After two years of design and build, my speakers are mostly finished, using active xover between mids/highs and bass, with passive between mids and highs. I am using the ESP 24db/octave boards, populated with quality parts, including LM4562.
Although I'm quite happy with the dynamics and detail this arrangement affords, I don't entirely like how things have turned out. When run full-range, the mid/high speakers are open and clear, with delicate, extended, spacious highs and tonally truthful mids. With the same amps and xovers in the circuit, there is plenty of detail (almost too much), but the midrange seems fake, with kind of a shallow or hollow tonality. Some of that may be due to xover point/slope issues, and I think I can fix that. But the highs seem "flat" and somehow truncated, with no sign of life or openness, and that is NOT a slope or xover point problem.
In addition to the designed bypass caps, I have added ceramic caps soldered right across the power pins of each opamp. I don't think that's an issue.
I have read all the threads, here and elsewhere (spending many hours), that discuss opamp sonic characteristics, and I know there are lots of different opinions with no clear consensus. I wonder if anyone has found an ideal combination of opamps in this or a similar application. I am on a limited budget, so I can't just buy 8 OPA627's or discrete opamps, unless that is absolutely the only alternative to achieve good sound.
If I were to try a couple 627's, would they yield the best results as input buffer, output buffer, or in the filter section?
Any other opinions or suggestions? I'm more than a little frustrated after all this work and expense, and I would appreciate your help.
Peace,
Tom E
Although I'm quite happy with the dynamics and detail this arrangement affords, I don't entirely like how things have turned out. When run full-range, the mid/high speakers are open and clear, with delicate, extended, spacious highs and tonally truthful mids. With the same amps and xovers in the circuit, there is plenty of detail (almost too much), but the midrange seems fake, with kind of a shallow or hollow tonality. Some of that may be due to xover point/slope issues, and I think I can fix that. But the highs seem "flat" and somehow truncated, with no sign of life or openness, and that is NOT a slope or xover point problem.
In addition to the designed bypass caps, I have added ceramic caps soldered right across the power pins of each opamp. I don't think that's an issue.
I have read all the threads, here and elsewhere (spending many hours), that discuss opamp sonic characteristics, and I know there are lots of different opinions with no clear consensus. I wonder if anyone has found an ideal combination of opamps in this or a similar application. I am on a limited budget, so I can't just buy 8 OPA627's or discrete opamps, unless that is absolutely the only alternative to achieve good sound.
If I were to try a couple 627's, would they yield the best results as input buffer, output buffer, or in the filter section?
Any other opinions or suggestions? I'm more than a little frustrated after all this work and expense, and I would appreciate your help.
Peace,
Tom E
Tom, this may not be what you want to hear, but I speak from personal experience. I suspect that what you are hearing as spacious highs, etc. are actually cone breakup modes that are adding pleasing colour to the sound when the drivers are run full range. I've been there, and been fooled too, and only realised what was going on when I ran distortion tests. So my advice, double check your measurements before starting to obsess on op-amp sound.
Thanks for your responses so far. I don't have any measuring equipment. The xovers are thoroughly broken in, as they have been playing for a month or so. I have been listening to music and a CD with steady test tones to try to set the levels. I know that is not ideal, but it's all I've got.
The level of mids to highs is fixed by a passive xover, designed by Madisound. I am completely satisfied with it. The mid/high sections sound excellent without the active xovers in the circuit, and it's not because of cone break-up!
The active xover is between bass and mids/highs. There are pots on board to adjust gain. I am also considering using fixed or stepped resistors to get rid of the pots in case they are a factor in the aspect of the sound I don't like.
These speakers do lots of good things, but I know they can be better. I have ordered some of the NE5532 to try out. I also ordered parts to try different turnover point and slope.
Peace,
Tom E
The level of mids to highs is fixed by a passive xover, designed by Madisound. I am completely satisfied with it. The mid/high sections sound excellent without the active xovers in the circuit, and it's not because of cone break-up!
The active xover is between bass and mids/highs. There are pots on board to adjust gain. I am also considering using fixed or stepped resistors to get rid of the pots in case they are a factor in the aspect of the sound I don't like.
These speakers do lots of good things, but I know they can be better. I have ordered some of the NE5532 to try out. I also ordered parts to try different turnover point and slope.
Peace,
Tom E
Attachments
Tom,
What kind of design planning went in to the active crossovers? Since you mentioned that you are using the ESP boards, I assume you just bought and built the boards without doing any planning of the crossover itself. In that case what you have is a textbook crossover, applied to a non-textbook situation.
You mentioned that you didn't take any measurements or don't have that capability, so I would not assume that the crossovers aren't messing with the phase and frequency response, and what you are hearing is an uneven frequency response and some peaks and nulls on and off axis. That is likely to be the problem you are experiencing. Because the mid/high midrange driver(s) are also rolling off somewhere in the vicinity, their response (amplitude and phase) is also thrown in to the mix and is further messing with the system response. It seems from what you have said that none of this was taken in to account as part of the active crossover section.
The right way to design a crossover is to measure the drivers, already mounted in the final enclosure, and use the measured frequency response and phase info to design the crossover. This is what Madisound does to design their passive crossovers, and as you have mentioned they sound great. It's no different with active crossovers or for lower frequencies.
Don't blame the op-amp or the potentiometer, or "break in" - these things are way, way down your list of potential problems here.
-Charlie
Do not underestimate the complexity of doing a active x/o that does non-standard transfer functions. A passive L-plus- parallel-LCR on the woofer leg of a system can work wonders if properly optimized. Doing the same x/o active will be quite challenging. Without proper measuring software to check your work I should not even think about it....
For examples check the Linear X/LEAP website.
Good Luck,
Eelco de Bode
For examples check the Linear X/LEAP website.
Good Luck,
Eelco de Bode
Charlie,
Indeed, these actives were built stock, with merely a guess at the turnover point (350Hz, as it stands), based on listening and the mid/high plots provided by Madisound. The slope is fixed (for now) by the design at 24dB/octave, and I hoped that such a steep slope would help avoid some of the phase/amplitude interaction to which you refer. I understand the proper design procedures include measuring, then planning the xover parameters. That is exactly what I paid Madisound to do.
I know now that the initial educated guess method didn't work. I'll continue experimenting. I cannot afford to invest in measuring equipment, and I don't know enough to obtain accurate results and proper implementation even if I did.
There are really two issues here, however, and I don't think they're related.
The hollowness and lack of body in the lower midrange is no doubt a result of the active xover function, with a possible minor contribution of the opamps I'm using. I really like some of what the active xover does for the system, so right now I am committed to making that work. It already sounds pretty good (better than my B&W's in many respects), and I would be satisfied with even minor improvement. Attaining perfection would certainly involve measuring and complex circuit design. Every system has compromises, and every listener learns to accept them.
The other issue is the nature of the high frequencies, and I think that is surely affected by the opamps or some other component or the circuit design itself. The mid/high units sound really excellent when run full-range. Of course, they don't make much bass, but the highs and mids are very clean and open. With the active xover in the circuit and everything else (amps, wire, etc.) the same, the highs collapse and are somewhat truncated. I find it hard to believe that the active xover turnover has much to do with that, since it's almost a full decade away (mids/tweets cross at appx 2900Hz). Am I wrong there? I'm not being stubborn; merely trying to clarify.
Peace,
Tom E
Indeed, these actives were built stock, with merely a guess at the turnover point (350Hz, as it stands), based on listening and the mid/high plots provided by Madisound. The slope is fixed (for now) by the design at 24dB/octave, and I hoped that such a steep slope would help avoid some of the phase/amplitude interaction to which you refer. I understand the proper design procedures include measuring, then planning the xover parameters. That is exactly what I paid Madisound to do.
I know now that the initial educated guess method didn't work. I'll continue experimenting. I cannot afford to invest in measuring equipment, and I don't know enough to obtain accurate results and proper implementation even if I did.
There are really two issues here, however, and I don't think they're related.
The hollowness and lack of body in the lower midrange is no doubt a result of the active xover function, with a possible minor contribution of the opamps I'm using. I really like some of what the active xover does for the system, so right now I am committed to making that work. It already sounds pretty good (better than my B&W's in many respects), and I would be satisfied with even minor improvement. Attaining perfection would certainly involve measuring and complex circuit design. Every system has compromises, and every listener learns to accept them.
The other issue is the nature of the high frequencies, and I think that is surely affected by the opamps or some other component or the circuit design itself. The mid/high units sound really excellent when run full-range. Of course, they don't make much bass, but the highs and mids are very clean and open. With the active xover in the circuit and everything else (amps, wire, etc.) the same, the highs collapse and are somewhat truncated. I find it hard to believe that the active xover turnover has much to do with that, since it's almost a full decade away (mids/tweets cross at appx 2900Hz). Am I wrong there? I'm not being stubborn; merely trying to clarify.
Peace,
Tom E
Please show some pics of the crossover enclosure with it's internals + PCB's both sides.
These ceramic caps could be the issue, if they have long legs.
PSU design could be an issue too, as well as grounding, singnal wires tracing etc.
LM4562 have low resistance inputs (bipolar) which could make some problems, if you choose high values of resistors in surounding crossover network.
LM4562 is quite fast (55mhz GBW) - that may cause oscilliations etc, as the boards you've got don't limit signal bandwith/are single sided/probably not designed for such fast opamps.
Try some FET-input opamps, OPA2134/2132 is the safe bet. It is relatively slow, FET, low distortion with reasonable driving capabilities opamp.
These ceramic caps could be the issue, if they have long legs.
PSU design could be an issue too, as well as grounding, singnal wires tracing etc.
LM4562 have low resistance inputs (bipolar) which could make some problems, if you choose high values of resistors in surounding crossover network.
LM4562 is quite fast (55mhz GBW) - that may cause oscilliations etc, as the boards you've got don't limit signal bandwith/are single sided/probably not designed for such fast opamps.
Try some FET-input opamps, OPA2134/2132 is the safe bet. It is relatively slow, FET, low distortion with reasonable driving capabilities opamp.
Last edited:
All you need is a cheap electret microphone and a cheap microphone amp. The software is free - ARTA for measurements ARTA Home and Jeff Bagbys's Excel sheets jbagby for simulation, or the demo version of LspCAD Untitled . You'll be stumbling in the dark without measurements.
I agree to that NE5532 has a warmer sound than LM4562, and I hope you like them. These OPAxxx JFET input opamps suggested above also have a warmer sound and they are less "sterile" with the downside of some treble "grit".
I guess all your filtering caps are polypropylene, and not electrolytic or ceramic.
I agree to that NE5532 has a warmer sound than LM4562, and I hope you like them. These OPAxxx JFET input opamps suggested above also have a warmer sound and they are less "sterile" with the downside of some treble "grit".
I guess all your filtering caps are polypropylene, and not electrolytic or ceramic.
I am also using a 24dB/Oct L-R active crossover to integrate a low frequency system with a 2-way speaker which has its own passive crossover from mid to high. IMO this board is much better than the ESP one in terms of layout, engineering & vfm:
Bare PC board for our EC24 stereo crossover
I am using LM4652s and I think they sound slightly more refined than NE5532s. However I reckon the series caps in the XO (4 + any I/O couplers in the hp network) will have more affect on the sound than the opamps. I am more than happy with the treble performance from this lot feeding my OW1s.
The benefits of the mid unit not having to handle lf far outweigh any slight losses in the active XO. Stick with it!
Watch out for dc offset if using 5532 in the low pass filter.
Bare PC board for our EC24 stereo crossover
I am using LM4652s and I think they sound slightly more refined than NE5532s. However I reckon the series caps in the XO (4 + any I/O couplers in the hp network) will have more affect on the sound than the opamps. I am more than happy with the treble performance from this lot feeding my OW1s.
The benefits of the mid unit not having to handle lf far outweigh any slight losses in the active XO. Stick with it!
Watch out for dc offset if using 5532 in the low pass filter.
Last edited:
Are you sure your op amp has the required bandwidth for frequencies up to 40KHz ?
Thanks for all your suggestions.
Dave, that board you linked certainly is more complex than the ESP design, but I'm not sure I'm interested in the extra features. What makes you say it has better engineering and layout? Unlike the ESP site, they tell you almost nothing about the circuit or design. Value is better with on-board regulation, but I am already kind of committed to the ESP boards, what with a power supply and all the xover parts. The ESP board also has balanced input, which is a nice feature for me.
Regarding parts selection: I am using WIMA poly film caps and PRP resistors at all signal path locations. I think those are very good parts, and I suspect they don't leave too much of a bad sonic signature. I have built monoblock amps for the mids/highs with one half of a stereo xover inside the same enclosure for each channel, so there are no IC's involved. Internal wiring is OCC solid core copper/teflon.
The PS is decent quality: it was used in the Hagerman Clarinet preamp. Right now I am using one for both xovers with an umbilical running between them. I plan to build another supply so I have one for each xover, but I expect that will not greatly affect the sound.
The opamp power supply is bypassed by Panny FM's, small ceramic caps on the PCB, and a ceramic cap soldered directly across the power pins of each opamp. The lead length is about as short as it can get!
I am currently considering reconfiguing the xovers as 3rd order Butterworth to find out what effect it has on the lower mids/upper bass.
Peace,
Tom E
Dave, that board you linked certainly is more complex than the ESP design, but I'm not sure I'm interested in the extra features. What makes you say it has better engineering and layout? Unlike the ESP site, they tell you almost nothing about the circuit or design. Value is better with on-board regulation, but I am already kind of committed to the ESP boards, what with a power supply and all the xover parts. The ESP board also has balanced input, which is a nice feature for me.
Regarding parts selection: I am using WIMA poly film caps and PRP resistors at all signal path locations. I think those are very good parts, and I suspect they don't leave too much of a bad sonic signature. I have built monoblock amps for the mids/highs with one half of a stereo xover inside the same enclosure for each channel, so there are no IC's involved. Internal wiring is OCC solid core copper/teflon.
The PS is decent quality: it was used in the Hagerman Clarinet preamp. Right now I am using one for both xovers with an umbilical running between them. I plan to build another supply so I have one for each xover, but I expect that will not greatly affect the sound.
The opamp power supply is bypassed by Panny FM's, small ceramic caps on the PCB, and a ceramic cap soldered directly across the power pins of each opamp. The lead length is about as short as it can get!
I am currently considering reconfiguing the xovers as 3rd order Butterworth to find out what effect it has on the lower mids/upper bass.
Peace,
Tom E
The 4652s are supposed to be unity gain stable, but they're also fast. I can't imagine building filters and trying to reach any conclusion about what's going on without being able to scope the circuit. Filters are one of the fussier applications where some opamps do far better than others, and some may require a non-standard tweak to avoid oscillations.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.