The 20 pin DIP is way too big, and has 0.400 center-to-center, a 14pin DIP is 0.300
If you're going to do a real hack, just skip the board and attach wire to the SMD and run it down to the board. I use 24ga telephone wire stripped out from 50 pair short scraps.
The old J-fet ICs require the 10K input resistor for protection, some newer types may not. Check the data sheets. I don't know how well those bi-polar types will work, they still need an input resistor to protect the internal clamp diodes (see the IC data sheet) and the filter part values may need to be sized to reduce input bias errors. The suggested OPA4227 is only 10^7 input impedance vs 10^12 on the TL and LF parts.
If you're going to do a real hack, just skip the board and attach wire to the SMD and run it down to the board. I use 24ga telephone wire stripped out from 50 pair short scraps.
The old J-fet ICs require the 10K input resistor for protection, some newer types may not. Check the data sheets. I don't know how well those bi-polar types will work, they still need an input resistor to protect the internal clamp diodes (see the IC data sheet) and the filter part values may need to be sized to reduce input bias errors. The suggested OPA4227 is only 10^7 input impedance vs 10^12 on the TL and LF parts.
Hi,
I have built this circuit and used it in my 3 way active speakers.
Am using OPA4228.
Its a very simple active crossover .
if u have a scope u can really get something out of it.
But this is used only when u know the crossover freq that u want.
xover freq pints are 341 and 3400.
Cost me hardly 10$ to build it and i got the OPA as a give away from a friend.
Cheers,
Sachi
I have built this circuit and used it in my 3 way active speakers.
Am using OPA4228.
Its a very simple active crossover .
if u have a scope u can really get something out of it.
But this is used only when u know the crossover freq that u want.
xover freq pints are 341 and 3400.
Cost me hardly 10$ to build it and i got the OPA as a give away from a friend.
Cheers,
Sachi
I love this xover, as a 2way, got some good sample IC's comming
(OPA4134UA) and Browndog adapters, I'm using only IC1 as high pass and IC5 as low pass both driven by IC1a. IC3 is totaly out of the circuit. When all was crossed over at 239hz it looked great on the scope and the 24db cut was acting fine both high and low pass.
Then I decided to use it at 160hz with the approprate resistor changes, the low pass cut nicely at 160hz, but the high pass just gives a little boost 1or2 db just before it starts to cut at 160hz, this did not happen at the 239hz it was flat up till the cut starts but at 160hz the high pass just boosts a little before it starts the cut.
Is there any explanation for this, as the only change is 2 resistors from 20k to 30k that go to ground at the input of IC1d and IC1c the caps remained the same 47nf. Is there some form of resonant circuit going on between the 30k and the 47nf that doesn't happen with a 20k resistor?
The low pass IC5 had no such problem with the 160hz setting
.
Here is the link to the circuit again.
http://www.siliconchip.com.au/cms/A_30278/article.html
Cheers George
(OPA4134UA) and Browndog adapters, I'm using only IC1 as high pass and IC5 as low pass both driven by IC1a. IC3 is totaly out of the circuit. When all was crossed over at 239hz it looked great on the scope and the 24db cut was acting fine both high and low pass.
Then I decided to use it at 160hz with the approprate resistor changes, the low pass cut nicely at 160hz, but the high pass just gives a little boost 1or2 db just before it starts to cut at 160hz, this did not happen at the 239hz it was flat up till the cut starts but at 160hz the high pass just boosts a little before it starts the cut.
Is there any explanation for this, as the only change is 2 resistors from 20k to 30k that go to ground at the input of IC1d and IC1c the caps remained the same 47nf. Is there some form of resonant circuit going on between the 30k and the 47nf that doesn't happen with a 20k resistor?
The low pass IC5 had no such problem with the 160hz setting
.
Here is the link to the circuit again.
http://www.siliconchip.com.au/cms/A_30278/article.html
Cheers George
You also need to change the 10K in the high pass to 15K. As it is now you have two sections of your filter at 160 Hz. and two at 239 Hz.
If you made this change and still have a bump it is likely due to the caps being out of tolerance. You alter the Q as well as the tuning if your caps are off. Sallen Key topology is fairly sensitive to component tolerances.
If you made this change and still have a bump it is likely due to the caps being out of tolerance. You alter the Q as well as the tuning if your caps are off. Sallen Key topology is fairly sensitive to component tolerances.
Hi George,
did you move the first 10k on IC5d to the IC1a output? or using as posted?
Try to match caps on both the high pass and low pass sides to better than 1% and resistors to better than 0.5%.
Note that where you have parallel caps, each can be rejects from the selection process as long as they correct each others' error, i.e. +4% and -4.9% effectively cancel out to give <-0.5%.
An advantage of using the parallel pairs is elimination of meter accuracy error.
You are talking about displaying the frequency vs attenuation graphs on scope. How are you achieving this? Precise details please.
did you move the first 10k on IC5d to the IC1a output? or using as posted?
Try to match caps on both the high pass and low pass sides to better than 1% and resistors to better than 0.5%.
Note that where you have parallel caps, each can be rejects from the selection process as long as they correct each others' error, i.e. +4% and -4.9% effectively cancel out to give <-0.5%.
An advantage of using the parallel pairs is elimination of meter accuracy error.
You are talking about displaying the frequency vs attenuation graphs on scope. How are you achieving this? Precise details please.
BobEllis said:
Why would the caps be out of tolerance now with the 30k's in place and not when the 20k's were there, same caps unchanged?
Here is what I've done, the high pass 4 x 2.2nf are 47nf and the 2x 20k are 30k. The low pass, 4x 10k changed for 15k.
The 47nf were the same ones when the xover was set for 239hz, then there was no slight rise in output before cutting at 239hz.
Attachments
Hi George,
you have upset the Q of the high pass filter. You must change both the resistor sets. 10k>15k and 20k>30k
There are two parts to the 2pole filter wrapped around each opamp stage. Consider the low pass first, the feedback part through the 47n//47n and the passive part using the second 10k+47n.
The same applies to the high pass. The feedback part is the 10k and the passive part is the second 2n2+20k. Notice the 2R, R, C and 2C in turquoise but one R has been omitted from the top 10k.
The ratios in both parts must be kept the same.
The alteration would have been much simpler if you had linked IC3d pin 14 to IC1a pin1. then you could have used the 47nF in their original locations and only needed to change the eight resistors. Are the resistors set up in the form of DILs? 8 pin DIL socket allow easy frequency changes.
That is the big disadavantage of unity gain S&K. interlinking of Q with F which in equal value S&K are sustantially separated but now gain changes with Q. However there is MFB which separates Q, F and gain to make changes easy but cost one extra component and a lot of unmatched component values (a difficult one for the beginner).
you have upset the Q of the high pass filter. You must change both the resistor sets. 10k>15k and 20k>30k
There are two parts to the 2pole filter wrapped around each opamp stage. Consider the low pass first, the feedback part through the 47n//47n and the passive part using the second 10k+47n.
The same applies to the high pass. The feedback part is the 10k and the passive part is the second 2n2+20k. Notice the 2R, R, C and 2C in turquoise but one R has been omitted from the top 10k.
The ratios in both parts must be kept the same.
The alteration would have been much simpler if you had linked IC3d pin 14 to IC1a pin1. then you could have used the 47nF in their original locations and only needed to change the eight resistors. Are the resistors set up in the form of DILs? 8 pin DIL socket allow easy frequency changes.
That is the big disadavantage of unity gain S&K. interlinking of Q with F which in equal value S&K are sustantially separated but now gain changes with Q. However there is MFB which separates Q, F and gain to make changes easy but cost one extra component and a lot of unmatched component values (a difficult one for the beginner).
Hi George,
I have just noticed that there is no DC blocking on the output of the filters nor on the output of the buffers. IC1d offset is blocked by the following 2n2 but all the other opamp offsets feed through to the outputs.
This may require careful checking of output offset and how it varies with temperature and time.
There is an inherent mismatch of impedance between the inverting and non-inverting input pins on all the opamps. Some opamps will cause variable output offsets due to the tolopology adiopted by the designer.
It may be of no consequence with many opamps. But worth checking.
I have just noticed that there is no DC blocking on the output of the filters nor on the output of the buffers. IC1d offset is blocked by the following 2n2 but all the other opamp offsets feed through to the outputs.
This may require careful checking of output offset and how it varies with temperature and time.
There is an inherent mismatch of impedance between the inverting and non-inverting input pins on all the opamps. Some opamps will cause variable output offsets due to the tolopology adiopted by the designer.
It may be of no consequence with many opamps. But worth checking.
Andrew T, as I said in a previous post, I have a couple of millivolts of offset at the output of each high and low pass amp and it is very stable, all sections are dc coupled from input to output, except for the caps it the high pass filter, the only thing i can think that's causing this rise before the cut in the high pass is a reactance of the 30k with something, it's small.
I measured it on the scope, with a signal of 3v peak to peak at say 1k sinewave I sweep down the the cut off at 160hz, all is flat until I get to about 200hz then it rises to to about 3.1v peak to peak then starts to fall away with the 24db cut comming in at 160hz, it's weird as this was not there when the filter was set for 239hz same caps just different resistors, also I disconected the lowpass incase it was interacting with it, so it is just the high pass on it's own that's doing it.
Hopefully it's the TL074's doing it and when I get the OPA4134UA in there it may go away, I'm sure the rise is not detectable as it is so small in level (3%), but it's there and it bugs me.
Cheers George
I measured it on the scope, with a signal of 3v peak to peak at say 1k sinewave I sweep down the the cut off at 160hz, all is flat until I get to about 200hz then it rises to to about 3.1v peak to peak then starts to fall away with the 24db cut comming in at 160hz, it's weird as this was not there when the filter was set for 239hz same caps just different resistors, also I disconected the lowpass incase it was interacting with it, so it is just the high pass on it's own that's doing it.
Hopefully it's the TL074's doing it and when I get the OPA4134UA in there it may go away, I'm sure the rise is not detectable as it is so small in level (3%), but it's there and it bugs me.
Cheers George
Sorry Bob E, you were right.
I see now where I've made the mistake and getting the rise in output before the cut on the highpass section, it's not mentioned in the main schematic on Silicon Chip but the feedback resistor also has to change slightly from 10k to 15k not just the r2's going to ground at the input of each section.
I only found this out after down loading the Linkwitz Riley filter calculator program.
http://sound.westhost.com/project09.htm#download_esp_lr
A part left out by Silicon Chip on their main schematic.
Cheers George
I see now where I've made the mistake and getting the rise in output before the cut on the highpass section, it's not mentioned in the main schematic on Silicon Chip but the feedback resistor also has to change slightly from 10k to 15k not just the r2's going to ground at the input of each section.
I only found this out after down loading the Linkwitz Riley filter calculator program.
http://sound.westhost.com/project09.htm#download_esp_lr
A part left out by Silicon Chip on their main schematic.
Cheers George
Sorry also to you Andrew.
Yeah I missed it in yours as well, as this is my first venture into active xovers I'm learning, and also seeing that projects have faults in their text (but one R has been omitted from the top 10k). Now all is perfect and behaving as planed, yet I hope the OPA4134UA will sound better in the highs.
The LF347's are doing a great job in the low bass, upper bass and lower mids, of my Martin Logan Monolith III's, but I have resevations about them in the top end, so long as the music is not too congested with highs they remain clean and distortion free (middle of the road type music), but when it gets complex I feel that the ML passive xover remained more in control, I'm hoping the OPA4134UA will fix this, if it doesn't I will keep the active for the bottom end as use the passive for the top end.
I feel that with the passive at 12db per octave @ 125hz and the active @ 24db 160hz they should mate well and be in the ball park to give a nice flat xover area, what do you think?
Cheers George
Hi George,
the passive crossover will already have some compensation built into the pass band and possibly just outside the pass band of each stage of the crossover.
You will need to mimic these compensations by adding extra stages to the active filter to make your speaker voice as they were when passive That's the purpose of the other sheets in MOXliteA that I posted). If you leave them as they are with just single pole or double pole slopes and no compensation you have done the easy 10% of turning the speakers into active. The difficult 90% relies on a lot of experimental listening and knowledgable adjustment.
One method I have seen proposed is to measure the voltage at the driver terminals for an audio signal sent in through the standard passive crossover. Then using the results of this for EACH driver to construct the active crossover to mimic the compensations found by measurement.
The complexity of all this extra 90% frightens me and I have avoided it for 20 years. Now I have a digital crossover that allows easy and nearly infinte adjustments to Passband Slope and Notch and Boost and Q and Frequency and Delay.
I am hoping this will allow me to voice the final crossover to sound just as nice if not better than the passive original. If I cannot make it better then I will have wasted a lot of effort (and money) for nought.
Without the compensations we can NEVER reach the standard of the passive crossover.
the passive crossover will already have some compensation built into the pass band and possibly just outside the pass band of each stage of the crossover.
You will need to mimic these compensations by adding extra stages to the active filter to make your speaker voice as they were when passive That's the purpose of the other sheets in MOXliteA that I posted). If you leave them as they are with just single pole or double pole slopes and no compensation you have done the easy 10% of turning the speakers into active. The difficult 90% relies on a lot of experimental listening and knowledgable adjustment.
One method I have seen proposed is to measure the voltage at the driver terminals for an audio signal sent in through the standard passive crossover. Then using the results of this for EACH driver to construct the active crossover to mimic the compensations found by measurement.
The complexity of all this extra 90% frightens me and I have avoided it for 20 years. Now I have a digital crossover that allows easy and nearly infinte adjustments to Passband Slope and Notch and Boost and Q and Frequency and Delay.
I am hoping this will allow me to voice the final crossover to sound just as nice if not better than the passive original. If I cannot make it better then I will have wasted a lot of effort (and money) for nought.
Without the compensations we can NEVER reach the standard of the passive crossover.
I feel that you may have misunderstood me,or I did'nt explain myself properly, If the it's the LF347 that is bugging me in the highs and the swap with OPA4134UA fixes the problem, all will be fine.
But if the OPA4134UA still doesn't achieve the control and separation that the Martin Logans passive high pass xover (125 2nd order) has I will go back to using it, and keep the active for the low pass, but seeing it's at 4th order at 160hz how will the crossover point look when these are together, will I have to lower the 4th order active to 125hz as well, or will I have to take even further down to 100hz?
Cheers George
But if the OPA4134UA still doesn't achieve the control and separation that the Martin Logans passive high pass xover (125 2nd order) has I will go back to using it, and keep the active for the low pass, but seeing it's at 4th order at 160hz how will the crossover point look when these are together, will I have to lower the 4th order active to 125hz as well, or will I have to take even further down to 100hz?
Cheers George
AndrewT said:Hi George,
I have gone back through all your posts but I can't find a description of your passive crossover.
Can you confirm what passive you have? rubbish!
(The passive 2nd order for the ML ELS pannels)
It's a 12db @ 125hz (cap 100uf in series and inductor 5mh to ground).
This is what I would like to know, if I keep the bass active 24db do I still blend the two different slopes together at the same -6db point (125), or will I get a sag or peak with the combinded slopes because of the different orders, that's about as simple as I can describe it.
Hear is a pic of the passive, the series 2ohm , 39uf and the lcr stay in regardless of using passive or active.
Cheers George
Hi george,
keep all of the high pass to the panel.
To enable lower distortion from the panel and a bit of extra power handling, you can activate the the woofer and send a matching signal to the retained high pass passive.
Since the existing is rolled off at 125Hz you can use exactly this for your active crossover.
The passive is 2pole 125Hz. Add a 2pole 125Hz to make this a 4pole giving a very slight reduction in very low bass and no effective increase in power handling.
Now supply the bass amp with a signal from a 4pole 125Hz crossover.
The two pole into the passive mid/treble may be close to Butterworth or not (I cannot tell). Adding a second 2pole Butterworth gives the opportunity to go Linkwitz Reilly.
Cascading two 2pole Butterworth for the low pass gives your L-R crossover. This is what you have in your crossover. just adjust your frequency to 125Hz.
There is an alternative.
Move the crossover point to 2octaves above the existing passband of the mid/treble unit. i.e. 500Hz. Now build a 4pole L-R Hi and Lopass crossover. This will give more power handling for the Mid/treble but could be asking too much of the bass unit.
There are only two ways to find out. Ask someone else how the modification sounded or do it yourself.
BTW. converting the bass to active will change the Q of the speaker and you will hear the difference whether for better or worse.
keep all of the high pass to the panel.
To enable lower distortion from the panel and a bit of extra power handling, you can activate the the woofer and send a matching signal to the retained high pass passive.
Since the existing is rolled off at 125Hz you can use exactly this for your active crossover.
The passive is 2pole 125Hz. Add a 2pole 125Hz to make this a 4pole giving a very slight reduction in very low bass and no effective increase in power handling.
Now supply the bass amp with a signal from a 4pole 125Hz crossover.
The two pole into the passive mid/treble may be close to Butterworth or not (I cannot tell). Adding a second 2pole Butterworth gives the opportunity to go Linkwitz Reilly.
Cascading two 2pole Butterworth for the low pass gives your L-R crossover. This is what you have in your crossover. just adjust your frequency to 125Hz.
There is an alternative.
Move the crossover point to 2octaves above the existing passband of the mid/treble unit. i.e. 500Hz. Now build a 4pole L-R Hi and Lopass crossover. This will give more power handling for the Mid/treble but could be asking too much of the bass unit.
There are only two ways to find out. Ask someone else how the modification sounded or do it yourself.
BTW. converting the bass to active will change the Q of the speaker and you will hear the difference whether for better or worse.
AndrewT said:
Thanks Andrew, seeing my bass driver now has the better ACI SV12 in place of the Eminence rubbish, the QTC with the passive was still at .9 because of the series impeadence of the xover, this was with both drivers, but now with the active it is .65 and sooo much better it's not funny.
The top end will wait till the OPA4134UA are in, if they don't still meet my expextations for transpaency and separation, I will go back to using the passive for the high pass, and bring the active low pass back to 125hz.
I will wait and see what the OPA4134UA will do, but to my ears the best opamp in the signal path in the mids and highs maybe no opamp at all .
Cheers George
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.