• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Sivga Phoenix Review (headphone)

flipflop

Addicted to Fun and Learning
Joined
Feb 22, 2018
Messages
927
Likes
1,240
@MediumRare @bobbooo
Let's not forget about
Relevant quote:
The use of a reference band of 200-400 Hz is based on an observation made in Part One (see section 4.8 of [1]). When asked to judge the spectral balance of each loudspeaker across 6 frequency bands, listeners referenced or anchored their judgments to the band centered around 200 Hz. One plausible explanation is that many of the fundamentals of instruments, including voice, fall within 200-400 Hz, and the levels of the higher harmonics are referenced to it.

So if the target had to be perceptually correct, it would need to be even higher, putting 90%+ of these headphones' response below it.
 

bobbooo

Major Contributor
Joined
Aug 30, 2019
Messages
1,479
Likes
2,079
So if the target had to be perceptually correct, it would need to be even higher, putting 90%+ of these headphones' response below it.

Sounds like a classic reductio ad absurdum to me ;)
Due to the completely different target shapes involved, I don't think you can just translate one-to-one from loudspeakers to headphones like that.
 

MediumRare

Major Contributor
Forum Donor
Joined
Sep 17, 2019
Messages
1,949
Likes
2,275
Location
Chicago
The EQ and sound signature will be the same. Amir references at 400Hz looks like on the left channel. Had Amir referenced to say 1kHz the 300Hz would appear higher but the only difference would be the position of the 0dB line.
Well, the sound signature is inherent in the physical device, but I disagree about the EQ, which is the practical effect of my comment. Note, Amir created several boosts and, while there was an improvement, mentioned a different approach may have worked better. An alternative would be, first, to cut the upper bass, which would have several advantages, including less distortion.
 

MediumRare

Major Contributor
Forum Donor
Joined
Sep 17, 2019
Messages
1,949
Likes
2,275
Location
Chicago
Relevant quote:
....
So if the target had to be perceptually correct, it would need to be even higher, putting 90%+ of these headphones' response below it.
Ah, but don't forget "and higher harmonics". The human voice is centered in the range 200 Hz to 1,000 Hz, same for the trumpet, violin, and to a large degree, guitar.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,392
Location
Seattle Area
Well, the sound signature is inherent in the physical device, but I disagree about the EQ, which is the practical effect of my comment. Note, Amir created several boosts and, while there was an improvement, mentioned a different approach may have worked better. An alternative would be, first, to cut the upper bass, which would have several advantages, including less distortion.
If you did that they would be turning the volume up and increase distortion across the board to get the same loudness. This is a zero sum game.
 

MediumRare

Major Contributor
Forum Donor
Joined
Sep 17, 2019
Messages
1,949
Likes
2,275
Location
Chicago
If you did that they would be turning the volume up and increase distortion across the board to get the same loudness. This is a zero sum game.
I might not be communicating clearly. To get closer to the target EQ I’d be lowering a narrow range instead of boosting more extreme bass and treble. Isn’t that theoretically the better way to get to flat?
 

Maiky76

Senior Member
Joined
May 28, 2020
Messages
440
Likes
3,706
Location
French, living in China
This is a review and detailed measurements of the Sivga Phoenix open-back headphone. It was sent to me by the company and costs US $255 on Amazon including Prime shipping.

I must say, this is one attractive looking headphone!
View attachment 126765

The padding and soft surfaces feel luxurious as well. Reminds me of the inside of an Audi automobile!

I absolutely love the fact that they indicate the channel designation both in large letters inside and on the sides:

View attachment 126764

On some headphones it literally takes me a minute or two to find the darn designation.

The headphone is also fairly light at just 300 grams. Cups are round as you see with a diameter of 58 mm. The depth is variable and at the shallowest end is 17 mm which is not a lot. If you have your ears poking out, it may touch dust cover.

The measurements you are about to see are made using a standardized Gras 45C. I searched for any and all measurements I could find online. Alas while a number of them are close to mine, none are using the exact fixture down to coupler and pinna. As you will see, I have confirmed the approximate accuracy of the measurements using Equalization and listening tests. Ultimately headphone measurements are less exact than speakers above a few kilohertz so keep that in mind as you read these tests.

Getting a good fit was super easy with both channels matching each other at my two targets on the first install.

Sivgo Phoenix Measurements
There is nothing more important than frequency response of a headphone as each is seemingly different and that difference leaves a very distinct character:
View attachment 126766

We almost have a match for our target between 100 and 500 Hz but above and below that range we have a shortfall. The higher frequency loss will likely translate into a dull sounding headphone. There is however enough bass for it to not sound light in that department.

Relative chart may be easier for some to understand even though it conveys the same information as above:
View attachment 126767

Basic emphasis here then is the region between 200 and 300 Hz.

Relative distortion shows what we see from typical headphones in bass but there is also some disturbance in 3 to 6 kHz:

View attachment 126768

Fortunately since we have good bit of upper bass energy, we won't be boosting that so what we see is what we get:

View attachment 126769

Group delay is not revealing:

View attachment 126773

Impedance is variable but low which means typical portable devices should be able to drive it:
View attachment 126774

That is especially so since the Phoenix is the most sensitive headphone I have measured so far:

View attachment 126775

Sivga Phoenix Listening Tests
First impression was not positive: the sound is dull with most emphasis in lower frequencies. I had to pull up the parametric EQ right away to get it to be more correct. This is a quick and dirty attempt at fixing the gross problems:

View attachment 126776

Broad corrections will probably benefit from mathematical correction than my manual one. Still, the improvement was dramatic in both balanced sound and better spatial imaging. Female vocals that were lost before now came out front and center as they should.

Conclusions
The measured performance of Phoenix headphone shows broad deviations from out preference target. Subjective listening tests confirmed the same resulting in strong amount of EQ. The outcome was positive but not super exciting.

While I like the look and feel of the Sivga Phoenix, the overall performance is not good enough for me to recommend it. If you own it, I highly recommend using equalization to get better fidelity.

-----------
As always, questions, comments, recommendations, etc. are welcome.

Any donations are much appreciated using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/
Hi,

Here are some thoughts about the EQ.

Notes about the EQ design:
  • The average L/R is used to calculate the score.
  • The resolution is 12 points per octave interpolated from the raw data (provided by @amirm)
  • A Genetic Algorithm is used to optimize the EQ.
  • The EQ Score is designed to MAXIMIZE the Score WHILE fitting the Harman target curve with a fixed complexity.
    This will avoid weird results if one only optimizes for the Score.
    It will probably flatten the Error regression doing so, the tonal balance should be more neutral.
  • The EQs are starting point and may require tuning (certainly at LF).
  • The range above 10kHz is usually not EQed unless smooth enough to do so.
  • I am using PEQ (PK) as from my experience the definition is more consistent across different DSP/platform implementations than shelves.
  • With some HP/amp combo the boosts and preamp gain need to be carefully considered to avoid issues
Good L/R match.

I have generated one EQ, the APO config file is attached.

Score no EQ: 67.6
Score Armirm: 76.5
Score with EQ: 96.6

Code:
Sivga Phoenix APO EQ Score Flat@HF 96000Hz
April292021-102224

Preamp: -4.9 dB

Filter 1: ON PK Fc 265 Hz Gain -4.9 dB Q 0.59
Filter 2: ON PK Fc 1463 Hz Gain 1.92 dB Q 0.56
Filter 3: ON PK Fc 2516 Hz Gain -1.5 dB Q 7.35
Filter 4: ON PK Fc 3268 Hz Gain 4.7 dB Q 3.35
Filter 5: ON PK Fc 4438 Hz Gain -4.55 dB Q 5.5
Filter 6: ON PK Fc 8996 Hz Gain -3.4 dB Q 8
Filter 7: ON PK Fc 12582 Hz Gain -5.3 dB Q 8

Sivga Phoenix Dashboard.png
 

Attachments

  • Sivga Phoenix APO EQ Score Flat@HF 96000Hz.txt
    402 bytes · Views: 95

Francis Vaughan

Addicted to Fun and Learning
Forum Donor
Joined
Dec 6, 2018
Messages
933
Likes
4,697
Location
Adelaide Australia
What if the total SPL is used - or even just by eye - to line up curves a closely as possible? In this case about 4 dB. Then we’ll see the EQ problem is quite different: too much upper bass, the rest is actually pretty good.

As Amir notes above, it is a zero sum game. No matter what you do with offsetting the reference, the final EQ will be the same, although you might get there with a different combination of parameters.
IMHO drawing the thick reference line in the difference to reference graph is just visual clutter. It adds nothing to the information presented and just confuses people. The fact that the reference line is chosen to intersect the data at a specific frequency just adds to the confusion. Nothing in the location of the reference line has anything to do with either the perceived sound or the final effective EQ chosen. The point of EQ is to get the line flat, not to get it to match an arbitrary reference line. Once it is flat, it will match the reference line by definition.
 
Last edited:

aldarrin

Member
Joined
Jun 13, 2020
Messages
79
Likes
106
??? The deficiencies are not from flat. The target curve is variable. The bit that was boosted in upper bass is something I actually ignored since it is very little above target.

I think he's talking about normalizing at a different frequency. Not sure what tools you're using, but in Crinacle's database it looks like this:
1619664200926.png


I wonder if you could minimize overall distortion by normalizing where it's highest. In this case it'd be around 50 Hz. That way you won't be boosting regions where the headphones produce the most distortion.
 

Oddeeo

Member
Joined
Mar 28, 2021
Messages
16
Likes
11
Location
Boston
Luxury feel like the interior of an Audi?

Lets hope the kind of muppets who aggressively, selfishly and carelessly drive new model Audi's around these parts don't go looking for heaphones then ;)
Ah, I found the cyclist! ;)
 

Robbo99999

Master Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
6,874
Likes
6,672
Location
UK
Hi,

Here are some thoughts about the EQ.

Notes about the EQ design:
  • The average L/R is used to calculate the score.
  • The resolution is 12 points per octave interpolated from the raw data (provided by @amirm)
  • A Genetic Algorithm is used to optimize the EQ.
  • The EQ Score is designed to MAXIMIZE the Score WHILE fitting the Harman target curve with a fixed complexity.
    This will avoid weird results if one only optimizes for the Score.
    It will probably flatten the Error regression doing so, the tonal balance should be more neutral.
  • The EQs are starting point and may require tuning (certainly at LF).
  • The range above 10kHz is usually not EQed unless smooth enough to do so.
  • I am using PEQ (PK) as from my experience the definition is more consistent across different DSP/platform implementations than shelves.
  • With some HP/amp combo the boosts and preamp gain need to be carefully considered to avoid issues
Good L/R match.

I have generated one EQ, the APO config file is attached.

Score no EQ: 67.6
Score Armirm: 76.5
Score with EQ: 96.6

Code:
Sivga Phoenix APO EQ Score Flat@HF 96000Hz
April292021-102224

Preamp: -4.9 dB

Filter 1: ON PK Fc 265 Hz Gain -4.9 dB Q 0.59
Filter 2: ON PK Fc 1463 Hz Gain 1.92 dB Q 0.56
Filter 3: ON PK Fc 2516 Hz Gain -1.5 dB Q 7.35
Filter 4: ON PK Fc 3268 Hz Gain 4.7 dB Q 3.35
Filter 5: ON PK Fc 4438 Hz Gain -4.55 dB Q 5.5
Filter 6: ON PK Fc 8996 Hz Gain -3.4 dB Q 8
Filter 7: ON PK Fc 12582 Hz Gain -5.3 dB Q 8

View attachment 126961
I do like your EQ's Maiky, for the service you are giving everyone, but I wouldn't use a high Q filter like Q8 high up the frequency range like you're doing at 8996Hz and 12582Hz (Filter 6&7).....the reason for this is that the response is unpredictable the higher up the frequency range you go due to anatomical variations person to person and also the variance associated with placing your headphone on your head in even a minutely different position.....so those high Q filters you devised are very likely to miss the peaks for that person and instead do more harm than good. If you want to target those two peaks I think you should use around Q4 (or lower) on your 8996Hz peak, and then above 10kHz use low Q filters below Q2 or more sensibly just use High Shelf Filters to "balance" the response above 10kHz.
 

MediumRare

Major Contributor
Forum Donor
Joined
Sep 17, 2019
Messages
1,949
Likes
2,275
Location
Chicago
??? The deficiencies are not from flat. The target curve is variable. The bit that was boosted in upper bass is something I actually ignored since it is very little above target.
I meant flat versus target.
 

dfuller

Major Contributor
Joined
Apr 26, 2020
Messages
3,338
Likes
5,054
Is this a planar or a dynamic driver? The peaking distortion + frequency response leads me toward planar but the impedance having something of a peak in the mid-bass makes me think dynamic...

Update: these are dynamic, which makes that 4k distortion spike even worse. I sort of expect that from planars at some level but good dynamic drivers don't have that problem.
 
Last edited:

Palfim

Member
Joined
Feb 3, 2021
Messages
30
Likes
16
AHH the sivga phoenix! On the outside it's one of the most beautiful headphones, plus they make alot advertisements on Amazon, always wonderd how they really sound.would have been too good to be true I guess, but sexyness sells ,haha!

So thanks for the review, very informative, was hoping this would pop up sometime. With so many reviews, it takes most of your time up I guess?
 
Top Bottom