I'm working on a three-way speaker system that uses Scanspeak drivers powered by an actively crossed-over LM3886 power stage. The electronics include a Linkwitz Transform and a 24db/octave crossover that can be switched to 12dB/octave. There is a remote power on section, balanced input and a couple other cool things going on.
The board has seven LM3886T devices on it: one for high end, two in parallel for midrange, and basically a BPA-200 for low end. The first iteration uses smaller format drivers and will use lower Vcc, resulting in only one mid range opamp and one BTL pair being needed (plus of course the tweet power). The next design will use a larger driver set using the same board, but will operate at +/- 35V and will need all seven opamps.
I'm currently working on the power supply section and am looking forward to auditioning several different designs, as it is my view that the power supply is the determining factor in overall sound quality. I'm starting with a 600VAC toroidal and paralleled 2,200uF caps bypassed with films.
The board has seven LM3886T devices on it: one for high end, two in parallel for midrange, and basically a BPA-200 for low end. The first iteration uses smaller format drivers and will use lower Vcc, resulting in only one mid range opamp and one BTL pair being needed (plus of course the tweet power). The next design will use a larger driver set using the same board, but will operate at +/- 35V and will need all seven opamps.
I'm currently working on the power supply section and am looking forward to auditioning several different designs, as it is my view that the power supply is the determining factor in overall sound quality. I'm starting with a 600VAC toroidal and paralleled 2,200uF caps bypassed with films.
Attachments
Very nice; I have a project planned to convert my existing 3-way mains into a self-powered, actively crossed system with 4th-order LW aligned crossover as well. Your idea sounds very similar to mine. I was also planning to use balanced inputs (differential signalling over CAT-5 cable) to supply their input signals.
You may be overestimating your power requrements, but that is not for me to judge. Anyway, you may find that even the larger drivers perform well with the smaller initial set of amps that you have planned. Try and see I guess (that's half the fun).
I am curious to hear your ideas on power supply design. My intent was to use a single transformer per channel (one each Left and Right) but to have the rectifiers and caps individual per amp-speaker (tweet, mid, woofer each have their own rectifiers et all). I believe that this would substantially reduce the interaction/coupling between the power supply rails for the different frequency bands' amps, vs. just using one set of rails for all amps. The PSRR of these chips is, after all, merely good, not exceptional.
You may be overestimating your power requrements, but that is not for me to judge. Anyway, you may find that even the larger drivers perform well with the smaller initial set of amps that you have planned. Try and see I guess (that's half the fun).
I am curious to hear your ideas on power supply design. My intent was to use a single transformer per channel (one each Left and Right) but to have the rectifiers and caps individual per amp-speaker (tweet, mid, woofer each have their own rectifiers et all). I believe that this would substantially reduce the interaction/coupling between the power supply rails for the different frequency bands' amps, vs. just using one set of rails for all amps. The PSRR of these chips is, after all, merely good, not exceptional.
Thanks macboy.
I am also concerned about inter-amplifier power coupling effects. I have a 600VAC xformer mounted in the bottom of each speaker so there are individual supplies for left and right by design. I was also considering seperate bridges for the bands but after thinking about it, I don't see that multiple rectifiers is a substantial benefit regarding isolation. I hope to get appreciable isolation gains through individual banks of capacitors for each band, possibly combining mid and high (not a lot of space in the speaker cabinet). What interests me the most is if there will be any audible difference in using reasonable caps and grossly expensive ones.
The main advantage of having several chips in parallel is that you are able to get the output impedance way low, and that's a good thing. I will be impressed if I find that i don't need more power than the single set of BTL's on 25V rails.
I did a short listening test of the Linkwitz Xform on a set of mockup cabinets to see what it sounded like before having a case company builf the real deal. I consider low-end directional and, more importantly, it has to be created in the same physical manner that the rest of the spectrum (or low end part at least) is. When 80-250Hz is coming out of one size box and driver and everything below that is coming out of a sub, I don't think that you can get the imaging right. So I wanted to try to get a three way system to be flat all the way down, understanding that the output wasn't going to be all that. I have to say that I am more than pleased with the overall response of the proto's. I do concert sound for a living and am used to tons of low end that isn't necessarily well-imaged with the rest of the system. The proto's are vastly different, and they are using a 7.5" woofer with something like 7mm Xmax! Of course there's no sternum crushing going on, but they still get loud, and they go way dooowwwwnnnn.
I am also concerned about inter-amplifier power coupling effects. I have a 600VAC xformer mounted in the bottom of each speaker so there are individual supplies for left and right by design. I was also considering seperate bridges for the bands but after thinking about it, I don't see that multiple rectifiers is a substantial benefit regarding isolation. I hope to get appreciable isolation gains through individual banks of capacitors for each band, possibly combining mid and high (not a lot of space in the speaker cabinet). What interests me the most is if there will be any audible difference in using reasonable caps and grossly expensive ones.
The main advantage of having several chips in parallel is that you are able to get the output impedance way low, and that's a good thing. I will be impressed if I find that i don't need more power than the single set of BTL's on 25V rails.
I did a short listening test of the Linkwitz Xform on a set of mockup cabinets to see what it sounded like before having a case company builf the real deal. I consider low-end directional and, more importantly, it has to be created in the same physical manner that the rest of the spectrum (or low end part at least) is. When 80-250Hz is coming out of one size box and driver and everything below that is coming out of a sub, I don't think that you can get the imaging right. So I wanted to try to get a three way system to be flat all the way down, understanding that the output wasn't going to be all that. I have to say that I am more than pleased with the overall response of the proto's. I do concert sound for a living and am used to tons of low end that isn't necessarily well-imaged with the rest of the system. The proto's are vastly different, and they are using a 7.5" woofer with something like 7mm Xmax! Of course there's no sternum crushing going on, but they still get loud, and they go way dooowwwwnnnn.
Banned
Joined 2002
WasteRunner,
I don't know if youv'e considered this, but for a bridged amp, you do not need gobs of capactiance between each rail and ground, since the speaker current will always flow from the V+ rail to the V- rail (never from + to gnd or gnd to -). So your big caps need only to be directly across the V+ and V- rails, and in fact this is preferred, because it keeps the speaker currents from influencing the ground voltage. This also gives you the benefit of greater energy storage in the caps for a given uF rating (because you doubled the voltage across the caps, and Energy is proportional to Voltage squared). You will of course need caps of appropriate voltage rating. You will still need small caps on each rail to gnd (I'd use maybe a couple hundred uF, more if the pre-amp opamps shared these rails).
I don't know if youv'e considered this, but for a bridged amp, you do not need gobs of capactiance between each rail and ground, since the speaker current will always flow from the V+ rail to the V- rail (never from + to gnd or gnd to -). So your big caps need only to be directly across the V+ and V- rails, and in fact this is preferred, because it keeps the speaker currents from influencing the ground voltage. This also gives you the benefit of greater energy storage in the caps for a given uF rating (because you doubled the voltage across the caps, and Energy is proportional to Voltage squared). You will of course need caps of appropriate voltage rating. You will still need small caps on each rail to gnd (I'd use maybe a couple hundred uF, more if the pre-amp opamps shared these rails).
I hadn't considered that macboy. It sounds like a good idea to decouple the bridged pairs to the supplies instead of the ground, since that is essentially were the laod current is ending up anyway. As far as doubling the effective onboard capacitance, I wonder how sharing the big EL caps between the two chips will combine with decoupling to ground. They share them in the design because, being out of phase, they draw off opposite caps and thus don't need individual ones. I wonder if that quadruples the effective capacitance or if the effects cancel. I'll have to think about it. To me, tricks like this is what makes superior performing devices.
What power supply arrangements have you tried and what differences have you noticed?
What is the safest way to share files in this forum without posting my email?
What power supply arrangements have you tried and what differences have you noticed?
What is the safest way to share files in this forum without posting my email?
Attachments
Today I was able to get everything going. The circuitry performance is pretty much what I was expecting, although I'm not certain that LM3886's sound the same as LM3875's. The relays that disconnect the AC mains from the bridge that feeds the LM3886T's work as expected. The crossover, processing and servos stay powered while the power section is disconnected when not in use to lower Iq draw. I've archived my CD collection in iTunes (lossless) and send digital out of my laptop to a Theta DAC, so the plan is to only have to turn on the laptop and get music (easiest for the G-fry, too). Connecting the power section is quiet, but powering it down causes an audible transient. I've built LM3875 amps that don't pop, so I need to figure out why this one does. Also, the noise floor is not low, and this is probably due to there being no true shielding for the circuit board, maybe lining the cabinet with foil will help, and/or grounding the heatsink.
The cabinet design and performance is somethign else entirely. I plan on using SMAART to get the actual measured acoustic response, but just by listening I can tell several things. The low end is not balanced in the sense that there is too much from 20 up to 60Hz or so. I would guess that this is either because the parameters of the driver are inaccurate or, more likely, the proto cabinet volume is larger than the target volume. The proto cab used scraps so I was somewhat limited to certain sizes, the resultant volume being different from the one the Linkwitz Transform values were chosen for. I shot for larger thinking that it would be easier to reduce the internal volume until it was right for the circuit.
I expected the midrange to require the most tweaking and that is proving to be the case. The goal is to get it to be flat from 300 to 6k. I cut a hole in the back of the cabinet which basically makes the space behind the driver a square tube, the plan being to cut out a long block of various materials to try to get a happy medium between a sealed back and open baffle. The circuit board has a stage of passive EQ that is probably going to be needed to reduce the output above 1k. Hopefully this little driver will be able to do 300Hz well enough to keep up with the top of the woofer.
I plan on letting the drivers break in before doing too much tweaking. I'll do a moderate amount of updating to this thread as things progress.
The cabinet design and performance is somethign else entirely. I plan on using SMAART to get the actual measured acoustic response, but just by listening I can tell several things. The low end is not balanced in the sense that there is too much from 20 up to 60Hz or so. I would guess that this is either because the parameters of the driver are inaccurate or, more likely, the proto cabinet volume is larger than the target volume. The proto cab used scraps so I was somewhat limited to certain sizes, the resultant volume being different from the one the Linkwitz Transform values were chosen for. I shot for larger thinking that it would be easier to reduce the internal volume until it was right for the circuit.
I expected the midrange to require the most tweaking and that is proving to be the case. The goal is to get it to be flat from 300 to 6k. I cut a hole in the back of the cabinet which basically makes the space behind the driver a square tube, the plan being to cut out a long block of various materials to try to get a happy medium between a sealed back and open baffle. The circuit board has a stage of passive EQ that is probably going to be needed to reduce the output above 1k. Hopefully this little driver will be able to do 300Hz well enough to keep up with the top of the woofer.
I plan on letting the drivers break in before doing too much tweaking. I'll do a moderate amount of updating to this thread as things progress.
Attachments
thanks for the interest. I'll attach a couple nice pictures. Remember they're prototypes and as such not to be judged cosmetically. I realize that these pix probably will drag this thread to the loudspeaker side, but I promise I will crack them open when I get a chance and show the chip amp innards. You can see the 12x5" heatsinks at the bottom and the signal and power cabling beneath that. No worries, its balanced! You can also see the midrange cutout on the back with the sliding panel. Kind of looks silly.
A quick listing of notes on the proto amp board in order to keep this in the correct forum:
I haven't cranked it up for a long time yet to test the limits of thermal dissipation on the sinks. These speaks are loud, loud, loud and with the well represented low end, things get out of hand fairly quickly so exercising the maximum power area will probably be easiest with a dummy load. As it is, with seven LM3886T's on each sink and 25V rails the sinks are just warm enough to be noticeable even with good output volume. The output DC offset averages between 0.5mV and 3mV, I was expecting less but c'mon, let's be realistic. I was concerned about how the servos would behave being powered on without the LM3886's on, but their outputs don't wander off or do anything wacky on power up, and I haven't populated the anti-latch up diodes yet either, haven't had a problem, but they will go in soon as I can find the right ones.
All and all, this project is turning out surprisingly well. There have been some frustrations, but I've been listening to these protos with ballpark gain set resistors, jankie midrange cutout, no baffle step compensation and one transformer and they are fairly convincing, which is a good thing becasue it is wrong to expect too much more than the out-of-the-box performance from anything. No SMAART data yet, I am trying to get my USB pre back from my monitor engineer friend whose been on the road.
At this point I am eagerly thinking of the larger versions, because I really want to see what the amp circuitry can do @ 35V rails. I also want to take these protos into larger rooms and see how they react. I am wondering if the excessive low end is from room gain as my apartment is somewhat long and narrow.
I have attempted to make the individual power stages as identical as possible because I think that the precision and coherence of the resulting system benefits from it, the main goal being to get the speakers to disappear. There are several things that I plan on trying as soon as I am familiar with the unaltered performance of the protos. Comments and suggestions welcome.
A quick listing of notes on the proto amp board in order to keep this in the correct forum:
I haven't cranked it up for a long time yet to test the limits of thermal dissipation on the sinks. These speaks are loud, loud, loud and with the well represented low end, things get out of hand fairly quickly so exercising the maximum power area will probably be easiest with a dummy load. As it is, with seven LM3886T's on each sink and 25V rails the sinks are just warm enough to be noticeable even with good output volume. The output DC offset averages between 0.5mV and 3mV, I was expecting less but c'mon, let's be realistic. I was concerned about how the servos would behave being powered on without the LM3886's on, but their outputs don't wander off or do anything wacky on power up, and I haven't populated the anti-latch up diodes yet either, haven't had a problem, but they will go in soon as I can find the right ones.
All and all, this project is turning out surprisingly well. There have been some frustrations, but I've been listening to these protos with ballpark gain set resistors, jankie midrange cutout, no baffle step compensation and one transformer and they are fairly convincing, which is a good thing becasue it is wrong to expect too much more than the out-of-the-box performance from anything. No SMAART data yet, I am trying to get my USB pre back from my monitor engineer friend whose been on the road.
At this point I am eagerly thinking of the larger versions, because I really want to see what the amp circuitry can do @ 35V rails. I also want to take these protos into larger rooms and see how they react. I am wondering if the excessive low end is from room gain as my apartment is somewhat long and narrow.
I have attempted to make the individual power stages as identical as possible because I think that the precision and coherence of the resulting system benefits from it, the main goal being to get the speakers to disappear. There are several things that I plan on trying as soon as I am familiar with the unaltered performance of the protos. Comments and suggestions welcome.
Attachments
Thank you, Upupa. I've noticed that in all the 'IMHO' murk on the web, you generally have intelligent criticism of what gets thrown out there, so I appreciate the thumbs up. I'm curious about what aspect of this project struck you as being done well. Wish I could show you the PCB layouts of this ultra-high performance DJ mixer that my company just came out with. Not a single electrolytic in the signal path and it uses external power supplies for 18V rails! It sounds amazing and we've had great responses to it. I'm particularly proud of the phono pre's in it.
I plan on A-B'ing the protos against my friend's Response 2.5's driven with a Spectral amp and Theta converters; easily one of the higher fidelity systems that I've heard outside a mixdown suite. My hypothesis is that, although the actual devices used for the amplification have an effect on the sound, this effect is in the noise compared to the precision of the passive devices, the response of the power supply and various other things. So, if you are in control of as much as possible from the DAC comzouta to the actual transducer gozhinta, you can pretty much accomplish anything you want to, performance-wise. The LMxxx chips are basically a moderately-sized output pair with a differential front-end. Genious.
I plan on A-B'ing the protos against my friend's Response 2.5's driven with a Spectral amp and Theta converters; easily one of the higher fidelity systems that I've heard outside a mixdown suite. My hypothesis is that, although the actual devices used for the amplification have an effect on the sound, this effect is in the noise compared to the precision of the passive devices, the response of the power supply and various other things. So, if you are in control of as much as possible from the DAC comzouta to the actual transducer gozhinta, you can pretty much accomplish anything you want to, performance-wise. The LMxxx chips are basically a moderately-sized output pair with a differential front-end. Genious.
Nice pics
Well done. For a prototype you did PDG. Looks like you used a lower grade birch. It makes a great substrate for veneer. Let me know if you need any help in that regard. I can help! Been a cabinet maker for almost twenty years and done nothing but fancy stuff for the bulk of it.
Mark
Well done. For a prototype you did PDG. Looks like you used a lower grade birch. It makes a great substrate for veneer. Let me know if you need any help in that regard. I can help! Been a cabinet maker for almost twenty years and done nothing but fancy stuff for the bulk of it.
Mark
Actually, Mark, the finish is the thing that we are really hung up on. I don't think that enough went into the proto cabs to do anything other than stain or paint them, but I am completely in love with the ProAc exotic wood finishes, bird's eye maple being my favorite. We had assumed that it was veneering of some type because of the edging quality. Is it possible to do that type of thing without considerable start up? I think I'd be willing to trade some of my design for cabinetry tips and sourcing of wood and the like, if you are interested. Thanks.
Let me run some of my ideas past you and see what you think. Sorry for the long-windidness.
In this design I'm most concerned with getting the mid range to be as flat as possible before resorting to electrical processing. I'm using the Scanspeak 12M/4631G covering from 300 to about 6k or so. Forcing this driver to do all that reigns in its output abilities somewhat, reducing the need for copious amounts of low end, hence the use of a smaller-than-usual low-end driver, which is the Scanspeak 18W/8545K. I realize that the Xmax being small (~6mm) reduces the amount of sub that this driver will produce, but I am guessing that by the time the 12M is just running out of 300Hz, the 18W will be doing the same at 25Hz.
I'm assuming that the 18W being in a sealed cab will reduce the need for mathematical design of the actual cabinet, aside from symmetry in the space surounding it and rigidity. Its the mid range that I'm most concerned with, and, to avoid internal reflections, had envisioned using an open-backed cabinet stuffed with foam of some kind, or possibly layers of increasing density, or something else entirely. Maybe it won't be possible without a sealed cabinet behind it. What are your thoughts?
In this design I'm most concerned with getting the mid range to be as flat as possible before resorting to electrical processing. I'm using the Scanspeak 12M/4631G covering from 300 to about 6k or so. Forcing this driver to do all that reigns in its output abilities somewhat, reducing the need for copious amounts of low end, hence the use of a smaller-than-usual low-end driver, which is the Scanspeak 18W/8545K. I realize that the Xmax being small (~6mm) reduces the amount of sub that this driver will produce, but I am guessing that by the time the 12M is just running out of 300Hz, the 18W will be doing the same at 25Hz.
I'm assuming that the 18W being in a sealed cab will reduce the need for mathematical design of the actual cabinet, aside from symmetry in the space surounding it and rigidity. Its the mid range that I'm most concerned with, and, to avoid internal reflections, had envisioned using an open-backed cabinet stuffed with foam of some kind, or possibly layers of increasing density, or something else entirely. Maybe it won't be possible without a sealed cabinet behind it. What are your thoughts?
Attachments
Help is on the way
I'm game for whatever you have for ideas. But this thread is in chip amps. And I personally hate it when threads meander. I'd be happy to discuss this in private e-mails. When design stuff gets going on the threads there are a couple of very knowledgable people and a great many who think they are. I've seen people go off track many times. Hapy to help. But please be a reasoning recipient
Mark
I'm game for whatever you have for ideas. But this thread is in chip amps. And I personally hate it when threads meander. I'd be happy to discuss this in private e-mails. When design stuff gets going on the threads there are a couple of very knowledgable people and a great many who think they are. I've seen people go off track many times. Hapy to help. But please be a reasoning recipient
Mark
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.