paradoxper
Headphoneus Supremus
Shut up and build it. Christ.
SELL. SELL. SELL.
SELL. SELL. SELL.
Raal Ribbon Headphones - SRH1A
- Thread starter once
- Start date
I'm waiting for resistors at the moment. They're a couple month lead time.
Aleksandar R.
Sponsor: RAAL 1995
OK, but making flat phase filter will not end up being flat phase at ear, but...Even if you don't compensate for the skewed phase response of headphones, I doubt it will sound bad in any way. Actually, it may sound better for other reasons, like resistor, connectors and wiring quality...
In any case, if you can, offer IIR, FIR and FIR with "analog phase". That would be the fairest thing to do and people will choose what's best for them.
You can get 10 of the worlds top acoustics and loudspeaker engineers in the world, and they can argue for days on the best way to do things. Good example here to show the results of this:
http://techtalk.parts-express.com/f...weeters-are-they-worth-it?p=850258#post850258
So this is why I can generate 5 filters based on FIR, and another 5 based on IIR in the Roon PEQ. And the end results will tell us the story rather than silly forum debates. Because in the end, all that matters is what the users of your headphones think. And them and only them can provide us with that data. the resistors, wiring and connector quality will be equal for all filters. So it will only be the filters that folks with this interface will be comparing.
Last edited:
Here's what the creator of Audiolense has to say about FIR vs IIR for speaker correction. I have no opinion to offer at this point as I'm end results driven. and we have yet to see the results.
"Eg OpenDRC which has limited FIR capacity gives more precise low frequency correction than anything IIR based that I've seen. With a PC you have plenty of headroom and low frequency resolution is a none-issue. Although not common it is possible to apply shorter filters at higher frequencies. In the new convolver I've made, the high frequency filters are shortened and do not have more filter coefficients and resolution than they need to to get the job done. I do not expect this to be a big differentiator, but I decided to do it anyway in order to make the convolver as efficient as possible.
IIR is IMO not a good technology for correcting driver responses, for the simple reason that it is very difficult to shape the frequency response of IIR's to undo what the driver and speaker cabinet typically does wrong. Therefore, with IIR, you always make something worse as you improve your focal issue. IIRs are really difficult to work with - and in my experience it is virtually impossible to do a precise correction manually by using parametric EQ. Those who have a half life worth of experience using IIR have "learned" that you should correct as little as possible above the schroeder frequency; This is a strongly held belief and understandable. But when they turn to FIR, some of them experience that this is no longer the case. There's a very telling discussion between me and Bob Katz a few years back on this forum where he does an 180* turnaround after a few intensive days of experimenting.
I have implemented Hypex' DSP in a speaker pair that I have in our summer house. I was unable to produce a correction with it that had the precision I knew was attainable. What worked best was the minimalist approach which is typical for those who use IIR. The crossover duties worked all right, though. I used Audiolense on top and got the results I always get.
I also did some IIR programming many years ago - because we were talking to a company who had hardware with IIR capacity. You can make a lot of shapes with a biquad if you control all four constants, which I did. So I was able to be more precise there than with a standard IIR GUI. But even though I could bent those biquads north and south, east and west I got nowhere near what I could do with Audiolense or the Tact RCS of the old days for that matter. There are also examples of Kii owners who use Audiolense or another FIR based correction to improve the sound quality...
Yes there is a difference between direct sound and reflections, but inside the ear canal everything has the same direction. Reflected sound merges with (slightly newer) direct sound. WIth a short sound and late reflection we get an echo that we are unlikely to improve. But when the reflection enters the ear while the tone is still playing it will modulate the direct sound: Stronger, weaker, phase shift.. We can and should do something about some of this The goal is not to correct the decay pattern, but to correct some of the negative things that some of the reflections does to the direct sound at certain frequencies - in other words what appears as direct sound in your ears - and is practically correctable.
In the bass and lower midrange there are typically a lot of global issues and when you correct the sweet spot, you get substantial improvements literally everywhere else too. From the hundreds and up this is done on a very broad band basis with emphasis on speaker correction - and perhaps not at all above a few kHz. The upper part of the correction is even less seat dependable.
Minimum phase target is the safest route, but there are people who prefer linear phase target or a hybrid.
The approach of putting up a mic, measure and invert doesn't have to be implemented as a brute force approach - and rarely is. And it isn't particularly seat dependant either. A couple of decades ago a guy named Mourjopolous proved that a realistic implementation of the inversion approach improved the sound in the sweet spot as well as in the rest of the room. And more refined implementations have surfaced since. A correction is much more sensitive to speaker placement than listening seat location. If you move the speakers a few inches you need to do a new measurement...
I have a pragmatic view on linear vs minimum phase crossovers. Audiolense supports both. A clear majority prefers linear phase and it tends to simulate a little better in the time domain after all the correction work is done. But there are exceptions.
There are limits to what can and should be done in every setup and every room, but those limits are not universal. The math is capable of correcting anything but perfect cancellations. I'm still waiting to see one of those. The bit depth of the FIR correction is a limiter compared to the unlimited theoretical capability, but hardly a bottleneck in practice. The speakers plus amps plus room arent perfectly linear and time invariant. The shortcomings would surface if we tried to eliminate all imprefections. And even if the hifi was perfect were we would get a "head in a vice" correction if we tried to correct everything. But nobody does. The best DSP-ers with the best DSP solutions are the ones who get closest to correcting what is worth correcting and leaving the rest as it is - in each case.
From a programmers point of view - FIR is much easier to work with and much more controllable. The only thing you can do with IIR but not FIR is to create a filter that rings forever, but you don't want that. There is absolutely nothing worth doing that requires IIR filters. And there are a lot of things worth doing where IIR produces substandard results compared to FIR.
In addition to everything above, Audiolense and a few other correction solutions are able to do some non-minimum phase improvements in the time domain. A lot of IIR proponents argue that this can't and/or shouldn't be done. But simulations and control measurments proves it can. And a vast majority of those who have Audiolense XO and the option to choose proves that time domain correction is regularly perceived as a sonic improvement."
"Eg OpenDRC which has limited FIR capacity gives more precise low frequency correction than anything IIR based that I've seen. With a PC you have plenty of headroom and low frequency resolution is a none-issue. Although not common it is possible to apply shorter filters at higher frequencies. In the new convolver I've made, the high frequency filters are shortened and do not have more filter coefficients and resolution than they need to to get the job done. I do not expect this to be a big differentiator, but I decided to do it anyway in order to make the convolver as efficient as possible.
IIR is IMO not a good technology for correcting driver responses, for the simple reason that it is very difficult to shape the frequency response of IIR's to undo what the driver and speaker cabinet typically does wrong. Therefore, with IIR, you always make something worse as you improve your focal issue. IIRs are really difficult to work with - and in my experience it is virtually impossible to do a precise correction manually by using parametric EQ. Those who have a half life worth of experience using IIR have "learned" that you should correct as little as possible above the schroeder frequency; This is a strongly held belief and understandable. But when they turn to FIR, some of them experience that this is no longer the case. There's a very telling discussion between me and Bob Katz a few years back on this forum where he does an 180* turnaround after a few intensive days of experimenting.
I have implemented Hypex' DSP in a speaker pair that I have in our summer house. I was unable to produce a correction with it that had the precision I knew was attainable. What worked best was the minimalist approach which is typical for those who use IIR. The crossover duties worked all right, though. I used Audiolense on top and got the results I always get.
I also did some IIR programming many years ago - because we were talking to a company who had hardware with IIR capacity. You can make a lot of shapes with a biquad if you control all four constants, which I did. So I was able to be more precise there than with a standard IIR GUI. But even though I could bent those biquads north and south, east and west I got nowhere near what I could do with Audiolense or the Tact RCS of the old days for that matter. There are also examples of Kii owners who use Audiolense or another FIR based correction to improve the sound quality...
Yes there is a difference between direct sound and reflections, but inside the ear canal everything has the same direction. Reflected sound merges with (slightly newer) direct sound. WIth a short sound and late reflection we get an echo that we are unlikely to improve. But when the reflection enters the ear while the tone is still playing it will modulate the direct sound: Stronger, weaker, phase shift.. We can and should do something about some of this The goal is not to correct the decay pattern, but to correct some of the negative things that some of the reflections does to the direct sound at certain frequencies - in other words what appears as direct sound in your ears - and is practically correctable.
In the bass and lower midrange there are typically a lot of global issues and when you correct the sweet spot, you get substantial improvements literally everywhere else too. From the hundreds and up this is done on a very broad band basis with emphasis on speaker correction - and perhaps not at all above a few kHz. The upper part of the correction is even less seat dependable.
Minimum phase target is the safest route, but there are people who prefer linear phase target or a hybrid.
The approach of putting up a mic, measure and invert doesn't have to be implemented as a brute force approach - and rarely is. And it isn't particularly seat dependant either. A couple of decades ago a guy named Mourjopolous proved that a realistic implementation of the inversion approach improved the sound in the sweet spot as well as in the rest of the room. And more refined implementations have surfaced since. A correction is much more sensitive to speaker placement than listening seat location. If you move the speakers a few inches you need to do a new measurement...
I have a pragmatic view on linear vs minimum phase crossovers. Audiolense supports both. A clear majority prefers linear phase and it tends to simulate a little better in the time domain after all the correction work is done. But there are exceptions.
There are limits to what can and should be done in every setup and every room, but those limits are not universal. The math is capable of correcting anything but perfect cancellations. I'm still waiting to see one of those. The bit depth of the FIR correction is a limiter compared to the unlimited theoretical capability, but hardly a bottleneck in practice. The speakers plus amps plus room arent perfectly linear and time invariant. The shortcomings would surface if we tried to eliminate all imprefections. And even if the hifi was perfect were we would get a "head in a vice" correction if we tried to correct everything. But nobody does. The best DSP-ers with the best DSP solutions are the ones who get closest to correcting what is worth correcting and leaving the rest as it is - in each case.
From a programmers point of view - FIR is much easier to work with and much more controllable. The only thing you can do with IIR but not FIR is to create a filter that rings forever, but you don't want that. There is absolutely nothing worth doing that requires IIR filters. And there are a lot of things worth doing where IIR produces substandard results compared to FIR.
In addition to everything above, Audiolense and a few other correction solutions are able to do some non-minimum phase improvements in the time domain. A lot of IIR proponents argue that this can't and/or shouldn't be done. But simulations and control measurments proves it can. And a vast majority of those who have Audiolense XO and the option to choose proves that time domain correction is regularly perceived as a sonic improvement."
Last edited:
A point regarding the full range driver speakers which leads to this kind of headphones like MySphere or SR1a.
A point regarding the full range driver speakers which leads to this kind of headphones like MySphere or SR1a.
Nice speakers! I bet they sound great. I actually ordered a dozen of the latest driver design from the former head engineer at Voxativ. Pretty amazing technology. 35-30k frequency response, and a 10mm xmax from a 5" driver! Not cheap though around $3000 a piece landed cost. Review on his complete speaker here:
https://6moons.com/audioreview_articles/camerton/
Yes, I am aware of Camerton. Imported in France by TotalDAC who introduced me to Voxativ.
Full range drivers have the advantage to avoid crossovers and can be enclosed in sealed cabinets avoiding many resonances. Sound is more natural and I see a good comparison with SR1a.
Another point of similitude are bass with low energy and I use a sub Sunfire Atmos XT265.
Voxativ in their Absolut system has programmed a specific DSP.
You can reach something like this with your new box + filters for the SR1a.
Full range drivers have the advantage to avoid crossovers and can be enclosed in sealed cabinets avoiding many resonances. Sound is more natural and I see a good comparison with SR1a.
Another point of similitude are bass with low energy and I use a sub Sunfire Atmos XT265.
Voxativ in their Absolut system has programmed a specific DSP.
You can reach something like this with your new box + filters for the SR1a.
Yes this is precisely what I plan with the Camerton speakers. Only using 64 bit floating point DSP. And the same technique will be used with these headphones. Some folks like to come up with theories based on their limited understanding of how things work. Others like to parrot what others have said, not having an ounce of understanding of their own that is rooted from hands on experience. I like to try every possible way, and let the end results do the talking. Einstein once said the only source of knowledge is experience. Makes sense to me.
Last edited:
See comments and questions below.
If that is the case then why is there even a product like the LOKI? So I use a DAP, either the iBasso 220 or the Questyle QPR1 via optical out to send flac files over to DAVE/HMS. So there is an equalizer built into the DAP’s software and is that what I should use to boost bass response? Will it be as effective as the LOKI between the DAVE and headphone amplifier?
Why is there Honda Civic's when we have S class Mercedes's? The best way to do EQ is with DSP. But not all DSP is equal in quality. And if you have an analog source, you must have a ADC/DAC combo to do DSP. And the quality of the ADC/DAC combo will determine your end result. So it can be expensive to do good EQ with an analog source. If your source is digital from a computer, there's no need for something like the Loki.
Using portable DAP's as source gear is something people do because of the portability. Not because it's the best way to do things. A better way to do things is using a computer running software like Roon, and using Roon's PEQ, or generate 3rd party convolution filters and load into Roon's convolution engine. The DSP built into these portable devices is mediocre at best.
Last edited:
Recently I was examining a digital interface device that had built in DSP. The device was advertised as having a digital dynamic range of 144dB. So enough to resolve up to 24/96 PCM. However what they failed to mention was if you utilized any of the DSP features at all, that dynamic range drops down to 100dB. Just enough to resolve redbook, but not high resolution audio which was a feature this device was marketed of being able to do. Most of these cheap DSP chips used in these portable DAP's do the same thing. But don't expect them to tell you this. So I recommend sending the audio out of these portable devices flat, unless you know for sure the chip they use for DSP is capable of the digital dynamic range of the music formats you listen to.
Oh and I forgot to mention when I questioned them if the device was capable of resolving all advertised formats with the DSP enabled they said "of course"! But when I examined the board and found the DSP chip they were using I knew it was a lie:
https://www.st.com/resource/en/datasheet/sta311b.pdf
Oh and I forgot to mention when I questioned them if the device was capable of resolving all advertised formats with the DSP enabled they said "of course"! But when I examined the board and found the DSP chip they were using I knew it was a lie:
https://www.st.com/resource/en/datasheet/sta311b.pdf
Last edited:
Yes, Roon has developed their software around the capacity of powerful processors used at high clock rates able to manage DSP functions without altering music quality. Reason also why they advise to use a networked DAC.
There was a discussion about the different clock frequency approaches between Antipodes and Roon.
Let’s know when your new box is ready.
Yes this level of DSP can't be done on portable devices that can fit in your pocket. And the portability advantage goes out the window when you connect these DAP's to a Chord Dave DAC, Mscaler, Loki, speaker amp, interface box, then headphones. Unless you have awful big pockets and a large battery strapped to your back. If not I recommend using your laptop or desktop computer as the source.
Regarding the interface, the resistor values I need are not available off the shelf from any of the distributors that offer them. So won't likely be until April when they arrive.
Last edited:
what do you mean by this??
Previously I did a very silly mistake, by applying a +4dB bass shelf and not compensating for it, which caused relatively frequent (digital) clipping. After applying a balancing -4dB compensation, I now don't experience clipping anymore, even when pushing beyond my usual listening levels (i.e. up to 85+dB
Users who are viewing this thread
