Towards a complete Linn Kaber Aktiv solution using an rPi and AV receiv

[oomzay]

I thought I'd share my journey towards a Linn Kaber "Aktiv" solution using CamillaDsp for the crossover and an off-the-shelf AV receiver as the multichannel amplifier. This will hopefully work with any beefy AV receiver with HDMI input and a direct/pure multichannel mode. But be warned - the AV receiver must be dedicated to this application. Selecting an unfiltered multichannel source or engaging the receiver's fancy effects modes will likely blow your tweeters! I will be keeping a capacitor inline with the tweeters as insurance.

I am new to x-over design so please pile in with any tips and hints. Hopefully this will lead somewhere that is useful to others wanting to drive active speakers on a budget!

A little background

I have owned two pairs of kabers since 2007 when I bought them very cheaply (~£125 a pair) and was going to use them for AV surround. In the event this didn't work out so they have only had part time jobs ever since. One is factory passive and the other is factory aktiv - but it came with passive x-over boards which I fitted into an external box - which is now proving very handy...

My eureka moment was when I realised that with camillaDSP/rPi I could feed all 6 (2 x 3) filtered channels over an HDMI cable to an AV receiver by pretending they were 2 x Front, 2 x Side and 2 x Rear channels, which seems rather obvious now!

Progress so far

• Passive stereo over HDMI - Trivial
  
• Passive Tri-amped - Stereo mixed to 2 x 3 channels - !!! MASSIVE IMPROVEMENT !!! - could just stop here!
  
• Active Tri-amped - proof of concept - Simple x-overs 80Hz LR2 & 2800Hz LR4 - Doesn't sound great but !!! THE CONCEPT WORKS !!!
  
• Search, beg or borrow an existing Linn Kaber DSP x-over  pipeline - No luck with this!
  

Future

• Check I am really getting 24bits over the HDMI? (How?)
  
• Get to grips with speaker measurement tools. (REW and friends?)
  
• Design and tune DSP kaber x-overs.
  
• Add room EQ to DSP pipeline.
  
• Optimise rPi boot time? (rPi is powered   from the AV switched socket)
  
• Write-protect the rPi. 
  
• Investigate ways to publish and qualify speaker x-overs pipelines with community/moOde
  
• ...
  

[Nutul]

Good to see interesting results already! :-)

[oomzay]

Update

Well this has been the most frustrating IT experience of my life! After initial very rapid progress I have hit problem, after problem, after problem just trying to get 6 channels of 24bit audio out of HDMI without random channel switching! I thought I was embarking on a journey to learn about x-over design, but so far I have been mostly battling with different HDMI drivers, HDMI driver bugs, ALSA, PULSEAUDIO, PIPEWIRE, Various OS variants, REW compatability, JAVA compatability... And the gruelling battle isn't over - I still have not escaped the random HDMI channel switching! But... I have had some great help from many, especially Pavel over at avnirvana and...

...WaHey!!! I finally have a (just) stable enough REW system to get initial measurements of the original Passive Linn Kaber drivers. These will serve as a baseline for the real task ahead.

I came here to share the frequency plots with you... but the technology has confounded me again! How do I upload an image here?

[TheOldPresbyope]

Update

...
I came here to share the frequency plots with you... but the technology has confounded me again! How do I upload an image here?

That would be cool. Upload an image as an attachment. Once uploaded you can insert it in your post.

Example:
  default-cover-v6.png (Size: 5.07 KB / Downloads: 130)

Regards,
Kent

[oomzay]

(@TheOldPresbyope - Thanks for the tip - I hadn't scrolled down and seen the attachments - just looking in the widgets above the edit box!)

Below is my initial plot of the Linn Kaber Passive drives with original crossovers on axis at ~90cm (averaged L&R speakers). 

For those that don't know about the Kabers they have 3 drives: Trebble, Mid and Bass. The Mid and Bass use identical drives. Manufacturer documents specify the xovers as 80Hz LR2 and 2800Hz LR4.

The plot is made with calibrated UMIK-1. Multiple measurements seemed surprisingly consistent - the 1/3 plots overlapped each other exactly for most of the active range. However there are lots of obstacles and reflective surfaces in the room so I will take better measurements outside when the sun next shines (but that could be months!).

   

So now the fun begins as I feel my way forward... Before I implement active xovers I want to understand if the passive ones really match expectations...

I don't think we can read much into the nulls at 100 & 200 Hz at this stage because as well as the hard floor in the room, the mic was on top of tall thin speaker with its back square to the speaker under test!!!

I was a little surprised how close the bass and mid are up to 100Hz because I always assumed they had different sized enclosures - but looking at this plot I now think the enclosures must be very similar.

I was most surprised by the -8dB drop at the mid-trebble x-over. Shouldn't that be closer to -6dB ideally?

Crossover experts please pile-in with your opinions!

[forty5cal1911]

(@TheOldPresbyope - Thanks for the tip - I hadn't scrolled down and seen the attachments - just looking in the widgets above the edit box!)

Below is my initial plot of the Linn Kaber Passive drives with original crossovers on axis at ~90cm (averaged L&R speakers). 

For those that don't know about the Kabers they have 3 drives: Trebble, Mid and Bass. The Mid and Bass use identical drives. Manufacturer documents specify the xovers as 80Hz LR2 and 2800Hz LR4.

The plot is made with calibrated UMIK-1. Multiple measurements seemed surprisingly consistent - the 1/3 plots overlapped each other exactly for most of the active range. However there are lots of obstacles and reflective surfaces in the room so I will take better measurements outside when the sun next shines (but that could be months!).



So now the fun begins as I feel my way forward... Before I implement active xovers I want to understand if the passive ones really match expectations...

I don't think we can read much into the nulls at 100 & 200 Hz at this stage because as well as the hard floor in the room, the mic was on top of tall thin speaker with its back square to the speaker under test!!!

I was a little surprised how close the bass and mid are up to 100Hz because I always assumed they had different sized enclosures - but looking at this plot I now think the enclosures must be very similar.

I was most surprised by the -8dB drop at the mid-trebble x-over. Shouldn't that be closer to -6dB ideally?

Crossover experts please pile-in with your opinions!

Begin by measuring them very near field as in 1 meter directly on axis to remove room modes and reflections from the equation. I would also use a sweep vs pink noise and gate the measurement to 50 ms you may have to go a bit higher for the bass driver. Then re-post your FR plots. That should give a cleaner look and the drivers direct FR.

My apologies I overlooked the statement that you had measured pretty much near field. I would scrap their crossover design altogether. I would start at LR 4th order at 3k and LR 4th order at 300.

[oomzay]

@forty5cal1911

Here a plots at 1m with 50ms windows. These are now the Aktiv speakers with no crossovers - But I do have 3R8 + 15uF in series with the tweeter as insurance for now. (3R8 might be a bit much)...

[forty5cal1911]

I would go ahead and build my xover filters in Camilla and map to their corresponding channels, then perform individual driver measurements to confirm acoustic vs electric xover points and adjust them as necessary. It's the acoustic xover point that matters not what you put in the DSP filter. Then EQ each driver to match your intended response including above and below the xover point if necessary. Example would be the 7db boost from ~160 to 300hz above the 100hz xover point in your passive midbass measurement, it would need to be flattened to follow the LR 4 rolloff.

Pipeline editor > Filters > Add

Type: BiquadCombo
subtype: LinkwitzRileyHighpass
order: 4
freq: 3000

Then on your Mixer tab add that filter to your intended channel. Same for the LR Lowpass and Highpass for the Midrange and the lowpass for your woofer.

REW can generate the EQ filters for you for import into Camilla but you have to give it the curve for each driver you want it to match.

[forty5cal1911]

You can use the attached spreadsheet to generate your individual drivers EQ targets based on your desired overall FR. Go to the 3-way House Curve worksheet.

[oomzay]

I would go ahead and build my xover filters in Camilla and map to their corresponding channels, then perform individual driver measurements to confirm acoustic vs electric xover points and adjust them as necessary. It's the acoustic xover point that matters not what you put in the DSP filter. 

Going one step at a time... To get the ball rolling I have used the manufacturers nominal x-over specs: 

• Bass (RED) - Lowpass 80Hz LR2
  
• Mid (ORANGE) - Lowpass 2800Hz LR4
  
• Trebble (BLUE) - Highpass 2800Hz LR4
  

Here is the measured response (@1m < 50ms):

   

Many questions! 

Does this response look like a "typical" starting point to you at this stage? They are quite old speakers but L&R give very similar results so any wear and tear has been symmetrical!

Some things caught my attention:

• The trebble is several dB lower than the mid. 
  
• The trebble falls off from 4K, not 2.8K
  
• The combined response (GREEN) at 60Hz is _lower_ than the individual Bass & Mid at 1m on mid-point. (Phase?)
  
• The bass + mid SPL varies by up to 10dB in the 100 to 1000Hz range (Seems a lot)
  
• At 40Hz we are down at least 15dB. Can Linn's claim of 40Hz for their Aktiv Kaber speakers be met?
  

What would you do next?

• Should I revise the 3R8+12uF tweeter protection? (I feel much safer with it there due to the vc4hdmi channel-switching bug and my own mistakes!)
  
• Should I EQ individual drivers and adjust their levels first or adjust the x-overs and then EQ the whole thing?
  
• What is sensible SPL goal for EQ (e.g. 3dB)?
  
• How low should I be aiming with these speakers?
  
• Should I be worrying about phases at the x-over points?
  

(I have played with house curves and EQ on the passive speakers - so I know "how", but not what exactly "what" I should be aiming for to get the best from these speakers!)

Thanks for patient support!

P.S. Linn Passive speakers don't have an 80Hz Highpass on the Mid driver - So I have followed that on DSP - The Bass and Mid drivers are meant to work together below 80Hz.