Michael Hanney, creator of the incredible Zwiftalizer tool, has released a site update with improved features which make it even easier to analyze your Zwift log files, research hardware performance and (new!) look at power source popularity.
Power Sources
This new report shows which brands of smart trainers and power meters are most popular in each country.
Hopefully this report will be as widely used as the Benchmarks report (which is currently based on 13,700 log samples)–but in order for that to happen, we need the Zwift community to upload their logs to Zwiftalizer so we have a large sample size. As Michael says:
Unfortunately I did not store any smart trainer or power meter make or model information before the 1.1 update, so the survey needs community participation to make the report worth while.
To add your power source data to the Zwiftalizer database, simply upload your logfiles like you’ve done in the past. Just make sure you keep the “Upload my system specs” box checked. Go to the ‘ANT+ Devices’ tab to see your smart trainer and power meter device name. All the major brands should be recognized, but if your device is not recognized a pop-up will ask you for its name.
ANT+ Section
The reformatted ANT+ charts make it easier to spot drop outs. AN easy to read description of your overall signal quality is also included, along with tips to improve (if necessary).
Sample with a drop out in the Kickr, and two basic sensors (HR and Cadence), but importantly – not the Power2Max power meter (top, green).
A drop out in the basic sensor (HR), but not the signal from the Favero BePro power meter pedals – it went low, but did not drop completely.
BTLE Section
The improved Bluetooth section now recognizes Wahoo and Tacx specific signal types including Kickr Snap. Michael says:
There isn’t much info in the log for BTLE devices so I don’t have much to go on. The plot shows the general pattern of messages logged but that’s about it. I… hope to do more with BTLE in future if I can include all brands of smart training, including Elite, CycleOps and Kinetic.
Benchmarks Section
Zwiftalizer’s Benchmarks section is my favorite feature of the site, because it shows how different hardware (CPU, graphics cards, etc) performs on Zwift. Michael has enhanced this feature with the following upgrades:
- Benchmarks can now be filtered by platform, resolution, profile, CPU brand, GPU brand and min FPS.
- For Mac and iOS fans: added model and release year for most iMac, Macbook Air, Macbook Pro, iPhone and iPad models.
- For PC builders: added links to search for CPU and GPU deals from Pricewatch, Tweakers, Newegg, eBay and Amazon. (Note: the links to Amazon are unashamedly affiliate links. So if you’re shopping for a computer or graphics card, you can support Zwiftalizer just by using the links!)
Benchmark Filter Examples
Want to know which Macs run smoothly at 1080 resolution with high quality effects?
Set filters to Platform = Mac, Resolution = 1080 (FHD), Profile = High, CPU = All, GPU = All, Min FPS >= 30
Want to know which iOS devices run 1080?
Set filters to Platform = iOS, Resolution = 1080 (FHD), Profile = All, CPU = All, GPU = All, Min FPS = All
Want to know which GPUs run the full Ultra resolution and effects at a minimum of 60 FPS?
Set filters to Platform = All, Resolution = 2160 (4K), Profile = Ultra, CPU = All, GPU = All, Min FPS >= 60
Want to find deals on that 4K capable GPU?
Click any of the Pricewatch, Tweakers, Newegg, eBay or Amazon links beneath the system name.
Other Stuff
This update also includes some small bug fixes, minor performance improvements, and a little bonus surprise flare if you agree to share (check ‘upload my stats’ for your reward).
Great work as always, Michael. Ride on!
The large gaps (always seem to be 90 seconds?) in the new ANT+ reports don’t seem to correlate to anything. The “ANT+ device searches” correlate to “ANT: Goto Search” in the Zwift logs but the logs also indicate that data is still being received on all channels even while searching so a device search doesn’t necessarily mean a drop out. These gaps in data don’t appear in the Zwift logs, the .fit files, or Strava so I think it’s a Zwiftalizer bug.
I also noticed that the graph for the trainer says “4-7 is normal” but mine only shows up to 4 like all other sensors.
My guess is you can only trust the Zwiftalizer ANT+ report if you have the same sensors as the developer.
Kyle, those large gaps are something I can fix. Sounds like I’m filtering too much data out now. I test with logs from about 30 users but there’s always going to be a device I haven’t seen before. If you want to send a log it would really help. Send me a DM @zwiftalizer for my email address.
Yes, not all ‘goto search’ entries are drop outs – (which I why I relabeled that chart) but all real drop outs do produce ‘go to search’ entries. This can be tested by removing the battery from a paired device mid flight.
It’s all a bit tricky really trying to produce a report of what happened when the log is nothing more than a debug output of what didn’t happen. Take the RxFail messages for lets say a power meter. I assume the software expects to receive a signal up to 8 times a second. Lets say it only receives 1 good one so 7 RxFails get logged – still a good second (but my aggressive filtering might make that second look like a bad one now). The other tricky part is if the device really does disconnect for a while and the software doesn’t log any RxFails at all. 8 minus nothing is 8. Good second or bad? There is no way to know. So I now look to see if a ‘goto search’ happened within the same timeframe and zero out the bar if one does. I am completely willing to accept that that might also be too aggressive. I’m open to suggestions for how else to solve that one.
So yes there is some guess work, but a lot of people still find it useful for trouble shooting their ANT+ signal strength.
For what it’s worth my devices are: PowerTap Pro+ hub, Garmin GSC 10 cadence sensor and Garmin HRM1B/G heart rate monitor chest strap.
Thanks for the reply, Mike. The ANT+ graphs interest me because I had given up on ANT+ a year ago and stuck to Bluetooth but I’m thinking of buying the Powertap P1s which only work “properly” with ANT+. Thus I went back and gave ANT+ another ago and seem to have completely fixed my drop outs but if I were to believe Zwiftalizer 1.1 things are worse than ever.
I’ve been thinking of posting my experience as a comment on Shane Miller’s ANT+ video to address some misinformation in the community like only WiFi channel 10 interfering (with 20MHz width it’s channels 8-12, and with 40MHz it’s 6-14) but I’m still working on some hunches (i.e. the dongle being too close to the trainer is also bad) and need to research more about ANT+’s frequency agility. My goal is to get all channels 2Hz (all my devices look like 4Hz in the logs) a drop out (i.e. go to search) is considered two seconds of no data but with frequency agility I think it means to try the next frequency which isn’t necessarily a bad thing (for comparison, Bluetooth hops frequencies 800 times per second). Frequency agility isn’t handled automatically like with Bluetooth so it has to be implemented in Zwift but since the Zwift logs specifically reference “channels” and “radio frequency set” I think they’re using it (this would be why sometimes simply restarting Zwift fixes things – Zwift sets different frequencies on startup).
Also worth pointing out that if frequency agility is indeed in use, all WiFi channels can interfere with ANT+ depending on what frequencies the Zwift developers have chosen to use.
I have a Tacx Bushido Smart and notice you do have an example log from one in your repo. My measurement for ANT+ signal quality is “Rx fails per channel per hour” and for me anything >4000 means drop outs but the test logs you have show things like >8000, >15000, etc per hour. Do these people find Zwift unusable?
To add to this, when using frequency agility it looks like you can only set three frequencies. This must be what is happening when Zwift logs “radio frequency set”. It’s not like Bluetooth where it adapts automatically.
So a real drop out must be three “Goto Search” within six seconds or 24 (at 4Hz) Rx fails on the same channel in six seconds.
It looks like a chunk of my comment above was removed. I think I used too many greater/less thans and it interpreted it as HTML.
I think it should have read:
…My goal is to get all channels less than 2000 Rx fails per hour and I already have my HR and cadence down to 600-1200 and am just working on getting my trainer down from 2000-3000.
I’ve only skimmed the ANT+ spec but it seems like for connections greater than 2Hz (all my devices look like 4Hz in the logs) a drop out (i.e. go to search)…
Thank you for taking the time to write that up Kyle. Fantastic insight & valuable feedback. I’d like to use your findings to change how I’m filtering if you don’t mind. I had not considered the Rx fails per hour value before. That’s a good one for giving an overall measure of quality. I like that a lot. I will take a closer look at my samples logs again. I know who each one belongs to and even though it’s been a while, I do remember some people having quite bad issues.
In my personal experience, with a permanent set-up and everything kept the same except the power source – the PowerTap Pro+ is by far the most stable signal. 4iiii Precision is also very good, but the kickr 2 was very average by comparison, perhaps because it was transmitting power and receiving resistance change message on the same channel.
I will need to amend my ANT+ tips. (I only mention channel 10 as one to avoid). Maybe just recommending channels 1 to 5 would be good enough.
The problem with the WiFi tips is that we don’t know what three frequencies Zwift uses for ANT+’s frequency agility feature. If they use frequencies like 2457MHz (channel 10), 2458MHz, 2459MHz, etc then avoiding WiFi channels that overlap is possible but instead we really should tell Zwift to use better frequencies.
If on the other hand they use frequencies like 2412MHz (channel 1), 2437MHz (channel 6), 2462MHz (channel 11) then changing the channel on your router is probably pointless.
It also worries me that the logs say “frequency set” and not “frequencies set” because they might not even be fully using the agility feature. We need more info from Zwift.
I need to read up on frequency agility. I thought ANT+ only worked on 2457MHz. I’ve just pushed an update by the way where each bar is the 2 second average of assumed sample rate (4HZ or 8HZ) minus RxFails. It takes longer to load, but should be a better representation. Please let me know if things look better for you. The way I am assuming the sample rate has a lot of room for improvement too.
Im still learning about it but found this quote: “ANT devices may use any RF frequency from 2400MHz to 2524MHz, with the exception of 2457MHz, which is reserved for ANT+ devices.”
It turns out that ANT+ is to ANT what GATT is to Bluetooth.
So it seems like frequency agility isn’t available to ANT+ but then why do the Zwift logs even mention setting the frequency of a channel if it can’t be changed? And for that matter why even call them channels?
Perhaps 2457MHz is just some master channel for finding ANT+ devices and then you can set your own three frequencies to communicate on?
Mike, I just tried out the updated Zwiftalizer and things seem to match up much better: https://drive.google.com/file/d/0BzTQ-TolcOhSa05fVnFtSkpLQTg/view?usp=sharing
At the start the Zwift logs say what channel a device ID is on and then using the knowledge that two seconds of no data on a channel is a drop out I seem to be able to reliably match “goto search” to dips in the graphs.
The only exception is with my trainer where there’s no obviously worse dips on the graph. My trainer appears to be 4Hz but is that graph still based on 8Hz? Perhaps that’s why I can’t see dips?
Some other things:
1) I’m pretty sure that ANT+ does only use 2457MHz, so avoiding channels 8-12 or 6-14 would be a valid recommendation.
2) Two seconds of no data (or 8 Rx fails at 4Hz on the same channel in two seconds) is a drop but because the Zwift logs don’t contain microseconds what you’ll sometimes see is 8 fails spread out over 3 seconds. I think you just have to assume that the channel with the most fails in the three seconds prior to “goto search” is the one with a drop out.
3) The effect of a “goto search” probably depends on the ANT+ dongle. I have a Garmin, a Suunto, and a Tacx dongle and the Garmin connects to devices <1s but the other two can often take 5-20s. Perhaps the reason why I'm not seeing total drop outs after "goto search" is simply because the Garmin reconnects so quickly?
Hi Kyle, that’s good to hear. Regarding 4HZ or 8HZ – the way I guess the sample rate for a channel needs improving. I basically just guess based on the max number of RxFails seen in any 2 second period. It’s a bit goofy because it assumes a lot of RxFails will inevitably happen – which may not be true – a really good signal will have a low max RxFails in a 2 sec bucket, which is probably why it thinks your trainer is being sampled at 4HZ not 8. I could just always use 8, but then the basic devices (cadence and hr) would look wrong. Or, I could use the device manufacturer ID and model ID to tell if the device is a smart trainer or power meter (8HZ). When I get that right I will be able to add in the overall RxFails as a percentage of expected signals over the duration of the activity. It’s going to get better. Your feedback has really helped. Feedback from the community has been great actually. I never mind being told where I’m off course. I only made this because I wanted a side project to use to learn reactjs and couldn’t bear the thought of making yet another todo list app 😉
on 2) I too noticed that fails can be spread unevenly over the 2 or 3 second windows. That’s why I was taking averages over n seconds, but clearly when I make n bigger it gets less useful.
on 3) I don’t know what is going on there. I have a Suunto movestick mini. My devices usually connect very quickly but my HR strap can sometimes take a while.
Is there something other than “goto search” that you count as a search? My latest ride only has one search in the logs but shows two on Zwiftalizer.
I’m doing the same to work out Hz, i.e. just assuming that the max Rx fails per second must be the Hz. 4Hz is definitely accurate for heart, speed, and cadence (a speed + cadence sensor is 8 Hz) but I can’t find any specific info about FE-C (resistance) or power. In fact this link states that power should only be 1-2Hz: https://www.thisisant.com/developer/ant-plus/device-profiles#521_tab
I’ve signed up for an ANT+ Adopter account and once they authorise me I should be able to download the ANT+ profiles PDFs which I hope will contain better info. I’ll post over on Github if I find anything useful.
The Suunto Moviestick Mini was the first ANT+ dongle I tried and I would highly recommend trying out the mini Garmin ANT+ dongle as it finds and connects to sensors much quicker. The Garmin dongle is a rebranded version of this OEM Dynastream dongle https://www.thisisant.com/developer/components/antusb-m whereas the Suunto seems to be a Chinese generic with no way to identify the true manufacturer.
To use a log from an iPhone or iPad, tap on the Zwift logo on the login screen 3 times. Select a log from the list on the left. Tap OK. Add to iCloud Drive or Dropbox. Then open zwiftalizer.com in Mobile Safari or Chrome. Tap where it says ‘Drop Log.txt or browse’ and select the log you saved on iCloud or Dropbox.