Basic questions on passive line-level XOs and with bi-amping

[NebuK]

Heyyas,

as i've build quite a few speakers by now and am currently looking forward to finishing my second diy amplifier i was planning to start out with something a little more modern. Each time when i'd bought those huge chunks of copper and those big resistors for the passive hi-level crossover i've thought "oh god, this is so outdated".

So i wanted to start out with something more modern. Designing and tuning the filters for a bi-amped (or multi-amped) system with active lo-level crossovers (e.g. one of the many nice linkwitz-riley 24db/oct filters) is my goal here. But then again - i thought - well, first things first. I need to "practice" and learn first.

So, what do to? Take two parts that are known to work, and put something own in there that's new - so i know it does work. My current plan is to take some cheap~ish, simple design for a two-way speaker and build it. Then i'd add some ready-made amplifiers and try toying around with passive crossovers on the lo-level side. Given this scenario i do have a few questions.

First off would be how the passive side crossovers which are presented with much greater (source to XO and XO to amp) impedance variety behave? Are a few OPAs at input and output necessary for impedance adjustment? Can i even take the plans and values for the hi-level passive crossovers as a starting point?

Then i really do wonder about amplifier output power. If i'd just plunge some not-all-so-bad-hongkong-class-d two-channel amp in each speaker i'd have the same power for tweeter and midbass - which doesn't sound right. As the midbass has a much greater moving mass, plays lower frequencies it needs more "power", right? So whats the best practice here? I could imagine having some "amp and XO box" in the middle of the two speakers which houses the XOs, and maybe three two-channel amps? One for each midbass where the two channels are Bridged (assuming the amps are BTL-able) and one where each channel drives one of the two tweeters?

I'm really really looking forward to your comments on this . I'd love to know your experiences with this topic as well as best-practice guidelines. Even "oh god, read THIS first you d**k" is welcome . Thanks in advance, Regards - NebuK

[Jonathan Bright]

Hi NebuK,
Low level passives are possible and have been done before. There is at least one thread on the topic on this site somewhere. I guy north of me in Australia was doing one a few years ago.
With the low level components you can't just use the same values as the high level units. Because the impedences are so much higher you will find that the capacitors have a much lower values and the inductors (coils) are much higher but because the power is so much lower the physical size and cost is much less. I suggest you read up on the idea of "time constants". That is important for this work. You can use LC circuits of the appropriate impedence between two buffers but often that is not necessary as the output impedence of the pre-amp and the input impedence of the power amps are adequately low and high respectively. Or you can just use capacitors. It depends on the slope etc. second order or 12db/octave slopes are pretty simple and often adequate.

I think you are a little bit confused on the power required for high and low frequency drive units. In the specifications for each speaker the figure to look at is the sensitivity of each unit. This is usually quoted as decibels per watt per meter. (or db/w/m) This will be around 90db. (Some a few db lower lower and some a few higher.) ((In Europe, on the Continent, I think the DIN standard is also used the idea of "working power". You may get that on the specification sheet. My memory is a bit slack these days but I think that was the power (in watts) needed to generate 96db.))

Now going back to db/w/m if your low frequency unit is say, 91db and your tweeter is 91db then you can use a similar power amp for each unit. Now this is a very simplified discussion. The amps have to have the same gain etc. There is a lot more to say.

You will NOT need six amps. Four will be fine for 4 speakers.

I think you should do a bit more research around the area of impedences and "time constants"......but I have always thought this was a good idea as you can get the benefits of bi-amping without adding extra distortion that comes from an active element such as an op-amp.

But you will have fun learning as you go........

And you might find that an active system starts to look more attractive. If you do go passive and low level with buffers etc its not long before you have the same number of parts as an active system. And there is a lot of material about those all over the place which will make life easier.

Try a Google search and see what happens.

Good luck. Let us know how you are getting on.
Cheers, Jonathan

[AndrewT]

Does a buffer before and after a passive CR or RC filter pass as a passive filter?
The filter is not inside a feedback loop. The filter is passive. The buffers do not amplify. The circuit will work well at line level and a wide range around line level.
I will say Yes.
There is no active stage wrapped around the passive filter stage.

If you want/need a single pole filter then the usual formula Freq = 1 / [ 2 Pi R C] is accurate only if the source impedance is zero and the load impedance is infinite.

Small errors in those source and load impedances do not change the filter by much, but it does change.

Lets' say your buffer or unity gain opamp has an output impedance of 0r6 and the R of the RC is 10k then the ratio is ~ 16000:1
I doubt this will be audible.
Similarly if the input impedance of the output buffer is 100k and the impedance of the C at the crossover frequency is ~5k then the ratio is ~20:1. Again I doubt this will be audible.

The CR values can be corrected to take account of non ideal source and load impedances, but if the effects are not audible then why bother?

If you want a 2pole filter then we are in trouble. The Qs of the two stages are both 1/sqrt(2) and when two single pole stages are cascaded the Q of the whole 2pole is found by multiplying the single pole Qs together.
i.e. you end up with a Q=0.5 using two single pole filters with a buffer in between to ensure those near idealistic conditions mentioned earlier.

But this experiment is relatively cheap and I guess you will learn a lot from it.
Go ahead.

BTW,
I have asked and so far not found a reply to " how do we alter the passive filter to take account of the non ideal impedances" such that we maintain the same Butterworth or other defined roll off and retain the same defined frequency?

[NebuK]

Hey,

thanks for your verbose answers first!

As for the buffers - if placed between preamp and power-amp i can see why i don't need impedance adjustment. In the case of the source beeing - at worst - some kind of soundcard (imagine a case where the owner of a system uses a DAC with integrated output buffer, a friend comes over and wants to play a song from his laptop/whatever - even thought the soundcard would be crappy it'd be nice if the filters wouldn't have to cope with the totally random impedances of bad soundcards?) i thought it might be nice... - but okay, for trying around the simples approach (e.g. the one without active elements) would be nice, you're right.

As for the amplifiers - i know that 4 would suffice, i was thinking of a practical and simple way to distribute the power. As "elliot" wrote in his bi-amping article on sound.westhost.com with a XO-frequency of ~1500-3000 hz (which would be around what is used in most midbass-tweeter-small-two-way-speakers) i'd need around 70% of the power for the midbass, only 30% for the tweeter. I was thinking of the most simple way to achieve something like that...

Speaking of that bi-amping article - i think i've understood a lot more after finding it (it somehow didn't turn up in my first attempts to read into the topic) .

@Andrew: i still do have some problems "understanding" (as in, by intuition) input- and output impedances and their effect on passive filters (or even generally). If i understand correctly what you're saying then a simple one-pole (6db/oct) filter would in no way profit from buffering as due to the impedances simply added as "virtual resistors" to input/output would change the equation slight enough so that the frequencies are shifted only marginally.

On the other hand, i don't really understand why the addition of a inductor - making it a 2nd order filter - makes matter so much worse?

Also, while reading the westhost site made me understand quite a few things i'm a little more confused in other areas. "Elliot" seems to talk about bi-amping with active XOs again. While my initial thought was that it'd be nice to be able to use a finished tuned passive hi-level crossover on the lo-side i now understand that this kind of crossover massively interacts with the speaker itself, with all of its so wheeeWAT-properties (such as inducance changing with voicecoil temp, etc.) and that there is probably no way to mimic this on the lo-level side with a power-amp in between.

On the other hand the approach doesn't seem to be too popular. While i've started to hack together a speaker to toy around with this approach i'm really thinking about what to use as a guideline or startingpoint for the passive lo-level filter. Deduce the crossover frequency from staring at the hi-level filter schematic and simply ignoring that a speaker (which behaves strange) is connected to those components?

Thank you so much for your answers already!
Regards
- Dario

[AndrewT]

Originally Posted by AndrewT

................If you want/need a single pole filter then the usual formula Freq = 1 / [ 2 Pi R C] is accurate only if the source impedance is zero and the load impedance is infinite.

Small errors in those source and load impedances do not change the filter by much, but it does change.

Lets' say ..................... if the input impedance of the output buffer is 100k and the impedance of the C at the crossover frequency is ~5k then the ratio is ~20:1. Again I doubt this will be audible...............

Originally Posted by NebuK

............I still do have some problems "understanding" ...... input- and output impedances and their effect on passive filters ........... If i understand correctly what you're saying then a simple one-pole (6db/oct) filter would in no way profit from buffering................. input/output would change the equation slight enough so that the frequencies are shifted only marginally.

I showed an example using 100k input impedance to the next stage.
What if the next stage has an input impedance of 10k?

[planet10]

TLS.org | Passive Line-Level Crossover

dave

[NebuK]

Woah, something like a guideline . Thank you a bunch.

Also, sorry for first starting to ask around and then letting the topic go to sleep ;/. So much to do atm. Still, planning for the speakers i'll test this on is almost finished .

Thanks and Regards
- NebuK

[NebuK]

Also - somehow i haven't yet stumbled upon the term "PLLXO" as a abbreviation for ... well, passive line level XOs. SUDDENLY SEARCHING TURNS UP TONS OF RESULTS. Thanks

[TimboinOz]

Or not. Ie that is a 1st order high-pass and a 1st order low pass. Hp @ ~<300Hz and a Lp at 6kHz.

I'm modelling a 3-way active system and I want to cascade a 1st order passive with the MR enclosure's own sealed roll-off which will be somewhere below 300 Hz. that will give a 3rd order high pass acoustic slope. the matching low-pass for the bass array will be a 1st order passive into a 2nd order active slope and then into another matching amplifier channel. The active crossover has gain.

I also want to roll in a ribbon tweeter off it's own amp (matching the MR amp) in triode mode at ~ 6khz -3db high pass, and will need to low-pass the MR amp and driver at 1st order -3db as well, to match.

IF it IS possible, how much gain will I lose in the MR leg with two PLLs?

TIA

Timbo in Oz

[philipreji]

I have tried passive line level as well as active line level crossovers.

(1) The active line level XO I built was the three-way 24 dB/oct L-R circuit described by Rod Elliot. The output sound after tri-amping was too "clean" - the "clinical" type.

(2) Then I made a two-way passive line level, 6 dB/Oct RC filter XO. The system was bi-amped. The sound was direct and there was good transient response - "lively" sound.

(3) Then I modified (2) by introducing an IC buffer between the source and the passive XO. The sound improved dramatically.

(4) Right now I am using speaker level 6 dB/Oct passive filters (woofer connected directly to the power amp output, and tweeter in parallel to the woofer, with just one good quality capacitor in series).

Configuration (1) is the best if the system is used at high volumes. For low and medium volumes both (3) and (4) are good.

Just my experience, with my amplifier and my speakers..

cheers,

Reji