Sub sat system design

[Dr.EM]

Hello, I may require a small system soon for a fairly low cost and obviously DIY came into mind! As an overview, I need it to sound decent for music, films, games etc, it will be used with a PC in a bedroom size room. Many systems like this exist commercially but most look to have pretty terrible performance.

I intend to use a pair of Hi-Vi B3N's for the sats. These are quite well regarded, are cheap and available plus look pretty cool . These will be in a sealed box, it needs to be small so I am working with 1.6L. This causes a bit of a hump but with the crossover applied it isn't very serious. Dimensions will be quite tall and slim to save desk space and stop driver being so near desk/so off axis. W-96mm, H-365mm, D-89mm.

The "sub" will be a Hi-Vi M3a in a ported box, 28L, 35Hz tuning (as reccomended by manufacturer in fact). This will be shaped a bit like the BOSE subs ( ) with rough dimensions W-225mm, H365-mm, D-433mm. A slot type port will likely be employed. The driver is relatively cheap and looks well made, obviously matches B3Ns aesthetically too

The crossover I plan to use 6th order 36db on the sub and 4th order 24db on the satellites, this keeps the changeover more natural as there is already 12db rolloff inherent in the small sealed satellites. Steep orders appear to produce better changeover and with it being quite high (120hz) on the sub will hopefully reduce the localisation
effects. Also keeps the little B3N's happy with the 24db rolloff

The crossovers will naturally be active and BSC + notch filters will also be included in the satellites chain to tailer the response.

Here are some FR graphs:



I am also looking at using all pass filters to improve the phase response between the sub and satellites. I don't know how relevant it is at these frequencies and with the sub positioned on the floor? If it could help integrate the sub and sats better then I'd need some help figuring out how to make the physical all pass filter that I have modelled in ISD Pro (part values etc).

Here is the phase response with the all pass bringing the phases together at crossover frequency:



and here without the all pass:



Would it help in practise?

Proposed amplifiers are these:

http://www.maplin.co.uk/Module.aspx?...=5292&doy=25m8

Two modules needed. One bridged to power the sub, the other used at 2 channel amp for the sats. I'm sure they arn't great but am working to as small a budget as possible! Open to suggestions on this but nothing too complicated if possible, all the op-amps will already be problem enough .

Suggestions/comments? Sorry for long post!

[richie00boy]

Neat little project. I'd be tempted to first try a 4th-order Linkwitz-Riley crossover which will have pleasant phase characteristics. The sub is easy just needs a textbook filter, but the satellite you can use the acoustic output to match with a tuned 2nd-order filter to make a true acoustic 4th-order with the right Q. You just need to know Fc and Qtc which are quite easy to measure.

At the moment I don't think you are properly taking the acoustic response into account, just doing a rough implementation considering filter order but not Q.

As to if higher order is better, well lower order can give a more natural sound, it depends on if you notice localisation or steep roll-off. Everybody is different.

If it helps I have layouts for filter PCBs on my website www.readresearch.co.uk also filter calculators.

[Dr.EM]

Thanks for the reply! Can't say I understand fully the effect of Q but I had another play around with it to see what I could do. Arrived at a filter Q of 0.555 pretty experimentally by what line looked straightest with the frequency set for 113.1Hz (the box Fsc). Box Q was 1.17 for the satellite but I don't know how that fits in? Anything like 1.17 minus 0.707 (the ideal/desired Q) to arrive at 0.463? Line didn't look so "good" like this though

I set the 4th order sub filter to try and meet the satellite at the -6db point on the graph. No idea if thats right, I think the LR filter should be 6db down at the set frequency? Then again, the sub has a hump to its response of about 1db at its worst (65hz).

My results so far...

[richie00boy]

Not far off. The Q multiplies, so if your satellite Q is 1.17 you need a 2nd-order filter Q of 0.5/1.17=0.427. Set Fo of the filter to Fc of the satellite. This acoustic combination should be -6dB at this frequency and is your crossover frequency.

Linkwitz-Riley Q is 0.5. To make the sub filter it's just 2 cascaded Butterworth 2nd-order filters with Fo at the crossover frequency.

[Dr.EM]

Ok, trying that this is what I get:



Which looked wrong to me because of the droop in the satellites response down toward the crossover, but then again mabye the sub will account for this and fill it back out again? Pity I can't see the summed response. The -6db point is correct now though at around 113Hz. The little B3N seems to cope alright with the 12db filter, never exceeding its X-max with the available power, though at 2mm it might be quite high in distortion. Don't know how you'd implement this with 36db and 24db slopes now though.

[richie00boy]

I think the sub low end response may be skewing the low-pass because they are close together. You chould check this by making the box extend down to 1Hz or something, then the crossover should look better.

If this is the case then this is where it gets difficult. I think you need to adjust the low-pass to give you an acoustic 4th-order Linkwitz-Riley, taking the sub box induced high-pass into account. Show use a graph of the sub without low-pass on.

[Dr.EM]

Ok, I see what you mean. The bass response looks like a hill and there is no flat range? Here is the reponse without the lowpass:



The main problem I had was the satellite having this droop in its lower end when using the updated filter Q. Is this normal? Does it actually cause a problem or does the sub fill it out?

[richie00boy]

When the sub is crossed over with the matching acoustic 4th-order Linkwitz-Riley response it will sum flat. As you have seen looking at the unfiltered sub response the response is not flat in the region where you want to cross over, so it's skewing the textbook response. Unfortunately working out the appropriate filter is not as simple as with the satellite. It involves trial and error. The problem is compounded by the humped sub response.

Maybe start by lowering the sub filter Q.

[Dr.EM]

Thanks again for your help!

I think this is something like it. I couldn't do a single 4th order so had to use two 2nd order to achieve this. I simply played with the Q factor so as to make the hump dissapear and maximal level be at 0db. I don't know if it's relevant but raising the crossover frequency makes the hump re-appear.

[richie00boy]

Looks better, but possibly Q needs lowering a bit more. Did you try lowering the filter frequency or just raising it?