more accurate Sony GNSS/GPS Chip for Garmin sports wearables

Reading Time: 4 minutes
Image | Sony

new Sony GNSS Chipset

Thank you @Mario

Interesting times ahead for sports-based GNSS accuracy.

Not only is there a brand new Sony GNSS chipset from Aug 2020 that claims to be more accurate, but there are also iterated versions of the older one that might explain changes we’ve seen in GNSS accuracy over the last two years as well as which provide inbuilt support for UDR (like Suunto’s FusedTrack). Indeed I’ve had it confirmed directly from a manufacturer that different models of sports watches do use different versions of these chips and that may well explain SOME of the variations in performance we’ve seen from the Sony chips across all the brands.

I will update THIS post if more detailed info comes to light from suppliers who I’ve asked to comment but it’s unlikely they will meaningfully comment.

Jump To


The Sony GNSS chip is widely used in sports wearables, although not in Wear OS and probably not in Apple Watch. The reasons it’s used are primarily twofold 1) it uses tiny amounts of power, 6mW, which lets the battery work for longer and 2) It supports multiple constellations like GPS+Galileo which at least give the marketing illusion of greater accuracy if nothing else.

I had assumed that the original Sony CXD5603GF chip was the one in universal, widespread usage. It might be…but I’m guessing it’s been silently superseded. Here’s the list of existing Sony models which, for our purposes, have superficially similar performance (specs, here)

  • CXD5603GF (2018 model)
  • CXD5604GF (UDR model, untethered dead reckoning)
  • CXD5605GF
  • CXD5606GF (UDR)
  • CXD5607GF
  • CXD5608GF (UDR)

I think almost all of us had been assuming that there were algorithm changes that fixed various performance issues and, whilst that might be partly true, I suspect the better-selling sports wearables have been silently changing the chipsets too and those chipsets simply might have better performance characteristics inherent in the hardware…or not. So, IIRC, the first Garmin to use the Sony chipset was the Forerunner 945 and I would bet that the Forerunner 745 announced this week uses a different, later one.

You will note that some models use UDR which incorporates motion sensor feedback into the positioning calculations and the keener ones amongst you will wonder why SBAS is not mentioned in the specs of any of these older models and yet Polar claims to use it (SBAS – positional data from fixed ground stations)

The Future of GNSS in Sports Wearables 2021

What Sony has just released is a GNSS sports Wearable chipset that works with dual-frequency bands. This does NOT MEAN GPS plus GALILEO nor GPS plus GLONASS. Each of the GPS, Galileo and BeiDou constellations operate over two frequency bands and when the new band frequency (and/or both bands together?) are used then greater accuracy can be obtained.

The bands have ‘L’ numbers and my reading of the Sony specs is that the new Sony chipset only works with L1 and L5 bands which effectively means GPS. Galileo and GLONASS do have multiple bands but I don’t know for sure if they can be commercially used and/or if the L1/L5 frequencies can be worked with – initially I’m assuming not.

So one possible future for us in 2021 is that Glonass and Galileo will be temporarily ditched in favour of multi-band GPS (L1 and L5). Of course, it’s never that simple.

Here is a selection from specs of the new Sony chipsets

Power Consumption  

1.5 GHz/1.2 GHz simultaneous reception





1.5GHz reception 6mW 7mW
1.2GHz reception 7mW 8mW


I’ve highlighted in bold the 6mW power consumption as this is the same as the current Sony chips require. As you can see to increase accuracy requires 50% more power (to 9mW).

But don’t forget that the current scenario of GPS+GLONASS or GPS+GALILEO also requires more power than GPS-only reception, from experience let’s say 20% more power.

Multi-band chipsets are not a new revelation, I’ve talked here about multi-band reception on a few occasions but that tech has really only been used in some smartphones.

Take Out

These chips are going to be used (I have had that effectively confirmed). There will probably be a software switch to enable/disable the greater accuracy mode.

Whether or not the new chip will materially increase accuracy remains to be seen but I would be hopeful of some limited improvements. Remember that using multiple constellations like GPS+GALILEO merely increases the likelihood of the base level of accuracy being achieved…we are talking now of increasing accuracy NOT making inaccuracy more consistent.

Will multiple constellations (GPS+Galileo) work together in some combination with multiple bands (L1+L5)? Yes (See comments below, maybe not on the Sony)

The current crop of sports wearables from Garmin, Polar and Suunto have more than adequate battery lives for most usages. Being able to sacrifice some battery life for greater performance will CERTAINLY appeal to some athletes – me, for a start ;-).

Spanner in the works?

There ARE still more & DIFFERENT improvements in GNSS accuracy that can happen in the relatively near future, perhaps even a step change over and above what Sony might enable here.

Sorry, that was a relatively hard-going piece. If you want to read something more entertaining with a GPS theme, then try this

Joshua Cheptegei | Gutted About his STRAVA GPS Track | 5000m World Record



Support this site with purchases at these partners - should click to a local choice in your country

30 thoughts on “more accurate Sony GNSS/GPS Chip for Garmin sports wearables

  1. Caveat: I’ve spent a bit of time learning about GNSS, but by no means a professional.

    Dual-band (and tripple-band, whenever that arrives) has the ability to determine location more precisely than single-band, because using two bands allows the GNSS receiver to estimate the ionospheric delay better. Also, dual-band allows the GNSS receiver to better detect if a signal is reflected off, say, a building than coming directly from the GNSS bird (think Urban canyons). This effect is called multi-path errors.

    In the medium blog below, the error circle (Dilution of Precision) of the dual-band Xiaomi Mi8 is almost half of that of the Samsung S8 (2.75meters vs 2.92meters).

    To improve on that, you need to integrate data from IMUs (this is basically what the UDR versions do).

    Note that the simple fact that you’re moving reduces the precision. TANSTAAFL (There Ain’t No Such Things As A Free Lunch).

    Using multiple GNSS systems gets you more data, which helps with more Precise Point Positioning (PPP), and getting shorter convergence times.

    For further reading:

  2. Galileo does have freely available dual band (as well as a higher resolution commercial encrypted signal IMS), and the whole constellation has it. GPS as I recall has freely available dual band on only the newest satellites. The military band will still not be available freely. I think it’s likely that the first dual band improvements will be seen with Galileo.
    One other aspect is the chipping rate, the bit rate of the signal modulating the carrier. This is higher on the new bands, which among other things makes it easy to eliminate multipath and simpler to lock the signal timing with the precision needed for better location.

    Interesting to see the powers needed. I’d be happy with taking a battery life hit for better tracks in the mountains.

  3. “Will multiple constellations (GPS+Galileo) work together in some combination with multiple bands (L1+L5)? Is that possible or desirable or EVEN MORE power-hungry? IDK.”
    U-blox about two years ago announced some chipset that are multi-constellation and multi-frequency and some with use of PPP, not for wearables but for other sectors (in particular the cars).
    It’s possible to download the tech data where they declare power consumption and, yes, they are even more power hungry.
    We discussed a bit about power consumption and possible adoption in the running wearables market two years ago, but with time things get better because newer chipset with new technology have always less power consumption with the same performance.
    So I guess that we will have for sure in the near future multiconstellation multifrequency running watch, the only question is when?

      1. I was referring to this datasheet of the gps chip of UBLOX , but unfortunately I can’t find any more the consumption of that chipset (it is used for the car market). As far as I can remember, using other constellations doesn’t increase a lot the power consumption (for example the use of GPS+GALILEO+BEIDOU has about the same power consumption of using GPS+GALILEO). In the opposite, with the use of the second frequency power consumption is about double. If they use a third frequency, power consumption is about three times. With the new Sony chipset it seems that the increase of power consumption using two frequency is not the double, but only 50% more (for example from 6 mW to 9 mW). This is a good new and I hope to see this new chipset in the Fenix 7 or FR955, even if I don’t think I will sell my FR945 and upgrade to the new one.

        1. I found power consumption of chip of Ublox at pag. 14, I don’t know how many frequencies they are using:
          tracking GPS only 68 mA 3 V -> 204 mW
          tracking GPS + GLONASS + GALILEO +BDS (Beidou?) = about the same = 85 mA 3V ->255 mW.
          It’s about twenty times the power consumption of the Sony chipset!!!

          1. Sure, 20x the power, but it can do 25Hz position updates (GPS only, “drops” to 20Hz when using all four GNSS). The Sony does 1Hz position updates. Different chips for different purposes.

            Let’s say you go insanely fast down a steep hill on a bike, and you go 90kph. That’s 25 m/sec. Running, even 36kph is ‘just’ 10m/s. How bad does your track look with a 1Hz update? 95% of the time, pretty good.

            The highest recorded top speed of a MotoGP motorcycle is just shy of 360kph (through the speed trap at the end of the straight of Mugello). That’s 100m in a single second. If you map the track of that bike, and do it even remotely accurately, you need at least 10Hz updates, if not more. For laptimes and intermediate timepoints, you’ll use interpolation to get down to 1/100th of a second.

            Sidenote: increasing your update rate indiscriminately will expose the system short-term multipath errors, which will throw the path off, e.g. when you drive under an underpass/bridge. The quality of the antenna (consistent phase-delay from all angles) becomes increasingly important. I have a nice laptimer/datalogger, which unfortunately has a shitty antenna design, so it *has* to be pointed straight up all the time to get a half-decent signal.

  4. Was reading about some hardware tech that was supposedly helping GPS chips maintain accuracy in more difficult environments. Can’t for the life of me remember what it was called but they had direct comparisons vs Garmin and specifically mentioned there would be devices featuring their tech by EOY 2020.

  5. Could it be that the new chipset with L5 is already in the FR745? If you look at the battery life in gps mode compared to the FR245 there is a big drop. (24h in FR245 and 16h in FT745 even if it is bigger and heavier and that usually means better battery life

    1. I just paid attention to the dimensions
      3.2×3.7×0.5mm (5.92) vs 7.0×8.0×1.4mm (78.4)
      if my maths is correct that’s a 13x bigger volume…less room for the battery. more consumption from the chip. it’ll take a brave sportswatch company to use that chip

      1. You’re comparing the GF and the GG chip. For a lot of industries, the GG is perfectly fine and puts less strain on your manufacturing (ultra-fine-line PCBs and extremely precise mounting equipment). The XFBGA packaging has a 0.4-0.5mm solder ball pitch vs the 0.8mm for the LFBGA package.

        STATS ChipPAC has some fore info on this. Otherwise, googling “xfbga packaging” and “lfbga packaging” turns up a lot of good links.

  6. @Kai “Sure, 20x the power, but it can do 25Hz position updates (GPS only, “drops” to 20Hz when using all four GNSS). The Sony does 1Hz position updates. Different chips for different purposes.”
    Two years ago TFK published an article about Polar team sport pods that had a gps chip with 10 Hz update rate: it was designed for football and team sport where the player were not running steady, but were doing short and fast sprint. The pods should be mounted in the back of the shirt, if I remember correctly. We were speculating that maybe Polar would have introduce this chip also in sport watches, but we were wrong, because I never heard any more about this 10 Hz chipset (10 signal reception in one second instead of 1 in one second).
    I found the power consumption of the Broadcom chipset of two years ago, it was in tracking mode 5mA x 3V = 15 mW, about 1/3 more of the new Sony chipset. This January Broadcom relasesed a second version of his dual frequency chipset, but I wasn’t able to find power consumption.

  7. “Will multiple constellations (GPS+Galileo) work together in some combination with multiple bands (L1+L5)? Yes (See comments below, maybe not on the Sony)”
    In my opinion the new Sony chipset will be able to do not only dual frequency, but also multi-constellation (gps+galileo or gps+beidou). We don’t have any data, but probably power consumption will not increase very much from using only one constellation or two.
    In the current watches with single-frequency the setting gps+galileo seems to use not so much more power than gps only.
    It will be interesting to see if Garmin will use also the chinese Beidou constellation: it seems that China in few years launched a lot of satellites, and probably Beidou performs better than Galileo (I’m not following very much how is Galileo doing, but it seems to me that Europe was not so fast like China in implementing their gps system)

Leave a Reply

Your email address will not be published. Required fields are marked *