What words of mine do you consider 'unncessarily rude' ? If you mean my reference to your 'crystal ball' that was based on your 'seeing' lots of IMD measurements in my future.
Abraxalito;
build on your hunches, and post some nice measurements that demonstrate your hunches as correct. Theory and practice are a long way apart.
Along the way hone your blind testing skills with a variety of tests to validate your ability to differentiate between systems with easily measured differences.
Obviously you are in no need of explanations, just realistic experience.
build on your hunches, and post some nice measurements that demonstrate your hunches as correct. Theory and practice are a long way apart.
Along the way hone your blind testing skills with a variety of tests to validate your ability to differentiate between systems with easily measured differences.
Obviously you are in no need of explanations, just realistic experience.
Nice measurements? Nice? It would seem to me that 'nice' measurements would be ones that one wants to see, that are in accord with one's preconceived (expectation) biases.
Indeed they are - practice leads to decent theory in my case. I started out with various theories about what would sound good and a lot of those have been overturned so far. So far my theories are working hard to keep up with practice.
Can't see why that notion would be at all useful. Care to elucidate more?
'Obviously' based on which particular observations you've made?
Nice measurements, nice wide bandwidth impulse response measurements. Explanations of deviations that aren't audible, deviation that are audible, and show deviations from target bandwidth.
What manner of evaluations have overturned which of your theories and replaced them with your current set of hunches?
What practices have led you to your current set of hunches?
What manner of evaluations have overturned which of your theories and replaced them with your current set of hunches?
What practices have led you to your current set of hunches?
You've brought hunches to the table. Show some work on how you prepared them. I'm not going to swallow what you haven't demonstrated in practice.
Nobody else is either.
Theoretical conjecture doesn't cut it.
Nobody else is either.
Theoretical conjecture doesn't cut it.
Twest: i agree, if you're making 3,4,5 way with an 8 stage limit, then you're likely to run out of stages. I'm not sure, however that greater than 8 th order (4stages) is ever a requirement.
But it is only opinion, no fact in here.
But it is only opinion, no fact in here.
Plenty of that is already shown on my blog. When you've reviewed that, perhaps you'll have questions or comments. By all means leave such on the blog.
Your claim that my hunches are wrong was just that, theoretical conjecture. How ironic 😀
I looked through your blog entries; I see no measurements beyond simulations.
You are going backwards with your high pole count antialiasing filters.
Stop band of -50dB is poor.
Tiny FIR filter blows this sort of performance away.
Getting even close to theoretical performance for so many capacitors and coils from real circuit is daydreaming. Precision of component values is one factor. Saturation of ferrite cores is problematic too.
All your metrics appear to be based on perceived sound quality. Where are blind ABX results? Where are measurements of real circuit performance?
Where are the elucidations to back up your above claims?
Saturation of ferrite cores which are gapped and subject to a maximum of 2mA through them - you sure set new records for daydreaming 😀
<afterthought> I shall say this once more - put your comments and criticisms of specific shortcomings on the blog. Handwaving terms like 'poor' and 'blows away' and 'problematic' won't cut any mustard.
Saturation of ferrite cores which are gapped and subject to a maximum of 2mA through them - you sure set new records for daydreaming 😀
<afterthought> I shall say this once more - put your comments and criticisms of specific shortcomings on the blog. Handwaving terms like 'poor' and 'blows away' and 'problematic' won't cut any mustard.
Last edited:
Hi you could consider also the Najda board that can do FIR and IIR with excellent results/flexibility and reasonnable costs. See http://www.diyaudio.com/forums/digital-line-level/215379-dsp-xover-project-part-2-a-56.html
quite easy to upgrade dacs output stages with whatever you want to
You mean XOs with steeper than 8th order slopes aren't needed? If so, my experience agrees. It wasn't until quite recently I had a good way of ABXing XOs but when I stepped through every alignment from first to eighth order and experimented with steeper stuff I found LR6 seemed to be optimum. Bessel alignments never sound quite as good to me as Linkwitz-Riley---maybe due to the properties of my room but I suspect more that my dipoles are optimized for small acoustic size, meaning their off axis response sums most consistently when drivers are in phase rather than summing flat on axis in quadrature---so I mostly looked at LR2 through LR8. My experience with this it's the usual sort of diminishing returns. LR2 to LR4 is a clear win for improved directivity (via lower crossover points enabled by reducing drivers' transition band xmax requirements) and reducing the width of the phase transition band when using warped phase playback. LR4 to LR6 has the same value proposition with a reduced increment. LR6 to LR8 was pretty much a wash. Couldn't move the crossovers down much so the directivity benefit was quite small and my subjective impression was the speaker started to sound less coherent as the transitions from one driver to the next were too fast to glom together psychoacoustically. Not the easiest thing to measure or ABX but there are plausible explanations around near and far reflections path and length spreads. LR10 and above couldn't really move the crossover points and made the speaker seem to come apart more than LR8 did.
If you're referring to processing per channel I'd also agree it's perfectly possible to build a good active speaker with a limitation of a total of 8 orders of filtering per channel. However, if you're trying to get as much out of the system as is feasible additional orders are beneficial. My current triamp uses a total of 18th, 24th, and 16th order for the tweeters, mids, and subwoofers, respectively. 6, 12, and 6 of those orders are LR6 XO as described above. The rest are all ABX tested with each biquad or first order patch showing subjective improvement from a sound in = sound out perspective. It also happens 7, 8, and 3 of the orders, respectively, perform acausal phase correction and are quite difficult to approximate in the analog domain.
A disclaimer: I've improved various parts of the implementation over the past couple years and mean to revisit LR8 XO as it's possible better management of out of band artifacts would flip the subjective preference.
Twest, i am a novice, and my meaning is simply that i cant see a situation where i would need to go beyond 8th order (or even 6th order). In the same way that I'm no longer sure that 2nd order is enough for driver protection (in the midrange or tweeter case).
3rd to 6th is probably the sweet spot. In my own first active efforts, i found 5th order was needed (though i would never have thought so, i was aiming for LR4)
3rd to 6th is probably the sweet spot. In my own first active efforts, i found 5th order was needed (though i would never have thought so, i was aiming for LR4)
i´m using 1st for 8" woofers , 2nd for 5" midrange and 4th for 1" ribbon tweeters , no protection on tweeters , with minidsp and nanodigi , so far so good.
I wouldn't dare omit a DC blocking capacitor (but that's just me). Maybe you have AC coupled output amplifiers instead Iduarte?
I have some small Fountek ribbons I'm planning to use in another active design, my first 3way...
I have some small Fountek ribbons I'm planning to use in another active design, my first 3way...
Last edited:
dont think so , it´s been this way for a year without any problem , i´m using 4 fontek cd1 with a 12W tube amp (as tweeter amp only , only need 5W really ) , crossed at 3.5k 4th order. I should put a cap but it´s been working fine , when i have the 1st problem i´ll do it.
Last edited:
Well with my present amplifier, whilst i don't get a turn on thud , i get a click as the capacitor in the power supply discharges.
The capacitor does very little else, except align phase slightly better ( according to simulations )
I'm also not sure of the failure mode of a tube amp. Perhaps its not an issue, since there's an output transformer, DC output would be negligible anyway.
DSP is interesting, but I'm not totally convinced its good enough yet. Just a preconception of mine, but DSP is a great design tool.
What id really like to do is design the filter in DSP and build in analogue ��
A discrete tube xo would be something to behold��
The capacitor does very little else, except align phase slightly better ( according to simulations )
I'm also not sure of the failure mode of a tube amp. Perhaps its not an issue, since there's an output transformer, DC output would be negligible anyway.
DSP is interesting, but I'm not totally convinced its good enough yet. Just a preconception of mine, but DSP is a great design tool.
What id really like to do is design the filter in DSP and build in analogue ��
A discrete tube xo would be something to behold��
Last edited:
33uF in series with the tweeter, active xo before the amp. Another transformer in the ribbon tweeter probably doubly isolates DC with your valve amp. I doubt you need a capacitor, the transformers would do quite a good job i imagine. Quite neat actually .....
Last edited:
i could never get the sound i have with a passive system but i respect people that think otherwise
- Status
- Not open for further replies.