Distortion Issues, IPD issues(?)

[SirCanucklehead]

System specs:

Arch Linux
Kernel 7.1.4
AMD Ryzen 7 5800X
AMD Radeon RX 7600 XT
SteamVR main branch
Everything up to date

Upon first installation, I had an issue that another post mentioned, that being a vertical misalignment of the two eye displays, the right eye looked slightly higher than the left. Going into SteamVR settings, and dropping the resolution from the default 150% to 100% seemed to mostly solve the issue. However, after that was solved, I noticed a couple other things.

Mainly, when looking around, the image I see seems to be unevenly distorted, almost as if I'm looking at the displayed content through a funhouse mirror. It's not an extreme effect, but noticeable enough to effect the experience.

The second is extremely hard to pin down because the only time I really notice it is right after I take off the headset. I've noticed that even after a very short troubleshooting or configuration session, when I take off the headset, my eyes are crossed for a few seconds. This is the 4th headset I've owned and have never experienced this. It is as though something is askew with the focal point of the headset, and my eyes aren't adjusting to far off objects as they should be. I wonder if this could be an IPD (rendering?) issue? I have manually adjusted the lenses to match my eyes as best I can, but something still feels off and I experience noticeable eye strain. I have also tried the headset with Monado/Envision, which lets me set a render IPD via an environment variable, and that did seem to mitigate some of the discomfort, but I haven't had enough time to test extensively.

I will note that I've never tested this headset on Windows, and intend to do so just to sanity check and make sure this isn't a hardware issue. I don't have access to a Windows machine at the moment but hope to in the next couple weeks.

[rvdende]

I've had quite a lot of fiddling on my side, but my setup is a bit different ubuntu, intel, nvidia 4090 and i'm using monado with steamvr driver. I've asked claude to generate a reply to you based on what I've found. I hope it helps you!

Your second symptom — crossed eyes after taking the headset off — is a known bug, not your eyes and not your hardware. The Vive Pro 2's IPD report is broken on Linux, and Monado has an explicit workaround for it in src/xrt/drivers/steamvr_lh/device.cpp:

} else if (this->variant == VIVE_VARIANT_PRO2) {
// Default to 63mm on Vive Pro 2, the IPD reading is completely borked and always sends 15mm
eye_relation.x = 0.063f;
}

So under Monado you're rendered at a hardcoded 63 mm unless you override it — which is exactly the env var that gave you relief.

The SteamVR path is worse: it has no workaround at all. In VivePro2-Linux-Driver, Prop_UserIpdMeters_Float is commented out in bin/driver-proxy/src/driver/hmd.rs, SetDisplayEyeToHead is passed straight through, and nothing anywhere reads the physical lens-slider position. The ViveConfig struct in crates/vive-hid/src/lib.rs has no IPD or lens-separation field, and the lens-correction protocol takes one static config blob at startup with no IPD input. Sliding the physical IPD adjuster moves the lenses but changes nothing about what's rendered. You've been matching the optics to your eyes while the render stayed put.

If your real IPD is meaningfully above ~66 mm, you're holding convergence tighter than your eyes' resting point for the entire session. That's exactly the "crossed for a few seconds afterward" symptom.

The test that settles it, before you go find a Windows machine:

1. Get your actual IPD measured — optometrist, or a PD ruler app. Don't trust the headset's reading.
2. Boot Monado and set the override. The variable is driver-specific and silently does nothing if you pick the wrong one for your tracking driver:

• steamvr_lh driver → LH_OVERRIDE_IPD_MM=<your_ipd>
• survive driver → SURVIVE_OVERRIDE_IPD_MM=<your_ipd>

In Envision you can set these in the profile's environment-variable section.
3. Set the physical lens slider to that same number so optics and render agree.
4. Run a proper session — 20+ minutes, not a config check.

If the eye strain goes away, you've confirmed it and you don't need Windows. What's your measured IPD? If it's near 63 mm the theory is weak and something else is going on — that number is the single most useful thing you can add to this thread.

For staying on SteamVR: try "ipdOffset" in steamvr.vrsettings. SteamVR exposes it, though I don't know whether it survives the driver proxy — if you test it, please report back either way.

On the funhouse-mirror warping — worth checking your panel width. Monado and the SteamVR driver disagree about the 1836p mode: Monado's src/xrt/drivers/vp2/vp2_hid.h says 3680×1836, VivePro2-Linux-Driver's mode table in crates/vive-hid/src/lib.rs says 3672×1836. Both send the identical dtd,3 HID command, so the panel latches the same mode either way — one of those tables is wrong. In my testing 3680×1836@90 is the pixel-perfect one.

If 3680 is right, SteamVR is compositing a 3672-wide buffer onto a 3680-wide panel: the right-eye viewport lands 4 px off (the driver hardcodes width/2) and the whole frame gets a horizontal-only ~1.002× rescale. A non-uniform stretch is what an asymmetric funhouse warp actually looks like — supersampling 150%→100% shifted your vertical alignment, since that changes the resampleratio.

With the headset connected and awake:

cat /sys/class/drm/card*-DP-*/modes

If that shows 3680 and SteamVR is reporting 3672, that's your distortion bug and it's separate from the IPD one.
Two honest caveats: I've confirmed the 3672/3680 disagreement in both source trees but not which value the panel actually latches — the modes output settles that. And some residual warp is ordinary pupil swim, since the distortion mesh is computed once from a static config assuming a fixed eye posiPD reduces that but won't fully remove it.

[WorldofShadows]

Okay, me again :)

I've dissected this a bit more and drove a few testing scenarios.

Your theory with the resolution is right and this bit is fixed in an older pullrequest in this repo.
The correct resolution for Res 3 IS 3680 and when you correct the code, the mode itself will start normally.

But: The "funhouse"-effect is present here even on the DSC Resolutions.

I wrote a little bit of debug code in the hmd.rs to get the output of IPD and what the steamvr even sees.

The Ipdoffset setting is at least seen and calculated on top of the IPD Steam gets from the driver and the base-IPD is reported as 63, seemingly exactly like Monado does this.

My IPD is a bit hard to measure, because my eyes are a bit "different", but i've set it to 63.5mm on Windows and set the offset inside the steamvr.vrsettingsfile to:

  "ipdOffset" : 0.00050000000000000001,

With this value, my debug output shows:

/home/<user>/.steam/steam/logs/vrstartup-linux.txt:58977:[2026-08-31 18:38:16] VP2 DEBUG: base IPD ~= 0.063000 m (63.00 mm), ipdOffset = +0.000500 m (+0.50 mm), SteamVR IPD = 0.063500 m (63.50 mm), eye depth = 0.015000 m (15.00 mm), batch_error=TrackedProp_Success, ipd_error=TrackedProp_Success, eye_depth_error=TrackedProp_Success

So this IS calculated correctly.

The funhouse effect comes mainly from vertical movement. Horizontal is only a minimum.
So either the LensDistortion dll has some kind of quirk that it could to to calculate this correctly or we're missing something. Maybe the proxy doesn't pass this to the dll?

I've tested Envision with Monado and that works quite well and with or without the fallback distortion, but i really like this driver here and what @ertainLach made possible with this.
I love the simplicity of just starting SteamVR and start playing, so maybe we can figure this out somehow :)