Tweeter line level passive XO
I'm really impressed with the bang for the buck with my 228rmb speakers, but have a theory they're rather being limited by their crossover. Especially the first order tweeter crossover which hasn't enough LF rejection.
In search of a steeper XO for the tweeter I've decided to go the whole hog and design a kick-*** filter that'll allow me to fully activate these puppies and see just how much they're limited by their electronics. kinku just pointed me to a filter design program called AADE which I've just finished using to design my first LC bandpass filter.
Its bandpass because I want to stick it straight after the DAC and it needs the anti-imaging function, meaning a steep low pass around 18kHz. The tweeter doesn't need this as they're fairly low on IMD, but the tweeter amp sure benefits from having no ultrasonics. Hence here is my first attempt at a tweeter bandpass filter, to go between the DAC and a dedicated tweeter amp.
For now I'll use the TDA1521 for tweeter duty but I have ideas how to go a little bit beyond that in terms of PSRR. That may well feature in a later blog post.... I'll now get on and start winding some coils for this
Update : here's the prototype build. I used the 18mm diameter pot cores so as to get higher Q values at 4kHz. Winding didn't take very long as the biggest value (1.24mH) only has 70 turns of 0.44 wire. Qs turned out to be plenty high enough, DCR for these is below 0.5ohm. Caps this time are NP0, paralleled 100nFs plus assorted smaller values because lower frequency Q is an aim and X5Rs have rather high ESR at lower freqs.
Update2 - added a munged, 10th order filter for those who aren't so comfortable with winding inductors. This has 5 inductors per channel and only two different values. -3dB points are 3.6kHz and 19kHz. I shall have a look around for off-the-shelf inductors that might be suitable but I'm not hopeful as all those I've so far seen don't have enough LF Q (by LF I mean 4kHz).
In search of a steeper XO for the tweeter I've decided to go the whole hog and design a kick-*** filter that'll allow me to fully activate these puppies and see just how much they're limited by their electronics. kinku just pointed me to a filter design program called AADE which I've just finished using to design my first LC bandpass filter.
Its bandpass because I want to stick it straight after the DAC and it needs the anti-imaging function, meaning a steep low pass around 18kHz. The tweeter doesn't need this as they're fairly low on IMD, but the tweeter amp sure benefits from having no ultrasonics. Hence here is my first attempt at a tweeter bandpass filter, to go between the DAC and a dedicated tweeter amp.
For now I'll use the TDA1521 for tweeter duty but I have ideas how to go a little bit beyond that in terms of PSRR. That may well feature in a later blog post.... I'll now get on and start winding some coils for this

Update : here's the prototype build. I used the 18mm diameter pot cores so as to get higher Q values at 4kHz. Winding didn't take very long as the biggest value (1.24mH) only has 70 turns of 0.44 wire. Qs turned out to be plenty high enough, DCR for these is below 0.5ohm. Caps this time are NP0, paralleled 100nFs plus assorted smaller values because lower frequency Q is an aim and X5Rs have rather high ESR at lower freqs.
Update2 - added a munged, 10th order filter for those who aren't so comfortable with winding inductors. This has 5 inductors per channel and only two different values. -3dB points are 3.6kHz and 19kHz. I shall have a look around for off-the-shelf inductors that might be suitable but I'm not hopeful as all those I've so far seen don't have enough LF Q (by LF I mean 4kHz).
Total Comments 17
Comments
-
So tweeters are fairly low on imd?
Say I want to simplify and omit the anti-imaging filter...
What tweeter amp would be best in that case? Any known amps, from the top of your mind?Posted 13th June 2014 at 06:07 AM by Alexandre -
Glad you picked up on that Alex - you can do more research on that at David Griesinger's website. He found amps worse than tweeters on generating IMD when researching audiability of ultrasonics.
If you want to simplify then go to the online Chebyshev Pi high pass filter calculator - you'll find it about halves the number of inductors.
I don't know any existing amps for tweeters - except perhaps that JLH amp I put up on my last blog post. As its classA it should be fairly decent. Class AB amps would probably need to have lots of caps added before gaining my seal of approval...Posted 13th June 2014 at 07:24 AM by abraxalito -
I've been interested in this kind of thing myself - an RC or LRC passive filter BEFORE the amplifiers, which would all be built into the speaker.
A few years ago, I heard a DIY speaker at my local club that used a battery powered class-A amp with a line-level passive crossover for the midrange and tweeter. These were augmented by a mains-powered, built-in subwoofer below 100Hz or so. The system was an MTM using two pro 8" midwoofers and a high-quality horn+compression driver combo, so the class-A amp (10W) was plenty to provide lots of SPL.
Although you can use a design program, you can also just look up values in textbooks for a resistance terminated network, and then scale all the component values as needed. Of course the design program makes everything very facile. It's a good idea to check the network's input impedance to make sure that it doesn't get too low and present a difficult load to the previous stage in the electronics chain (e.g. the preamp).Posted 13th June 2014 at 03:39 PM by CharlieLaub -
Richard, I imagine it´s really hard to get such a filter right. Too many tolerances interacting. This is wizardry man
But once you get it right it must be very good indeed. I wanted to show you this Ed Meitner interview:
Quote:"With 44.1, the filter, the so-called oversampling filter or the interpolator, is a phase linear device, right? So you have a ringing before the transient. And you have a ringing after. Now the ringing before the transient, our hearing system is anticipating and we hear that as a noise modulation, ever so slightly, but grating. I remember whenever we played with CD here, we always went back to a one FS system with an analogue filter"
The DSD/SACD Revolution, Part II:Posted 13th June 2014 at 06:37 PM by Alexandre -
@Charlie - so how did the speaker with the PLLXO sound? Its my idea of a no-compromise speaker, using a PLLXO with amps optimized for their respective drive units.
The main problem with PLLXOs is the working impedance, which is one big reason I put this straight after the DAC, where the impedance is low (50ohms). At normal line levels you'll have problems sourcing (most probably winding) inductors big enough so as not to unduly load the driving kit. I have an example elsewhere on my blog of a passive line-level filter - its an ultrasonic filter but already the inductors are into the 10's of mH. Imagine bringing this down to 3kHz or so for a speaker XO....
@Alex - actually if you compare getting passive filters working against active filters, passive wins hands down. All the work is done in getting the right values and Q's, but then when I've wired them up they work first time. Active filters always seem to run into some problem with oscillation or supply decoupling or grounding errors.
Interesting quote from Ed Meitner, thanks. I wonder why he hasn't noticed the noise modulation inherent in DSD yet?Posted 14th June 2014 at 12:02 AM by abraxalito -
I´m not familiar with dsd, but I would imagine commercial interests play a role too...
He also said the "one FS converter" was the sanity check.
(If anyone reading this is wondering what "one FS" means: it´s the same as non-oversampling!)Posted 14th June 2014 at 12:25 AM by Alexandre -
OK, now I´m seriously interested in the passive filters
There´s so much to inductors (core material, air gaps, multi-filar windings, getting the Q right - do you add the resistors to tune Q?)
The problem is that I would have to wait one month or more if I ordered cores from china... No local sources that are affordable.Posted 14th June 2014 at 12:41 AM by Alexandre -
Getting the Q right - well actually this is really about aiming for the highest possible Q. When I first started playing with passive filters that was my biggest problem - although low freq Q didn't matter too much (I was building AAFs with cutoffs near 20kHz, the cutoff freq is where Q matters most) I couldn't get the 20kHz Q high enough no matter how hard I tried.
Then I discovered ferrites with air gaps, and haven't looked back. Getting HF Qs in the region of 200 isn't a big problem now, and rarely do I need such a high value. I've only used multifilar windings (like Litz wire) on my classD amp where I needed low loss at 300kHz. I'm using wire with about 0.44mm diameter for this PLLXO and the LCR meter shows the losses are less than 50% higher at 20kHz than at 1kHz, so this isn't a problem. But going higher in diameter I'll avoid for where I need decent HF Q.Posted 14th June 2014 at 01:28 AM by abraxalito -
Hmmm, there´s one test I´ve got to do now, curiosity arised... My only source is PC and I upsample to 88.2 or 96KHz inside J.River... Now I want to know what sort of filter it employs and wheter or not it adds pre-ringing. Need to run some step functions and look at the output.
Posted 14th June 2014 at 05:31 AM by Alexandre -
At a guess - almost certainly it'll use a linear phase FIR to upsample, so there will be pre-ringing and post-ringing. But the ringing is almost certainly ultrasonic unless you can hear above 20kHz.
I think Mouser has some ferrite cores, problem is with those that they weren't gapped ones last time I looked. A bit tricky to get into grinding down ferrite - apparently its possible with the right kind of emery paper but I've never tried that. Or perhaps a diamond nail-file would do the trick?Posted 14th June 2014 at 09:57 AM by abraxalito -
These are looking good: 20pcs Pot Core P26 16 4600UH 25 5mm x 8mm Ferrite New Military Surplus | eBay
Too bad the bobbins are not included.
Do you make the gap on the center of the core only?
(I could use stacked paper to make a gap, but that would leave a gap on the ouside as well...)Posted 14th June 2014 at 06:30 PM by Alexandre -
Yeah they look promising - you can use sellotape or paper to make the gap and it doesn't matter about having the gap on the outside as well, it just means you've effectively got twice the thickness of the paper (or sellotape) as your gap. Probably you'll need a gap between 0.2 and 0.5mm to get these to having a useful AL value (the cores I use are AL=250nH).
Having no bobbins is a bigger drawback than having no gap I fear. You could try using a wooden dowel and wrap some stiff paper around it as a kind of former but then there's no way to constrain the windings at each end.
Progress update - I wound all the coils and I've finished wiring up one channel. It worked first time - I connected the siggen (fortunately it has 50ohm output impedance) and my AC voltmeter to the other end. At 1kHz, no output but stepping up the frequency to 4kHz, bingo! Another 'works first time' passive filterPosted 15th June 2014 at 12:16 AM by abraxalito -
Alex - I've been over to that interview with Ed Meitner and I realized I must have read it before because its so obvious Ed's been drinking DSD-flavoured Koolaid. The kinds of excuses he comes up with for why he thinks DSD beats PCM read like post-purchase rationalizations of someone who's spent a lot of money on a new toy and is talking themself up so as not to admit making a mistake. There's just so much religious nonsense - like
To convert audio into PCM is a very alien thing,
Imagine what happens at your zero crossing. You have all those bits flipping.
You have minimal resolution at zero crossing, whereby with DSD you have maximum resolution and on and on.
So now look at a PCM signal at zero crossing, and all you’ve got at that moment where it crosses zero is you have zero-bit resolution. The only resolution you’ve got is dither.
Then later on he's claiming DSD is no worse for jitter than PCM is. I just have to laughPosted 15th June 2014 at 10:46 AM by abraxalito -
I agree, that is just ridiculous. What´s alien is the high levels of noise, that have to be dealt with in DSD.
At least he admitted the non-oversampling dac was the sanity check!Posted 16th June 2014 at 02:23 AM by Alexandre -
Yeah the high level of noise is the thing from outer space as regards audio technology.
I think Thorsten has done some back of the envelope kinds of calculations (maybe on AudioAsylum?) about DSD's resolution. If you bear in mind that its 64X OS but there are only two levels then in one sample interval, how many different analog levels can be created with a succession of 64 0 or 1's? The answer is nothing like 16bit PCM - if you have all 64 as 1's that's the positive extreme, all 64 as 0's that's the negative. In between - well flip one bit from 1 to 0 this gives 63 1s and 1 0. What's the difference in level when averaged out? One part in 64 ISTM, therefore 6bits. Of course going to lower frequencies you get more bits in the averaging so the resolution seems to go up with lower freq, maybe this is why DSD sounds best in the bass?Posted 16th June 2014 at 03:09 AM by abraxalito -
Richard, would this be the material about IMD?
https://www.davidgriesinger.com/intermod.ppt
Quoting David Griesinger:
"The various tweeters tested – 3 metal dome tweeters and one soft dome tweeter – produce insignificant amounts of intermodulation products below 20kHz when driven by ultrasonic signals.
Amplifier distortion can produce distortion products below 20kHz that are audible (with difficulty) in the absence of other signals below 20kHz.
But with a high quality amplifier these distortion products are not audible in the presence of even extraordinary ultrasonic sources such as rattling keys.
Unless the amplifier is driven into clipping."Posted 16th June 2014 at 03:33 AM by Alexandre -
Yeah you found it. I found that introducing my passive filter after the TDA1387 DAC really gave a kind of 'holographic' feel to the soundstage which I'd not experienced before. I wrote about it on WBF here - Digital that sounds like analog
Posted 16th June 2014 at 03:49 AM by abraxalito