On top of my head, they are both ~1% accurate, at a higher power such as V02 (lets use 350w for the math)
You could be doing 350w, the srm might read the +1% and output 353.5w and the neo could read -1% and be 346.5. In this case BOTH are within specs and you have a 7w difference (if the headunit rounded down/up it could be a 9w difference).
The higher the power the more the drift window is acceptable.
Seems like you are spot on with the big ring, I wouldn't think its reasonable to beat 1-2w difference at higher powers.

(seems like even the recording BLE vs ANT+ could come into play, the other day I was surprised to find 2w difference between Android ant+ & Garmin ant+ on the same power meter for a 20min interval at 300w *shorter and harder will be more volatile*)

Did you disable auto-zero, and manually re-calibrate the SRM when you switched chainrings? Because the SRM will shift its zero setting a little when you change chainrings (and also as a function of how much you cross-chain).

---

Less is more.

I noticed a similar thing with my Neo. The fact is it's just not as accurate as some make it out to be. Mine behaves similar to yours - in smaller gears the actual power is lower than reported by the Neo. I always stay in the big chainring because of this, which is a shame because small chainring would be quieter.

^^This is why I use the small chain ring because it’s quieter. But if the large chainring is the only accurate one I will stay in that gear.
Seriously?? Am I suppose to recalibrate when I shift the front derailleur. I find that hard to believe that would account for the 10 watt difference. And chain line is pretty straight. Same middle cog in the back. Just different front chain ring.

10w difference compared to what.... your crank or pedal based system .. how about drivetrain loss ?

10w compared to my srm. The srm crank reports 10 watts lower in the small chain ring compared to the neo. The large chain ring is within 1-2 watts. This is only in erg mode.
The larger chainring should have less friction so should read lower.
I believe this difference is due to the speed the fly wheel is rotating maybe?

I did a whole lot of testing using the dcrainmaker analyser between the neo and my Quarq. Infact my two neos and 2 of my 3 quarqs... I only say that to add a bit of robustness to what I saw/say in that it takes out the uncertainty that comes from comparing two things and not knowing which is right, or even that either is right.

I found that in teh end the neo was accurate, allowing for drivetrain loss. However, I started off until I did some geek stuff being very suspicious. The issue for me came down to the lack of calibration on the neo being confused with not needing it to warm up. What I was seeing was the neo reporting about that 10-15 watt above the quarqs for first 10-15mins as it warmed up then settling in to be 10watts below (which I was happy with for drivetrain losses). I didn't get the chainring differences on the quarq you're reporting from the SRM. Logically here that can't be the neo, as aside from differing drivetrain efficiency the neo simply measures power from rear axle torque.

I would really recommend you spend a bit of time going through your data on the analyser site. Should help you to understand what's what. Note that I also found some significant differences in what was being shown on screen from different head units (garmin 510/935 and Zwift via Ant). Now the Zwift isn't an issue here, but the odd thing was that the post ride data matched perfectly, but the averaging / smoothing for the onscreen is clearly done differently which means if you look at the data from two different sources on two different head units you may also be thinking it's the power sources that are differing, when in reality there's a second layer of 'error' that's being added from the differing head units. Note I'm not talking about one showing 3s average and the other 10s or instantaneous.

These trainers are not measuring torque at the axle, they're measuring flywheel speed at a given resistance setting. Tacx claims the Neo is more accurate because it uses magnetic resistance rather then a belt, but it's still using flywheel speed to estimate power. It's no more of a "true" power meter than any other smart trainer.

The reason gearing matters is because for a given gear the combination of resistance and flywheel speed used to estimate power will be different.

I did this, and found the inconsistencies between the Neo and my power meter to be be worse on the small chainring, so I use the large chainring. (There's also a bit more inertia when using the large chainring, which I prefer; but with the Neo in ERG mode there's less inertia difference than with other trainers).

OK, I see where you're coming from. I don't fully agree with some of your logic, but there's a bit of semantics in here that doesn't really matter for the bigger issue of trying to get you confident in the two data sources you have, or identitify there is a fault in one and enable you to get that fixed.

Just to clarify I wasn't suggesting the neo was doing anything different to other trainers, just that it was doing it different to the crank based solution. What I failed to probably explain is that if you were to change the gearing on front and rear derailers to provide the same "wheel speed" (as in if riding on the road) then the Neo wouldn't / couldn't know if you were on the inner or outer ring. You mentioned discrepancies between the rings that can ONLY be an issue on the SRM. There's a separate source of error that would be the 'slope' of the neo (ok there isn't any calibration so there is no slope, but explaining the gradient of the power as opposed to the intercept/offset) where it had different errors for low rotational speed compared to higher rotational speed. So testing for this would be not to use ERG mode, but constant resistance ('speed' varies power) for two or three different resistance levels.

Are you OK to post the links to the analyser data? I'm interested, and maybe that I can see where you're coming from easier with the data.

I'm just a bit confused at the moment as you've got two bits of data, where equally either could be right / wrong, and indeed there's something odd with the chainrings that my view is more likely to be SRM related, and yet you seem to focus on the Neo being in error (thread title). Don't get me wrong, I thought my Neo was dodgy to start with too, until I got a second one (wife got my 'old' one) and had the same data. And you may well have a dodgy neo. In which case then of course be good to get that demonstrated and see Tacx sort it. Or perhaps send your SRMs back for calibration.

I don't have the dcr analyzer data anymore. Just to be clear though, I wasn't talking about equivalent gearing for big vs small chainring, where the "wheel speed" would be the same. I agree that should make no difference. I was talking about using the same middle cog (for straight chainline) in both cases. So for the big chainring "wheelspeed" was higher.

The one thing I will say for the Neo, it seems to be pretty consistent. Comparing it to my power meter I don't see big fluctuations during a workout. It's off by a bit, but consitently so. I still prefer to use the power meter for recording though, partly because it's better are recording sudden changes in power and also for consistency.

First a couple comments: I test power meters all the time using both static and dynamic loads, and my Neo has been very very accurate. In almost all cases, if there's a difference, it's the bike power meter that's off, not the Neo. Once I get the bike power meter reading properly I've found that it will read the same as my Neo. (I do this as a second or third check in addition to using calibrated weights and also roller testing.)

Second, there's this:
I think this is the SRM reading differently in the two chainrings, not the Neo. Try either using calibrated weights or roller testing to verify. Also I'd bet that the chainring close to the Neo (the large one) is the one that's right. The difference of 10w is probably not friction - that's a ton of friction. It's probably a calibration problem with the SRM and the chainrings.

I did quite a bit of chain tests in the same gear (58x16 or so IIRC) using my SRM and Neo and the delta was quite consistent at only a few watts (<5W @ 250W). I then tried to see if I could test cross-chaining losses and had to put the Neo at a simulated 5% (IIRC) slope and changed to a bigger cog in the rear (36T or something like that). The Neo then started reading something like 15-20W lower than my SRM (in the same chain ring - running 1x). Either cross-chain losses were huge (which I don't believe) or one of the powermeters (or both) were off - and my guess would be the Neo since the speed, inertia, and torque was different, which I would guess would have a larger impact on the Neo (and subjectively the SRM felt right). I gave up on measuring cross-chain losses so didn't dive further into it.

I did quite a bit of testing with Neo and Wahoo kickrs.
By far Neo is the most consistent and precise unit.

I compared with P1 and Garmin pedals.
I have seen similar behaviour: Neo @300W in small ring reports few watts higher (9w on P1 and 6W on Garmin) Big ring Neo shows 2W higher and 1W lower on Garmins)
When testing at 100W difference is pretty much the same but it makes much higher %

But this is consistent even after 3-4 hour ride. What is interesting: when looking at simple ride power average of entire ride it is super close within 1-3W from power meter.

I think this is super close since it impossible to find 2 power meters showing exactly the same power.

Wahoo kickers were off by 35W in worst case scenario... And one unit had massive power drift, At the end of 2 hours ride I was doing 25W more...

How do you know which is “right”? My vector2 is always 7-10w higher than Neo across a ride of all different intensities. If I do intervals around 330w I might be 10–12w.

I just use Neo when zwift racing because everyone says it’s faster. But would be nice for “free speed” if I switched it to the Vector2s

Wouldn't that difference simply be explained by drivechain losses? Vectors measure power at the pedal before any drivechain losses, Neo measures it at the cassette after any losses. Losses for a well-maintained set up are typically in the 2-3% range, which would translate pretty well to the 7-10 watts you're seeing.

If you want free speed just drop your weight/height in Zwift ;-) (just kidding, although I'm sure plenty of people are doing this)

Your test is bogus. You need to record the power measured by the Neo and compare your SRM readings to that, not to the power target.