If I were you, I would just try with those two resistors and see where it gets you. Nothing to lose. 🙂
You don't have control over the actual source material, which, if had any post production done on it at all, has passed through linear PCM (not just DSD). Converters do not have 24 bit performance even if they result in 24 bit words, and there is no acoustic space in the world that has 24 bit performance, no mics with 24 bit performance, and so on and so on. 24 bit end result audio is a complete myth. DSD is more destructive than pure, and on we go.
You're at a good 16 to 18 bits in the raw material no matter what you do. Flac is just coding, no better than linear PCM, just more efficient. Bit depth doesn't define quality, and neither does sampling rate. These are consumer concepts driven by marketing hype and subjective, biased opinion. They aren't real.
Speakers are always THE huge limiting factor, massively and destructively modifying your audio. Speakers are worse than high bit-rate MP3 coding. Or any coding. Or any other step in a fully analog chain. Or any chain. Everything else in the system is like a straight wire by comparison. You can worry all you like about distorion caused by summing (which is an imaginary property), but your speaker is THE signal modifier. Always, and it doesn't matter what it is. It's just a question of degree.
Outside...you've lost control of half of the speaker. Half of every speaker's acoustic system is the acoustic space its in.
You got it in the first few posts. You can do it a number of ways, the result is the same. The only difference is how summing fits your application, and interfaces with your gear, which you STILL have not defined!
Yes, that is it, and that is completely wrong. Summing to mono isn't your problem to solve. Your entire quality limiter is the speaker and speaker system. Your problem is how to best interface a mono sum to equipment around it. But you haven't clearly defined that equipment yet, and you invented your own solution regardless.
And measure...measure...measure. You can't hear music with high distortion amounts because of masking, etc. Listening won't do it if you want to really know. Especially fully sighted (biased) listening.
You can't do this. All summing of any two signals will have interaction by definition, but no additional distortion...again, by definition. The "Internet" is full of opinions.
Sorry that you don't have a full grasp of this or what is important. You've invented your own solution based on whatever conception you have of how this works, and haven't shared enough information for anyone to assist. Heck your solution could be the best one, but without full disclosure of everything, nobody can ever tell.
Mono summing never preserves anything. I creates and entirely new waveform which is the result of adding two other waveforms together. No preservation is involved.
You're using a term that doesn't apply here. Masking is the concept of strong frequencies making adjacent weaker ones difficult or impossible to hear. It's the concept upon which lossy coding is built. Cancellation...the addition of two opposite signals...will always happen in a sum. It's not avoidable, and is dependant on the signals being summed.
Electronics aren't the problem here.
Use a Dome filter that maintains a 90 degree difference between L and R before summing.
When L+R is summed to mono the L-R information is lost. When L+R are summed in quadrature, using a Dome filter, the L-R information is preserved.
Quadrature Summing Filter Construction Information
https://www.proaudiodesignforum.com/forum/php/viewtopic.php?t=1319
Some demo files are in this thread. "The Studer "90°" Dome Filter Stereo to Mono Quadrature Summer"
https://proaudiodesignforum.com/forum/php/viewtopic.php?t=1055
"Paperback Writer," in the above thread, provides a good example of the differences between L+R and Quadrature summing.
The blue trace is for in-polarity inputs (L+R), the green is for out-of-polarity inputs (L-R).
Last edited:
Simple jfet mixer can easily serve as stereo to mono.
https://www.electronics-lab.com/project/fet-audio-mixer-3/
Attachments
Roughly 40 years ago I preached the quadrature summing with a pair of 90 degree networks idea as THE definitive mono sum. But it turns out, I was wrong. I'll explain why in a bit.
First, some things need clearing up.
This is a rather incorrect statement because it implies a general L-R cancellation, which doesn't happen, and which in reality only happens under very specific conditions. All L-R is not lost in a simple mono sum. In fact, very little is, if any. And it's not the real reason for quadrature summing anyway.
L-R is in reality a "difference" between L and R signals, and exists under many conditions, and in all stereo mixes, but rarely is lost in a L+R mono sum. For example, L-R exists when a signal is in one channel only, such as L, and not R. Simple math: 1(L) + 0(R) = 1L and 1(L) - 0(R) = 1L. The sum and difference are the same, so even though there is a strong L-R signal, it won't be lost when summed to mono. In fact, the specific conditions where L-R is lost in L+R summing is only when L and R are the same signal at the same amplitude, but with one inverted. 1(L) + -1(R) = 0 Any difference in level between the two, or phase between the two results in an L-R that is no longer zero. When L and R are equal but opposite polarities, we can say they are out of phase. In practice, having an exactly equal but out of phase signal or sound in any stereo mix is to be avoided because it won't actually reproduce well on a pair of speakers, and will sound odd in headphones, and remains quite challenging to cut into a lacquer master for vinyl. So recovering a perfect L-R signal in a mono mix isn't an issue often at all, certainly not the reason to consider quadrature summing.
There is a problem. Consider that many main lead instruments and vocals are panned center, which makes them equal signals in level, phase, and waveform. When you simply do L+R that signal will add to +6dB, which could throw off the balance of a mix because sounds panned hard left or right do not build at all when summed. Arguably a mono sum could upset the overall mix. This sounds like it could be a huge problem, but since the dawn of stereo records and FM stereo radio there has been a keen interest in making sure there is at least some "mono compatibility" in mixes. Anybody creating a serious mix will check their mix in all of the most likely listening methods, including mono, unbalanced stereo, low bandwidth audio systems, noisy environments, etc. The amateurs won't, but all pros do.
So there is a risk of a build-up of center panned material when mixed to mono. That was my interest in the 90 degree network, and that does sort of help, but instead of the build up being +6dB, it would be +3dB (not zero), so not a complete fix. Back when I looked into this, I stole the idea from the old Dolby Stereo matrix used in theaterical films, where a pair of all-pass filters 90 degrees apart solved the problem of panning to and from the surround channel (actual L-R) without a null in the middle of the pan. But in practice, using that for mono is only a little better relating to the 6dB center build-up, and not a total fix.
And here's the elephant in the room. That was all fine if you were trying to get 4 channels out of two, like in the Quad days and Dolby Stereo Optical track days. But we live in a time of obsession with phase shifts, most notably that of anti-alieasing filters, but also of speakers, amplifiers and even wire (yikes). If we, or some of us, are obsessing about in band audio phase shifts that are actually inaudible, we might just be very concerned about the massive phase shifts that take place in multi-stage all-pass filters that cover the entire audio band. The audible effects of those things has been well documented, and is even put to purposful work if there is a need to destroy waveform assymetry (like in AM radio). But audible it is, and we'd have to weigh that against whatever possible small advantage there may be in creating a mono sum.
Ultimately I rejected the idea for some of the above reasons, but mostly because attempting to create a universal fix for mono summing stereo recordings is really a fools errand. I say that because it's all very subjective as to what is "right", but also, completely outside our ability to do anything about it if it's not. In my design, I had a simple L+R sum, an quadrature sum, and also an L only and R only for recovering the best groove wall on a mono record. In the end, I didn't use quadrature summing, and it didn't catch on commercially, even though introduced in the late 1960s, and easily achievable in later years.
The Beatles examples are very cherry-picked. And if you study what early Beatles stereo was, and why it was so strange, you find that these things were mostly intended to be mono releases anyway, with stereo as an afterthought. You will always find an example of something that sums well with quadrature summing. But you'll also find examples of material that sums just fine without it.
And so, it's back to a pair of resistors.
I encourage others to listen to the sound files in the linked to threads and hear the difference for themselves. In some cases the differences or small or non-existent, in others they are not. There is also an example for illustration using tones where one tone completely disappears when mono'd.
Using the extreme early Beetles examples elements that are hard-panned get pulled up in the mix when 90 degree summed where in mono they get pushed down into it. The overall instrument and vocal balance of the mix becomes closer to the stereo presentation whereas the mono presentation doesn't.
The completely out-of-polarity example is often found in synth patches and is the reason that I explored this option as a corrective tool to be used prior to mixing and mastering. Another example are phaser/flanger EFX where a stereo output is derived through sum and difference. When mono'd these effects completely disappear. Disappearing elements in mono is a frequent complaint and yes out-of-polarity elements is poor recording technique but it does happen far more often than many recording and mixing engineers would like.
I do have clients that use these to fold-down stereo for their mono vinyl lathes.
Using the extreme early Beetles examples elements that are hard-panned get pulled up in the mix when 90 degree summed where in mono they get pushed down into it. The overall instrument and vocal balance of the mix becomes closer to the stereo presentation whereas the mono presentation doesn't.
The completely out-of-polarity example is often found in synth patches and is the reason that I explored this option as a corrective tool to be used prior to mixing and mastering. Another example are phaser/flanger EFX where a stereo output is derived through sum and difference. When mono'd these effects completely disappear. Disappearing elements in mono is a frequent complaint and yes out-of-polarity elements is poor recording technique but it does happen far more often than many recording and mixing engineers would like.
I do have clients that use these to fold-down stereo for their mono vinyl lathes.
My customers disagree with you.
yeah even i misspelled "Beatle's" ...i used two t's
what ic's are used in the filter section?
i'd like to give this a try.
what ic's are used in the filter section?
i'd like to give this a try.
Last edited:
This proves my point. Regarding as subjectively better mono remix, quadrature summing and simple summing are not universal, they have to be applied on a per-case basis.
The application here, as badly defined as it is, strongly implies a universal and constant remix method. You can use either, the results will vary. So why go through the extra effort to apply quadrature summing?
Exactly.
Sure, but if the final mix is to be released to the general public, that would be a problem to solve in the mix. Complete out-of-polarity signals are not presented reliably, or sometimes at all, on two channel stereo speaker systems.
Again, exactly my point, and not new or unique to modern music. It's the reason for custom mono mixes. It's all a compromise that can only be chosen based on knowlege of end use. In other words, there's no universal mono application that results in perfect mono on every track.
Ok, there's a special case. I'm not going into the stats of how many clients do that and how much of their real audience is mono vinyl, but yes, that's completely valid. But again, no mono sum is once and done for all material. Or at least, shouldn't be. It puzzles me that someone mastering for a specific mono release wouldn't do a special mono mix.
I'm not sure they do, because nobody would apply any mono method universally without auditioning the results and responding to it. Unless they're rank amateurs.
That's exactly what I said. It was conceived as a production tool for the production process to correct a problem on a specific element before it gets mixed. Or, if discovered by the mastering engineer, sending it back to a remix stage to repair the particular problem element. Usually a keyboard or disappearing out-of-polarity bass due to poor decisions in production.
And I never said that it was a universal fix to provide mono for the consumer. That's why it provides selection of stereo (bypass), L+R conventional mono and I+Q sum.
In listening tests though I've never heard an I+Q sum sound worse than the mono. In many cases they are similar and in other cases I+Q provided the closest to the stereo mix balance. Not saying one is right or wrong - just that they can be different and the results to my tastes favor I+Q over mono because it sounds more like the stereo mix.
A custom mono mix may not be available or there is no budget for it. People having their work cut in mono may be using it for club playback.
I never said that there was a universal mono application that results in perfect mono. You assumed that or made it up.
There you go again. I never said it was. And I already answered the statement about special mono releases.
No one said that they don't audition the results. That's their job. Why do you keep making things up?
You remind me of why I almost never post here. I would have been a lot better of had I not shared this with the community because some singular dude wants to argue, mischaracterize and make up statements I never made.
"That's exactly what I said. It was conceived as a production tool for the production process to correct a problem on a specific element before it gets mixed. Or, if discovered by the mastering engineer, sending it back to a remix stage to repair the particular problem element. Usually a keyboard or disappearing out-of-polarity bass due to poor decisions in production.
And I never said that it was a universal fix to provide mono for the consumer. That's why it provides selection of stereo (bypass), L+R conventional mono and I+Q sum."
This thread isn't about a production tool at all. It's a once/done mono mix for a portable speaker system.
My repeated comments about your solution not being universal all relate to the point of the OP.
"No one said that they don't audition the results. That's their job. Why do you keep making things up?"
I simply said nobody would apply a mono mix without auditioning it. I didn't say they did. I'm not making things up, you're misquoting me.
"You remind me of why I almost never post here. I would have been a lot better of had I not shared this with the community because some singular dude wants to argue, mischaracterize and make up statements I never made."
I'm sorry if I upset you. You brought up quadradure summing, and some might think your point was that it's the universal end-all solution, better than simpling mono summing. And I have tried it, decades ago. I'm applying some logic to the OP's request, and pointing out why quadrature summing, in his case, is probably not required. He implied "absolute best audio..." etc.
And I never said that it was a universal fix to provide mono for the consumer. That's why it provides selection of stereo (bypass), L+R conventional mono and I+Q sum."
This thread isn't about a production tool at all. It's a once/done mono mix for a portable speaker system.
My repeated comments about your solution not being universal all relate to the point of the OP.
"No one said that they don't audition the results. That's their job. Why do you keep making things up?"
I simply said nobody would apply a mono mix without auditioning it. I didn't say they did. I'm not making things up, you're misquoting me.
"You remind me of why I almost never post here. I would have been a lot better of had I not shared this with the community because some singular dude wants to argue, mischaracterize and make up statements I never made."
I'm sorry if I upset you. You brought up quadradure summing, and some might think your point was that it's the universal end-all solution, better than simpling mono summing. And I have tried it, decades ago. I'm applying some logic to the OP's request, and pointing out why quadrature summing, in his case, is probably not required. He implied "absolute best audio..." etc.
90° cancels 270° the same as 0° cancels 180°. But in the unlikely case that the two channels are exactly out of phase and equal amplitude, the correct result is zero. It is just as likely that the relationship between channels is opposite whatever phase combination you might choose to mix them with. Recordings are created assuming the mono mix is (L+R)/2.
depending on what your music source is there may be a simple setting for STEREO to MONO in your media player. I know Foobar has it.
Precisely, but it's far less likely that 90/270 would actually occur because there's no simple mechanism to do that. There is to get 180 out. I still suggest that it is also a major mixing error to include anything significant material exactly 180 out when the release is intended for two channel stereo, but quadrature summing would fix that exact cancellation.
We're still down a rabbit hole here, as the OP's application is a fixed mix for playing undefined stereo material in a single speaker.
Well, that was a lot of information i didn't know...
So, there is in fact no universal way to convert stereo to mono. There is basically no best way. Now i understand.
So, in the end, like most of you told me, i'm using two resistors.
Now, i will actually talk about the entire system. It will be LONG:
So, the L and R channels coming from the DAC are now summed togheter with just two resistors. Then buffered with a unity-gain OpAmp which output is connected to a potentiometer. Then the variable voltage given by the potentiometer is buffered with another unity-gain OpAmp that drives the three bandpass filters. That potentiometer is the master volume potentiometer.
Now, here are the reasons why they are all band-pass filters:
First bandpass filter (30 - 125Hz)
The subwoofer bandpass filter. The lower frequencies below 30Hz are attenuated, because they can cause my subwoofer to overheat. I tested that at continuous 30Hz, and it handled it. But at 25Hz, it drew much more current, and it could overheat. So i decided to attenuate the frequencies below 30Hz. The upper 125Hz is the maximum recommended frequency i can feed to this subwoofer according to its datasheet.
Second bandpass filter(125 - 4000Hz)
The midranges bandpass filter.
The lower 125Hz is the minimum recommended frequency i can feed to the midranges, according to the datasheet.
The upper 4000Hz is the maximum recommended frequency i can feed to the midranges according to the datasheet.
Third bandpass filter(4000 - 20000Hz)
The tweeters bandpass filter.
The lower 4000Hz is the minimum recommended frequency i can feed to the tweeters, according to the datasheet.
The upper 20000Hz frequency low pass is present to block higher frequencies that could cause damage to the amplifier. It has happened to me once. It is some sort of protection.
I selected this combination of speakers and then matched the band pass filters for them.
After each bandpass filter there is a gain stage with a gain of 1.5, and volume potentiometer in the end.
So one potentiometer controls the bass volume, the other one the mids volume, and the last one the treble volume.
The gain stages are added so that the power amplifiers can reach their maximum power. (Voltage coming from the bandpass filters is not enough)
I designed the three bandpass filters to give a near perfect flat frequency response. Their summed frequency response has just small variations of less than 0.5dB.
The phase shift they cause is also not very drastic and won't result in a lot of phase cancellation.
Theese filters are made by OPA1612 OpAmps, polypropilene capacitors (KYET) and very low thermal noise SMD resistors (± 10 ppm / °C) (0805 package).
There are also decoupling capacitors of 100nF close to every OpAmp [Murata capacitors].
The main dual psu capacitors are two 220uF 25V nichicon FG series.
This dual psu is completely isolated from the main psu that powers the POWER amplifiers.
The audio source come from the QCC5125 Bluetooth module, which psu is also completely isolated from both the main PSU (the one that powers the power amplifiers) and the dual psu for the OpAmps. This keeps all noise at bay, expecially the classic noise caused by Bluetooth module. In the end, the noise of all the pre-amplifiers and Bluetooth module is basically 0.
Now, i'm also addressing that i changed that module with another one that has a better DAC. The audio coming out of the QCC5125 is digital data, which needs to be fed into a DAC to obtain the analog audio signals. The used DAC is now PCM5102, which can drive loads as low as 1Kohms with no additional distortion.
Now, the audio outputs from the three bandpass filters go to the inputs of three power amplifiers.
But this is not the only audio source of the speaker. There are also RCA inputs. There are relays that will disconnect the L and R channels from the Bluetooth module and connect the ones from the RCA connectors at their place. (This of course happens at the input of the there band-pass filters pre-amplifier.
I removed the DC blocking capacitors on the bandpass filters, because they are already present on the power amplifiers. They are 10uF 25V nichicon FG series.
Sub bandpass output goes to L and R inputs of first power amplifier.
Mid bandpass output goes to L and R inputs of second power amplifier.
High bandpass output goes to L and R inputs of third power amplifier.
The first power amplifier drives the dual VC subwoofer.
The second power amplifier powers the two midranges.
The third power amplifier powers the two tweeters.
What are the amplifiers?
Well, i said this speaker was 400W RMS... But now, i'm upgrading it to a 1300W RMS speaker.
Well... I think this is not just a portable speaker anymore. It is a club speaker. Intended mainly for outside use.
600W RMS (Max volume) - 1 x Subwoofer
300W RMS (Max volume) - 2 x Midranges
50W RMS (Max volume) - 2 x Tweeters
Total is 1300W RMS at maximum volume. But that is if all the volume knobs are at maximum, which won't be the case.
Theese knobs will be adjusted to make a good sound spectrum.
Yes, the bandpass filters summed frequency response is basically flat, but not at the speakers. That's why there are theese three separate volume knobs for bass, mid, treble.
And finally, let's talk about the amplifiers.
They are based on the TPA3255 from TI.
300W x2 RMS at 10% THD into 4ohms.
260W x2 RMS at 1% THD into 4ohms.
Yes, i have to admit, not very good performance actually, but it is acceptable for me.
They are powered from the main power supply which is 53V. It is a powerful metal-shielded boost converter, with 22000uF of input capacitance and 6600uF of output capacitance. All capacitors are nippon-chemicon ones. There are also a lot of ceramic capacitors for higher frequencies noise suppression. This boost converter is powered from a 25.6V (nominal) battery.
But also, each power amplifier has its own capacitors (2 x 2200uF).
The 53V voltage is stepped down to 24V with a buck regulator, and then stepped down again to 5V for the bluetooth module and 15V for the dual PSU system for all the OpAmps.
The bluetooth module and OpAmps are powered by isolated modules.
To reduce noise from the 24V switching buck regulator there are interference suppression inductors and capacitors.
Then the 24V goes directly to two voltage isolators (B2424s-3W) that will give me two totally isolated 24V rails. Since they are both isolated, i can connect them to form the dual rail supply for the OpAmps.
Then, because theese DC-isolators modules are switching ones, there are L7815 and L7915 linear regulators after them + noise suppression inductors and capacitors, to give the OpAmps a nice noise-free supply.
At the same time, the 24V supply is also connected to a 12V linear regulator, then to a B1212S-2W voltage isolator.
After that voltage isolator, there is L7805 regulator + noise suppression inductors and capacitors, to also give the Bluetooth module a noise-free supply.
Now you might argue that theese circuits are NOT totally isolated from each other, which is true. They still share the same common ground, because the audio output from the Bluetooth module is connected to the input of the three bandpass filters, and the outputs of the pre-amplifiers are connected to the inputs of the power amplifiers.
But still, the voltage isolators break the ground loops and this removes all the noise, expecially the one generated from the bluetooth module.
So, system ready. I power this up and listen closely to the tweeters, after making sure the volume is at 0...
There is basically no noise. Almost silence. I measured the noise floor from all the speakers combined to be less than 15dB at one meter distance from them. Awesome.
And no, the amplifiers are not in mute or standby mode. This is the actual noise floor of the amplifiers.
The speakers are mounted like this:
The subwoofer is in its own SEALED box.
The two midranges and tweeters are in their own SEALED box. Of course that does not make effect on the tweeters. They are mounted there to be close to the midranges.
The subwoofer box is 'on the ground' and the midranges + tweeters box is stacked on top of it, with vibration - absorbing foam between them (otherwise, wood on wood would cause rattling) They are both cubes with 34.5x34.5x34.5cm internal diameter.
They are made with a kind wood... That i don't know how to say in english, sorry. It is one of the most common woods used for speakers enclousers, and sounds excellent.
It sounds really good to me. It all works perfectly fine, and then, all of a sudden... That stereo to mono coversion puzzled me, but now i discovered that there is really nothing else i can do here.
It will sound basically the same no matter what summing method i try. So, in the end, i just summed the L and R channels with two resistors.
What i thought was that there might have been a way to make the L and R summing better, but at this point, i say i'm satisfied with the resistors method.
Oh and, last thing...
You might call me crazy... There is a 25.6V 1kWh battery inside of the sealed box where the midranges and tweeters are mounted... A lithium Iron phosphate one, which is much more durable than normal lithium ion ones, and also safer.
I know, this affects a lot the resonance frequency of the midranges chamber...
And, all the rest of the circuits are mounted on the inside surface of the subwoofer box... This does not seem to negatively affect the subwoofer either.
To me the speaker sounds very good. I might have gotten lucky. The battery is situated exactly in the middle.
I can always compensate errors by adjusting the volume knobs. Not totally, but decently.
In the end the total weight of the speaker would be around 20kg which is a lot actually. But remember that this is a club speaker. It is transportable, not portable. It has detacheable wheels for easier transportation. It can be easily put in a car hood. I'm planning to use it at the beach (Definetly not at 1300W RMS with strangers around). But i will test it at max power now that there isn't anyone.
I will make another smaller, less powerful, more portable speaker another day.
In the end i decided to make the BIG one first.
The battery is 1kWh for decent durability. It charges in 12 hours with a 19 / 24V power supply. It has a balancing and protection BMS and of course, a fuse.
And that is my entire project.
I think it seems very bad for you, but for me it is enough. I am not striving to absolute perfection.
I should have not written 'absolute best audio'. Because, actually... I probably never heard 'absolute best audio'.
I played that speaker to a lot of people, including my audiophile friend.
He says it sounds very good, but of course not as good as professional equipment. All the others say it sounds amazing.
To me... It sounds really good.
So, in the end, for me, it's a win!
Sorry, everyone... I lost myself in glass of water. Such insignificant changes in that stereo to mono conversion... Won't basically change anything in my setup.
I said everything in words because i don't have the schematics at hand for now.
If someone wants, i can publish the schematic of the bandpass filters and power supply for the OpAmps & Bluetooth module. All the other circuits are bought and i dont have their schematics.
So, there is in fact no universal way to convert stereo to mono. There is basically no best way. Now i understand.
So, in the end, like most of you told me, i'm using two resistors.
Now, i will actually talk about the entire system. It will be LONG:
So, the L and R channels coming from the DAC are now summed togheter with just two resistors. Then buffered with a unity-gain OpAmp which output is connected to a potentiometer. Then the variable voltage given by the potentiometer is buffered with another unity-gain OpAmp that drives the three bandpass filters. That potentiometer is the master volume potentiometer.
Now, here are the reasons why they are all band-pass filters:
First bandpass filter (30 - 125Hz)
The subwoofer bandpass filter. The lower frequencies below 30Hz are attenuated, because they can cause my subwoofer to overheat. I tested that at continuous 30Hz, and it handled it. But at 25Hz, it drew much more current, and it could overheat. So i decided to attenuate the frequencies below 30Hz. The upper 125Hz is the maximum recommended frequency i can feed to this subwoofer according to its datasheet.
Second bandpass filter(125 - 4000Hz)
The midranges bandpass filter.
The lower 125Hz is the minimum recommended frequency i can feed to the midranges, according to the datasheet.
The upper 4000Hz is the maximum recommended frequency i can feed to the midranges according to the datasheet.
Third bandpass filter(4000 - 20000Hz)
The tweeters bandpass filter.
The lower 4000Hz is the minimum recommended frequency i can feed to the tweeters, according to the datasheet.
The upper 20000Hz frequency low pass is present to block higher frequencies that could cause damage to the amplifier. It has happened to me once. It is some sort of protection.
I selected this combination of speakers and then matched the band pass filters for them.
After each bandpass filter there is a gain stage with a gain of 1.5, and volume potentiometer in the end.
So one potentiometer controls the bass volume, the other one the mids volume, and the last one the treble volume.
The gain stages are added so that the power amplifiers can reach their maximum power. (Voltage coming from the bandpass filters is not enough)
I designed the three bandpass filters to give a near perfect flat frequency response. Their summed frequency response has just small variations of less than 0.5dB.
The phase shift they cause is also not very drastic and won't result in a lot of phase cancellation.
Theese filters are made by OPA1612 OpAmps, polypropilene capacitors (KYET) and very low thermal noise SMD resistors (± 10 ppm / °C) (0805 package).
There are also decoupling capacitors of 100nF close to every OpAmp [Murata capacitors].
The main dual psu capacitors are two 220uF 25V nichicon FG series.
This dual psu is completely isolated from the main psu that powers the POWER amplifiers.
The audio source come from the QCC5125 Bluetooth module, which psu is also completely isolated from both the main PSU (the one that powers the power amplifiers) and the dual psu for the OpAmps. This keeps all noise at bay, expecially the classic noise caused by Bluetooth module. In the end, the noise of all the pre-amplifiers and Bluetooth module is basically 0.
Now, i'm also addressing that i changed that module with another one that has a better DAC. The audio coming out of the QCC5125 is digital data, which needs to be fed into a DAC to obtain the analog audio signals. The used DAC is now PCM5102, which can drive loads as low as 1Kohms with no additional distortion.
Now, the audio outputs from the three bandpass filters go to the inputs of three power amplifiers.
But this is not the only audio source of the speaker. There are also RCA inputs. There are relays that will disconnect the L and R channels from the Bluetooth module and connect the ones from the RCA connectors at their place. (This of course happens at the input of the there band-pass filters pre-amplifier.
I removed the DC blocking capacitors on the bandpass filters, because they are already present on the power amplifiers. They are 10uF 25V nichicon FG series.
Sub bandpass output goes to L and R inputs of first power amplifier.
Mid bandpass output goes to L and R inputs of second power amplifier.
High bandpass output goes to L and R inputs of third power amplifier.
The first power amplifier drives the dual VC subwoofer.
The second power amplifier powers the two midranges.
The third power amplifier powers the two tweeters.
What are the amplifiers?
Well, i said this speaker was 400W RMS... But now, i'm upgrading it to a 1300W RMS speaker.
Well... I think this is not just a portable speaker anymore. It is a club speaker. Intended mainly for outside use.
600W RMS (Max volume) - 1 x Subwoofer
300W RMS (Max volume) - 2 x Midranges
50W RMS (Max volume) - 2 x Tweeters
Total is 1300W RMS at maximum volume. But that is if all the volume knobs are at maximum, which won't be the case.
Theese knobs will be adjusted to make a good sound spectrum.
Yes, the bandpass filters summed frequency response is basically flat, but not at the speakers. That's why there are theese three separate volume knobs for bass, mid, treble.
And finally, let's talk about the amplifiers.
They are based on the TPA3255 from TI.
300W x2 RMS at 10% THD into 4ohms.
260W x2 RMS at 1% THD into 4ohms.
Yes, i have to admit, not very good performance actually, but it is acceptable for me.
They are powered from the main power supply which is 53V. It is a powerful metal-shielded boost converter, with 22000uF of input capacitance and 6600uF of output capacitance. All capacitors are nippon-chemicon ones. There are also a lot of ceramic capacitors for higher frequencies noise suppression. This boost converter is powered from a 25.6V (nominal) battery.
But also, each power amplifier has its own capacitors (2 x 2200uF).
The 53V voltage is stepped down to 24V with a buck regulator, and then stepped down again to 5V for the bluetooth module and 15V for the dual PSU system for all the OpAmps.
The bluetooth module and OpAmps are powered by isolated modules.
To reduce noise from the 24V switching buck regulator there are interference suppression inductors and capacitors.
Then the 24V goes directly to two voltage isolators (B2424s-3W) that will give me two totally isolated 24V rails. Since they are both isolated, i can connect them to form the dual rail supply for the OpAmps.
Then, because theese DC-isolators modules are switching ones, there are L7815 and L7915 linear regulators after them + noise suppression inductors and capacitors, to give the OpAmps a nice noise-free supply.
At the same time, the 24V supply is also connected to a 12V linear regulator, then to a B1212S-2W voltage isolator.
After that voltage isolator, there is L7805 regulator + noise suppression inductors and capacitors, to also give the Bluetooth module a noise-free supply.
Now you might argue that theese circuits are NOT totally isolated from each other, which is true. They still share the same common ground, because the audio output from the Bluetooth module is connected to the input of the three bandpass filters, and the outputs of the pre-amplifiers are connected to the inputs of the power amplifiers.
But still, the voltage isolators break the ground loops and this removes all the noise, expecially the one generated from the bluetooth module.
So, system ready. I power this up and listen closely to the tweeters, after making sure the volume is at 0...
There is basically no noise. Almost silence. I measured the noise floor from all the speakers combined to be less than 15dB at one meter distance from them. Awesome.
And no, the amplifiers are not in mute or standby mode. This is the actual noise floor of the amplifiers.
The speakers are mounted like this:
The subwoofer is in its own SEALED box.
The two midranges and tweeters are in their own SEALED box. Of course that does not make effect on the tweeters. They are mounted there to be close to the midranges.
The subwoofer box is 'on the ground' and the midranges + tweeters box is stacked on top of it, with vibration - absorbing foam between them (otherwise, wood on wood would cause rattling) They are both cubes with 34.5x34.5x34.5cm internal diameter.
They are made with a kind wood... That i don't know how to say in english, sorry. It is one of the most common woods used for speakers enclousers, and sounds excellent.
It sounds really good to me. It all works perfectly fine, and then, all of a sudden... That stereo to mono coversion puzzled me, but now i discovered that there is really nothing else i can do here.
It will sound basically the same no matter what summing method i try. So, in the end, i just summed the L and R channels with two resistors.
What i thought was that there might have been a way to make the L and R summing better, but at this point, i say i'm satisfied with the resistors method.
Oh and, last thing...
You might call me crazy... There is a 25.6V 1kWh battery inside of the sealed box where the midranges and tweeters are mounted... A lithium Iron phosphate one, which is much more durable than normal lithium ion ones, and also safer.
I know, this affects a lot the resonance frequency of the midranges chamber...
And, all the rest of the circuits are mounted on the inside surface of the subwoofer box... This does not seem to negatively affect the subwoofer either.
To me the speaker sounds very good. I might have gotten lucky. The battery is situated exactly in the middle.
I can always compensate errors by adjusting the volume knobs. Not totally, but decently.
In the end the total weight of the speaker would be around 20kg which is a lot actually. But remember that this is a club speaker. It is transportable, not portable. It has detacheable wheels for easier transportation. It can be easily put in a car hood. I'm planning to use it at the beach (Definetly not at 1300W RMS with strangers around). But i will test it at max power now that there isn't anyone.
I will make another smaller, less powerful, more portable speaker another day.
In the end i decided to make the BIG one first.
The battery is 1kWh for decent durability. It charges in 12 hours with a 19 / 24V power supply. It has a balancing and protection BMS and of course, a fuse.
And that is my entire project.
I think it seems very bad for you, but for me it is enough. I am not striving to absolute perfection.
I should have not written 'absolute best audio'. Because, actually... I probably never heard 'absolute best audio'.
I played that speaker to a lot of people, including my audiophile friend.
He says it sounds very good, but of course not as good as professional equipment. All the others say it sounds amazing.
To me... It sounds really good.
So, in the end, for me, it's a win!
Sorry, everyone... I lost myself in glass of water. Such insignificant changes in that stereo to mono conversion... Won't basically change anything in my setup.
I said everything in words because i don't have the schematics at hand for now.
If someone wants, i can publish the schematic of the bandpass filters and power supply for the OpAmps & Bluetooth module. All the other circuits are bought and i dont have their schematics.
Last edited:
The point is that if you shift both channels the same amount then what was originally 180° apart is still 180° apart.
Dude,
To be honest, you wrap yourself around the axel over odd things. You are concerned about sound purity, as you convert stereo to mono. I mean you are kind of committed to living with the big distortion of going from stereo to mono.
The reality is no electronic component is perfect. The sound that goes in and the sound that comes will not be perfect copy. It is all about trade offs and how to pick and choose the best combination of design compromises and trade offs.
As the designer, the engineer, have made the decision you want a mono version of your stereo signal. And you want to retain absolute phase. If it were me, I probably would not be distracted by the absolute phase. Looks like you are, then a 2nd amp to flip it back into phase is the thing to do. How much distortion does a unity gain inverting amp add to a signal? I am thinking, "Not much. At not so much that I would notice." In your case, hopefully less distortion than your signal being out of absolute phase.
Yes, but it also matters how that shift is accomplished. There is little chance that signal path could be shifted a consistent full bandwidth 90/270. That would take the filters described. However, there are many possibilities for 0/180. Both would cancel in a sum, the former is unlikely to occur. The filters suggested move the phase relationship of both channels a relatively consistent 90 degrees, so 0/180 no longer occurs at all. It becomes 90 degree difference, no longer 180. And 0/0 becomes 0/90.