No one is pulling rank on you here. And, believe me, there are some genuine stars here, especially on the electronics front.
I think what you don't appreciate is the scale of those charging spikes and a 50% increase in VA rating isn't really going to make up for the I squared element, especially if you insist on staying with 20,000uF (directly against my advice). You would probably get away with this in a domestic amp with 10,000uF of capacitance but my impression is that this is going to be taxed very much more heavily in normal use. Also, while I don't know where you are going to bring in your mid bass driver, but if your sub is going to be a genuine sub, then a 4" driver is going to have to be pushed very hard to produce any sort of decent spl with only ~50 sq cm of radiating area. (This also makes choosing a horn loaded tweeter seem a very strange choice as just about any tweeter is going to manage the 84 or 85dB/W your mid is able to produce - bear in mind that every 3dB of spl is going to double your power. Personally I would have chosen an 8" high efficiency, low Q, paper driver that starts off with ~220 sq cm of area for this sort of application, especially if it's going to be used outdoors where sound just evaporates. I am very much not a PA expert, but you can see why they often refer to 10" drivers as midranges!) I don't know what you are planning on doing with shaping the response of the sub but if you are going to use a Linkwitz transform then that REALLY eats power! An additional half octave of extension is going to cost you 4x the power, and a whole octave, 16x. You can do it the REL way which is, kinda, more flexible but ultimately the maths is the same; an impassable relationship between power and extension. So, on both sides, your amplifier is going to be pushed rather heavily and, I would suspect, is going to just start the evening at 3/4 volume, with not much room above it left to go (especially on perceived spl).
I'll also add that it's my experience that puny transformers with large value reservoir capacitors tend to make for a very loose and muffled bass end, and the general dullness usually pervades the whole amplifier. You can't really just exchange one for the other, any more than you can vary the values of a capacitor/inductor pair in a crossover at will. What I might try in your position is to reorganise those caps into a CRC, putting something like 0.5R between two banks of 10,000uF if you insist on using them all. This should give you a bit more smoothing (the current draws will not be kind to the smoothness of the rail voltages, at all) and it should be kinder to your transformer - though, frankly, my first move would be to delete 10k of the 20k entirely.
I'd like to give you a simulated figure for what those charging spikes are but, in spite of extensive searching, I couldn't find my power supply model. It is more than 5 years old but I'm still surprised not to have found it. I did have a go with one of the models supplied with LTspice (filename starts with P1_Bridge_rectifier...) but it's very simplistic as a transformer model, is missing any resistive losses, only models one capacitor, and wasn't responding to changes in that C as I expected; I think mainly because there are large inductances that don't actually show up in that way IRL. It also uses Schottky diodes, which give a slightly wider window for charging current to flow. It did have the advantage of being a 110V supply but that doesn't really make up for the other shortcomings. Nonetheless it shows spikes of circa 20A towards the start (I haven't adjusted it for an actual start) and, more generally, draws of +8A peak to -5A for a current load of 3.5A RMS at 50Hz. If my memory is serving me correctly, this is all very, very benign in comparison to what I modelled 6 to 8 years ago, where the spikes were much higher and of much shorter duration. I can't remember the exact peaks they reached (though maybe I'll be able to find the schematics on another computer) but I remember them being alarming and one of the takeaway lessons from that work (alongside what ESR does) was that, contrary to what might seem reasonable intuition, additional smoothing capacitance was not a universal panacea and in fact compounded your problems.
As it happens, a little over 10 days ago I started work on a project to accurately model transformers in linear supplies, partly with a view to finding out why there seem to be "audio" transformers and others that appear not to work well at all. It's more to optimise their use in amplifiers and perhaps extract some important parameters, than to give information on heating and power needs, but it should adapt to those criteria very readily. I'm starting by characterising all the transformers I have lying around, including some from (often defunct) manufacturers with a reputation for being good in audio applications. I feel it's a very overlooked aspect of audio and I'm unconvinced that even those manufacturers with a reputation for being good on power supplies have a full understanding of what's going on. You will probably have built your project before I get a decent way in, but I should be able to tell you what those charging spikes look like and what their magnitude is fairly early on. Certainly they are enough to spray RF all over your amplifier if your leads to the capacitors are too long, which is not what you would expect from the waveforms I saw on the LTspice model I lifted.
I think what you don't appreciate is the scale of those charging spikes and a 50% increase in VA rating isn't really going to make up for the I squared element, especially if you insist on staying with 20,000uF (directly against my advice). You would probably get away with this in a domestic amp with 10,000uF of capacitance but my impression is that this is going to be taxed very much more heavily in normal use. Also, while I don't know where you are going to bring in your mid bass driver, but if your sub is going to be a genuine sub, then a 4" driver is going to have to be pushed very hard to produce any sort of decent spl with only ~50 sq cm of radiating area. (This also makes choosing a horn loaded tweeter seem a very strange choice as just about any tweeter is going to manage the 84 or 85dB/W your mid is able to produce - bear in mind that every 3dB of spl is going to double your power. Personally I would have chosen an 8" high efficiency, low Q, paper driver that starts off with ~220 sq cm of area for this sort of application, especially if it's going to be used outdoors where sound just evaporates. I am very much not a PA expert, but you can see why they often refer to 10" drivers as midranges!) I don't know what you are planning on doing with shaping the response of the sub but if you are going to use a Linkwitz transform then that REALLY eats power! An additional half octave of extension is going to cost you 4x the power, and a whole octave, 16x. You can do it the REL way which is, kinda, more flexible but ultimately the maths is the same; an impassable relationship between power and extension. So, on both sides, your amplifier is going to be pushed rather heavily and, I would suspect, is going to just start the evening at 3/4 volume, with not much room above it left to go (especially on perceived spl).
I'll also add that it's my experience that puny transformers with large value reservoir capacitors tend to make for a very loose and muffled bass end, and the general dullness usually pervades the whole amplifier. You can't really just exchange one for the other, any more than you can vary the values of a capacitor/inductor pair in a crossover at will. What I might try in your position is to reorganise those caps into a CRC, putting something like 0.5R between two banks of 10,000uF if you insist on using them all. This should give you a bit more smoothing (the current draws will not be kind to the smoothness of the rail voltages, at all) and it should be kinder to your transformer - though, frankly, my first move would be to delete 10k of the 20k entirely.
I'd like to give you a simulated figure for what those charging spikes are but, in spite of extensive searching, I couldn't find my power supply model. It is more than 5 years old but I'm still surprised not to have found it. I did have a go with one of the models supplied with LTspice (filename starts with P1_Bridge_rectifier...) but it's very simplistic as a transformer model, is missing any resistive losses, only models one capacitor, and wasn't responding to changes in that C as I expected; I think mainly because there are large inductances that don't actually show up in that way IRL. It also uses Schottky diodes, which give a slightly wider window for charging current to flow. It did have the advantage of being a 110V supply but that doesn't really make up for the other shortcomings. Nonetheless it shows spikes of circa 20A towards the start (I haven't adjusted it for an actual start) and, more generally, draws of +8A peak to -5A for a current load of 3.5A RMS at 50Hz. If my memory is serving me correctly, this is all very, very benign in comparison to what I modelled 6 to 8 years ago, where the spikes were much higher and of much shorter duration. I can't remember the exact peaks they reached (though maybe I'll be able to find the schematics on another computer) but I remember them being alarming and one of the takeaway lessons from that work (alongside what ESR does) was that, contrary to what might seem reasonable intuition, additional smoothing capacitance was not a universal panacea and in fact compounded your problems.
As it happens, a little over 10 days ago I started work on a project to accurately model transformers in linear supplies, partly with a view to finding out why there seem to be "audio" transformers and others that appear not to work well at all. It's more to optimise their use in amplifiers and perhaps extract some important parameters, than to give information on heating and power needs, but it should adapt to those criteria very readily. I'm starting by characterising all the transformers I have lying around, including some from (often defunct) manufacturers with a reputation for being good in audio applications. I feel it's a very overlooked aspect of audio and I'm unconvinced that even those manufacturers with a reputation for being good on power supplies have a full understanding of what's going on. You will probably have built your project before I get a decent way in, but I should be able to tell you what those charging spikes look like and what their magnitude is fairly early on. Certainly they are enough to spray RF all over your amplifier if your leads to the capacitors are too long, which is not what you would expect from the waveforms I saw on the LTspice model I lifted.
Thanks. Here's some back-story, if you're interested. If not, just skip to the bottom.
You have absolutely valid concerns re: applicability of small drivers and low-power amps. I'll address that first:
Generally, for PA stuff, you do indeed need lots of SPL. ... Except when you don't. There are certain times when I bring my main rig to events, and it's just... coasting. Not even coasting, drifting. Like just flickering the first LED on my mixer's level meter. One such example is an outdoor event that spans a few rows of a parking lot. Think farmers market. In those cases, I don't need tons of SPL. I need background music that won't get in the way of conversation. Which leads to the other problem -- my existing setup fills an area around the speakers, but then might not be contributing anything meaningful further away. Similarly, if I'm trying to cover something like a banquet hall. One or two tables right in front of the speaker(s) are getting their left ear pummeled, while the people in the back might not even be able to tell what song is playing. So, I'm trying something different. Think of sound reinforcement in a shopping center: Lots of speakers to cover a larger area, but none of them playing very loud. This is what I'm going for with this build. They may indeed start the evening at 3/4 volume -- but then they'll stay there. That would be perfect. There are events where I know for certain, there is NO tendency for the faders to creep upward over time. This system would not be appropriate for a rock show, and I know that. That's OK. I have other hardware for that.
You're also right about the driver selection. It is definitely arbitrary. Kind of. For the tops, it was 100% driven by whimsy. I came up with this design years ago, because I had a few of the Goldwood horn tweeters spare from a different project. They're well regarded for being remarkably decent for a $2 driver. I thought it would be amusing to (as I stated a couple of posts ago) build a mini-me clone of my Yamaha Club V monitors. The horn was selected because it looks like the compression horn from a PA/stage monitor. That's it. The 4" woofer fit the baffle, and was a paper cone. That was it's selection criteria. They just barely work together, and even still, the crossover region isn't quite as clean a hand-off as I would like, but it passed the "do I stop noticing speakers and just enjoy the music when I play something through this?" test. It didn't have to be practical, it was just for fun. I spent maybe a few hours on a Friday in WinISD and FreeCAD, and a few hours on Saturday turning some scrap off-cuts into a prototype enclosure. The result happened to work well enough, and I'm thrilled with that.
The sub was similar -- it's also a repurposed experiment. I thought the Tang Band 6.5" drivers were neat, and picked up a few of the neodymium variants back before they got super expensive. (I think I paid $40 or $60 each for them. This was over a decade ago.) The box is 0.7 ft^3, tuned to 30Hz. When I built it, I thought, wow that sounds quite nice. Given the choice of keeping the test enclosure, or tossing it into the fire pit of our next backyard hangout, I think I'll go with the former.
Then, I got this idea that .. hmm .. maybe these would work for those small gigs where I just need a little bit of unobtrusive sound. I paired the sub and the one prototype top that I had already built with a little desktop class D amp, used software to handle the crossover and level matching, stuck them in my semi-open dining / living area, and parked myself about 20 ft away. At around 85% volume on the (supposedly) 60w/ch amp -- which, again, was hamstrung on the "tops" channel to match the "sub" channel's sensitivity -- it was loud enough that my ears were ringing a little bit after half an hour or so. Kinda seems like it ought to be enough for what I want to do with it. I guess we'll see.
So that's that. I'm taking a big swing here building a small army of these, but if it doesn't work out, that's alright. I'll find a more sedate domestic use for them. They'll be great in my office and the garage... my mom's living room... etc..
...
OK, now the technical stuff.
First, you mentioned a LInkwitz Transform. I think that might have been a misunderstanding, or maybe just conjecture with a warning that it won't work well. For the record, I'm not using a Linkwitz Transform. There IS a fixed 4th-order Linkwitz-Riley low-pass filter on the sub at 85Hz, and a complimentary LR24 HP filter on the mains output that is selectable between "off", "12dB" which just taps between the two series stages (a bit unorthodox Q, but close enough for jazz), or "on". There is an EQ section, but it's effective on the whole signal, not just the sub, and was not intended to make up for rolloff in a small box. It's just there as a tone control. At max +6dB, it's not too spicy.
Obviously, I'm in over my head with the transformer bit. I don't know what to do about that. I don't have tons of space, so I can't just chuck in the largest toroid that I can get my hands on and call it a day. So I'm going to have to find out what is juuuuust big enough. Since I already have a 250VA, I might just need to go ahead and cobble together a prototype PCB with just the amp ICs (and their immediate life-support system), and see what it is, and isn't, capable of. I would much rather design it properly, but If YOU are getting results you don't trust with the LTspice model you have, I don't see any hope for me to try and get this right on the first try. I wouldn't know if I'm looking at a false positive, or a false negative, or if it's dead on. So I might just have to build an underpowered one, and find out how many orders of magnitude too-small it really is, by running sustained tests at 25%, 50%, 75% .. see when the toroid starts getting warm, then try a larger size and run the tests again. I might end up with a few spare toroids, but that wouldn't be the worst problem to have, and I genuinely don't know how to solve this otherwise.
I'm starting to catch on with what you're saying about the capacitor bank. You're right, it's not immediately intuitive. My very basic sims were just geared to finding out how much capacitance I needed to keep the ripple low watermark above the clipping threshold. Maybe I need to try a different approach. I've been trying to keep the operating voltage as low as possible, but I have a few volts headroom still. I could try bumping that up to the next highest secondary voltage, and provide more margin for ripple, which would allow for a smaller bulk capacitance. Does that sound reasonable? Or am I still on the wrong path?
If I end up needing 1000VA to get something in the neighborhood of 120W of total output, then I need to give up and just buy a switching supply. That's out of my league to design.
You have absolutely valid concerns re: applicability of small drivers and low-power amps. I'll address that first:
Generally, for PA stuff, you do indeed need lots of SPL. ... Except when you don't. There are certain times when I bring my main rig to events, and it's just... coasting. Not even coasting, drifting. Like just flickering the first LED on my mixer's level meter. One such example is an outdoor event that spans a few rows of a parking lot. Think farmers market. In those cases, I don't need tons of SPL. I need background music that won't get in the way of conversation. Which leads to the other problem -- my existing setup fills an area around the speakers, but then might not be contributing anything meaningful further away. Similarly, if I'm trying to cover something like a banquet hall. One or two tables right in front of the speaker(s) are getting their left ear pummeled, while the people in the back might not even be able to tell what song is playing. So, I'm trying something different. Think of sound reinforcement in a shopping center: Lots of speakers to cover a larger area, but none of them playing very loud. This is what I'm going for with this build. They may indeed start the evening at 3/4 volume -- but then they'll stay there. That would be perfect. There are events where I know for certain, there is NO tendency for the faders to creep upward over time. This system would not be appropriate for a rock show, and I know that. That's OK. I have other hardware for that.
You're also right about the driver selection. It is definitely arbitrary. Kind of. For the tops, it was 100% driven by whimsy. I came up with this design years ago, because I had a few of the Goldwood horn tweeters spare from a different project. They're well regarded for being remarkably decent for a $2 driver. I thought it would be amusing to (as I stated a couple of posts ago) build a mini-me clone of my Yamaha Club V monitors. The horn was selected because it looks like the compression horn from a PA/stage monitor. That's it. The 4" woofer fit the baffle, and was a paper cone. That was it's selection criteria. They just barely work together, and even still, the crossover region isn't quite as clean a hand-off as I would like, but it passed the "do I stop noticing speakers and just enjoy the music when I play something through this?" test. It didn't have to be practical, it was just for fun. I spent maybe a few hours on a Friday in WinISD and FreeCAD, and a few hours on Saturday turning some scrap off-cuts into a prototype enclosure. The result happened to work well enough, and I'm thrilled with that.
The sub was similar -- it's also a repurposed experiment. I thought the Tang Band 6.5" drivers were neat, and picked up a few of the neodymium variants back before they got super expensive. (I think I paid $40 or $60 each for them. This was over a decade ago.) The box is 0.7 ft^3, tuned to 30Hz. When I built it, I thought, wow that sounds quite nice. Given the choice of keeping the test enclosure, or tossing it into the fire pit of our next backyard hangout, I think I'll go with the former.
Then, I got this idea that .. hmm .. maybe these would work for those small gigs where I just need a little bit of unobtrusive sound. I paired the sub and the one prototype top that I had already built with a little desktop class D amp, used software to handle the crossover and level matching, stuck them in my semi-open dining / living area, and parked myself about 20 ft away. At around 85% volume on the (supposedly) 60w/ch amp -- which, again, was hamstrung on the "tops" channel to match the "sub" channel's sensitivity -- it was loud enough that my ears were ringing a little bit after half an hour or so. Kinda seems like it ought to be enough for what I want to do with it. I guess we'll see.
So that's that. I'm taking a big swing here building a small army of these, but if it doesn't work out, that's alright. I'll find a more sedate domestic use for them. They'll be great in my office and the garage... my mom's living room... etc..
...
OK, now the technical stuff.
First, you mentioned a LInkwitz Transform. I think that might have been a misunderstanding, or maybe just conjecture with a warning that it won't work well. For the record, I'm not using a Linkwitz Transform. There IS a fixed 4th-order Linkwitz-Riley low-pass filter on the sub at 85Hz, and a complimentary LR24 HP filter on the mains output that is selectable between "off", "12dB" which just taps between the two series stages (a bit unorthodox Q, but close enough for jazz), or "on". There is an EQ section, but it's effective on the whole signal, not just the sub, and was not intended to make up for rolloff in a small box. It's just there as a tone control. At max +6dB, it's not too spicy.
Obviously, I'm in over my head with the transformer bit. I don't know what to do about that. I don't have tons of space, so I can't just chuck in the largest toroid that I can get my hands on and call it a day. So I'm going to have to find out what is juuuuust big enough. Since I already have a 250VA, I might just need to go ahead and cobble together a prototype PCB with just the amp ICs (and their immediate life-support system), and see what it is, and isn't, capable of. I would much rather design it properly, but If YOU are getting results you don't trust with the LTspice model you have, I don't see any hope for me to try and get this right on the first try. I wouldn't know if I'm looking at a false positive, or a false negative, or if it's dead on. So I might just have to build an underpowered one, and find out how many orders of magnitude too-small it really is, by running sustained tests at 25%, 50%, 75% .. see when the toroid starts getting warm, then try a larger size and run the tests again. I might end up with a few spare toroids, but that wouldn't be the worst problem to have, and I genuinely don't know how to solve this otherwise.
I'm starting to catch on with what you're saying about the capacitor bank. You're right, it's not immediately intuitive. My very basic sims were just geared to finding out how much capacitance I needed to keep the ripple low watermark above the clipping threshold. Maybe I need to try a different approach. I've been trying to keep the operating voltage as low as possible, but I have a few volts headroom still. I could try bumping that up to the next highest secondary voltage, and provide more margin for ripple, which would allow for a smaller bulk capacitance. Does that sound reasonable? Or am I still on the wrong path?
If I end up needing 1000VA to get something in the neighborhood of 120W of total output, then I need to give up and just buy a switching supply. That's out of my league to design.
Last edited:
Alright, I had a look, and between Digi-Key, Mouser, and Parts-Express, my only option for "more than 24V" is "25V" -- and I wonder if that's even significant, or lost in the difference between two manufacturers' measurements.
The next step up is 30V, which RMS->Peak - Vfwd is already at 41V, with no margin for 10% high line. pffff.... So much for that idea.
I have two options for larger capacity. Digi-Key has a 500VA @ 24-0-24. It's too big. It won't fit. Parts Express has a 330VA at 25-0-25. It will fit -- just barely.
The next step up is 30V, which RMS->Peak - Vfwd is already at 41V, with no margin for 10% high line. pffff.... So much for that idea.
I have two options for larger capacity. Digi-Key has a 500VA @ 24-0-24. It's too big. It won't fit. Parts Express has a 330VA at 25-0-25. It will fit -- just barely.
You are not going to need a 1kVA transformer, that's for sure. From your description this is looking more like a domestic system that is going to be played loud - with the reminder that 6dB is 4x power. 250VA is likely to be fine, especially if you are taking multiples of these to each venue. You might like to try and point them at each other, rather as though they were walls, so you get a bit of extra radiation resistance, which translates into higher efficiencies. Hell, 160VA might just scrape by on a satellite if you know it's not going to be strained. But neither into 20,000uF. (By all means try it and let me know. I'm open to all new data, even anecdotal data.)
will cont below...
will cont below...
If you want a smidgeon more or less voltage then it might very well be worth playing with the bridge arrangement. You can go for another bridge recifier to get another diode drop or reduce your diode drop by using a set of Schottky diodes. If you are going to be drawing the currents we've been talking about then it's almost never going to reach its nominal voltage.
back to my original reply...
back to my original reply...
So if this is going to be effectively a domestic system then you can indulge in the same cost-cutting liberties that all manufacturers do - like work out the cooling needed and then divide it by 4. 😀😀😀
The LR4 is my pet hate crossover alignment for a plethora of reasons, but one is particularly relevant here, and that is the 3dB dip in power response at crossover. That's half power! In normal 2-ways this reduces the band playing to mannequins dancing along your mantlepiece, but here you are sacrificing power at just the point that gives you slam and power in the bass. Your choice of crossover point is also very close to the peak average spectral output, which is around 120Hz, which is currently being handled by your weakest link driver. In fact it's probably being saved a little by the LR4 (I can't remember exactly how wide this dip is because I never use L-Rs, ever) and it would probably take a little bit of work to see how a change affects the power needs if changing the alignment, but I would seriously consider moving it up to 120Hz, or above, to remove some of the excursion demands on the 4". At least share them in some way.
As I write this I realise I am not at all clear how this is being done. Is it two amplifiers, with one on the bass (sub) and one for a passively crossed over two-way top - both driven from the same supply - and an active x-o in between? And does each amplifier have two chips in parallel?
While I remember, a little tip that will definitely help with the cooling. Have the fan on at all times, but at an inaudible level. This can preempt demands for cooling and increase your thermal limits in practice. In many cases, by the time a fan switches on it is almost inevitably going to go to cut out at some point later on. This can avoid that by keeping the basal temperature lower, though still warm enough to not have the newly switched on sound. You basically save a margin of heat capacity so that amount more energy is needed before it starts to reach unretrievable levels, by which time the event is usually over.
The LR4 is my pet hate crossover alignment for a plethora of reasons, but one is particularly relevant here, and that is the 3dB dip in power response at crossover. That's half power! In normal 2-ways this reduces the band playing to mannequins dancing along your mantlepiece, but here you are sacrificing power at just the point that gives you slam and power in the bass. Your choice of crossover point is also very close to the peak average spectral output, which is around 120Hz, which is currently being handled by your weakest link driver. In fact it's probably being saved a little by the LR4 (I can't remember exactly how wide this dip is because I never use L-Rs, ever) and it would probably take a little bit of work to see how a change affects the power needs if changing the alignment, but I would seriously consider moving it up to 120Hz, or above, to remove some of the excursion demands on the 4". At least share them in some way.
As I write this I realise I am not at all clear how this is being done. Is it two amplifiers, with one on the bass (sub) and one for a passively crossed over two-way top - both driven from the same supply - and an active x-o in between? And does each amplifier have two chips in parallel?
While I remember, a little tip that will definitely help with the cooling. Have the fan on at all times, but at an inaudible level. This can preempt demands for cooling and increase your thermal limits in practice. In many cases, by the time a fan switches on it is almost inevitably going to go to cut out at some point later on. This can avoid that by keeping the basal temperature lower, though still warm enough to not have the newly switched on sound. You basically save a margin of heat capacity so that amount more energy is needed before it starts to reach unretrievable levels, by which time the event is usually over.
Interesting point. A cursory glance through parametric search yields I can source a Schottky diode rated at 10A, Vf of 570mV, maybe a little tight at 45V reverse voltage. Not really sure what the prevailing wisdom is on how close I can ride that line -- the standard rectifier was rated at 600V reverse, so it wasn't even worth considering.
That would bump me up from 32.9V to 33.37V. Not a lot of margin, but it might be worth doing anyway. Every little bit counts.
That would bump me up from 32.9V to 33.37V. Not a lot of margin, but it might be worth doing anyway. Every little bit counts.
I'm not the best person to ask on this subject, but DIYAudio is definitely the best place. This should fit the bill and is not much more than jellybean Schottkys, if there is such a thing. Somewhere recently I was looking at a schottky and its Vfwd went vertically from 0.2V. 🙂 It did curve eventually. I wonder if I downloaded the datasheet? I usually do.
https://www.ween-semi.com/sites/default/files/2018-11/BYV34X-600.pdf
https://www.ween-semi.com/sites/default/files/2018-11/BYV34X-600.pdf
Sorry, this is the one I meant to post. That first one isn't a Shottky.
https://www.vishay.com/docs/88721/sb520.pdf
https://www.vishay.com/docs/88721/sb520.pdf
The plan is for an active 2.1 system, 4x LM3886 in total -- 2x parallel for the sub, one per channel for the stereo output.
The sub's filter is fixed @ 85Hz, with two cascaded SK 2nd-order active filter stages.
The "2.0" half has a switch-selectable filter. Either bypassed (full-range), 12dB/oct. (tapped from the mid-point of the cascaded filters), or 24dB/oct. (after both filters). The 2-way tops use a passive XO.
All the filter stages are Q=0.5, which I'm now realizing is not correct for LR4. I should've been using Q=0.7 (Butterworth) for a proper LR4. Hmm.. I'm glad you mentioned that.
The top half is definitely playing things a bit fast-and-loose, since the Q will be a compromise on one or the other slope. If I were doing this in DSP, I would have a variable filter cutoff, 1st/2nd/3rd/4th-order, and options for Q of 0.5 or 0.7. If I were only ever going to use THESE tops with THIS sub, placed in close proximity, at the same level, I would just build a filter with whatever attributes yielded a correct acoustic response and make it fixed as well.
I know this is not ideal, but I think there are going to be so many variables that, in reality, the alignment is never going to be correct, and it probably doesn't really matter that much because there are likely going to be more significant acoustic issues anyway. I did at least implement a polarity switch (which just bypasses an inverting op-amp at the end of the filter chain), so I can, at minimum, select from two sub-optimal choices with the least phase error. 🙂
...
Re: Fans: I'm inclined to agree with this. I've considered whether to use a 2-stage speed control, or just go ahead and run the fan at full speed. I'm not sure it will matter. I don't think I really need that much airflow. Anything is going to be better than passive convection, which is the norm. I'm putting this in an enclosed chamber with vent holes, and shoe-horning 4x ICs on one ~6x6x1" aluminum sink, so I can't take TOO much for granted, but I think I'll be able to use a quiet enough fan (probably blower, technically) that, if music playing, you're not going to ever hear the fan.
The sub's filter is fixed @ 85Hz, with two cascaded SK 2nd-order active filter stages.
The "2.0" half has a switch-selectable filter. Either bypassed (full-range), 12dB/oct. (tapped from the mid-point of the cascaded filters), or 24dB/oct. (after both filters). The 2-way tops use a passive XO.
All the filter stages are Q=0.5, which I'm now realizing is not correct for LR4. I should've been using Q=0.7 (Butterworth) for a proper LR4. Hmm.. I'm glad you mentioned that.
The top half is definitely playing things a bit fast-and-loose, since the Q will be a compromise on one or the other slope. If I were doing this in DSP, I would have a variable filter cutoff, 1st/2nd/3rd/4th-order, and options for Q of 0.5 or 0.7. If I were only ever going to use THESE tops with THIS sub, placed in close proximity, at the same level, I would just build a filter with whatever attributes yielded a correct acoustic response and make it fixed as well.
I know this is not ideal, but I think there are going to be so many variables that, in reality, the alignment is never going to be correct, and it probably doesn't really matter that much because there are likely going to be more significant acoustic issues anyway. I did at least implement a polarity switch (which just bypasses an inverting op-amp at the end of the filter chain), so I can, at minimum, select from two sub-optimal choices with the least phase error. 🙂
...
Re: Fans: I'm inclined to agree with this. I've considered whether to use a 2-stage speed control, or just go ahead and run the fan at full speed. I'm not sure it will matter. I don't think I really need that much airflow. Anything is going to be better than passive convection, which is the norm. I'm putting this in an enclosed chamber with vent holes, and shoe-horning 4x ICs on one ~6x6x1" aluminum sink, so I can't take TOO much for granted, but I think I'll be able to use a quiet enough fan (probably blower, technically) that, if music playing, you're not going to ever hear the fan.
Oh, I forgot to address the filter point. I could try and move the Fc around and see what happens. I do actually want to run some measurements, just for my own curiosity, on the summed response, with the final enclosures, and the final passive XO, and the actual amp PCB more or less finished. The preamp will be a different PCB -- mostly due to how I need to pack things into the space I have, and how various parts like ICs and knobs want to be oriented relative to their PCB.
During my test, everything seemed totally happy with a filter point of 85Hz, tops vented. No signs of distress from the 4" woofer. My model says I should be OK. It's close to Xmax, but should be under the line until I push them past their thermal comfort zone anyway. Full-range, they're OK at low power, but unsurprisingly, can't be pushed as far. I haven't actually ran them until they bottomed out though. I will at some point, because I want to know where their limits are. But.. SO FAR ... it seems they are as capable as I need them to be.
So, I have some flexibility on filter point, and can fiddle with that to see if something sounds better than what I chose. (Which was chosen half by "that's about right" and half "this uses part values I'm already using elsewhere.")
During my test, everything seemed totally happy with a filter point of 85Hz, tops vented. No signs of distress from the 4" woofer. My model says I should be OK. It's close to Xmax, but should be under the line until I push them past their thermal comfort zone anyway. Full-range, they're OK at low power, but unsurprisingly, can't be pushed as far. I haven't actually ran them until they bottomed out though. I will at some point, because I want to know where their limits are. But.. SO FAR ... it seems they are as capable as I need them to be.
So, I have some flexibility on filter point, and can fiddle with that to see if something sounds better than what I chose. (Which was chosen half by "that's about right" and half "this uses part values I'm already using elsewhere.")
Why is the sub fixed? Can't you just change the value of the capacitors or resistors? What's the in box resonance of the 4" and what's its Q?
If I'm reading these graphs right, and if I'm going to be seeing multi-A peaks during the charging cycle, this is more like 0.5V fwd as well. You get 0.3V up to about 1A, then the line starts curving further to the right, and by 5A, it looks like it's up to half a volt. Unless I'm mistaken anyway, which is possible. 🙂
If the two responses are going to be sub-optimal then I would have thought that the LR alignment, unfathomably fashionable though it is, which relies on one response overlaying the other, would be the very worst combination you could have. Your enemy at the moment, I would have thought, is excursion in that 4". Vented could be a solution to that but you have to be very bloody clever. The more usual solution is to seal the cabinet and make it as small as reasonably possible. I would take it to small enough for a Q of 1 and shove a 1st Order in front of it to make a HP Butterworth. Then do your filter on the LP as a 3rd Order Butterworth at the same frequency, which you may as well do round a single op amp.
Incidentally, you don't need DSP for any of this. All this low frequency stuff translates direct from theory into action. You don't even need to take a microphone out for it! (Which is a good thing because accurate LF measurements are a pig.) 🙂
Edit: I have a niggle at the back of my mind that there may be a really clever way to do this, which also includes reducing the amount of power the mid has to take. I'm sure I've done this is some other work, but I may have to sleep on it for it to come to me.
Incidentally, you don't need DSP for any of this. All this low frequency stuff translates direct from theory into action. You don't even need to take a microphone out for it! (Which is a good thing because accurate LF measurements are a pig.) 🙂
Edit: I have a niggle at the back of my mind that there may be a really clever way to do this, which also includes reducing the amount of power the mid has to take. I'm sure I've done this is some other work, but I may have to sleep on it for it to come to me.
In a nutshell -- because parts.
I built a single LM3886 amp a few years back ... pretty much copied the datasheet typical application circuit, and I don't even remember what the toroid in there was rated at, but it never gets pushed, so it hasn't ever been a concern. I used it in a tapped horn (same driver, actually) and paired it with some computer speakers for my desk. It has a variable lowpass filter, but because multi-gang pots are harder to source, it's only 2nd-order. I don't really like running subs with slopes that shallow, so I decided this one was going to be 4th-order. But, that makes it a lot harder to implement a variable Fc. And if I wanted it to be a one-knob variable filter for both HPF and LPF? pfff... Sure. Bourns, can you get me a quote on qty 1 of an 8-gang pot, please? haha Maybe it's better now, or I could try hunting down something from one of the sites that sprung up to help out the guys in Group DIY. But when I was looking at those during that first build, they were basically unobtanium unless you were willing to buy 1000 of them.
Right now, it's "fixed" in that it is set in hardware. But, of course, I can swap resistors and caps all I want while I'm testing. It's not what I would call convenient, and definitely not something I can do out in the field, but for testing? Sure.
The tops have an impedance peak at 34Hz and 97Hz, according to WinISD. The excursion is at its peak at 85Hz, dips way down at 60Hz, and goes unloaded below that. By 50Hz, it's beyond Xmax, and by 40Hz, the cone is laying on the ground in front of the speaker. If you switch the HPF to 12dB @ 85Hz, the excursion peaks at 100Hz, and is below Xmax (4mm) @ 43W. It looks much the same at 24dB/oct., except with less than 1mm below 70Hz.
I'm actually not sure how to calculate box Q.
EDIT: If you are interested in satisfying your own curiosity, the box has a net usable volume of about 215 in^3, tuned to 60Hz, with 2x round ports 0.804" diameter (3/4" PVC pipe ID), 5.4" long, with an end correction factor of 0.85. The driver parameters have changed a tiny bit in the old and current datasheet for the Dayton Audio DS115-8 driver, but I'm using Qes 0.46, Qms 2.1, Fs 55.2, Mms 7.9g, Re 5.8 ohms, Le 0.8mH, Sd 54.1 cm^2, Xmax 4.1mm, Pe 35W. Sorry for the mix of metric and imperial -- I'm using whatever the relevant parts align most naturally with.
Last edited:
That sounds about right. What do the ones you found do? It's a damn sight better than your average, probably automotive, 35A bridge (of the sort that I generally use).
We are on probably the very best forum in the world for people who have used Schottkys. More people have probably tweaked with them here than in all the audio manufacturers put together. 🙂 And I'm not one of those. I don't mind being wasteful and the main problem of bridges, which is ringing, has been solved. Have a look for a thread on it. Or if you can't find one, ask.
I kinda picked one at random from a list of similar options, and ended up with the Vishay MBR1045. The graph is very similar to the one you posted. Digi-Key's parametric search had it rated as 0.57V @ 10A, but it's only 0.3V @ 1A. Similar.
I lost an afternoon once reading about bridge rectifier ringing. I still don't quite understand it. I think the problem is the change from forward-conducting to reverse-biased and blocking, turning on a dime when the rectified mains waveform cross the threshold of where the caps are charged to. Effectively going open-circuit at somewhere just under the peak of the waveform. Even if I did manage to glean the specifics of the problem, I haven't quite figured out "who cares?" yet. I don't know if this is a reliability concern for the diodes, or introduces noise in the supply that the PSRR can't reject, or if it's emitting RF noise that could be inductively coupled into low-level signal traces, or what. So, yeah, I probably should lose another afternoon and see if things get any clearer. 🙂
I lost an afternoon once reading about bridge rectifier ringing. I still don't quite understand it. I think the problem is the change from forward-conducting to reverse-biased and blocking, turning on a dime when the rectified mains waveform cross the threshold of where the caps are charged to. Effectively going open-circuit at somewhere just under the peak of the waveform. Even if I did manage to glean the specifics of the problem, I haven't quite figured out "who cares?" yet. I don't know if this is a reliability concern for the diodes, or introduces noise in the supply that the PSRR can't reject, or if it's emitting RF noise that could be inductively coupled into low-level signal traces, or what. So, yeah, I probably should lose another afternoon and see if things get any clearer. 🙂
Why is a 3rd Order slope too shallow?
I haven't looked at WinISD for years. Is that the one written by Jeff Bagby? If so, he knows what he is doing. Very much so. Does it not cover sealed boxes?
I've been meaning to do a YT tutorial on calculating Qs in low frequency loading but still haven't got round to it. It's a much needed thing but is actually v simple. Q goes up in direct proportion to frequency. It also needs to be adjusted from the datasheet if there are any resistances in series and the Qe recalculated. 1/Qt = 1/Qe + 1/Qm, just like resistors in parallel. I prefer viewing it in the R1xR2/(R1+R2) format. You don't have any resistances so you can skip that bit. The VAS (equivalent volume) is the suspension compliance recalibrated for a diaphragm of that area - so, remove the speaker suspension and it would be the same stiffness if you placed it in a box of that volume. Because 2pi.f = sqrt(k/m), double or halve the mass or stiffness/compliance and the resonant frequency moves by sqrt(2). When you put the driver (with its suspension, 🙂 ) in a box, those volumes are like two resistors in parallel as well. So a VAS of 120 litres going in a box of 60 litres will give an effective 40 litres - 60x120/80=40. That is 3x as stiff as as the suspension alone, so the resonant frequency will go up by a factor sqrt(3) from what it was in free air. If it was 25 it will become 25 x 1.732 = 43Hz. If its Q in free air was 0.32 then it will now be 0.32 x 43/25 (which also happens to be the sqrt(3), or 1.732, we saw earlier, giving .554.
If you are doing it the other way around, and aiming for a volume to produce a target Q, you just solve a little equation. If your target Q is 0.707 and in free air it is 0.353 (for ease) then you'll need double the free air resonance and so 4x the stiffness of your VAS, so an effective target of 30 litres (using the example beforehand). You then say, what volume in combination with 120 litres will give me 30L? So, using V for volume, V x 120/(V + 120) = 30 ; 120V = 30V + 3600 : 90V = 3600 : => V = 40. And that's it.
I haven't looked at WinISD for years. Is that the one written by Jeff Bagby? If so, he knows what he is doing. Very much so. Does it not cover sealed boxes?
I've been meaning to do a YT tutorial on calculating Qs in low frequency loading but still haven't got round to it. It's a much needed thing but is actually v simple. Q goes up in direct proportion to frequency. It also needs to be adjusted from the datasheet if there are any resistances in series and the Qe recalculated. 1/Qt = 1/Qe + 1/Qm, just like resistors in parallel. I prefer viewing it in the R1xR2/(R1+R2) format. You don't have any resistances so you can skip that bit. The VAS (equivalent volume) is the suspension compliance recalibrated for a diaphragm of that area - so, remove the speaker suspension and it would be the same stiffness if you placed it in a box of that volume. Because 2pi.f = sqrt(k/m), double or halve the mass or stiffness/compliance and the resonant frequency moves by sqrt(2). When you put the driver (with its suspension, 🙂 ) in a box, those volumes are like two resistors in parallel as well. So a VAS of 120 litres going in a box of 60 litres will give an effective 40 litres - 60x120/80=40. That is 3x as stiff as as the suspension alone, so the resonant frequency will go up by a factor sqrt(3) from what it was in free air. If it was 25 it will become 25 x 1.732 = 43Hz. If its Q in free air was 0.32 then it will now be 0.32 x 43/25 (which also happens to be the sqrt(3), or 1.732, we saw earlier, giving .554.
If you are doing it the other way around, and aiming for a volume to produce a target Q, you just solve a little equation. If your target Q is 0.707 and in free air it is 0.353 (for ease) then you'll need double the free air resonance and so 4x the stiffness of your VAS, so an effective target of 30 litres (using the example beforehand). You then say, what volume in combination with 120 litres will give me 30L? So, using V for volume, V x 120/(V + 120) = 30 ; 120V = 30V + 3600 : 90V = 3600 : => V = 40. And that's it.
You don't need to waste an afternoon. Jan, who answered some of your questions earlier, will sell you the edition of Linear Audio that covered it. Or maybe the article by itself is available on his Linear Audo website. Either way it provides the solution and the download may even direct you to subsequent discussions on places like here which could cover various people's implementations of it and their subjective impressions.
LOL. No, it's no better now and you're still on a hiding to nothing. And tracking is far more of a problem than it is on any volume control; especially if you get particular about things being accurate. I did have an idea of using slippable belts once but it started to seem far too convoluted. I may save the idea for a dual mono pre-amp - where you don't have to take measurements to see if everything is in the right place. 🙂
I can't use cubic inches - not even in engines. Actually, especially not in engines. I need a VAS to add to those figures. That poor driver! 55sq cm and an Xmax of 4mm. It's a midrange! Is there any reason not to use it as such? Are they going to be too far apart from the bass? I'm feeling for it like a pet that needs rescuing. How is that Xmax calculated? Is it just peak to peak or is there a Faital style addition of 1/3 gap height? I'm guessing it's just voice coil height minus gap height. If not, what are those two things?
And can you put up a datasheet for the Tang Band? Is there a reason not to use it above 85Hz?