Kali Audio IN-8 Studio Monitor Review

[tuga]

Sure it is as most experienced acousticians don't recommend filling narrow bass dips fully.

Is there a threshold frequency above which you can start filling dips?
How much should you boost?

Currently I'm only trimming the top of the peaks below 100Hz.

 

[thewas]

Sounds good, often the psychoacoustic smoothing of REW or Acourate gives a good control check (but not good to design the filters themselves) of how much correction makes sense, it also depends on the width of the dips and their root cause.

 

[QMuse]

Is there a threshold frequency above which you can start filling dips?
How much should you boost?

Currently I'm only trimming the top of the peaks below 100Hz.

Peaks should be trimmed throughout the entire range as all of them are some kind of resonance.

Narrow dips in LF range (20-300Hz) coming from room modes should be left untouched, wider dips can be filled up to 5-6dB.

In 300-800Hz range I'm using the same technique as in the 20-300Hz range but with smaller gain and lower Q. Carefull and precise measurement is needed in this range as this is where room and speakers are equally affecting measured response.

Above 800Hz it's about speaker's correction and adjusting tonal balance with very low Q filters.

 

[musictracer]

FYI Kali Audio received the long waited 3d party Frequency/Distortion analysis plots just yesterday.
They are in the process of publishing them in their web site, however they already shared an image in their FB page to get an idea of how they look. I guess it has been quite smoothed (probably 1/6 smoothing) but they look quite balanced, apart from the dip in the highs we are all aware of. Any thoughts?

 

[detlev24]

Not sure why they only mention the 2nd and 3rd harmonic distortion, instead of THD [which would consist of H2-H9]...

It also seems the measurement was taken at 85 dB SPL - their reference level. I do not understand the right % numbers then: at height of 85 dB it should read ~100% equivalent [the solid black line would be 100%], instead of something below 4%. Probably, they took two different graphs and threw them over each other with no meaningful correlation of the dB vs. % numbers.

To me, the applied smoothing looks a bit more like 1/3, but we will not know without further confirmation.

 

[anmpr1]

Not sure why they only mention the 2nd and 3rd harmonic distortion, instead of THD [which would consist of H2-H9]... It also seems the measurement was taken at 85 dB SPL - their reference level. I do not understand the right % numbers then: at height of 85 dB it should read ~100% equivalent [the solid black line would be 100%], instead of something below 4%. Probably, they took two different graphs and threw them over each other with no meaningful correlation of the dB vs. % numbers.

FWIW their spec sheet shows 85dB as the loudness for a "listening distance at 2.8 meters continuous SPL with 20dB dynamic headroom". So they are claiming 105dB SPL, but don't list a distortion value for that loudness. They do list THD at 90dB, which ranges from 0.8% to 1.4% depending on frequency.

I still say that if one is interested in this sort of product they should borrow it for a couple of weeks from their local guitar store, and if it doesn't meet their needs turn it in for something else. Loudspeaker specs/measurements are all over the place anyway.

 

[JohnYang1997]

Not sure why they only mention the 2nd and 3rd harmonic distortion, instead of THD [which would consist of H2-H9]...

It also seems the measurement was taken at 85 dB SPL - their reference level. I do not understand the right % numbers then: at height of 85 dB it should read ~100% equivalent [the solid black line would be 100%], instead of something below 4%. Probably, they took two different graphs and threw them over each other with no meaningful correlation of the dB vs. % numbers.

To me, the applied smoothing looks a bit more like 1/3, but we will not know without further confirmation.

In speaker measurements THD will have interference from noises. 2nd and 3rd are much more comprehensive and accurate.
The left scale is for sound level/frequency response. The right percentage scale is for the distortion.

 

[LeftCoastTim]

FYI Kali Audio received the long waited 3d party Frequency/Distortion analysis plots just yesterday.
They are in the process of publishing them in their web site, however they already shared an image in their FB page to get an idea of how they look. I guess it has been quite smoothed (probably 1/6 smoothing) but they look quite balanced, apart from the dip in the highs we are all aware of. Any thoughts?

Hmm. So are they going to put "frequency response : +3dB/-8dB 20-20kHz" on their specs? Or maybe it will be +/- 6dB.

 

[detlev24]

In speaker measurements THD will have interference from noises. 2nd and 3rd are much more comprehensive and accurate. [...]

Maybe I got you wrong, but per definition THD is the ratio of the sum of powers of all harmonic components, to the power of the fundamental frequency [e.g., Wikipedia]. The sum of positive harmonic components can never be below its single addends. Hence, if we talk about noise floor and THD would show such an interference, this means, it "crosses" noise floor and shows either the same, or a lower level. Consequently, all harmonic components must be at an even lower level in this case and thus, would experience higher masking by noise floor than THD.

No; THD would provide less good-looking = higher % and furthermore, our hearing does tolerate - especially - H2 quite well: high amounts of H2 might even be perceived to be "pleasant"; whereas this is not true for odd-order harmonics. Neumann has a good overview of measurement descriptions on their website.

Please correct me, if I am wrong.

 

[Maiky76]

Hi All,

I am looking at buying some studio monitor paired with a MINIDSP HD as do-it-all speaker for the family.

The IN8 look very interesting as they use a coaxial driver that helps with the POWER response in the midrange (Mid to TW Xover) and they seem to have decent LF response compared to the Kef R3 with a similar configuration.
In addition if you factor everything they are rather affordable and compact.
Latter, multiple subs may be added.

I have (mostly) read the thread and decided to have a look at the amount of EQ the IN8 may require and how to design this EQ.
As always, one need to have a close look at the DI to decide what to EQ and what not to EQ.
It is a very interesting exercise.

It is just my first crack at it but i just wanted to share the results as to really do appreciate the effort put into publishing the data, kudos to @amirm.
This is just theoretical as I don’t own the speaker and one would have to probably add LF EQ for Room compensation.

But again if we start with a “flat and smooth on/off-axis” that’s 70-75% of the Predicted Preference Rating (p464 of @Floyd Toole book, 1st edition)

1. As a starting point I used the LW flat target but doing so up to 16k will result in way too much HF above 5k.
   Nothing new just look at the DI again… Also a it seems that one must be cautious around 1kHz.
   Similar to https://www.audiosciencereview.com/...udio-monitor-review.10897/page-50#post-379642
2. Then I added the PIR target with 9dB drop from 20 to 20kHz, as a preferred target (History of Harman Headphone Target Curve, slide 5 http://www.juloaudio.sk/Umiestnenie_reprosustav/History of Harman Target Curve.pdf)

The equation I used is:
Target_PIR = 99.57-1.31947*log(f);

One can offset the target up and down by changing the 99.57 value, of course without changing the slope (-1.31947).
Note that is does not mean, I believe, that one should compromise the LW/on-axis FR to reach that goal but instead select the speakers (Spinorama data, controlled directivity etc.) that come close to that and then devise the EQ to “polish off” the final result.

1. I used Genetic algorithms to fit the LW target from 50 to 5000Hz and simultaneously the PIR target from 40 to 16kHz.
   Probably weighting should be applied there but I did do NOT that, just equal weight except for the range limitation imposed on the LW.

Some biquads may be omitted to keep the number compatible with the DSP and leave space for in-room EQ
Also the high Q @5200Hz might be more pleasing to eye than needed especially with the production tolerances.

Proposed EQ:
Type Freq Q GaindB
Highpass: 25.25, 0.00, 1.74,...
Peaking: 168.5, 1.27, 3.89,...
Peaking: 245.0, -1.10, 0.50,...
Peaking: 331.0, 2.50, 4.00,...
Peaking: 1250.0, -1.22, 6.00,...
Peaking: 1905.5, 1.41, 6.00,...
Peaking: 2512.0, -2.36, 4.00,...
Peaking: 2800.0, 1.84, 1.36,...
Peaking: 4734.5, 0.62, 2.00,...
Peaking: 5200.0, 1.61, 11.0,...
Peaking: 6836.0, -1.10, 5.00,...
Peaking: 6856.0, -0.89, 3.50,...
Peaking: 9700.0, 0.73, 2.75,...
Peaking: 9704.0, 1.39, 3.75,...
Peaking: 14825, -0.84, 4.00,...

I decided to boost the LF a bit as the weight of the LF response in the Predicted Preference Rating is 25 -30% (p463 of @Floyd Toole book, 1st edition). Why not?
The first boosting HighPass Biquad can be omitted but I believe it should be fine since the speaker has built-in limiters/compressors .
It should help when the user is not playing vey loud, the limiter will just kick in earlier when the volume goes up.
Alternatively decreasing the over all sensitivity of the system by 6dB should work but at the price of a lower max SPL.
Not an issue for me as I don’t intend to play loud.

here is the data compared to @QMuse EQ
https://www.audiosciencereview.com/...udio-in-8-studio-monitor-review.10897/page-40

The output of the optimizer looks very similar to @QMuse EQ...

And a zoom for the LW and PIR

Please feel free to comment.

Next step:
Directly work on the LW, PIR and predicted rating.
If anyone could point me towards the exact equation that is used @MZKM I could try.


Cheers All

M

 

[Absolute]

My comments are limited to a caution concerning EQ based on single-sample measurements. They may or may not be usable for other samples depending on tolerances from manufacturer and tolerances from the driver supplier.

Also, not all dsp runs the same mathematical Q, so that might throw you off a bit. Minidsp runs with Proportional Q I think and some others run with Constant Q.

https://website-maat-digital-downlo...ublic/docs/Understanding_EQ_Architectures.pdf

 

[QMuse]

Proposed EQ:
Type Freq Q GaindB
Highpass: 25.25, 0.00, 1.74,...
Peaking: 168.5, 1.27, 3.89,...
Peaking: 245.0, -1.10, 0.50,...

I decided to boost the LF a bit as the weight of the LF response in the Predicted Preference Rating is 25 -30% (p463 of @Floyd Toole book, 1st edition). Why not?

Everything you showed looks good to me except these 2 filters - IMHO opinion filters below 300Hz are part of room EQ and not speaker EQ so they should be based on actual in-room measurements preferably done with MMM spatially averaged over the listenint position.

 

[Absolute]

Everything you showed looks good to me except these 2 filters - IMHO opinion filters below 300Hz are part of room EQ and not speaker EQ so they should be based on actual in-room measurements preferably done with MMM spatially averaged over the listenint position.

Unless there are resonances? In that case you would need to deal with the resonance first, then room EQ. Or are resonances that low not a concern? Generally speaking, not in this instance.

 

[QMuse]

Unless there are resonances? In that case you would need to deal with the resonance first, then room EQ. Or are resonances that low not a concern? Generally speaking, not in this instance.

There are no resonances in Kali's response below 300Hz, but speaking generally, it is impossible to predict how the room will influence the response below 300-400Hz. In that range room and speaker are becoming one entitiy so you don't separate speaker's response from room. You measure it in-situ and simply correct the response.

So, to answer your question, I would correct speaker only downto 300-400Hz even if Klippel would found resonances in the range below, as you never know if they would come handy in particular room or not. And if not, then you'll correct them.

 

[Maiky76]

Hi,

Thanks for the comments.

@Absolute
"My comments are limited to a caution concerning EQ based on single-sample measurements. They may or may not be usable for other samples depending on tolerances from manufacturer and tolerances from the driver supplier."

I do understand that and actually hereafter in the overlay of the data from Kali reproduced here by @musictracer and the NFS data just for reference.
The top two curves are matched at LF and the two bottom ones are matched at mid/HF.

Other than the trough at 100Hz which seems to be a measurement artifact (LP6, LP8 and IN8 all exhibit the same), the difference is less than 1.5dB although I had to apply a liberal amount of smoothing on the NFS data… better remeasure the LW with the very units that I may end up with.

@Absolute
"Also, not all dsp runs the same mathematical Q, so that might throw you off a bit. Minidsp runs with Proportional Q I think and some others run with Constant Q."

The idea is to make sure the MINIDSP HD has enough EQ capabilities before committing to the purchase.
I intend to use the advance programming (a and b coefficients directly from Matlab) so the Q definition should not be a problem, if i am not mistaken.
It should also enable me to use the “Crossover” block as 4 additional PEQ for a total of 24 PEQ per channel + FIR.

I might subsequently play with phase correction via the FIR which I believe would be 1x2048 taps per channel and potentially free up some more IIR by including them into the FIR - only for the upper range as @96kHz 2048 taps is not much.


@QMuse
"Everything you showed looks good to me except these 2 filters - IMHO opinion filters below 300Hz are part of room EQ and not speaker EQ so they should be based on actual in-room measurements preferably done with MMM spatially averaged over the listenint position."

MMM? That’s the plan! 24 PEQ - about 13 max for the speaker correction leaves 11 or more PEQ per channel for the LF correction which should be enough.

Cheers All
M

 

[Absolute]

Seems like you have this under control! There's plenty of stuff that you can see with your eyes that you don't hear with your ears, so I'll think you'll be fine with that many points

 

[Maiky76]

Hi,

More on the predicted score for the EQ:
https://www.audiosciencereview.com/...nt-sinad-discussion.10818/page-12#post-418108

Cheers
M

 

[sweetchaos]

Audioholics posted a review of Kali LP-8.

 

[Xyrium]

Nice. I wonder if that rising response above 15k is another sample to sample variation, or some sort of measurement fluke. I was looking for a cheap pair of speakers to replace the 305's on my electronic drum kit. Hopefully Kali can get their QC straightened out for v2, and perhaps address the diffraction issue causing the dip in the midrange that Amir measured. The reported hiss won't bother me while I'm wailing away at 95dB.

 

[617]

Nice. I wonder if that rising response above 15k is another sample to sample variation, or some sort of measurement fluke. I was looking for a cheap pair of speakers to replace the 305's on my electronic drum kit. Hopefully Kali can get their QC straightened out for v2, and perhaps address the diffraction issue causing the dip in the midrange that Amir measured. The reported hiss won't bother me while I'm wailing away at 95dB.

Top octave in a waveguide coax design like this is always going to be a bit sensitive to mic location during measurements. I wouldn't let that particular issue deter you. I guarantee this speaker will be a huge step up for e drums.