Hi y'all
I need to build some linkwitz-riley active filters for my speakers. The problem comes down to this: I can only get hold on ne5532, ne5534 (for the low, band & high pass) at affordable prices in my area (& i do need lots of them). The thing is that they don't do well unity gain so i saw that building a discrete buffer was better and even cheaper. Since i saw some buffer schematics on the net i thought to adapt them to my needs. So here it goes. I found this schematic
http://users.verat.net/~pedjarogic/audio/
and the one that i included (i found it somewhere un this forum, so it's the property of it's maker, not mine).
My intent was to use the npn/pnp superpairs instead of 2SK170BL and replace all FETs with BJTs (BC550c, BC560c, again the only good cheap ones that i can easily find) even lots in parallel to reduce noice (don't think it's needed).
I don't have any drawing program, not to mention electronical schematic drawing on the box from wich i'm posting in order to get my ideas easily out so please be patient with me.
So what do you think ?
I need to build some linkwitz-riley active filters for my speakers. The problem comes down to this: I can only get hold on ne5532, ne5534 (for the low, band & high pass) at affordable prices in my area (& i do need lots of them). The thing is that they don't do well unity gain so i saw that building a discrete buffer was better and even cheaper. Since i saw some buffer schematics on the net i thought to adapt them to my needs. So here it goes. I found this schematic
http://users.verat.net/~pedjarogic/audio/
and the one that i included (i found it somewhere un this forum, so it's the property of it's maker, not mine).
My intent was to use the npn/pnp superpairs instead of 2SK170BL and replace all FETs with BJTs (BC550c, BC560c, again the only good cheap ones that i can easily find) even lots in parallel to reduce noice (don't think it's needed).
I don't have any drawing program, not to mention electronical schematic drawing on the box from wich i'm posting in order to get my ideas easily out so please be patient with me.
So what do you think ?
Attachments
I would say that using an NE553x or building a discrete op-amp, the cost is such a small part of the project / not high value to ignore the difference. NE5532 is stable at unity gain but NE5534 requires a compensation capacitor if the gain is less than 3. Details are in the datasheet.
I would use chips myself as the performance is very good and it will take up much, much less board space than discrete. This would be important to me if I was using a lot of op-amps as you seem to be doing.
If you really want to use discrete, check out Moamps MOX project and the realisation of the design by JensRasmussen. There is also a discrete op-amp on the Pass DIY site.
I would use chips myself as the performance is very good and it will take up much, much less board space than discrete. This would be important to me if I was using a lot of op-amps as you seem to be doing.
If you really want to use discrete, check out Moamps MOX project and the realisation of the design by JensRasmussen. There is also a discrete op-amp on the Pass DIY site.
What about Rout >10K
It might get nasty if i were to use NEs with low value resistors inside the filter chain to reduce parasitic noise. NEs (most opamps) don't really like to see that low impedance at their outputs. With discretes, the sky is the limit (now think of a constant current of about 10 mA going through the BJTs, the output voltage about 2V on around 10K resistance, the output current would be 0.2 mA, even a 10V if you are bad enough would produce a 2 mA output current which compared to the 10 mA class A regime of the buffer it's only 20% variation, not bad i say).
The NE also has a high imput current (might or might not matter in this situation).
Then, what about the maximum frequency response ?
The NE can do 140 KHz, the BC550/560 can do 250 MHz.
And if you have 6 NE in series (in buff.+2x12db/oct lowpass + 2x12db/oct highpass + out buff.) for the mids, then what would the total noise be ? I'm inclined to think that such a stage made of bipolar buffers might be less noisy and with less distortion at high signal levels than one made with NEs.
You are welcomed to comment.
It might get nasty if i were to use NEs with low value resistors inside the filter chain to reduce parasitic noise. NEs (most opamps) don't really like to see that low impedance at their outputs. With discretes, the sky is the limit (now think of a constant current of about 10 mA going through the BJTs, the output voltage about 2V on around 10K resistance, the output current would be 0.2 mA, even a 10V if you are bad enough would produce a 2 mA output current which compared to the 10 mA class A regime of the buffer it's only 20% variation, not bad i say).
The NE also has a high imput current (might or might not matter in this situation).
Then, what about the maximum frequency response ?
The NE can do 140 KHz, the BC550/560 can do 250 MHz.
And if you have 6 NE in series (in buff.+2x12db/oct lowpass + 2x12db/oct highpass + out buff.) for the mids, then what would the total noise be ? I'm inclined to think that such a stage made of bipolar buffers might be less noisy and with less distortion at high signal levels than one made with NEs.
You are welcomed to comment.
If you use low value resistors to keep the noise down you will end up with large capacitors, possibly requiring values above 1 uF or so which are hard/expensive to obtain as quality and tight tolerance parts.
The input current of the NE553x is probably lower than that you will be able to attain with discrete design using bipolar transistors, especially if you start trying to run them at high currents as you seem to want to do.
Just because a BC550/560 has an Ft of 250 MHz does not mean it's usable up to that frequency in isolation and even less so when to make a feedback amplifier such as an op-amp.
You will probably find that an NE553x will be lower noise or as low as anything you can do with BC550/560 transistors.
I think you will only find the noise of any decent op-amp a problem if you have like >20dB gain and are amplifying like <10mV signals.
The NE553x is used in top recording studio consoles with 100s of op-amps in, many of them in the signal path. They amplify microphone level signals and still the overall noise is very low. It will simply not be a problem in a relatively basic active crossover.
The input current of the NE553x is probably lower than that you will be able to attain with discrete design using bipolar transistors, especially if you start trying to run them at high currents as you seem to want to do.
Just because a BC550/560 has an Ft of 250 MHz does not mean it's usable up to that frequency in isolation and even less so when to make a feedback amplifier such as an op-amp.
You will probably find that an NE553x will be lower noise or as low as anything you can do with BC550/560 transistors.
I think you will only find the noise of any decent op-amp a problem if you have like >20dB gain and are amplifying like <10mV signals.
The NE553x is used in top recording studio consoles with 100s of op-amps in, many of them in the signal path. They amplify microphone level signals and still the overall noise is very low. It will simply not be a problem in a relatively basic active crossover.
I guess the crossover filters you have in mind are Sallen-Key with Av=1, allowing use of unity gain buffers.
The best of Pedja's buffers is probably this one, my second choice would be this one especially if your output load is above 10k. I would avoid the npn/pnp pair buffers unless you have a 50MHz oscilloscope to track down stability problems.
I have built a variant of Pedja's "best" buffer and find it to be essentially transparent if not loaded too heavily. My version uses a "fetlington" or cascaded FET source follower/NPN emitter follower instead of just the FET follower at the 2SK170BL position.
Crossovers are line level so you don't need low noise FETs like 2SK170BL, but high transconductance (gm) is desirable for low output impedance. The "fetlington" allows use of lower gm FETs since an NPN has higher transconductance than FETs, an NPN biased at 1mA has a gm of 38mS, or a follower output impedance of 1/gm or 26 ohms.
The best of Pedja's buffers is probably this one, my second choice would be this one especially if your output load is above 10k. I would avoid the npn/pnp pair buffers unless you have a 50MHz oscilloscope to track down stability problems.
I have built a variant of Pedja's "best" buffer and find it to be essentially transparent if not loaded too heavily. My version uses a "fetlington" or cascaded FET source follower/NPN emitter follower instead of just the FET follower at the 2SK170BL position.
Crossovers are line level so you don't need low noise FETs like 2SK170BL, but high transconductance (gm) is desirable for low output impedance. The "fetlington" allows use of lower gm FETs since an NPN has higher transconductance than FETs, an NPN biased at 1mA has a gm of 38mS, or a follower output impedance of 1/gm or 26 ohms.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.