Chips for active cross overs

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I want to build some bi-amped studio monitor type speakers.

I'm thinking of a conventional design. There will be a balanced XLR input that goes to a buffer then to a 24db/octave low and high pass filters. Each filter drives a conventional LM3386 chip amp that connects to either a woofer or tweeter.

Question about the filters: Before I wire up a cascade of TL074 op amps and many precision passive parts is there some "filter chip" available that would make the design simpler? I figure there are so many bi-amped monitors on the market that someone might be offering an integrated active crossover chip. Although the TL074 does come in a quad package so it is not to bad.

Are any better op amps than TL074? It was the preferred part last time I looked, years ago.
 
Last edited:
Taking a look at the THD/IMD, CMRR, PSRR, and GBP specs on the 49990 (follow the link above and click on datasheet) should answer your question. For DC servos JFET devices like the TL071 family are attractive for their combination of low input bias, decent linearity, and sufficient gain bandwidth for the application. But for main signal path use I'm not aware of any JFET parts with linearity, bandwidth, or rejection abilities comparable to the 49710 family or 49990.

Granted, the 49740 is 10 times the price of a TL074. For a stereo LR4 crossover that's an extra USD 16, USD 20 if you include the 49710/TL071 to provide a fifth op amp for diff to single ended conversion in this case. Not a big deal, IMO, when building in DIY quantities.
 
Filter Chips

Yes, there are lots filter chips that could be used. Many are switched capacitor types that I would avoid... but there is one standout:

I have used and I can highly recommend the Burr-Brown UAF42. Each IC contains a second order HP+LP stage plus one uncommitted op-amp. The IC has on-chip 0.5% (IIRC) tolerance capacitors. They can be cascaded for higher order or for multi-way crossovers. External resistors set crossover frequency and Q across any values that you would need for audio work. I recommend using the inverted configuration. Specs are good. This is much better IMHO than building your own using op amps because of the on-chip capacitors and the ability to re-use the IC (it is a DIP14, so just put it in a socket) in various ways/projects.

-Charlie
 
Taking a look at the THD/IMD, CMRR, PSRR, and GBP specs on the 49990 (follow the link above and click on datasheet) should answer your question. For DC servos JFET devices like the TL071 family are attractive for their combination of low input bias, decent linearity, and sufficient gain bandwidth for the application. But for main signal path use I'm not aware of any JFET parts with linearity, bandwidth, or rejection abilities comparable to the 49710 family or 49990.

Granted, the 49740 is 10 times the price of a TL074. For a stereo LR4 crossover that's an extra USD 16, USD 20 if you include the 49710/TL071 to provide a fifth op amp for diff to single ended conversion in this case. Not a big deal, IMO, when building in DIY quantities.

Thanks. I was the one to ask this question. I can buy TL074 in surface Mount SOIC packages for $0.21 each. That is five cents per op amp. I see a 49740 for $6.50 in SOIC-14 package or $13 per channel, not bad (but $0.42 is better.)

But I'm still surprised that there are no x-over chips. I wonder what is inside all the M-audio, Mackie, Focal, and so on bi-amped monitor speakers. I can't imagine they use 8 or 10 op amps and two chip amps per speaker and still sell and their price. I'd bet there is a highly integrated part in there.

As an example this is bi-amped with 18db x-over and retails for $150
Buy KRK Rokit Powered 5 Generation 2 Powered Studio Monitor (Each) | Powered Monitors | Musician's Friend
 
Last edited:
I think your expectations may be a bit high. Even fairly high end pro audio stuff I've taken apart relies on large numbers of op amps like the NJM4565. Haven't ever looked inside a Rockit but my experience is 18dB slopes are usually line level passive crossovers that are marketed as active solely on the basis the crossover is located in front of the power stage. Most likely just a two channel 8 SOIC input buffer with strings of 0402 resistors and caps between it and the power stages. So far the stuff I've taken apart that was marketed as LR4 has actually had Sallen-Keys in it, but it wouldn't suprise me if some of it's just RCRCRCRC.
 
Hi,

the LME offers just very few usable data. Non of it can be marked with "This is good for Audio". Just think of the ridicolously high OL amplification numbers.
At least 120dB more than necessary. So one can assume that more parts/stages are necessary to generate this level of amplification.
Indeed some of the best sounding CRs I listened to used simple unitygain SallenKey-filters and JFET-pairs as active decvices.
In filter circuitry input impedance most often varyies strongly with frequency -think of caps. A situation where JFET-Inputs are preferrable over bipolar inputs. Especially lowpass filter circuits don´t need those superfast OP amps which create problems of their own like layout, power supply etc.
The UAF42 can be mocked up by a simple quad-OP with tightly tolerated caps and Resistors. The filters generated with this chip are of 2nd order so they must be cascaded to fulfill the rquirement.
A well working alternative which needs less parts is the Hawksford Riley structure which can be taylored to the demands, asks for the lowest part count and achieves very high symmetry between LP and HP (kind of subtracting-filter). With 1 quad and 1 Dual OP and 4 caps (which don´t need to be tightly tolerated) You can build a complete 2way CR of 4th order (-24dB/opct).

jauu
Calvin
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.