twice, three times the number of units so; the leap of faith, it costs half, one third.
If I were to make a political statement I would make it on the internet. I would face enormous costs; my internet connection costs $100/month, my political statement would cost me $1200 in the first year.
The math, the simple arithmetic, is truly horrible. The reality is worse.
There is an old English saying; in for a penny, in for a pound. That is, perhaps, one of the best definitions of amortization that there is in english.
1 Like
The black chip SoC is a NordicSemi chip… 5.2 BLE @2 Mbps and supports NFC/2.4 Ghz protocols with a 64 MHz ARM® Cortex-M4F chip. It also appears (I could definitely be wrong) that the PCB was made by HyunWoo PCB. Wonder what it would take to actually hack it.
It also looks that Dexcom patented earlier this year several transmission protocols for either the G6 or upcoming G7 transmitters. There is a NFC version of the transmitter in the below patent.
2 Likes
Ah ha! I think it’s actually the older (2017) version - the NRF52832 which has 5.0 BLE but is otherwise identical to the NRF52805:
The old (52832) version costs $1.716/unit in lots of 35,000, $4 for single chips, the new 52805 is $1.04/unit in lots of 35,000 and $2.64 singly. The difference between the two, apart from the update to BlueTooth 5.2 is, well that the new unit has 37.5% of the memory of the old, which should make no difference to DexCom.
My guess is that the unmarked chip is a custom ASIC designed to handle the sensor itself and enhance the shelf life. A separate ASIC can ensure that the rest of the board doesn’t get switched on until the first time a sensor is connected; we know that Dexcom couldn’t guarantee a transmitter life over about 120 days but that they do guarantee a transmitter shelf life of 365 days, matching the sensor shelf life.
Most likely some of the more obvious pads connect to the 7 documented debug pins; indeed there are 7 pads in a row at the bottom of the picture 
I doubt Dexcom bothered to obscure or turn off the debug capability, why would they? If so then maybe the whole thing could be re-purposed, but why; there are lots of alternatives and even the Nordic version is widely available at low cost:
That one even has USB! Many of these things have ported, easily installable, versions of Python and quite a lot of RAM (unlike the Dexcom SoC).
Ha! Trying to complete with the Libre I think… The UK doesn’t seem to pay for CGMs but will pay for the Libre, not necessarily a good thing to kurdow to.
1 Like
I suspect you are spot on @jbowler. I was dredging through Dexcom patents registered in 2020 and was surprised how many were there. There is definitely a ramp up of something new, especially as they are patenting NFC comms.
2 Likes
Indeed; I only (briefly) looked at the intro to the one you posted a link to. I don’t ever read patents; we pay three times the damages if we read them. Kind of like cutting someone else’s trees down in Oregon; you only pay for the tree if you “didn’t know”, but you pay three times if you can be proved to a jury to have known this wasn’t your tree. Since trees, which take 100 years to produce where I live, are worth below one third of their value it’s a worthwhile risk, the same is not true of patents because…
These are noise patents. They’re intended to throw enough brown stuff against the wall that something will stick. I hate those; there is one with my name on it filed by my employer at the time and it is utterly contrary to my philosophy (note: not politics, philosophy, this is something very personal to me). They work by effecting an enhanced version of the copyright law (I am very much in favor of copyright); in effect Dexcom is saying, “You can’t copy this”, but with a much broader definition of “copying” than copyright law allows.
I kind of sympathize with the poor corporations, I certainly sympathize with my previous employer, who got royally royalled by another company with a lot of patents. The story is that said second company dumped a thousand patents on my employers desk then, when those had been worked through, a thousand more. Copyright law is worse of course. A simple thing cut into two is just a mess.
1 Like
As you pointed out @jbowler there is a multitude of “noise” which generally happens when companies surround primary patents with a defensible patent “moat”. This would account for the “noise” as the noise is purely a distraction.
Hopefully this “noise” is pertaining to G7 hardware.
1 Like
Crazy, here in the PNW you can make Christmas trees in 3 years if you pick the Doug Fir, but of course almost no one buys them here because the other species look much better but take 5-7 years to grow. The vaunted Doug Fir can grow to 30 feet and be harvested for lumber in 30 years. I am always amazed when we go up to the big mountain and they have trees that are only 4 feet high that are 250 years old because they are covered in snow most of the year.
5 Likes
They sound beautiful. We used to cut our own Christmas trees down in the forest as a child but I have to say, as an adult, I wouldn’t dare do it. Maybe if it was a Christmas tree farm but not a tree in the wild struggling to survive what life has to throw at it.
4 Likes
Yes.
2 Likes
It is a little easier to cut them down when just in Oregon there are approximately 40,000 square kilometers of forest. Add Washington and Northern California to the party = Lots of trees. Also Christmas trees are rarely over 7 years old otherwise they get too tall. It isn’t like you are cutting down a 100 foot tall one. Just the baby brother.
4 Likes
As long as it’s just the baby brother, okay, for Christmas sake! !
1 Like