# RuntimeError: Matrix is badly conditioned **URL:** \ **Category:** Support & Discussions\ **Tags:** preprocessing\ **Created:** [August 7, 2024, 7:54pm UTC](https://mne.discourse.group/t/runtimeerror-matrix-is-badly-conditioned/9177 "2024-08-07T19:54:33Z")\ **Posts on this page:** 10\ **Page:** 1
**Author:** ![mschendel](https://avatars.discourse-cdn.com/v4/letter/m/2acd7d/32.png) [@mschendel](https://mne.discourse.group/u/mschendel)\ **Post date:** [August 7, 2024, 7:54pm UTC](https://mne.discourse.group/t/runtimeerror-matrix-is-badly-conditioned/9177/1 "2024-08-07T19:54:34Z")
Hello, I am able to run the prepare\_emptyroom, but when I try the maxfilter command on the prepared data, I get this error. I tried a set of 7 subject data files and 7 empty room files, and each created the error. - MNE version: 1.7.0 - operating system: Ubuntu 22.04.4 LTS Here is some of the data I’m working with: empty\_ready is the empty room raw file that went through ```python In [9]: empty_ready = mne.preprocessing.maxwell_filter_prepare_emptyroom(empty_raw, raw =raw, bads ='union' , verbose=True) In [10]: empty_ready.info Out[10]: head transform dig: 135 items (3 Cardinal, 4 HPI, 128 Extra) events: 1 item (list) experimenter: MEG (meg) highpass: 0.1 Hz hpi_subsystem: 3 items (dict) line_freq: 60.0 lowpass: 330.0 Hz meas_date: 2024-01-10 19:25:58 UTC meas_id: 4 items (dict) nchan: 310 proj_id: 1 item (ndarray) proj_name: test projs: mag_ssp_upright.fif : PCA-v1: off, mag_ssp_upright.fif : PCA-v2: ... sfreq: 1000.0 Hz subject_info: 6 items (dict) > In [11]: empty_proc = maxwell_filter(empty_ready,origin = (0,0,40), int_order=8, ext_order=3, calibration=cal_data, cross_talk= ct_data, st_duration =6, st_correlation=0. ...: 95 , coord_frame = 'head', head_pos=head_pos) ...: Maxwell filtering raw data 102 of 102 magnetometer types replaced with T3. Bad MEG channels being reconstructed: ['MEG0443', 'MEG1622'] Processing 204 gradiometers and 102 magnetometers Using fine calibration sss_cal_3046_2009-04-29.dat Adjusting non-orthogonal EX and EY Adjusted coil positions by (μ ± σ): 0.5° ± 0.4° (max: 2.4°) Using origin 0.0, 0.0, 40000.0 mm in the head frame Processing data using tSSS with st_duration=6.0 /export/research/analysis/human/jstephen/shared/programs/python/anaconda_072020/new/envs/mne1.7/lib/python3.12/site-packages/mne/preprocessing/maxwell.py:2349: RuntimeWarning: invalid value encountered in scalar divide f"bits/sample ({100 * I_tots[lim_idx] / max_info:0.1f}% of peak " Using 95/95 harmonic components for 0.000 (80/80 in, 15/15 out) --------------------------------------------------------------------------- RuntimeError Traceback (most recent call last) Cell In[9], line 1 ----> 1 empty_proc = maxwell_filter(empty_ready,origin = (0,0,40), int_order=8, ext_order=3, calibration=cal_data, cross_talk= ct_data, st_duration =6, st_correlation=0.95 , coord_frame = 'head', head_pos=head_pos) File :12, in maxwell_filter(raw, origin, int_order, ext_order, calibration, cross_talk, st_duration, st_correlation, coord_frame, destination, regularize, ignore_ref, bad_condition, head_pos, st_fixed, st_only, mag_scale, skip_by_annotation, extended_proj, verbose) File /export/research/analysis/human/jstephen/shared/programs/python/anaconda_072020/new/envs/mne1.7/lib/python3.12/site-packages/mne/preprocessing/maxwell.py:406, in maxwell_filter(raw, origin, int_order, ext_order, calibration, cross_talk, st_duration, st_correlation, coord_frame, destination, regularize, ignore_ref, bad_condition, head_pos, st_fixed, st_only, mag_scale, skip_by_annotation, extended_proj, verbose) 384 logger.info("Maxwell filtering raw data") 385 params = _prep_maxwell_filter( 386 raw=raw, 387 origin=origin, (...) 404 extended_proj=extended_proj, 405 ) --> 406 raw_sss = _run_maxwell_filter(raw, **params) 407 # Update info 408 _update_sss_info(raw_sss, **params["update_kwargs"]) File /export/research/analysis/human/jstephen/shared/programs/python/anaconda_072020/new/envs/mne1.7/lib/python3.12/site-packages/mne/preprocessing/maxwell.py:688, in _run_maxwell_filter(raw, skip_by_annotation, st_duration, st_correlation, st_only, st_when, ctc, coil_scale, this_pos_quat, meg_picks, good_mask, grad_picks, head_pos, info, _get_this_decomp_trans, S_recon, update_kwargs, ignore_ref, reconstruct, copy) 660 def _run_maxwell_filter( 661 raw, 662 skip_by_annotation, (...) 686 # The time it takes to recompute S and pS themselves is roughly on par 687 # with the np.dot with the data, so not a huge gain to be made there. --> 688 S_decomp, S_decomp_full, pS_decomp, reg_moments, n_use_in = _get_this_decomp_trans( 689 info["dev_head_t"], t=0.0 690 ) 691 update_kwargs.update(reg_moments=reg_moments.copy()) 692 if ctc is not None: File /export/research/analysis/human/jstephen/shared/programs/python/anaconda_072020/new/envs/mne1.7/lib/python3.12/site-packages/mne/preprocessing/maxwell.py:1259, in _get_decomp(trans, all_coils, cal, regularize, exp, ignore_ref, coil_scale, grad_picks, mag_picks, good_mask, mag_or_fine, bad_condition, t, mag_scale, mult) 1257 msg = "Matrix is badly conditioned: %0.0f >= 1000" % cond 1258 if bad_condition == "error": -> 1259 raise RuntimeError(msg) 1260 elif bad_condition == "warning": 1261 warn(msg) RuntimeError: Matrix is badly conditioned: 18132836229591732 >= 1000 ``` I found the [previous thread](https://mne.discourse.group/t/maxfilter-error-ill-conditioned-matrix/1491) about this error, which seemed to indicate issues with head position measurement. I checked the head position from the participant data: ![M87104027_head_pos](https://global.discourse-cdn.com/free1/uploads/mne/original/2X/3/333327fa967d9a1da280cef804f910fd3835c084.jpeg) Is there anything else I can check to troubleshoot this? Any other info I can provide? Does this head position include too large a change? Thanks for your help, Megan ---
**Author:** ![mschendel](https://avatars.discourse-cdn.com/v4/letter/m/2acd7d/32.png) [@mschendel](https://mne.discourse.group/u/mschendel)\ **Post date:** [August 23, 2024, 4:14pm UTC](https://mne.discourse.group/t/runtimeerror-matrix-is-badly-conditioned/9177/2 "2024-08-23T16:14:37Z")
Hello, Just wanted to try to see if anyone can help troubleshoot this issue. I will try to attend the office hours next week. Here is another example, where the head position looks pretty fine to me from the original run: ```python mne.viz.plot_head_positions(head_pos, mode="traces") ``` ![M87104027_visit2_run1_headpos](https://global.discourse-cdn.com/free1/uploads/mne/original/2X/a/a054874947fdadb48e9aaa62d5bf4fe46d327d5b.jpeg) ```python In [16]: empty_proc = maxwell_filter(empty_ready,origin = (0,0,40), int_order=8, ext_order=3, calibration=cal ...: _data, cross_talk= ct_data, st_duration =6, st_correlation=0.95 , coord_frame = 'head', head_pos=head ...: _pos) ...: Maxwell filtering raw data 102 of 102 magnetometer types replaced with T3. Bad MEG channels being reconstructed: ['MEG0443'] Processing 204 gradiometers and 102 magnetometers Using fine calibration sss_cal_3046_2009-04-29.dat Adjusting non-orthogonal EX and EY Adjusted coil positions by (μ ± σ): 0.5° ± 0.4° (max: 2.4°) Using origin 0.0, 0.0, 40000.0 mm in the head frame Processing data using tSSS with st_duration=6.0 /export/research/analysis/human/jstephen/shared/programs/python/anaconda_072020/new/envs/mne1.7/lib/python3.12/site-packages/mne/preprocessing/maxwell.py:2349: RuntimeWarning: invalid value encountered in scalar divide f"bits/sample ({100 * I_tots[lim_idx] / max_info:0.1f}% of peak " Using 95/95 harmonic components for 0.000 (80/80 in, 15/15 out) --------------------------------------------------------------------------- RuntimeError Traceback (most recent call last) Cell In[16], line 1 ----> 1 empty_proc = maxwell_filter(empty_ready,origin = (0,0,40), int_order=8, ext_order=3, calibration=cal_data, cross_talk= ct_data, st_duration =6, st_correlation=0.95 , coord_frame = 'head', head_pos=head_pos) File :12, in maxwell_filter(raw, origin, int_order, ext_order, calibration, cross_talk, st_duration, st_correlation, coord_frame, destination, regularize, ignore_ref, bad_condition, head_pos, st_fixed, st_only, mag_scale, skip_by_annotation, extended_proj, verbose) File /export/research/analysis/human/jstephen/shared/programs/python/anaconda_072020/new/envs/mne1.7/lib/python3.12/site-packages/mne/preprocessing/maxwell.py:406, in maxwell_filter(raw, origin, int_order, ext_order, calibration, cross_talk, st_duration, st_correlation, coord_frame, destination, regularize, ignore_ref, bad_condition, head_pos, st_fixed, st_only, mag_scale, skip_by_annotation, extended_proj, verbose) 384 logger.info("Maxwell filtering raw data") 385 params = _prep_maxwell_filter( 386 raw=raw, 387 origin=origin, (...) 404 extended_proj=extended_proj, 405 ) --> 406 raw_sss = _run_maxwell_filter(raw, **params) 407 # Update info 408 _update_sss_info(raw_sss, **params["update_kwargs"]) File /export/research/analysis/human/jstephen/shared/programs/python/anaconda_072020/new/envs/mne1.7/lib/python3.12/site-packages/mne/preprocessing/maxwell.py:688, in _run_maxwell_filter(raw, skip_by_annotation, st_duration, st_correlation, st_only, st_when, ctc, coil_scale, this_pos_quat, meg_picks, good_mask, grad_picks, head_pos, info, _get_this_decomp_trans, S_recon, update_kwargs, ignore_ref, reconstruct, copy) 660 def _run_maxwell_filter( 661 raw, 662 skip_by_annotation, (...) 686 # The time it takes to recompute S and pS themselves is roughly on par 687 # with the np.dot with the data, so not a huge gain to be made there. --> 688 S_decomp, S_decomp_full, pS_decomp, reg_moments, n_use_in = _get_this_decomp_trans( 689 info["dev_head_t"], t=0.0 690 ) 691 update_kwargs.update(reg_moments=reg_moments.copy()) 692 if ctc is not None: File /export/research/analysis/human/jstephen/shared/programs/python/anaconda_072020/new/envs/mne1.7/lib/python3.12/site-packages/mne/preprocessing/maxwell.py:1259, in _get_decomp(trans, all_coils, cal, regularize, exp, ignore_ref, coil_scale, grad_picks, mag_picks, good_mask, mag_or_fine, bad_condition, t, mag_scale, mult) 1257 msg = "Matrix is badly conditioned: %0.0f >= 1000" % cond 1258 if bad_condition == "error": -> 1259 raise RuntimeError(msg) 1260 elif bad_condition == "warning": 1261 warn(msg) RuntimeError: Matrix is badly conditioned: 17743322965512802 >= 1000 ``` Please let me know if I can provide any additional info. Thanks, Megan ---
**Author:** ![mschendel](https://avatars.discourse-cdn.com/v4/letter/m/2acd7d/32.png) [@mschendel](https://mne.discourse.group/u/mschendel)\ **Post date:** [August 23, 2024, 6:19pm UTC](https://mne.discourse.group/t/runtimeerror-matrix-is-badly-conditioned/9177/3 "2024-08-23T18:19:15Z")
Some additional info: if I try to replicate my participant processing as I did using MEGIN’s command line maxfilter, I get the same error. So I’m just not using maxwell\_filter correctly, it seems. My original maxfilter command: ```python /neuro/bin/util/maxfilter-2.2 -v -force -autobad 60 -badlimit 20 -movecomp inter -st 6 -corr 0.95 -f M87104027+Study20240405+avmmn_visit2_run1_raw.fif -o /M87104027_avmmn_visit2_run1_tsss_mc.fif -origin 0 0 40 -frame head -format short -hp M87104027_avmmn_visit2_run1_tsss_mc_headpos.asc ``` What I tried in mne (I used the asc file created from my command line maxfilter-2.2 run for the head\_pos): ```python In [12]: raw_proc = maxwell_filter(raw,origin = (0,0,40), int_order=8, ext_order=3, calibration=cal_data, cros ...: s_talk= ct_data, st_duration =6, st_correlation=0.95 , coord_frame = 'head', head_pos=head_pos) Maxwell filtering raw data 102 of 102 magnetometer types replaced with T3. No bad MEG channels Processing 204 gradiometers and 102 magnetometers Using fine calibration sss_cal_3046_2009-04-29.dat Adjusting non-orthogonal EX and EY Adjusted coil positions by (μ ± σ): 0.5° ± 0.4° (max: 2.4°) Using origin 0.0, 0.0, 40000.0 mm in the head frame Processing data using tSSS with st_duration=6.0 /export/research/analysis/human/jstephen/shared/programs/python/anaconda_072020/new/envs/mne1.7/lib/python3.12/site-packages/mne/preprocessing/maxwell.py:2349: RuntimeWarning: invalid value encountered in scalar divide f"bits/sample ({100 * I_tots[lim_idx] / max_info:0.1f}% of peak " Using 95/95 harmonic components for 0.000 (80/80 in, 15/15 out) --------------------------------------------------------------------------- RuntimeError Traceback (most recent call last) Cell In[12], line 1 ----> 1 raw_proc = maxwell_filter(raw,origin = (0,0,40), int_order=8, ext_order=3, calibration=cal_data, cross_talk= ct_data, st_duration =6, st_correlation=0.95 , coord_frame = 'head', head_pos=head_pos) File :12, in maxwell_filter(raw, origin, int_order, ext_order, calibration, cross_talk, st_duration, st_correlation, coord_frame, destination, regularize, ignore_ref, bad_condition, head_pos, st_fixed, st_only, mag_scale, skip_by_annotation, extended_proj, verbose) File /export/research/analysis/human/jstephen/shared/programs/python/anaconda_072020/new/envs/mne1.7/lib/python3.12/site-packages/mne/preprocessing/maxwell.py:406, in maxwell_filter(raw, origin, int_order, ext_order, calibration, cross_talk, st_duration, st_correlation, coord_frame, destination, regularize, ignore_ref, bad_condition, head_pos, st_fixed, st_only, mag_scale, skip_by_annotation, extended_proj, verbose) 384 logger.info("Maxwell filtering raw data") 385 params = _prep_maxwell_filter( 386 raw=raw, 387 origin=origin, (...) 404 extended_proj=extended_proj, 405 ) --> 406 raw_sss = _run_maxwell_filter(raw, **params) 407 # Update info 408 _update_sss_info(raw_sss, **params["update_kwargs"]) File /export/research/analysis/human/jstephen/shared/programs/python/anaconda_072020/new/envs/mne1.7/lib/python3.12/site-packages/mne/preprocessing/maxwell.py:688, in _run_maxwell_filter(raw, skip_by_annotation, st_duration, st_correlation, st_only, st_when, ctc, coil_scale, this_pos_quat, meg_picks, good_mask, grad_picks, head_pos, info, _get_this_decomp_trans, S_recon, update_kwargs, ignore_ref, reconstruct, copy) 660 def _run_maxwell_filter( 661 raw, 662 skip_by_annotation, (...) 686 # The time it takes to recompute S and pS themselves is roughly on par 687 # with the np.dot with the data, so not a huge gain to be made there. --> 688 S_decomp, S_decomp_full, pS_decomp, reg_moments, n_use_in = _get_this_decomp_trans( 689 info["dev_head_t"], t=0.0 690 ) 691 update_kwargs.update(reg_moments=reg_moments.copy()) 692 if ctc is not None: File /export/research/analysis/human/jstephen/shared/programs/python/anaconda_072020/new/envs/mne1.7/lib/python3.12/site-packages/mne/preprocessing/maxwell.py:1259, in _get_decomp(trans, all_coils, cal, regularize, exp, ignore_ref, coil_scale, grad_picks, mag_picks, good_mask, mag_or_fine, bad_condition, t, mag_scale, mult) 1257 msg = "Matrix is badly conditioned: %0.0f >= 1000" % cond 1258 if bad_condition == "error": -> 1259 raise RuntimeError(msg) 1260 elif bad_condition == "warning": 1261 warn(msg) RuntimeError: Matrix is badly conditioned: 15210861592565222 >= 1000 ``` Thanks for any eyes on this, Megan ---
**Author:** ![richard](https://yyz2.discourse-cdn.com/free1/user_avatar/mne.discourse.group/richard/32/15_2.png) [@richard](https://mne.discourse.group/u/richard)\ **Post date:** [August 23, 2024, 9:40pm UTC](https://mne.discourse.group/t/runtimeerror-matrix-is-badly-conditioned/9177/4 "2024-08-23T21:40:41Z")
@larsoner Would you have any idea what’s going on here? ---
**Author:** ![larsoner](https://yyz2.discourse-cdn.com/free1/user_avatar/mne.discourse.group/larsoner/32/3_2.png) [@larsoner](https://mne.discourse.group/u/larsoner)\ **Post date:** [August 26, 2024, 4:18am UTC](https://mne.discourse.group/t/runtimeerror-matrix-is-badly-conditioned/9177/5 "2024-08-26T04:18:50Z")
Typically this is triggered by implausible (or very far out of helmet) head positions. But your position estimates look reasonable. I think it’s `origin = (0,0,40)`. Keep in mind the units are SI meters in MNE, not millimeters like in `maxfilter`. Can you see if `origin=(0., 0., 0.04)` fixes the problem? ---
**Author:** ![mschendel](https://avatars.discourse-cdn.com/v4/letter/m/2acd7d/32.png) [@mschendel](https://mne.discourse.group/u/mschendel)\ **Post date:** [August 26, 2024, 3:27pm UTC](https://mne.discourse.group/t/runtimeerror-matrix-is-badly-conditioned/9177/6 "2024-08-26T15:27:12Z")
Hi all, YES. Thank you for noticing that. I read [preprocessing doc](https://mne.tools/stable/generated/mne.preprocessing.maxwell_filter.html) , but I didn’t see anything about meters, there. It would be great to add it to info about the origin parameter. I also don’t see anything in the [tutorial article.](https://mne.tools/stable/auto_tutorials/preprocessing/60_maxwell_filtering_sss.html#sphx-glr-auto-tutorials-preprocessing-60-maxwell-filtering-sss-py) If I look closely in the output, using origin = (0,0,0.04), I see ```python Using origin 0.0, 0.0, 40.0 mm in the head frame ``` and if I use origin (0,0,40), I see: ```python Using origin 0.0, 0.0, 40000.0 mm in the head frame ``` …too bad I didn’t notice that sooner. Anyway. That’s sorted. Thanks for your help! Megan ---
**Author:** ![mscheltienne](https://yyz2.discourse-cdn.com/free1/user_avatar/mne.discourse.group/mscheltienne/32/827_2.png) [@mscheltienne](https://mne.discourse.group/u/mscheltienne)\ **Post date:** [August 27, 2024, 8:19am UTC](https://mne.discourse.group/t/runtimeerror-matrix-is-badly-conditioned/9177/7 "2024-08-27T08:19:18Z")
The documentation you link states for the `origin` argument: > origin : array\_like, shape (3,) | str > Origin of internal and external multipolar moment space in meters. […] There is nothing to change here. Mathieu ---
**Author:** ![richard](https://yyz2.discourse-cdn.com/free1/user_avatar/mne.discourse.group/richard/32/15_2.png) [@richard](https://mne.discourse.group/u/richard)\ **Post date:** [August 27, 2024, 11:50am UTC](https://mne.discourse.group/t/runtimeerror-matrix-is-badly-conditioned/9177/8 "2024-08-27T11:50:27Z")
Maybe there should be a warning if the values are unusually large ---
**Author:** ![mschendel](https://avatars.discourse-cdn.com/v4/letter/m/2acd7d/32.png) [@mschendel](https://mne.discourse.group/u/mschendel)\ **Post date:** [August 28, 2024, 4:00pm UTC](https://mne.discourse.group/t/runtimeerror-matrix-is-badly-conditioned/9177/9 "2024-08-28T16:00:47Z")
Hi all, You’re right. I read it, but I didn’t notice. It would be nice to include a hint in the error output to check head position and head origin. Thanks again for your help! Megan ---
**Author:** ![system](https://global.discourse-cdn.com/free1/uploads/mne/original/1X/85cc6bd2b69cb698a166dc6d880fb550510d0144.jpeg) [@system](https://mne.discourse.group/u/system)\ **Post date:** [September 4, 2024, 4:01pm UTC](https://mne.discourse.group/t/runtimeerror-matrix-is-badly-conditioned/9177/10 "2024-09-04T16:01:17Z")
This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.