Analysis of the Devialet Phantom I & II measurements
The Devialet Phantom speakers are small but powerful, with a unique design and excellent bass. They are expensive, but their sound quality is top-notch.I ordered a pair of Phantom speakers a few years ago (the large model in 2017). The sound was amazing, but I had some problems with the software and the linking box. The box would often disconnect the speakers, and the software was buggy. I hope these issues have been fixed since then.
The latency was also relatively high by default, but there was a way to reduce it. Ultimately, I decided to return the speakers because of the software and connectivity issues.
Recently I noticed that Devialet released a set of measurements for both speakers and wanted to look at them and compare with quasi anechoic measurements done by napilopez.
I used the data to generate the spinorama and derive some information.
Here is the marketing blurb from Devialet for the Phantom I:
and for the Phantom II:
They definitely have great marketing, let’s see if the measurements back it off.
They look sleek: Phantom I is the big brother:
Phantom I
Performance
Harmann/ Olive scores:| Speaker | Score | w/sub |
| Phantom I (reference) | 5.7 | 6.3 |
| Phantom I (eq lw) | 6.2 | 7.0 |
| Phantom I (eq score) | 6.8 | 7.6 |
This is good and at least part of the marketing is correct. For the maximum output, it is advertised for 103dB (this model, they have more powerful ones) but they do not tell at which frequency and if that’s peak or continuous, I would guess peak. The data from the GLL file are not helping to answer this.
Anechoic measurements
Let’s start with the standard CEA2034 graph:
Things are awesome below 2kHz with a flat response and an incredible 22Hz for -3dB point. Then we have a directivity error AND a significant drop (3dB) on all curves that should be audible.
Looking at the dispersion, the normalized horizontals shows the same thing: excellent below 1kHz, great below 4k and then messy.
We have a good horizontal and vertical patterns, the round head and the coaxial design are going good:
The horizontal and vertical contours shows the beaming and the weakness where the directivity changes:
contour"]https://www.spinorama.org/speakers/...vialet/vendor/SPL Vertical Contour.webp[/IMG]
Due to the beaming you will need to have your listening position in a 35° cone in order to have smooth treble.
More graphs are available on spinorama.org.
With an EQ
Let’s flatten the listening window first.
As expected the first eq linearize the listening window which is post EQ in the ±3dB band. A possibly unwelcome is that now the PIR does not go down enough and the speaker may be bright(er).The histograms show nice improvements on all curves. Note that for each curve, we have two histograms: one is full range and the other focus on the midrange. To understand if the EQ is having a positive impact, look at the EQ (orange) bars, they must be higher than the blue (reference) ones.
The following EQ will cost you 5dB of headroom:
Code:
EQ for Devialet Phantom I computed from Vendors-Devialet data
Preference Score 5.72 with EQ 6.18
Generated from http://github.com/pierreaubert/spinorama/generate_peqs.py v0.24
Dated: 2023-09-10-13:30:29
Preamp: -5.5 dB
Filter 1: ON PK Fc 22 Hz Gain +3.28 dB Q 5.22
Filter 2: ON PK Fc 76 Hz Gain -0.68 dB Q 2.43
Filter 3: ON PK Fc 109 Hz Gain +1.29 dB Q 2.02
Filter 4: ON PK Fc 447 Hz Gain -0.52 dB Q 0.57
Filter 5: ON PK Fc 1103 Hz Gain +1.54 dB Q 3.17
Filter 6: ON PK Fc 2813 Hz Gain +4.91 dB Q 5.01
Filter 7: ON PK Fc 3676 Hz Gain +3.76 dB Q 4.22Let’s flatten the predicted in-room response / score:
Code:
EQ for Devialet Phantom I computed from Vendors-Devialet data
Preference Score 5.72 with EQ 6.78
Generated from http://github.com/pierreaubert/spinorama/generate_peqs.py v0.24
Dated: 2023-09-10-13:36:59
Preamp: -4.5 dB
Filter 1: ON LS Fc 22 Hz Gain +4.16 dB Q 0.52
Filter 2: ON PK Fc 68 Hz Gain -0.57 dB Q 0.47
Filter 3: ON PK Fc 109 Hz Gain +1.64 dB Q 2.31
Filter 4: ON PK Fc 1141 Hz Gain +1.53 dB Q 3.21
Filter 5: ON PK Fc 2721 Hz Gain +4.33 dB Q 5.90
Filter 6: ON PK Fc 3676 Hz Gain +2.09 dB Q 4.87
Filter 7: ON PK Fc 12246 Hz Gain -1.62 dB Q 0.95EQ’s side by side
The 2 EQs are marginally different: they are trying to flatten the LW. Since the speaker is beaming a lot above 8kHz, the optimizer tries to compensate and give away some flatness on the LW to optimize for a flat PIR.
This EQ is able to optimise the score significantly:
Phantom II
Let’s now have a look at the smaller one.Performance
Harmann/ Olive scores:| Speaker | Score | w/sub |
| Phantom II (reference) | 7.0 | 8.1 |
| Phantom II (eq) | 7.3 | 8.3 |
This is excellent! For the maximum output, it is advertised for 95dB (this model, they have more powerful ones) but they do not tell at which frequency and if that’s peak or continuous. The data from the GLL file are not helping to answer this. That’s not a lot and it means they are perfect for near field listening or maybe mid-field aka 1 meter to 2.5 meter.
Anechoic measurements
Let’s start with the standard CEA2034 graph:
Things are awesome below 1kHz with a flat response and then they is a small (2dB) deep that should be audible. Top end is bright. The On Axis is very good, good enough to be a studio monitor with some light corrections. The directivity doesn't have significant error.
Looking at the dispersion, the normalized horizontals shows the same thing: excellent below 1kHz, great below 4k and then messy.
We have a good horizontal and vertical patterns, the round head and the coaxial design are going good:
The horizontal and vertical contours shows the beaming and the weakness where the directivity changes:
contour"]https://www.spinorama.org/speakers/...vialet/vendor/SPL Vertical Contour.webp[/IMG]
If you look closely at the contrours, it is possible that the measurements are swapped out.
Due to the beaming you will need to have your listening position in a ±35° cone in order to have smooth treble.
More graphs are available on spinorama.org.
With an EQ
Since the Listening Window is already very good, let’s focus on the Predicted In-Room Response:Let’s flatten the predicted in-room response / score:
Now that’s awesome:
Code:
EQ for Devialet Phantom II computed from Vendors-Devialet data
Preference Score 7.06 with EQ 7.35
Generated from http://github.com/pierreaubert/spinorama/generate_peqs.py v0.24
Dated: 2023-09-10-14:03:17
Preamp: -2.5 dB
Filter 1: ON LS Fc 30 Hz Gain +2.87 dB Q 0.90
Filter 2: ON PK Fc 62 Hz Gain -0.66 dB Q 3.73
Filter 3: ON PK Fc 76 Hz Gain -1.28 dB Q 2.17
Filter 4: ON PK Fc 342 Hz Gain -0.71 dB Q 0.41
Filter 5: ON PK Fc 1220 Hz Gain +0.32 dB Q 4.56
Filter 6: ON PK Fc 1594 Hz Gain +2.18 dB Q 2.37
Filter 7: ON PK Fc 7669 Hz Gain +1.88 dB Q 0.37Comparison with quasi anechoic measurement
As often, we see a very good concordance between the two methodologies:
Comparison between Phantom I and Phantom II
Phantom II is significantly flatter than the I. They both have relative minor issues in the 1kHz to 6kHz range. It is likely audible on the I and difficult to hear on the II.
Conclusion
Harmann/ Olive scores:| Speaker | Score | w/sub |
| Phantom I (reference) | 5.7 | 6.5 |
| Phantom I (eq lw) | 6.2 | 7.0 |
| Phantom I (eq score) | 6.8 | 7.6 |
| Phantom II (reference) | 7.0 | 8.1 |
| Phantom II (eq) | 7.3 | 8.3 |
The Devialet Phantom speakers are a mixed bag with a hint of greatness. Objectively, they are very good speakers, with impressive output and bass. However, some people may find the top end to be too harsh and the dip in the frequency response may be audible on the Phantom I. A digital signal processor (DSP) can be used to tame the top end.
The Phantom II looks excellent without EQ and awesome with it. For comparison, that’s the same level of quality as a KEF R5.
It is limited in terms of output but it would be perfect for use on a desk. It can almost compete with the best in class high-end studio monitors like the Genelec 8341 or Neumann KH120 II.
The Phantom speakers are not cheap, but they are unique and stylish speakers that may be easier to integrate into modern design houses.
