Thursday, March 4, 2021

DFA a1 agreement using Polar H10, ECG, HRV logger

Preface - missed beat artifact correction can lead to a bias in the DFA a1 measurement especially at low values below .5.  Now that the mystery has been solved about why may of us were getting these missed beats (Ant+ transmission errors), it's time to take another look at how close the agreement is between devices and methods to determine a1.  Why is this important?  Since a1 is dimensionless, my value of .5 represents the same fractal correlation pattern as yours does at .5.  But, if my device or method of calculating this is not as accurate, then I can get a misrepresentation of the physiologic state behind the a1 number.  With this in mind the following post will look at the "agreement" of DFA a1 values during an incremental ramp using 3 different "measurement types".  The ramp was the usual one I do, 130 to 230 watts over 20 minutes.  

The HRV was recorded  wearing a Polar H10 by a Garmin watch and an ipad using HRV logger.  I also wore a Movesense medical module ECG recorded at a sample rate of 500 Hz in the high chest position to get optimal waveform and signal voltage  

  • The Garmin recording (H10) was processed by Kubios software
  • the Movesense ECG was processed by Kubios
  • HRV logger did it's own processing from the H10 device. 

Therefore we have the following:

  • Polar H10 to Kubios
  • Polar H10 to HRV logger
  • ECG to Kubios

It's important to make sure that the Polar H10 is "ECG accurate" given what we saw in our Sensors study, since the H7 can "under-read" the a1 at uncorrelated levels.  The great thing about using bluetooth as the transmission protocol is the much reduced missed beat artifact rate. 

The results:

ECG waveform quality - Excellent:

Definitely can't complain about that.

Entire hour session tracking HR (gray), ECG a1 (black), H10 a1 via Kubios (red), H10 a1 via Logger (green).

Several observations:

  • The black and red (ECG, H10) appear very close with the exception of a short section in the middle area that contained some artifact in the ECG.
  • The H10 does not appear to be under reading compared to the ECG.  Of course, this ECG has not yet been "validated" but the waveform does look like it should.
  • The HRV logger seems to track a bit higher than the other two, something I've seen before.  The preprocessing and computation methods are different which must account for the differences since there were virtually no artifacts in the H10 or logger files.
  • The logger may have difficulty with the very low DFA a1 values.

A closeup of the ramp shows this even better:

  • The HRVT (HRV threshold felt to be near the AeT) would be about the same with the H10 or the ECG.
  • The HRV logger HRVT intersects the .75 DFA a1 about 1.5 minutes later and since the ramp is 5 watts per minute, only about 8 watts higher.
  • But, if one were tracking DFA a1 at the .25 region, the logger (green) would read a1 higher.

Correlation of DFA a1 by device type - (artifact area values removed)

Regression comparison of ECG vs H10 a1 values

Bland Altman analysis of the H10 vs the ECG:

  • This is showing us that the H10 and the ECG have excellent agreement.  Since it is very difficult to match the time stamps exactly (could have been off 2-3 secs), the scatter does not surprise me.
  • The bottom line here is that we now have much more confidence that the H10 and Movesense ECG are equivilant measurement devices for the DFA a1.

Advantage of the ECG:

For many researchers, elite athletes and older users (me), having a single lead ECG can be valuable.  Here is an example of a pair of atrial premature complexes that were present later on in the session.  This is not a dangerous situation, but high degrees of this or other arrhythmia seen under more stressful scenarios (marathon run, extreme heat, dehydration) can be documented with the Movesense as a chest belt unit:


What about HRV logger and accuracy?

It is possible that some users will get closer results using the Logger to Kubios?  I looked at a friends results recently and he had much better agreement with Kubios and the HRV logger.  This data was recorded with an H10 to both a Garmin headunit and the Logger (both bluetooth) during a bike incremental ramp (with the power in green, logger blue, kubios red):

  • The Logger in blue closely matched the H10.
  • But, few a1 values were below .5.  Would the low values have matched as well?


  • The Polar H10 with bluetooth transmission is a reliable, artifact resistant device for DFA a1 recording.
  • The Movesense Medical module closely matches and agrees with the H10.  It has the advantage of tracking arrhythmia and manually correcting noise artifact in Kubios.
  • The HRV logger app does approximate that of Kubios, but not exactly.  In my case it over estimates the DFA a1, but in other individuals it might be different.  
  • I've started to notice some people posting potential discrepancies with their AeT and the DFA a1 based HRVT using the HRV logger.  In addition some report not being able to get a1 down to very low values despite high intensity.
  • If in doubt, take a look at the RR series in Kubios to confirm the HRV logger values.  As time goes on, we hopefully will have a better handle on why this occurs and the means to compensate.



  1. Bruce, what exactly should I look in Kubios (free) regarding RR series ? I did cycle test and a treamill one (fixed speed increasing grade, to eliminate ground force impact) cannot get a result in runalyze...always seem to have LT1 at 160HR, which is not possible.
    Maybe I'm the kind of person who cannot get agreement on the data

  2. Take a look at this one -
    I used 2 minute windows in Kubios, one for each stage and plotted them out. There is definitely a learning curve for kubios.