sss said:as long as u dont let your mind to fool u .all thiose amps will sound the same when used correctly .
almost in all studio equipment u can fint the 5532 amps . adding another one wont do any difference
hey , but thats just my opinion
You mean my circuit design & layout are more important than the IC used ? It is what I assume too but need to check though.
I always build circuits using IC sockets to replace parts easily so it won't be hard for me to give a try to different opamps. Cost is another question...
I've read on other forums people report quality differences between NE5534 provided by different manufacturers : is it a problem with the building process ? In fact, does a number part only refers to the schematic used ?
Also, talking about sound differences, some report that SE/SA5534 sound better than NE5534 : what is the difference between SE/SA and NE, apart from temperature certification ?
I'm quite sceptical about these differences...
Last noob question : how is it possible to class-A bias an NE5534 externally ? Many people talk about this but I didn't find any schematic
That could be true, within reason. As long as your opamp meets the basic gain-bandwidth product requirement and has a slew rate high enough for audio and is stable at the gain you use, you might be hard pressed to tell a difference.
A major problem with opamp swap listening tests is the time between listening experiences. To be somewhat valid, they should be conducted blind, switching between two circuits. (Not looking to get into a testing procedure argument)
Then you run into the "just plop another opamp in there" troubles. A circuit/layout optimized for one opamp may be totally unsuitable for another. The NE5532 will be quite happy driving 1,000 ohm loads, where the OPA2134 will have a bit of trouble. Putting an OPA627 on a SMD single to DIP double adapter and plugging it in is likely to cause oscillation troubles.
So, there isn't a fair way to test except make an optimized circuit for each of your proposed chips and listen to each.
As for diffferent brands sounding different, it is quite possible that they have different circuits internally although they meet the same performance specifications. Sometimes the tooling is transferred/licensed, but manufacturing techniques vary. Again, the same time lapse issues come into play - auditory memory isn't long term for details.
A major problem with opamp swap listening tests is the time between listening experiences. To be somewhat valid, they should be conducted blind, switching between two circuits. (Not looking to get into a testing procedure argument)
Then you run into the "just plop another opamp in there" troubles. A circuit/layout optimized for one opamp may be totally unsuitable for another. The NE5532 will be quite happy driving 1,000 ohm loads, where the OPA2134 will have a bit of trouble. Putting an OPA627 on a SMD single to DIP double adapter and plugging it in is likely to cause oscillation troubles.
So, there isn't a fair way to test except make an optimized circuit for each of your proposed chips and listen to each.
As for diffferent brands sounding different, it is quite possible that they have different circuits internally although they meet the same performance specifications. Sometimes the tooling is transferred/licensed, but manufacturing techniques vary. Again, the same time lapse issues come into play - auditory memory isn't long term for details.
Hi,
You can use the NE´s well for filter applications. Noise figures, distortion figures and bandwidth are more than adequate for audio. I personally prefer FET-OPs because they are better suited to higher impedances which will always occur in filter networks, but thats really more a matter of personal preference than usefulnes of the device. Layout and circuit design are of greater importance than the choice of OP-amp (in this quite basic application!). A good small sized low inductance power supply line layout and proper decoupling is worth alot. Have an eye on stability with low closed-loop gains and stability against capacitive loading. Then nearly any of the afore mentioned OPs will do fine
There´s nothing to say against using double and quad-ops in Filters. I prefer them against single OPs because of lesser space and size usage and therefore less problems with proper supply and stray effects. I had a rather humorous moment when reading about not using double OPs and the quality of different OPs for first class circuits on one hand and using sockets on the other
jauu
Calvin.
oh, just a last thought: "Will anybody hear what OP You actually used, if You don´t tell him? Noooooooooooo!"
You can use the NE´s well for filter applications. Noise figures, distortion figures and bandwidth are more than adequate for audio. I personally prefer FET-OPs because they are better suited to higher impedances which will always occur in filter networks, but thats really more a matter of personal preference than usefulnes of the device. Layout and circuit design are of greater importance than the choice of OP-amp (in this quite basic application!). A good small sized low inductance power supply line layout and proper decoupling is worth alot. Have an eye on stability with low closed-loop gains and stability against capacitive loading. Then nearly any of the afore mentioned OPs will do fine
There´s nothing to say against using double and quad-ops in Filters. I prefer them against single OPs because of lesser space and size usage and therefore less problems with proper supply and stray effects. I had a rather humorous moment when reading about not using double OPs and the quality of different OPs for first class circuits on one hand and using sockets on the other
jauu
Calvin.
oh, just a last thought: "Will anybody hear what OP You actually used, if You don´t tell him? Noooooooooooo!"
youyoung21147 said:
www.tangentsoft.net/audio/opamp-bias.html
Also read the link to Jon Risch, bottom of the tangentsoft page.
There are small heatsinks you can glue on top of the opamp. Type: ICK-8L from Fischer (K from Kühlkorper
) Looks pretty too.
jcx said:
What is a "buffer" CFA ? I think the opamps mentionned don't need buffering between the active filter stages (low enough output impedance) don't they ?
You mean I'd obtain better results if using the MFB topology rather than Sallen-key ? Is it true for any opamp ?
In my circuit, I'll be using a 2d order highpass and lowpass : these circuits reverse the phase of the signal, so I was thinking of using an opamp as non-inverting buffer, and to use another in MFB configuration for the active filter itself. Is it a good idea ? Will the phase shift be again 0° or 360° ?
Lol, for sure
Professional designers seem to know it and hardly use expensive parts in their circuits because they know it does very little difference, and also know the customer doesn't know how cheap the parts are 
Hi,
the MFB has two advantages that I am aware of:-
1. it has independant control of gain and damping (Q).
2. it is less sensitive to component tolerance and there are extensive articles (appnotes) to correct high frequency errors.
Both of these advantages are irrelevant for a 2pole audio filter.
Use the equal value S&K or the unity gain S&K for your filters.
If you were going to use 4pole Butterworth then there is considerable advantage in using MFB.
BTW 4pole Linkwitz Reiley is just two 2pole filters cascaded to achieve that characteristic.
A composite opamp uses two specialised opamps cascaded to remove most of the errors inherently built into single chip opamps.
The buffer, in this context, is a high current opamp (often a line driver) hung on the end of the volt amp stage with a lowish voltage gain but inside the overall loop gain of the composite.
Read Walt Jung for comprehensive design of these.
Has anyone seen other authors on the composite route?
the MFB has two advantages that I am aware of:-
1. it has independant control of gain and damping (Q).
2. it is less sensitive to component tolerance and there are extensive articles (appnotes) to correct high frequency errors.
Both of these advantages are irrelevant for a 2pole audio filter.
Use the equal value S&K or the unity gain S&K for your filters.
If you were going to use 4pole Butterworth then there is considerable advantage in using MFB.
BTW 4pole Linkwitz Reiley is just two 2pole filters cascaded to achieve that characteristic.
A composite opamp uses two specialised opamps cascaded to remove most of the errors inherently built into single chip opamps.
The buffer, in this context, is a high current opamp (often a line driver) hung on the end of the volt amp stage with a lowish voltage gain but inside the overall loop gain of the composite.
Read Walt Jung for comprehensive design of these.
Has anyone seen other authors on the composite route?
http://lei.fis.uc.pt/pdfs/Com001.pdf
Is it the "cascaded thing" on the schematic page 2 ?
If yes, I don't see the advantage over a simple buffer, because anyway, the active filter itself doesn't use composite amps. My buffer will work at a gain of one to four, so maybe the gain in errors could be lost in transparency ?
Is it the "cascaded thing" on the schematic page 2 ?
If yes, I don't see the advantage over a simple buffer, because anyway, the active filter itself doesn't use composite amps. My buffer will work at a gain of one to four, so maybe the gain in errors could be lost in transparency ?
composite amps can be used in the filter sections, the output amp can drive low Z filter components when you want low noise/hi dynamic range without the heat coupling to the input op amp
the output amp can also provide lots more loop gain, once the global negative feedback is high enough the remaining error is the common mode nonlinearity at the input - inverting topologies keep the dif pair input at 0 V (virtual ground) so there are no common mode errors
it can add up to a lot more parts when you try fancy compensation to maximise loop gain
the output amp can also provide lots more loop gain, once the global negative feedback is high enough the remaining error is the common mode nonlinearity at the input - inverting topologies keep the dif pair input at 0 V (virtual ground) so there are no common mode errors
it can add up to a lot more parts when you try fancy compensation to maximise loop gain
I think I'll stick with the simple opamp for this first project.
I will only take care of the PSU, a discrete Zener+BD137 regulator and good filtering caps next to the opamps should do the trick. I assume the circuit shown on the AD797 datasheet is good enough to filter the PSU (100nF + (16µF,1 ohm) tantalum near the IC.
BTW what do you call common mode nonlinearities ? Is it the eventual pollution in the ground rail ? How to avoid it if it is that ?
I will only take care of the PSU, a discrete Zener+BD137 regulator and good filtering caps next to the opamps should do the trick. I assume the circuit shown on the AD797 datasheet is good enough to filter the PSU (100nF + (16µF,1 ohm) tantalum near the IC.
BTW what do you call common mode nonlinearities ? Is it the eventual pollution in the ground rail ? How to avoid it if it is that ?
common mode input voltage is the average of the V at the op amp +,- inputs
the input common mode voltage follows the signal in positive gain op amp circuits - like the Sallen-Key filter
op amps are supposed to amplify the difference between the +,- inputs by a really big number and the output should have Zero dependence on the average voltage at the +,- inputs
Common mode gain is a error term in a real op amp's approximation to the ideal op amp
the +,- inputs also have a parasitic impedance to the supply rails (ac/small signal ground ) that can be nonlinear so at the ppm level you can see some distortion with input common mode voltage that is "outside" of the negative feedback loop and cannot be corrected by the negative feedback
by using inverting circuits the +,- inputs have nearly zero common mode voltage swing and don't show this distortion
really low impedance source and feedback network impedance reduces the common mode error effect
balancing the input impedance (over frequency in filter circuits) helps by causing equal and canceling distortion at the +,- op amp inputs
really low impedance filter network components mean your op amp outputs are drawing more current and heating up causing distortion from internal thermal feedback as the output stage heat diffuses across the op amp silicon die to the input transistors
modern op amps reduce the thermal feedback by careful chip layout but the effects on intermod distortion is still measurable
low feedback network impedance also causes more ground current and can cause power supply coupled distortion - particularly with the output stage operating in Class B at higher V levels into a low impedance feedback network
The AD797 wouldn't be my 1st choice for a general purpose active filter op amp - the low noise spec can only be achieved with <100 Ohm source/filter impedances which are usually considered impractical due to the huge capacitor sizes required
Also the AD797 isn't exactly unity gain stable - source and feedback resistance have to be tailored to avoid problems at low gains with the AD797
the input common mode voltage follows the signal in positive gain op amp circuits - like the Sallen-Key filter
op amps are supposed to amplify the difference between the +,- inputs by a really big number and the output should have Zero dependence on the average voltage at the +,- inputs
Common mode gain is a error term in a real op amp's approximation to the ideal op amp
the +,- inputs also have a parasitic impedance to the supply rails (ac/small signal ground ) that can be nonlinear so at the ppm level you can see some distortion with input common mode voltage that is "outside" of the negative feedback loop and cannot be corrected by the negative feedback
by using inverting circuits the +,- inputs have nearly zero common mode voltage swing and don't show this distortion
really low impedance source and feedback network impedance reduces the common mode error effect
balancing the input impedance (over frequency in filter circuits) helps by causing equal and canceling distortion at the +,- op amp inputs
really low impedance filter network components mean your op amp outputs are drawing more current and heating up causing distortion from internal thermal feedback as the output stage heat diffuses across the op amp silicon die to the input transistors
modern op amps reduce the thermal feedback by careful chip layout but the effects on intermod distortion is still measurable
low feedback network impedance also causes more ground current and can cause power supply coupled distortion - particularly with the output stage operating in Class B at higher V levels into a low impedance feedback network
The AD797 wouldn't be my 1st choice for a general purpose active filter op amp - the low noise spec can only be achieved with <100 Ohm source/filter impedances which are usually considered impractical due to the huge capacitor sizes required
Also the AD797 isn't exactly unity gain stable - source and feedback resistance have to be tailored to avoid problems at low gains with the AD797
So if I understand well, it is better to use MFB topology for the filter part, and inverting configuration for the buffers to avoid common mode distorsion ?
I will use resistors in the 2.4k ~ 16k for the filter sections. I think it is a good compromise between resistor noise and impedance.
At first, I'll use NE5534AN opamps, with the PSU filtered like suggested in the AD797 datasheet. Maybe I'll try tweaks later
I will use resistors in the 2.4k ~ 16k for the filter sections. I think it is a good compromise between resistor noise and impedance.
At first, I'll use NE5534AN opamps, with the PSU filtered like suggested in the AD797 datasheet. Maybe I'll try tweaks later
Choices
Hi,
as well as the choice of OP should fit the application the choice of filter should be taken after Your needs. Order of the filter, precision, Q-value, parts count etc, etc play a role. If You need for e.g. highest symmetry, lowest parts count and ease of tuneability of the High- and Low-pass a LR-Hawksford-topology could be the best. If You need a simple standard circuit Sallen-Key could do the job. If You need independent tuneability of the filter parameters a state-variable-filter might be the right option. As with most things in life, filter design means to optimize a compromise thing to Your needs.
It will be very helpful for You to use a CAD-prog like Circuit maker (student version as free download) or Switcher-CAD or filter designers on the websites of AD, LT, TI.
Oh, btw. the value of the resistors is mostly not the big prob because those values are fixed and from the datasheet of the OP You can estimate what values You can choose. The capacitors are more problematic, because their impedance can vary from open circuit to short circuit. So You have to keep an eye on the biasing of the OP as well as its stability against low amplification factors.
jauu
Calvin
Hi,
as well as the choice of OP should fit the application the choice of filter should be taken after Your needs. Order of the filter, precision, Q-value, parts count etc, etc play a role. If You need for e.g. highest symmetry, lowest parts count and ease of tuneability of the High- and Low-pass a LR-Hawksford-topology could be the best. If You need a simple standard circuit Sallen-Key could do the job. If You need independent tuneability of the filter parameters a state-variable-filter might be the right option. As with most things in life, filter design means to optimize a compromise thing to Your needs.
It will be very helpful for You to use a CAD-prog like Circuit maker (student version as free download) or Switcher-CAD or filter designers on the websites of AD, LT, TI.
Oh, btw. the value of the resistors is mostly not the big prob because those values are fixed and from the datasheet of the OP You can estimate what values You can choose. The capacitors are more problematic, because their impedance can vary from open circuit to short circuit. So You have to keep an eye on the biasing of the OP as well as its stability against low amplification factors.
jauu
Calvin
Well, the filters I'm designing are quite simple :
the first one is a 2d order LR lowpass and highpass (just to send low frequency signal of my speakers to a subwoofer). This one shouldn't give me too many problems, I've already begun to build it.
the second one is for an anticipated project :
- LR 2d order lowpass (to woofer)
- LR 2d order highpass + LR 4th order lowpass (to midranger)
- LR 4th order highpass (tweeter)
I don't plan much tweaking on the components, only changing the resistors values in the filter sections to vary cutoff frequencies. Q will stay the same (0.5 for the 2d order and 0.707 for the 4th order). The capacitor values will be fixed as well.
What exactly is the hawksford topology ? I have found this paper http://www.essex.ac.uk/ese/research...s/C14 A family of circuit topologies LR-4.pdf
Is there anything like this in the schematics shown ?
For my second active filter, should I use one buffer per filter section or one could do the job for the 3 filter sections ?
Otherwise, assuming I'll use NE5534 opamps, is the Sallen-key topology more suitable for my prupose, or MFB due to common-mode distorsion ?
Concerning CAD, I use the AD java application, and also FilterPro from TI. I'll give a try to circuit-maker, it was precisely the kind of soft I were searching for ! Thx for suggesting this to me
the first one is a 2d order LR lowpass and highpass (just to send low frequency signal of my speakers to a subwoofer). This one shouldn't give me too many problems, I've already begun to build it.
the second one is for an anticipated project :
- LR 2d order lowpass (to woofer)
- LR 2d order highpass + LR 4th order lowpass (to midranger)
- LR 4th order highpass (tweeter)
I don't plan much tweaking on the components, only changing the resistors values in the filter sections to vary cutoff frequencies. Q will stay the same (0.5 for the 2d order and 0.707 for the 4th order). The capacitor values will be fixed as well.
What exactly is the hawksford topology ? I have found this paper http://www.essex.ac.uk/ese/research...s/C14 A family of circuit topologies LR-4.pdf
Is there anything like this in the schematics shown ?
For my second active filter, should I use one buffer per filter section or one could do the job for the 3 filter sections ?
Otherwise, assuming I'll use NE5534 opamps, is the Sallen-key topology more suitable for my prupose, or MFB due to common-mode distorsion ?
Concerning CAD, I use the AD java application, and also FilterPro from TI. I'll give a try to circuit-maker, it was precisely the kind of soft I were searching for ! Thx for suggesting this to me
Hawksford
Hi,
guess You meant Q: 0.7 for 2nd and 0.5 for 4th order filters.
What I like in the Hawksford topolgy is that the Highpass signal just passes through one OP regardless of the steepness of the filter.
Another big advantage is that You don´t need costly devices with tight tolerances, because High- and lowpass always work absolutely symmetric.
You need only the smallest number of frequency response forming devices e.g 2(4) caps and 2(4) Rs for a 2nd(4th) order filter.
Each integrator can be tuned independentely.
So kind of a near perfect solution? Well the problem is the response of the drivers themselves. While the filter works very much as in theory, the combination of filter + driver doesn´t need to. If the driver themselves behave good (You might use active biquadratic equalization) this topolgy is a very interesting one.
Since the caps in the integrators feedback loop place the OP in unitygain at higher freqs, You have to design the whole filter with unitygain stable OPs. Use OPs with low Input offsets. Dual- or quad-OPs are fine. I personally like the OPA604/2604, OPA2134/4134 or the newer AD8610/8620.
A 2nd order Hawksford. Integrators U2/C1/R2 and U3/C2/R3 determine filter response
its frequency response and phase
its group delay
A 4th order Hawksford. Integrators U2,3,5,6/C1,2,3,4/R2,3,4,5 determine filter response
its frequency response and phase
its group delay
jauu
Calvin
Hi,
guess You meant Q: 0.7 for 2nd and 0.5 for 4th order filters.
What I like in the Hawksford topolgy is that the Highpass signal just passes through one OP regardless of the steepness of the filter.
Another big advantage is that You don´t need costly devices with tight tolerances, because High- and lowpass always work absolutely symmetric.
You need only the smallest number of frequency response forming devices e.g 2(4) caps and 2(4) Rs for a 2nd(4th) order filter.
Each integrator can be tuned independentely.
So kind of a near perfect solution? Well the problem is the response of the drivers themselves. While the filter works very much as in theory, the combination of filter + driver doesn´t need to. If the driver themselves behave good (You might use active biquadratic equalization) this topolgy is a very interesting one.
Since the caps in the integrators feedback loop place the OP in unitygain at higher freqs, You have to design the whole filter with unitygain stable OPs. Use OPs with low Input offsets. Dual- or quad-OPs are fine. I personally like the OPA604/2604, OPA2134/4134 or the newer AD8610/8620.
A 2nd order Hawksford. Integrators U2/C1/R2 and U3/C2/R3 determine filter response
An externally hosted image should be here but it was not working when we last tested it.
its frequency response and phase
An externally hosted image should be here but it was not working when we last tested it.
its group delay
An externally hosted image should be here but it was not working when we last tested it.
A 4th order Hawksford. Integrators U2,3,5,6/C1,2,3,4/R2,3,4,5 determine filter response
An externally hosted image should be here but it was not working when we last tested it.
its frequency response and phase
An externally hosted image should be here but it was not working when we last tested it.
its group delay
An externally hosted image should be here but it was not working when we last tested it.
jauu
Calvin
Hi,
I think the overall Q for both the 2pole and 4pole Linkwitz riley are 0.5.
The 2pole will be formed from a pair of cascaded single pole Butterworth.
The 4pole will be formed from a pair of cascaded 2pole Butterworth.
Each of the Butterworth will be Q=sq root(1/ 2).
You,
can you explain why you have chosen just 2pole for the bass and mid crossovers?
I think the overall Q for both the 2pole and 4pole Linkwitz riley are 0.5.
The 2pole will be formed from a pair of cascaded single pole Butterworth.
The 4pole will be formed from a pair of cascaded 2pole Butterworth.
Each of the Butterworth will be Q=sq root(1/ 2).
You,
can you explain why you have chosen just 2pole for the bass and mid crossovers?
Wao this Hawcksford topology is rather complicated ! it needs 5 opamps where I only need 2 in a MFB or Sallen-Key filter, I think I won't use that... It could be useful in very steep filters, but maybe not for my purpose.
Thank you very much anyway for those nice graphs !
The overall Q of both filters is 0.5, I was only meaning each 2-pole section in the 4th order filter had a 0.707 Q.
Why only 12dB/oct for the bass/midrange ? well, because it means one opamp less in the signal path, and to have a smoother transition between the woofer and midranger, as far as both drivers are in acoustic phase. But maybe am I going wrong this way ?!
Thank you very much anyway for those nice graphs !
The overall Q of both filters is 0.5, I was only meaning each 2-pole section in the 4th order filter had a 0.707 Q.
Why only 12dB/oct for the bass/midrange ? well, because it means one opamp less in the signal path, and to have a smoother transition between the woofer and midranger, as far as both drivers are in acoustic phase. But maybe am I going wrong this way ?!
The acoustic roll off is what matters, not electrical.
Remember that the mid's roll off is the combined effect of the filter and its response in its subenclosure - you may end up with a fourth order acoustic roll off even though you only have a second order electrical filter. You might need a fourth order LP on the woofer to match the mid roll off with a second order filter.
As long as the woofer doesn't have a breakup that is not suppressed by at least 30 dB and the mid can handle the low end load, a 12dB/octave electrical slope is fine. Many metal cone drivers have breakup modes that are 15 dB or more higher than the midband response. With a second order filter improperly placed you could actually end up with rising response to well above the nominal crossover point.
I try to suppress the woofer breakups by at least 40 dB - so a notch filter is often required. Don't be so afraid of op amps in the signal path. IMHO they are far less a problem than unsuppressed breakups and their associated distortion.
Remember that the mid's roll off is the combined effect of the filter and its response in its subenclosure - you may end up with a fourth order acoustic roll off even though you only have a second order electrical filter. You might need a fourth order LP on the woofer to match the mid roll off with a second order filter.
As long as the woofer doesn't have a breakup that is not suppressed by at least 30 dB and the mid can handle the low end load, a 12dB/octave electrical slope is fine. Many metal cone drivers have breakup modes that are 15 dB or more higher than the midband response. With a second order filter improperly placed you could actually end up with rising response to well above the nominal crossover point.
I try to suppress the woofer breakups by at least 40 dB - so a notch filter is often required. Don't be so afraid of op amps in the signal path. IMHO they are far less a problem than unsuppressed breakups and their associated distortion.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.