I use the same work flow. However, delays and all kinds of compensations can be made with opamps using all pass filters, shelving filters and gyrators. In short, IIR filters can be translated into electronic circuits.
Hi,
I agree to the above mentioned workflow.
It only does´nt solve the issues of the implementation of the analog xover.
I made very good experiences using simple JFET-Buffers as active subassembly.
Already the two-NJFET Buffer with one JFET used as Source-follower and the second as ccs, connected to the first JFET´s source suffices.
2SK170, their new Dual replacement LSK389, BF862, 2N4391 are suitable candidates.
Better THD-figures and higher possible supply lines can be achieved with cascodes, preferably with a second pair of NJFETs.
The 2N4392 and 2N4391 may be used to cascode all above mentioned JFETs.
If higher current capabilities are required the JFET-Buffers may be easily beefed up with a small power transistor, like in the Calvin-Buffer (CFP-configuration).
These Buffers used in passive Filters or with feedback in unity-gain Sallen-Key structure achieve excellent sonic results.
But what is equally important but still most always overlooked in active filter design is simplicity.
By this I mean, that almost all active xovers (and certainly all digital IIR-filters) just combine textbook filters.
Those require a perfectly linear load transfer function of H=1.
Better candidates add textbook equalizer stages to cope with the always existant non-linearities of the real world load.
All together combines to a multitude of filter stages and a myriad of parts all within the signal path.
A decent passive xover on the other hand is designed with a non-linear load in mind and taylors the filter to the load using the least amount of parts.
Here, filtering and equalizing are combined, reducing parts number and stage-number count.
If one applies this technique to active filters one can reduce the number of parts involved considerably.
But one has to recognize that common terms like Linkwitz, Butterworth, Chebysheff loose their meaning almost completely.
For example: I need for my ESLs (and most any other hybrid-ESL) a steep highpass, a soft peaking lift and two Notches to achieve the desired amplitude response.
This required using just four of the above mentioned JFET buffers, of which two are positioned in the direct signal path and two are used as gyrators in a shunt configuration to emulate the inductances of the two notches.
Using true inductances the number of Buffers would reduce to two, of which the second functions already only as cable driver.
In essence: just one single JFET sourcefollower is required to generate the complete 4-filter response!
Of course such a filter is hardly flexible any more -similar to a passive xover within a speaker- but the sonic advantage is great.
Believe me or not, but that transports Music to Your ears, not HiFi.
Summed up these three points assure a extremely well sounding xover imho:
- precise evaluation of filter response
- reduction of the number of filter-, equalizer- and active stages
- simple discrete Buffer stages like Diamonds (bipolar) or cascoded/boosted JFETs (interestingly all come just with local but no global feedback).
jauu
Calvin
I agree to the above mentioned workflow.
It only does´nt solve the issues of the implementation of the analog xover.
I made very good experiences using simple JFET-Buffers as active subassembly.
Already the two-NJFET Buffer with one JFET used as Source-follower and the second as ccs, connected to the first JFET´s source suffices.
2SK170, their new Dual replacement LSK389, BF862, 2N4391 are suitable candidates.
Better THD-figures and higher possible supply lines can be achieved with cascodes, preferably with a second pair of NJFETs.
The 2N4392 and 2N4391 may be used to cascode all above mentioned JFETs.
If higher current capabilities are required the JFET-Buffers may be easily beefed up with a small power transistor, like in the Calvin-Buffer (CFP-configuration).
These Buffers used in passive Filters or with feedback in unity-gain Sallen-Key structure achieve excellent sonic results.
But what is equally important but still most always overlooked in active filter design is simplicity.
By this I mean, that almost all active xovers (and certainly all digital IIR-filters) just combine textbook filters.
Those require a perfectly linear load transfer function of H=1.
Better candidates add textbook equalizer stages to cope with the always existant non-linearities of the real world load.
All together combines to a multitude of filter stages and a myriad of parts all within the signal path.
A decent passive xover on the other hand is designed with a non-linear load in mind and taylors the filter to the load using the least amount of parts.
Here, filtering and equalizing are combined, reducing parts number and stage-number count.
If one applies this technique to active filters one can reduce the number of parts involved considerably.
But one has to recognize that common terms like Linkwitz, Butterworth, Chebysheff loose their meaning almost completely.
For example: I need for my ESLs (and most any other hybrid-ESL) a steep highpass, a soft peaking lift and two Notches to achieve the desired amplitude response.
This required using just four of the above mentioned JFET buffers, of which two are positioned in the direct signal path and two are used as gyrators in a shunt configuration to emulate the inductances of the two notches.
Using true inductances the number of Buffers would reduce to two, of which the second functions already only as cable driver.
In essence: just one single JFET sourcefollower is required to generate the complete 4-filter response!
Of course such a filter is hardly flexible any more -similar to a passive xover within a speaker- but the sonic advantage is great.
Believe me or not, but that transports Music to Your ears, not HiFi.
Summed up these three points assure a extremely well sounding xover imho:
- precise evaluation of filter response
- reduction of the number of filter-, equalizer- and active stages
- simple discrete Buffer stages like Diamonds (bipolar) or cascoded/boosted JFETs (interestingly all come just with local but no global feedback).
jauu
Calvin
Last edited:
Calvin,
Simplicity is a virtue and everything you can take out of the signal train is gain. However, 5 good op amps in a row, the maximum amount I have in my xovers, are inaudible on speakers. Mooly did a recent test, SY has done and published a test (with pretty crummy op amps), and this seems to be a fair conclusion.
I just wonder. Have you measured the distortion produced by your simple buffer topology filters?
Simplicity is a virtue and everything you can take out of the signal train is gain. However, 5 good op amps in a row, the maximum amount I have in my xovers, are inaudible on speakers. Mooly did a recent test, SY has done and published a test (with pretty crummy op amps), and this seems to be a fair conclusion.
I just wonder. Have you measured the distortion produced by your simple buffer topology filters?
Yes, like thé active crossover of Nelson pass. Someone else also proposed à cascoded jfet buffer but i do notre remember in wich thread...
Hi
When the spacing is less than ¼ wl, two drivers couple into one new source of radiation, they feel each other’s radiation pressure and exhibit mutual coupling and can combine into one more powerful source if within a horn. Once one reaches about ½ wl, they radiate as independent sources and produce an interference pattern which can be recognized in a polar measurement as a series of lobes and nulls.
One can see the effect of each when one inverts one of the two sources, when close coupled, the effect is a large reduction is radiated sound via cancelation while with the larger spacing, it is the pattern of lobes and nulls which changes orientation with very nearly the same acoustic power being radiated instead of canceling out.
Since in most loudspeakers the sources are too far apart to combine coherently in to one new source, the object is to have the main lobe pointed forward at the listener and not the nulls.
The ideal crossover slope is governed in part by the raw response who’s magnitude and phase are added to the crossovers magnitude and phase in the result the sum of the upper and lower band governs what the resultant crossover looks like. In other words, if the raw driver has a 2nd order high pass acoustic response, adding a 4th order high pass filter in that region results in a 6th order acoustic response in magnitude and phase.
An active crossover (particularly a loudspeaker controller) has the advantages that one has many crossover slopes to choose from and changing the frequency is easy plus equalization is trivial
. As long as the loudspeakers response is "minimum phase" any eq you add with not only correct the amplitude response but also the phase response.
hope that
helps
Tom Danley
Danley Sound Labs
Appreciation
Tom,
Thanks kindly for taking the time to answer my questions. It does indeed
help quite a bit.
Sincerely,
Scott
Tom,
Thanks kindly for taking the time to answer my questions. It does indeed
help quite a bit.
Sincerely,
Scott
Hi,
It´d be hard to tag a certain number onto a certain circuit, as THD relates to so many different parameters like signal amplitude, supply lines, load impedance, frequency, etc.
IIrc Erno Borbely has written something about the THD theme.
The simple 2-NJFET Buffer (ccs-loaded source follwer) is already low enough in THD to cause any worry, as long as the load impedance remains high enough and the load current stays below ~1/10 the idle current.
The 4-NJFET (cascoded ccs loaded cascoded source follower) reduces THD by up to a factor of 10 into appropriate loads at the cost of lower clipping voltage.
The beefed up 2-NJFET/bipolar-CFP shows extremely low THD into middle to high load impedances and low values into low impedances.
The 4-NJFET/bipolar CFP, also known as Calvin Buffer is extremely low over the full load impedance range, but it´s clipping limit is clearly ~+-5V lower due to the cascoding.
All four buffer´s ccs may be transformed into modulated current sources, thereby offering higher current drive capabilty, as they transit from SE mode to PP-mode.
Using JFETs is advantageous infilter circuits due to their lack of Base current/very high impedance they don´t load/affect filter stages.
Also beeing depletion mode devices the Biasing is easier, requiring less parts, making cirucits simpler.
jauu
Calvin
Yes, but lost those data years ago due to computer crash.
It´d be hard to tag a certain number onto a certain circuit, as THD relates to so many different parameters like signal amplitude, supply lines, load impedance, frequency, etc.
IIrc Erno Borbely has written something about the THD theme.
The simple 2-NJFET Buffer (ccs-loaded source follwer) is already low enough in THD to cause any worry, as long as the load impedance remains high enough and the load current stays below ~1/10 the idle current.
The 4-NJFET (cascoded ccs loaded cascoded source follower) reduces THD by up to a factor of 10 into appropriate loads at the cost of lower clipping voltage.
The beefed up 2-NJFET/bipolar-CFP shows extremely low THD into middle to high load impedances and low values into low impedances.
The 4-NJFET/bipolar CFP, also known as Calvin Buffer is extremely low over the full load impedance range, but it´s clipping limit is clearly ~+-5V lower due to the cascoding.
All four buffer´s ccs may be transformed into modulated current sources, thereby offering higher current drive capabilty, as they transit from SE mode to PP-mode.
Using JFETs is advantageous infilter circuits due to their lack of Base current/very high impedance they don´t load/affect filter stages.
Also beeing depletion mode devices the Biasing is easier, requiring less parts, making cirucits simpler.
jauu
Calvin
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.