Desktop speaker with Markaudio Alpair 5.2

I had them on the shelf already quite some time, had some ideas in my head for even longer, but the past easter weekend I finally got to building the follow up to my Desktop Speaker with the 1st gen Alpair 5.

Let's begin with the end result:

attachment.php




more info in the follwing posts.






*) And yes, it's the gen2 version, that is still "current" with dealers here in Europe
 

Attachments

  • PP2021 klaar.jpg
    PP2021 klaar.jpg
    303.9 KB · Views: 1,325
The first puzzle with these small drivers is "how to get decent bass". If it is to be satisfying without sub, imho the low end should extend to at least 55Hz. Which is half of the driver's fs...

In many of my small driver designs I did this with some form of a TL, but I decided to go another route this time. If the Qt of the driver is in the range of 0.5 - 0.7 a so called 5th order bass reflex (BR with a series capcitor) can give good results. So I fired up LspCAD... starting point is a cab og a volume of ~3* the Vas and a tune of 0.6* fs and a series cap of 200000/(fb*Z). After some tweaking I found that a volume slightly below 4 liters, a 55Hz tune and a 666µF cap would result in a nice smooth response
 

Attachments

  • HPBR response.jpg
    HPBR response.jpg
    46.2 KB · Views: 310
Last edited:
then the question of how to fit those 4 liters in something for on the desktop. Since I still had a couple of feet from a flatpanel pc monitor I chose to make the speakers wide and shallow in stead of the usual slim and deep. wide is relative of course, 20cm is double the minimum. keeping the golden ratio height is 32cm, and depth follows from the volume and panel thickness.

port dimension: to keep construction of the slot port practical I choose the slot depth to be the same as some wood I have available (so I can stick some wood in it to ensure it keeps dimensions during gluing). in this case 15mm (thickness of the baffle), width is 9cm, the simulation gave 25cm length for the 55Hz tune, I made it 24cm for the wall will "make" it longer. after construction tune was measured to be 53Hz, so a even bit shorter will be OK as well.

panel thickness I chose 9mm Baltic birch plywood (except baffle) because I had this laying around. baffle is 15mm plywood (of unknown origin)
 

Attachments

  • paasproject 2021 3D.png
    paasproject 2021 3D.png
    240.9 KB · Views: 322
  • paasproject 2021.skp.zip
    246.1 KB · Views: 56
Last edited:
so I brought in the measurement gear

attachment.php

we see the baffle hump around 1300Hz (yellow line), almost exactly as predicted by The Edge , nice if you only listen to vocal music, but otherwise worth treating with a simple notch (green line). top end is as per datasheet, slight peaky, but nothing to worry about.



btw: what is odd is the dip/peak combo around 2½kHz. this is a driver issue, I measured it also with the driver outside the box, see the 2nd attached pic for close range measurement of the driver alone I do recall some discussion some 5+ years ago on here about this or another MA driver, but I could not find it (maybe it was in the now removed sponsor MA subforum?)
 

Attachments

  • pp2021 - meting vrijstaand met eq .png
    pp2021 - meting vrijstaand met eq .png
    108.7 KB · Views: 728
  • alpair5 los closerange.png
    alpair5 los closerange.png
    84.4 KB · Views: 51
so, now time for a proper listening. to make it easier to make adjustments, I inserted a minidsp 2x4HD so I could easily and quickly experiment with different corrections without soldering.

the 1300Hz hump was easy
attachment.php


but even with that the sound was too thin, too cold. Tenor voices and violas especially were not OK. Normally with most of the MA drivers, the built in BSC (as Mark calls it) will result in a nice balanced response without the need for baffle step correction, but in this case the response of the BR5 system is added, and that shows (see 2nd post) a slowly dropping response already from 500Hz and down. so I started to experiment with some low shelfs in the DSP until I was happy with the sound
attachment.php
 

Attachments

  • notch 1234Hz -4.5dB q4.5 .PNG
    notch 1234Hz -4.5dB q4.5 .PNG
    27.6 KB · Views: 700
  • shelf 666Hz 3.3dB q1 .PNG
    shelf 666Hz 3.3dB q1 .PNG
    27.6 KB · Views: 715
the next step was to translate this into a passive filter. an LR series with a single LCR shunt should be able to do the job. sims gave an inductor of 0.56mH and resistor of 2.2 ohm for the BSC and an LCR of 1.8mH/10µF/3R0.

so soldering iron was heated and listening again. the bass sounded a bit on the heavy side now. so time to experiment with some of the values in the LR. The resistor was OK, but varying the L value (see the graph for simulated changes) made quite a difference: yes even changes small as 0.5dB are very audible. the 0.56mH coil made the bass too thick, the 0.39mH resulted in lack of warmth, the 0.47mH was just right.
after a couple of hours I decided on the 0.47mH inductor.

so the final filter is:
attachment.php

note that R1021 represents the sum of the actual resistor and the dc resistance of the coil.


and this is the actual filter (no room for a print or other fancy stuff)
a 14gauge Janzen wax coil for the LR, a thin wire air coil (baked) for the L in the LCR since a high Rdc (2R2) is no problem here
attachment.php
 

Attachments

  • filterschema_final.jpg
    filterschema_final.jpg
    29.9 KB · Views: 695
  • vergelijk spoelen.jpg
    vergelijk spoelen.jpg
    53.1 KB · Views: 76
  • filter.jpg
    filter.jpg
    219.4 KB · Views: 670
Last edited:
Just so: the Gen3 unit is the Gen2 with a copper VC replacing the CCAW coil . Created at the behest of the Japanese distributor and his listening group of friends, who apparently felt a copper coil provided a more consistent tonality across the range. It didn't result in a significant parameter or FR shift, so they're drop-in replacements.

Nice job with the design BTW. :)
 
Last edited:
in the below schematic the 666uF for the HP is not indicated, to be "inserted" close to the driver. btw: while the exact value of this cap can be varied based on yr own listening, don't omit it, it is an essential part of the concept

the next step was to translate this into a passive filter. an LR series with a single LCR shunt should be able to do the job. sims gave an inductor of 0.56mH and resistor of 2.2 ohm for the BSC and an LCR of 1.8mH/10µF/3R0.

so soldering iron was heated and listening again. the bass sounded a bit on the heavy side now. so time to experiment with some of the values in the LR. The resistor was OK, but varying the L value (see the graph for simulated changes) made quite a difference: yes even changes small as 0.5dB are very audible. the 0.56mH coil made the bass too thick, the 0.39mH resulted in lack of warmth, the 0.47mH was just right.
after a couple of hours I decided on the 0.47mH inductor.

so the final filter is:

note that R1021 represents the sum of the actual resistor and the dc resistance of the coil.


and this is the actual filter (no room for a print or other fancy stuff)
a 14gauge Janzen wax coil for the LR, a thin wire air coil (baked) for the L in the LCR since a high Rdc (2R2) is no problem here