# Hifiman HE400i Review (planar headphone)

#### pkane

##### Major Contributor
Forum Donor
But an average *is* a low pass filter. Look up any text on DSP: https://www.researchgate.net/publication/299485551_Average_Filtering_Theory_Design_and_Implementation#:~:text=A special implementation of a,that carrying high frequency distortion.

"A special implementation of a low pass algorithm is the averaging filter. It calculates the output sample using the average from a finite number of input samples. The averaging filter is used in situations where is necessary to smooth data that carrying high frequency distortion. "

Basically, an average is a crappy low pass filter in that you have no control over its cut off frequency or its strength. It gets used because it is so fast and easy to compute.

Let's test that. In my measurements I am showing the left and right response. People often average those two and build a filter based on that. Problem with that scheme is that the filter is no longer match either channel now.

Harman did a study of their Room EQ where they tested optimizing for one location versus a few locations around the seating position. The one optimized for one location won in listening test versus weighted averaged of multiple locations.

In my amplifier SINAD rating bar graphs I compute the geometric mean. That way, you have a real value of an amplifier that has as many products below it, than it has above it. Make an average and it doesn't stand for any and can be easily skewed by a single very high or very low value. This is actually a problem when I was testing headphones with B&K 5128. It is common to get a single outlier and with it, generate really bad "average." It is best to use GeoMean or discard that sample.

Anyway, we really don't need any of these arguments. Basic intuition shows that the two graphs that were show are essentially the same and could not come from two different designs:

If I breath on my fixture I can get one or the other. No way any kind of crude signal processing like averaging bail you out of the inaccuracy of the measurements. And the small sample size to boot.

Inaccuracy of measurement due to random variables can certainly be improved by averaging. No need to argue here, but, again, S/N of any signal with noise can be improved by averaging and has nothing to do with headphone measurements but with noise statistics.

#### Robbo99999

##### Major Contributor
Forum Donor
You have no idea what the pdf of these systems are. You have no idea how exacting the measurements are. You have no idea what the variance is. Averaging things doesn't bail you out of any of this. Remember, I am measuring devices with this fixture. I am telling you it doesn't remotely produce data to that resolution. If you want to believe otherwise, fine. I don't have the motivation to keep explaining it.
That's fine we can disagree, I still stand by my points made in my last couple of posts.

#### DualTriode

##### Addicted to Fun and Learning
The software tools we are speaking of here use FFT. With averaging noise variation is canceled out, or in other words is filtered , distortion not being random is not filtered and is more apparent due to averaging the noise. Real time THD+N is unchanged. S/N remains the same

#### pkane

##### Major Contributor
Forum Donor
The software tools we are speaking of here use FFT. With averaging noise variation is canceled out, or in other words is filtered , distortion not being random is not filtered and is more apparent due to averaging the noise. Real time THD+N is unchanged. S/N remains the same

You misunderstood the nature of noise in my message. It had nothing to do with sound.

#### DualTriode

##### Addicted to Fun and Learning
You misunderstood the nature of noise in my message. It had nothing to do with sound.

You do not understand noise.

#### dazedandconfused

##### Member
I just registered here to thank you Amir. That EQ has just brought my HE400i back into use. I've mostly been using my modified T50RP, SHP9500's (with SRH1840 pads) and DT770's as the HE400i was a bit too bright. I tried an EQ from Headfi (using 13 bands) that was an improvement but I found some female vocals a bit strident still. Your EQ has fixed that. I still may buy some Sundara's later this year, perhaps you might have reviewed them by then .

OP

#### amirm

Staff Member
CFO (Chief Fun Officer)
Welcome to the forum. Thanks for putting a smile on my face that my reviews/EQ are doing some good. I must say, I personally am so surprised every time I eq a headphone and realize how much better they can sound. And how much easier it is to do that than with speakers.

#### Nichdroid

##### Member
Welcome to the forum. Thanks for putting a smile on my face that my reviews/EQ are doing some good. I must say, I personally am so surprised every time I eq a headphone and realize how much better they can sound. And how much easier it is to do that than with speakers.
Same here, I've tried a couple of profiles, including Oratory's ones, but apart from your's none of them satisfied me; some sounded to much bass boosted, some other with strangeness in treble...

#### bobbooo

##### Major Contributor
That doesn't seem quite right to me. I would have thought that if you did a number of different measurements on the same headphone and took an average of them, then I would think you'd be able to resolve differences between headphones of the difference we see between the HE400i 2020 version and the HE4XX.....I'm fairly certain they're not just the same model of headphone underneath, so you'd expect to see differences between them and by extension I think the trend of difference between them would be that which Oratory has pictured & which bobbooo showed in his post:
View attachment 102118

For instance, this is a slew of measurements from Oratory on HD600:
View attachment 102119
The variation between the different measurements in the above graph is pretty stable up to 4kHz, and the average is gonna be pretty representative over that range. The main differences between the 400i 2020 version and the 4XX as seen in the first graph in the post is happening between 500Hz and 5kHz (most of which is the highest area of accuracy on the graph) so certainly over most of that range I think it's possible to resolve an average 1-2dB "real" difference between these two headphone models with no issue. If you rely on just one measurment, then no you can't resolve the difference with any confidence, but with enough measurements I think there's validity in the difference seen.

Totally correct. There are differences in the frequency response of the original HE400i, HE400i (2020 version) and HE4XX that are greater than measurement variation. Let's take another look at Oratory's measurements of all three in detail:

That's a difference in the sub-bass of 5.7 dB between the 4XX and original 400i (2016), and 2.5 dB between the 4XX and 400i (2020).

Here at 2 kHz (around where our ears are most sensitive), we have a difference of 3.1 dB between the 4XX and 400i (2020).

And we have a difference of 4.5 dB between the 4XX and original 400i around 5.6 kHz.

Now let's take a look at Crinacle's measurements (right channel, 3 reseat measurements, on a similar GRAS rig):

So at 2 kHz that's a measurement deviation around the mean of ±0.5 dB for the original 400i and ±0.2 dB for the 2020 version.

And a measurement deviation at 5.6 kHz of ±0.8 dB for the original 400i and ±0.6 dB for the 2020 version.

Now let's look at Rtings' frequency response consistency measurements of the original 400i:

At 2 kHz we have a measurement deviation around the mean of ±0.3 dB.

And at 5.6 kHz we have a measurement deviation of around ±0.5 dB.

These figures are in general agreement with Crinacle's. Rtings calculate an average deviation of 0.38 dB between 20 Hz and 10 kHz, among the top 10 of all open-back headphones they've tested, and say this about the frequency response consistency:
The HiFiMan HE-400i have a great frequency response consistency...The treble delivery is also very consistent across multiple re-seats, with the maximum deviation below 10KHz being less than 2dB.

All the above measurement deviations are less than the measured differences between the three HE4** versions from Oratory's measurements at the beginning of this post. These differences are indeed real, and not within measurement variance.

Similarly for unit variation, Oratory said this about the HE4XX (this was well over a year ago, so he may have measured even more by now, my emphasis):
I measured 3 specimens by now, including yours and one that Olaf (from OluvsGadgets) brought to the lab.
I built an average from all of them and use this for the EQ PDFs. They are all pretty close together actually, so much so that it doesn't make much sense to create an EQ setting specifically for one specimen only. Unit variation was excellently low on the three specimens I tested. Quite surprised by this!

Let's compare these headphones to a few other Hifiman models:

As the graph shows, at any one frequency up to 10 kHz, none of these headphones have responses that differ from the HE4XX by more than the HE400i (2016 or 2020 version). If the latter three (or just first and last) are the same within measurement variance, does that mean all these headphones are also the same within measurement variance? No, that would be absurd. Hifiman obviously have a 'house sound', but there are definite sonic differences between these headphones. In the bass, the HE400i models are the outliers from the median response with the worst bass extension, and at around 500-600 Hz too where they exhibit significant resonances that the others, including the HE4XX, don't. Between 1 and 2 kHz the HE4XX is the outlier, on average closer to the Harman target than any of the others. Between 2.5 and 5 kHz, the HE400i (2020) is the outlier, furthest from the target. And around 5.6 kHz the HE400i (2016) is the outlier, furthest from the target with a deep notch there. These are all real differences, not within measurement variance. The HE4XX's frequency response not only follows the Harman target the closest on average among these 5 headphones, but of all the Hifiman models Oratory's measured, which is evidenced by it having the highest calculated preference rating among them of 88/100. According to the best objective, scientific model we have, the HE4XX is Hifiman's best-sounding headphone. (It also happens to be one of their cheapest.)

Now let's compare the above frequency response differences to these more familiar headphones to many from Sennheiser:

These measurements are much closer to each other than any of the HE4XX, HE400i (2020), or HE400i (2016) are to each other, with only 1-3 dB difference in the upper mids / treble. Does that tell you that they are the same headphone? No. And that would still be a no even if the differences are within measurement variation. Why? Because this measurement variation is not just random error from the measuring equipment. The variation is intentional and desired, in order to simulate and account for actual statistical variation in real-world use of the headphones by the consumer, be that unit, positional, or channel variation, which are all averaged to account for this real-world variation. In other words, this measurement variation is a feature not a bug! It may be the case that one channel of one particular HD600 unit at one particular seating position might measure the same as one channel of one particular HD650 unit at another particular seating position, but statistically, on average, the measurements above for each headphone will be closest to what you will perceive listening to two channels in amalgam, with your particular unit, placed (for most somewhat randomly) on your head, and that's what matters to the listener. If you want more treble, then on average you will get this choosing the HD600 over the HD650, just like on average, you will get a headphone that more closely follows the Harman target if you choose the HE4XX over either of the HE400i models, as shown by industry standard, professionally made measurements.

But wait, we don't have to just stare at graphs to show these headphones are indeed different. There is at least one very obvious physical difference between them - the pads.

HE4XX's Focus A pads on the left, HE400i (2020)'s Focus pads on the right:

And the original HE400i's pads (which at least on the outside look the same as the above Focus pads from the 2020 version):

Here's what Oratory (a professional acoustic engineer of course who helps design headphones as part of his job) says about pads (my emphasis):
Pads certainly affect the sound a lot more than people would often think. I'd go as far as to say they're as important as the actual driver.
And it's not just visible material differences on the outside that can have a big effect. On pads' acoustic influence, he goes on to say:
That depends on more than just the material itself.

You can make well sealing earpads out of velours, for example - if the core material is closed-cell.
You can also make leaking earpads out of leather (perforated leather being the most obvious example, but also leakage in the stitching and bonding area).

The effect of earpads is rather tricky to predict. Personally I'll say it's even trickier than the effect of drivers.

The perforated leather inner edge of the HE400i models' pads could then possibly be the (partial) cause of their weaker bass extension to the HE4XX, the latter's pads featuring tightly woven material in its place which may provide a better seal. Other differences may explain the weaker still bass of the original HE400i to the 2020 version (e.g. differing internal pad or earcup material/structure, driver differences etc.). Of course as Oratory says though, pad effects are tricky to predict so the obvious physical differences between the Focus and Focus A pads of the HE400i and HE4XX respectively could have any number of acoustic consequences for the headphones.

So what's the lesson here? Don't be fooled into thinking two headphones are sonically the same within measurement error just based on somewhat similar frequency response, a photo of the back of the driver and the same initial number of their model name. It's both unfair to manufacturers and misinformative to consumers to entertain such rumors founded on zero concrete evidence. The default position should be innocent until proven guilty i.e. headphones sold as different models are indeed sonically different, unless strong evidence suggests otherwise. Hopefully this finally puts an end to the persistent baseless conjecture that the HE4XX are the same as the HE400i (either version). There are both physically visible (e.g. pad), and measurable acoustic differences between them, that exceed measurement variation. Oh, and a belated Happy New Year to everyone This one can't be worse than the last, right? Right??

Last edited:

#### Robbo99999

##### Major Contributor
Forum Donor
@bobbooo , nice analysis, compelling, I agree!

#### Maiky76

##### Senior Member
I'm still working on the preference rating, as I'm getting 77 (Oratory got 80 for 2016; I'll digitize his data and see what that gets me).
EDIT: Digitizing his data also gets me 77 (and the 2 are more similar than the HD800S, which Amir’s had a more substantial 10kHz dip).

EDIT: Using the same 1/12 octave smoothing, I get 80.

Here are some thoughts about the EQ.
Data provided by @amirm, I just don't understand why the highest frequency is 18775.xxxHz.

• The average L/R is used to calculate the score.
• The resolution is the correct 12 points per octave interpolated from the raw data
• A Genetic Algorithm is used to optimize the EQ.
• The starting point to to fit the Harman curve: EQ Fit
• The Score is then added in the Optimizer. The EQ Score is designed to MAXIMIZE the Score WHILE fitting the Harman target curve with a fixed complexity.
In the general case they will be rather similar but not identical, 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.
Typical LF response for this type of transducer.
Not of effort to EQ it: LF boost and a bit of mid boost.

I have generated two EQs, the APO config files are attached.
Score no EQ: 80.1
Score with EQ Fit: 88.4
Score with EQ Score: 90.4

Code:
``````Hifiman HE400i APO EQ Fit 96000Hz
January072021-142528

Preamp: -7.6 dB

Filter 1: ON PK Fc 24 Hz Gain 6.64 dB Q 0.37
Filter 2: ON PK Fc 1834 Hz Gain 7.87 dB Q 1.4
Filter 3: ON PK Fc 3210 Hz Gain -2.07 dB Q 1.76
Filter 4: ON PK Fc 5298 Hz Gain 3 dB Q 4.54
Filter 5: ON PK Fc 8498 Hz Gain -6.31 dB Q 9.5

Hifiman HE400i APO EQ Score 96000Hz
January072021-142528

Preamp: -7.6 dB

Filter 1: ON PK Fc 23 Hz Gain 6.4 dB Q 0.37
Filter 2: ON PK Fc 1834 Hz Gain 7.88 dB Q 1.4
Filter 3: ON PK Fc 3199 Hz Gain -2.1 dB Q 1.76
Filter 4: ON PK Fc 5297 Hz Gain 3 dB Q 3.73
Filter 5: ON PK Fc 8412 Hz Gain -6.81 dB Q 12.3``````

#### Attachments

• Hifiman HE400i APO EQ Score 96000Hz.txt
304 bytes · Views: 54
• Hifiman HE400i APO EQ Fit 96000Hz.txt
303 bytes · Views: 47
OP

#### amirm

Staff Member
CFO (Chief Fun Officer)
Data provided by @amirm, I just don't understand why the highest frequency is 18775.xxxHz.
Stuff above 10 kHz is not reliable let alone anything above 18 kHz. I will look to see if it didn't go to 20 kHz for the sake of it.

#### Robbo99999

##### Major Contributor
Forum Donor
Here are some thoughts about the EQ.
Data provided by @amirm, I just don't understand why the highest frequency is 18775.xxxHz.

• The average L/R is used to calculate the score.
• The resolution is the correct 12 points per octave interpolated from the raw data
• A Genetic Algorithm is used to optimize the EQ.
• The starting point to to fit the Harman curve: EQ Fit
• The Score is then added in the Optimizer. The EQ Score is designed to MAXIMIZE the Score WHILE fitting the Harman target curve with a fixed complexity.
In the general case they will be rather similar but not identical, 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.
Typical LF response for this type of transducer.
Not of effort to EQ it: LF boost and a bit of mid boost.

I have generated two EQs, the APO config files are attached.
Score no EQ: 80.1
Score with EQ Fit: 88.4
Score with EQ Score: 90.4

Code:
``````Hifiman HE400i APO EQ Fit 96000Hz
January072021-142528

Preamp: -7.6 dB

Filter 1: ON PK Fc 24 Hz Gain 6.64 dB Q 0.37
Filter 2: ON PK Fc 1834 Hz Gain 7.87 dB Q 1.4
Filter 3: ON PK Fc 3210 Hz Gain -2.07 dB Q 1.76
Filter 4: ON PK Fc 5298 Hz Gain 3 dB Q 4.54
Filter 5: ON PK Fc 8498 Hz Gain -6.31 dB Q 9.5

Hifiman HE400i APO EQ Score 96000Hz
January072021-142528

Preamp: -7.6 dB

Filter 1: ON PK Fc 23 Hz Gain 6.4 dB Q 0.37
Filter 2: ON PK Fc 1834 Hz Gain 7.88 dB Q 1.4
Filter 3: ON PK Fc 3199 Hz Gain -2.1 dB Q 1.76
Filter 4: ON PK Fc 5297 Hz Gain 3 dB Q 3.73
Filter 5: ON PK Fc 8412 Hz Gain -6.81 dB Q 12.3``````

View attachment 104296
Nice, interesting. I noticed you used a high Q filter of around 10 to 12 Q on your fifth filter at the sharp peak around 8400Hz. I think it would be better to use a lower Q filter this high up the frequency range (& just generally also) due to the individual headphone seating & variance from person to person that would make high Q narrow filters non-relevant and potentially damaging to the sound. I think you could have just used something like Q2 (or Q3) placed directly on that peak thus giving yourself more chance of actually catching that peak (as the peak might move from person to person or from one headphone seating to the next)......also using Q2 on that peak would be fine for the frequency response because it doesn't matter if the dip at 10kHz goes any lower, and as a side effect it will also lower the little raised area just to the left of that peak.

#### watchnerd

##### Grand Contributor
@amirm I assume the test results are with the original earpads?

I replaced mine because my ears got too hot with the originals, and I'm wondering what impact that might have on the EQ.

#### keebz28

##### Addicted to Fun and Learning
Forum Donor
@amirm I've been scouring and reading all the measurements on these for a while and I am finally buying the 2020 version. I didnt find any changes to the design of the drivers on the 2020 version from what i could find. Definitely going to use EQ profile in SoundSource software for these when i get them. One question i do have is, is a balanced source a necessity to bring out the best in these? I have Topping E30 paired to a Sabaj PHA3 tube amp but it does not have a balanced out.

#### Robbo99999

##### Major Contributor
Forum Donor
@amirm I assume the test results are with the original earpads?

I replaced mine because my ears got too hot with the originals, and I'm wondering what impact that might have on the EQ.
I think it could make a massive difference to how the headphone would measure, and hence whether the EQ would be valid or viable. Generally, if you're replacing the pads with aftermarket pads that are not of the genuine original spec then you can assume the EQ is invalid.

#### watchnerd

##### Grand Contributor
I think it could make a massive difference to how the headphone would measure, and hence whether the EQ would be valid or viable. Generally, if you're replacing the pads with aftermarket pads that are not of the genuine original spec then you can assume the EQ is invalid.

Which the description says:

"Package contains two pieces of earpads. They are SUNDARA's stock earpads and also compatible with all HE-series headphones (including HE400S, 400i, 300, 400, 500, 4, 5, 6, 5LE)."

But I agree they can still have a massive impact on the sound.

#### Robbo99999

##### Major Contributor
Forum Donor

Which the description says:

"Package contains two pieces of earpads. They are SUNDARA's stock earpads and also compatible with all HE-series headphones (including HE400S, 400i, 300, 400, 500, 4, 5, 6, 5LE)."

But I agree they can still have a massive impact on the sound.
Yeah, sure, what I meant is that if they are a different design of earpad to the ones measured then it will effect the frequency response, and then the EQ is not valid anymore....to a lesser or greater degree depending what & how much is changed within the response, but you don't really know until you measure it with the different earpads.....but my approach would be to assume the EQ is invalidated by using different earpads.

#### watchnerd

##### Grand Contributor
Yeah, sure, what I meant is that if they are a different design of earpad to the ones measured then it will effect the frequency response, and then the EQ is not valid anymore....to a lesser or greater degree depending what & how much is changed within the response, but you don't really know until you measure it with the different earpads.....but my approach would be to assume the EQ is invalidated by using different earpads.

Yes, of course.

Which is the whole reason why I asked Amir if the measurements used the original earpads.

I thought that was implied in what I wrote in the question.

Replies
117
Views
21K
Replies
199
Views
26K
Replies
28
Views
9K
Replies
37
Views
10K
Replies
17
Views
3K