Background: Use of peak-picked ECG is generally not recommended for heart rate variability (HRV) calculations because the apparent R-wave peak moves around within the QRS due to the peak-picking operation. This study quantifies the RR error in algorithms used to recover high resolution RR intervals from peak-picked ECG. Methods: Twenty minute samples of single lead ECG (n = 1000) were generated by the Physionet HRV ECG simulator. Each ECG record had a random combination of heart rate standard deviation and additive noise. ECG was decimated to 250sps from 1000sps and then peak-picked to the final sample rate of 125sps. The second peak picking method preserved the phase of the chosen peak for near linear 250sps reconstruction. High resolution RR interval recovery was based on up-sampling. The first method chose R-wave peaks. The second method used template matching to align beats. RR interval reference was the sequence of R-wave markers from the ECG simulator. HRV standard deviation of RR intervals (SDNN) was calculated for each record using 125sps decimated and peak-picked methods as input and R-wave peaks and template matching for RR interval recovery. Bland-Altman analysis was used to assess bias of SDNN error across the SDNN range from 10 to 90ms. Results: The bias of SDNN error at low SDNN values was pronounced in two cases, when up-sampling was not used in the RR recovery (0.5ms SDNN error in the SDNN range of 10ms) and for peak picked input and R-wave peaks used for RR recovery (0.2ms error). Template matching dramatically reduced the SDNN error at low levels of SDNN (0.01ms error for SDNN 10ms range). Conclusion: Low SDNN error may be possible using peak picked ECG. More study is required using ECG datasets with high resolution ECG and abnormal HRV.