Virtual Shifting Dropouts - Especially on Tacx NEO 2T - Likely Cause

When the Zwift Ride loses its BLE connection, Zwift immediately disables virtual shifting, even though the Tacx NEO 2T remains connected and ready. When Zwift reconnects to the controller, it reinitialises the controllers and enables virtual shifting again. This window is where the dropouts occur.

A fix sounds simple enough: preserve the virtual-shifting state during a temporary controller disconnect. But we don’t have insight into the Zwift code.

I’ve spent quite a while chasing these disconnects, but I can’t see an obvious commonality in the logs as to what’s actually kicking them off. What I can see very clearly is what Zwift does once a disconnect occurs.

When the controller disconnects, the Tacx NEO 2T connection hasn’t dropped, nor are there any corresponding BLE/TCP/UDP errors affecting the trainer:

Zwift Ride xxxx: status=not ready

Tacx NEO 2T: status=ready

[ZwiftProtocol] Disabled virtual shifting on Tacx Neo 2T

A few seconds later, Zwift reconnects to the controller, goes through reinitialisation, and virtual shifting is re-enabled. The reinitialisation appears to fix the previous irritant where virtual-shifting style and button layout could be reset to default after a reconnect. That’s progress, although the reconnection now may take longer.

Why disable virtual shifting the moment the controllers go offline? A brief disconnect isn’t a command from the user to shut virtual shifting off. If virtual shifting was enabled before the controller disappeared, and the user hasn’t changed anything, why change that configuration at all?

This is particularly painful if you’re using a Tacx NEO. Since there isn’t a physical flywheel, when virtual shifting is disabled the resistance disappears and you spin out, only for the resistance to slam back in when virtual shifting is re-enabled. My suspicion is that trainers with a heavier physical flywheel may mask some of this because of their flywheel inertia, but I haven’t tested that. Another unknown is how different firmwares on different devices may handle this event.

Just to cover the basics, I’ve tested different Bluetooth setups, including a TP-Link UB500 on an extension lead, turning off Race Mode, and disabling the ANT+ transmission. Some of these changes may appear to reduce the frequency of dropouts, but the problem is intermittent enough that I’m not going to claim cause and effect from that.

I can’t spot any consistent sign immediately before these disconnects in the log, so I’m not pretending to have a root cause. A non-scientific observation is that it happens more often in a busy world, on busy rides, and when transitioning to different road surfaces such as road to wooden bridge. But it’s not a repeatable test.

In the end, what is puzzling is that a transient loss of an input device is being treated as a change to the state of a trainer that has remained connected throughout.

@shooj - can you please run this by your developers.

Hi @Tardy_Tortoise
There’s a lot of information to parse here. Before next steps, I want to ask for clarification.

I want to be clear that this is main issue? That you expect resistance to remain constant even if the controllers lose signal?

This language is confusing to me - since the controllers send the signals to increase / decrease resistance, being unable to change resistance would make sense should the controllers lose signal.

You mentioned disabling the ANT+ transmisison during your troubleshooting. Unless you have a heart rate monitor or some other device that broadcasts ANT-only, we recommend unplugging the ANT dongle from your computer as it complicates the signal bonding. Are you disabling in OS settings, or by physical removal?

First, for clarity, there have been various reports regarding intermittent issues with virtual shifting. It’s more noticeable on the Tacx NEO, that’s all. I’m trying to help by explaining what is actually going on, and why the helpful suggestions regarding adapters and race-mode may alleviate the frequency, but they are not the final solution.

For simplicity, let’s work from the basis that the Zwift platform (in this case running on Windows 11) is at fault. Just bear with me.

The latest update fixed an issue where virtual shifting would drop the shifting style mid-ride. If you were using Shimano sequential and changed gears with just the right shifter, suddenly your up/down buttons were only up, and you needed to use the left paddle for down. This is part and parcel of the same problem. It’s exactly the same root cause – Zwift dropping and then automatically reconnecting the controllers.

Yes, absolutely it should, at least for some grace period. What’s more likely, someone powering off their controllers mid ride not via the UI, or a temporary disconnection for whatever reason?

It doesn’t matter if the controller blipped for a second or the Zwift application dropped the connection in error. Zwift suddenly dropping resistance to zero mid-ride, and then the trainer having to overcompensate when Zwift recovers and re-enables virtual shifting a second or two later, is not a good user experience.

Yes, but which is a better error-recovery scenario in the case of a temporary disconnection: leave resistance as it is, or suddenly reduce resistance to zero, causing the user to spin out and have high cadence and no power, which is where we are today?

Bear in mind, if you reduce the resistance to zero and a second or two later Zwift reconnects (as it does), then the trainer will overcorrect (high resistance) and the user then bogs down until things stabilise.

The Tacx NEO trainers have the ability to disable their broadcasting of ANT+ signals, which reduces the processing load on the trainer and helps reduce potential interference.That’s all.

Zwift can, and should, handle ANT+ and Bluetooth devices at the same time, but I take your point regarding user confusion. Regardless, this issue isn’t one of bonding; it happens with or without an ANT+ dongle in the mix.

Bottom line, I’m explaining why the problem occurs. I will pass it on to support as you suggest.

Could this be the reason why the last time I attempted to use virtual shifting my Neo2T it got stuck on shifting up into the largest gear?
(which unfortunately happened on a 20% grade…)

I assumed it was a sweat ingress issue and the button got “stuck”; not a communication error.


I haven’t used them since for the very obvious reason of that being enough of a bad experience to not want to use them again; and assuming they were faulty on a physical level.

It may be, Andrew.

Because of the nature of the interaction with the Tacx trainers, you get a blip as the resistance drops then, when Zwift reestablished the connection, the trainer tries to recover, but it will now overshoot from where it was. So, if you’re going up a hill, that’s going to be like hitting a brick wall.

The good news for anyone who has a similar issue, after I submitted a ticket, Zwift support were very attentive in routing it to the correct department.

Hi there, I just bought the same combination and have exactly the same problems. Is the problem solved after contacting Zwift? I’m at the point of throwing everything in the garbage. It’s so annoying buying such an expensive trainer and then having these issues…

Sadly, it is complicated as there are several moving parts and potential problems.

If you look at the log, or use https://trainerdx.com/, do you see Bluetooth disconnects just for the ride controllers?

If so, that is the issue with virtual shifting being disabled then re-established a second or so later. I have been looking deeper and mechanical flex from the controllers, or even big paws, could cause some button bounce from the pairing/power buttons. Zwift have kindly sent a replacement set of handlebars which I am trying, but that is only part of the problem.

There are also some virtual shifting bugs that are being discussed between Garmin and Zwift. But the problem there is that by not using established standards or interfaces, each vendor has had to implement virtual shifting in their own firmware, which inevitably leads to inconsistent experiences.

I have had several email exchanges with Zwift staff. They have added some extra logging messages which will help pin things down some more. Zwift are still working to deliver fixes, but I am not privy to release information.