Budget active 3 way crossover - buy, eval, mod

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I've always seen these active filter / cross over boards for sale from the far east vendors. Most advertise they are Linkwitz-Riley and may even add a data point or two (ie. XO points @300Hz and @3Khz) and maybe the supply voltage (AC15). Everything else is omitted, but the price looks attractive ($40 CDN. Certainly better than me designing a PCB, ordering all the parts, and assembling it.

It's not trendy, using op amps instead of a mini-DSP or an RPI+DAC. However, I have a project in mind, and I wanted to play around with an active filter design. I don't care what the specific XO points are, as I'll probably change them anyways. I wanted a board full of potential (2nd order LPs and HPs, occasionally cascaded.) to start with. So here goes.
 

Attachments

  • DSCN2375 reduced.jpg
    DSCN2375 reduced.jpg
    736.9 KB · Views: 1,049
Last edited:
power supply

I need a power supply. Normally these boards call for a transformer as they have a rectifier, filter, and regulator on board (probably). Unfortunately, transformers are heavy (shipping) and expensive so I'm going to use a small switch mode power supply. I can get a small SMPS for $10 that provides +/-15VDC @0.55A. Given the board has 9 op amps, and they use typically 10mA per op-amp, I should need 100mA approx. So plenty of headroom with this supply.
 

Attachments

  • DSCN2376 reduced.jpg
    DSCN2376 reduced.jpg
    556.6 KB · Views: 1,012
powering it up

A quick check on the polarity of the power connections, and the outputs (+/-15VDC) of the SMPS, then remove all the op-amps before connecting power to the board.

Hmm,.... no lights come on for the XO board. There is +13.6VDC and -13.2VDC from the boards regulator. It doesn't seem to be regulating very well. Current draw is 80mA. I could live with those voltages, but it should be better.

I trace/buzz out the circuit, and it appears to be a pass transistor with a shunt regulator (not my choice). The series pass transistor is working but the shunt regulator is set at too high a voltage for my +/-15VDC input. Not a problem we'll set it lower to +/-12VDC. I don't have a set of precision resistors to replace the resistors used to set the shunt voltage. So instead I placed a 22K resistor in parallel with R5 and another in parallel with R8 to lower their resistances. Both barnacles are placed on the backside. The shunt is there so I might as well use it.

I now have +/-12.1VDC board supply rails and the lights are on. Current draw is 110mA.
 

Attachments

  • active 3way feb 02 20180001 reduced.jpg
    active 3way feb 02 20180001 reduced.jpg
    207.5 KB · Views: 994
  • active 3way feb 02 20180002 reduced.jpg
    active 3way feb 02 20180002 reduced.jpg
    217 KB · Views: 989
  • DSCN2374 reduced.jpg
    DSCN2374 reduced.jpg
    535 KB · Views: 963
Last edited:
input stage

A quick buzz/trace out of the input stage reveals its DC coupled, singled ended, with some gain, and a noise filter.

There are enough component connection points there to float the input connector and make it into a differential input.
 

Attachments

  • active 3way feb 02 20180004 reduced.jpg
    active 3way feb 02 20180004 reduced.jpg
    175.9 KB · Views: 276
  • active 3way feb 02 20180003 reduced.jpg
    active 3way feb 02 20180003 reduced.jpg
    176.2 KB · Views: 334
The one and only
Joined 2001
Paid Member
I've always seen these active filter / cross over boards for sale from the far east vendors. Most advertise they are Linkwitz-Riley and may even add a data point or two (ie. XO points @300Hz and @3Khz) and maybe the supply voltage (AC15). Everything else is omitted, but the price looks attractive ($40 CDN. Certainly better than me designing a PCB, ordering all the parts, and assembling it.

Looks pretty nice from the photo, what appear to be polypropylenes and
Silmic electrolytics.
 
Filter cross over points

After assuring myself it wouldn't blow up my sound card. I ran a quick test through REW to plot the FR. No surprise it crosses at 300Hz and 3Khz. It also appears (no surprise) to be an LR4, pass flat with -24db/octave attn.
 

Attachments

  • FR tests.jpg
    FR tests.jpg
    89.5 KB · Views: 377
Last edited:
I think you are showing that each filter is using a gain block.
If that is the case then you don't have a unity gain S&K.
Instead you have an equal component value S&K.
This gives a gain of ~+8dB when set up for LR4.
Fortunately ECV filters have separate control of frequency and Q (Damping)
Look up an ECV S&K design app (Ti have a few).
You will find that each (2pole) Butterworth must be set up with Q values that are specific to it's location in the string of filters.
It is worth checking that the PCB uses this correctly.
 
I am surprised that the Low Pass roll-off is not showing an attenuation limit.
This occurs when the output tap is taken from the opamp output pin.
There is a solution: move the tapping point on the Low Pass filters from the output pin to the input pin.
It looks like they may have done that. Can you confirm?
 
I am surprised that the Low Pass roll-off is not showing an attenuation limit.
This occurs when the output tap is taken from the opamp output pin.
There is a solution: move the tapping point on the Low Pass filters from the output pin to the input pin.
It looks like they may have done that. Can you confirm?

I just listed the first input stage to make sure I could connect my sound card to it. I still have to map out the actual filter stages and configurations. I suspect the actual filter stages are unity gain.

I'm using a PC soundcard, and the levels were set midway as I just needed to check where the crossover points occurred. The FR graph only shows the top 80db of the curves so you don't see the actual floor. Once I get it setup properly I'll post more graphs related to performance.

Interesting idea about the op-amp tap points. I would have expected to always use the output pin. Thanks. I'll post the schematic(s) after I map it out.
 
Floorplan and connector map

The silkscreen on the pcb has a number of different terminologies (ie. #1, right, treble, bass, low, etc.) and some are wrong. The midrange output labels are reversed.

So I mapped out the board. I use Channel#1 or Channel#2 for the stream, and [L,M,H) for the frequency range, and (LP, HP) for the filter type.
 

Attachments

  • active 3way floorplan reduced.jpg
    active 3way floorplan reduced.jpg
    204.8 KB · Views: 522
Really surprised to find DonVK here, after all his rather sophisticated omni experiments and optimizations.

May I kindly ask you what you have in mind with experimenting with a PCB layout, that clearly only provides for standard, non-optimized textbook transfer function filter sections?

Or is this a curiosity-only driven exercise?

Eelco
 
Surprised? why? I'm an audio-omnivore :). I enjoy listening to those Omnis nearly every day. I learned a lot from that exercise. It's a 2way passive LR4.

The current idea was related to active speakers and I needed a cross over. So I ordered a few items like this 3way for curiosity and evaluation. There's no documentation, so I posted it, in case others were interested. I suspected this 3way design was a "cut and paste" but you never know with this stuff. No promises made, so no promises broken :) So far I'm happy with it, its a vanilla filter design.

One use for it is performance testing against the passive XO in my Omni. The same amp, same speaker design, but compare XO before amp to XO after amp. The second use is a possible active n-way cylinder that needs multiple driver sizes to create very wide directivity and low diffraction. The speaker EQ and room EQ would be done in s/w in an RPi3B, but I still need multiple outputs for the amps.
 
Crossover filter blocks

The filter blocks are standard LR4, the same form as the original SL ones at Active Filters The derivations and formulas are on that SL site as well. Its a great site.

I've provided the layout and schematic from trace/buzzing out the pcb. These are the filter blocks along the top row of the pcb. The lower row is exactly the same so I won't repeat it. Given the symmetric structures and repeated values it would be easy to adjust the frequencies. Given this board is a field of 2nd order LPs and HPs it would be easy to reconfigure and tap out what you need. All the intermediate filter stages are unity gain, the only gain change occurs at the initial input stage.
 

Attachments

  • active 3way filter blocks0001 reduced.jpg
    active 3way filter blocks0001 reduced.jpg
    223.8 KB · Views: 496
  • active 3way filter blocks0002 reduced.jpg
    active 3way filter blocks0002 reduced.jpg
    209.6 KB · Views: 507
  • active 3way filter blocks0003 reduced.jpg
    active 3way filter blocks0003 reduced.jpg
    290.4 KB · Views: 487
You use 33nF 20off, for two channels.
Measure them.
Select the lowest and the highest and pair them off to give ~66nF.
Select the next L + H and pair them. And again till you have 4 pairs, all near 66nF
The middle group of 12 off 33nF are hopefully close enough to being the same, you just insert them where singles are required.

Same for 3n3F

Once you have all those capacitors selected, you don't want to change them.

To change frequency you change the resistors. Since these are 1%, then selection is much less important.
But use DiL sockets on the PCB. Solder all the resistors to DiL sockets/headers. Then plug them in to set the frequency.
Make up a spare set of soldered socketed resistors to change frequency.
I see the Low Pass and the High Pass have the setting resistor arranged in pairs. That suits sockets.
The bandpass is a mix of singles and pairs. Might be better using SiL pins and sockets.
 
Last edited:
All good suggestions AndrewT. Matching pairs using "select on test" certainly would work and I agree that reprogramming via resistors changes is the preferred route. Its just easier to get variety in resistor values. I'm just not at the point where I know exactly what I want it programmed to.

I plan on doing some performance measurements, now that I know what the circuitry is. In particular noise levels and harmonic distortion.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.