Hi. I'm designing rev2 of a circuit that will be used between a minidsp 2x4 HD and an Icepower 200AS2. In the first revision i had a volume pot bewteen the last opamp and the amplifier, but this introduced an increasing amount of noise with attenuation. This time i want to feed the amplifier directly (or via a 100r) from the last opamp to prevent this noise. I still need to attenuate the signal, mainly to lower the noise generated by the minidsp. However, i want as few opamps in the signal path as possible, so i wondered if there are any reasons why i shouldn't add an L-pad between the buffer stage and the crossover-section, as drawn below. I've checked that both slope and phase is unaffected as long as i keep the values of the l-pad below 1k. I will be using NE5532.
Hi,
I don't get your configuration: you use a minidsp and then some analog xover? Where is located your amp in your schemo? As i understand it you drive the driver with a 5534 through a transformer?
Why use a minidsp so? An hybrid implementation ( low/mid xover digital then passive for mid/high)?
Anyway, using an analog attenuator ( a pad) in front of power amp (and after dsp's dac) is what i use. It doesn't lower noise floor in my case ( in reality i think it worsen it a bit) but it brings things that i was in need of (but i doubt this will be of any interest for a domestic application) and a way to control overall system level.
I implemented it using the calculator from this project as inspiration:
https://www.amb.org/audio/delta1/
One thing with this kind of passive attenuator: use a buffer as close as possible to the pad or you'll have noise issue ( these are relatively high impedance attenuator so don't expect to be able to drive anything with it. )
I don't get your configuration: you use a minidsp and then some analog xover? Where is located your amp in your schemo? As i understand it you drive the driver with a 5534 through a transformer?
Why use a minidsp so? An hybrid implementation ( low/mid xover digital then passive for mid/high)?
Anyway, using an analog attenuator ( a pad) in front of power amp (and after dsp's dac) is what i use. It doesn't lower noise floor in my case ( in reality i think it worsen it a bit) but it brings things that i was in need of (but i doubt this will be of any interest for a domestic application) and a way to control overall system level.
I implemented it using the calculator from this project as inspiration:
https://www.amb.org/audio/delta1/
One thing with this kind of passive attenuator: use a buffer as close as possible to the pad or you'll have noise issue ( these are relatively high impedance attenuator so don't expect to be able to drive anything with it. )
Last edited:
Sorry, i should have been more specific. Hybrid is correct, i'm using a passive xo low/mid, and active mid/high. The transformer in the circuit is just a crude way to simulate the amplifier in VituixCAD.
This attenuation has been a great source of frustration. Maybe there is no way around using another opamp for the attenuation only, ideally i should have a pot in there as well to make things more flexible.
This attenuation has been a great source of frustration. Maybe there is no way around using another opamp for the attenuation only, ideally i should have a pot in there as well to make things more flexible.
Ok got it now.
I would not place your pad in front of your filtering cell ( second opamp and parts around) because there is chance the attenuator Z disturb your filtering cell, rather i would put it between your dac output ( minidsp) and your xover input buffer ( first opamp). It fullfill your wish not to induce other active stage. 😉
Your minidsp analog output stage should be enough to drive it and buffering would take place too. One stone, 2 birds.
The r2r switched attenuator i linked replace a pot and give up to 128 ( accurate and trackable between chanels and constant impedance input so don't change frequency response with steps) steps . But i didn't bothered with that: i made a mechanically switched one ( using lorlin 3x4) with 3 position of attenuation: one for full spl out ( for high dynamic source material and critical listening) so no attenuation and 2 other position corresponding to 'regular level' ( when family is present something in the -12db range) and late night ( not to disturb in the -18db range).
I do fine level setup in digital ( up to 12db att, i've got digital level control at input of my dsp/ my 'source' is digital so no analog between computer and dsp). There should be a way to make same thing with a minidsp i think.
A passive Low/mid and active mid/high. You mixed up both no? If not it is unusual approach. Could you explain why you made this choice ( if not mixed up).
I would not place your pad in front of your filtering cell ( second opamp and parts around) because there is chance the attenuator Z disturb your filtering cell, rather i would put it between your dac output ( minidsp) and your xover input buffer ( first opamp). It fullfill your wish not to induce other active stage. 😉
Your minidsp analog output stage should be enough to drive it and buffering would take place too. One stone, 2 birds.
The r2r switched attenuator i linked replace a pot and give up to 128 ( accurate and trackable between chanels and constant impedance input so don't change frequency response with steps) steps . But i didn't bothered with that: i made a mechanically switched one ( using lorlin 3x4) with 3 position of attenuation: one for full spl out ( for high dynamic source material and critical listening) so no attenuation and 2 other position corresponding to 'regular level' ( when family is present something in the -12db range) and late night ( not to disturb in the -18db range).
I do fine level setup in digital ( up to 12db att, i've got digital level control at input of my dsp/ my 'source' is digital so no analog between computer and dsp). There should be a way to make same thing with a minidsp i think.
A passive Low/mid and active mid/high. You mixed up both no? If not it is unusual approach. Could you explain why you made this choice ( if not mixed up).
Last edited:
Ok makes more sense to me.
I think you got it too from your circuit pov. Yes a bridging input is what you target ( input Z way higher than out Z). 47khz is not a bad target value ime.
Have you calculated/simulated your buffer input impedance?
Depending on how you implement your circuit it could be interesting to use your buffer inverting to have a virtual ground input ( input on neg rathef than +).
I think you got it too from your circuit pov. Yes a bridging input is what you target ( input Z way higher than out Z). 47khz is not a bad target value ime.
Have you calculated/simulated your buffer input impedance?
Depending on how you implement your circuit it could be interesting to use your buffer inverting to have a virtual ground input ( input on neg rathef than +).
I haven't calculated it, as i assume the ne5532's input is high enough to not affects things that much. Any specific advantages with an inverting circuit? Could you provide an example?
An inverting configuration will have a noise penalty. But perhaps you're not that concerned about noise or you'd not be using 11k in your S-K low-pass. Incidentally you can save a resistor by making your L-pad out of 2 * 22k then you can dispense with the first 11k in the S-K section.
Noise is one of my main concerns. The circuit is taken from this site: https://sound-au.com/project09.htm
Will these values cause extra noise?
The L-pad starts to affect frequency response if i use higher value resistors. When R1 is 1k, the response is not affected.
Will these values cause extra noise?
The L-pad starts to affect frequency response if i use higher value resistors. When R1 is 1k, the response is not affected.
Yes opamp input Z will be high impedance but it make sense to know the value to implement your attenuator.
http://www.uneeda-audio.com/pads/
Abraxalito, noise penalty by which phenomenom? The virtual ground could be contaminated or something else?
http://www.uneeda-audio.com/pads/
Abraxalito, noise penalty by which phenomenom? The virtual ground could be contaminated or something else?
The noise of a 5532 is roughly equivalent to that generated by a 1k resistor. An 11k resistor generates SQRT(11) times more noise. If you want to do justice to your 5532's low noise qualities it would be better to scale your 11k resistors down to 1k1 and your caps up in value by the same factor (10).
The (thevenin equivalent) resistance of the L-pad adds to the first 11k resistor but my suggestion was ditching that entirely (replace with a wire). Then your low-pass section will behave as if driven by a zero ohm source.
The (thevenin equivalent) resistance of the L-pad adds to the first 11k resistor but my suggestion was ditching that entirely (replace with a wire). Then your low-pass section will behave as if driven by a zero ohm source.
If we wish to maintain the 100k input impedance and use an inverting configuration then the 100k resistor has to go between the source and the -ve input (virtual GND). Thus the noise contribution of 100k is added to the source. Of course we could go lower for the input impedance but loading of the source becomes an issue. I agree the inverting configuration is nice in that it could incorporate the 6dB attenuation but the noise penalty of a 100k series R seems too severe in this application.
Thank you very much for explaining this! So i should always strive to keep series resistans low in order to minimize noise. Back to the drawing board.
Where would you place an attenuation circuit here? Before, between or after input and output buffers?
Where would you place an attenuation circuit here? Before, between or after input and output buffers?
If I was doing this I'd put the attenuator at the output but I'd use an autoformer rather than an L-pad to keep the noise under control and have low enough impedance to get decent drive of any cable. You're showing a trafo in your schematic, that would do a dandy job of attenuation if you can adjust the turns ratio.
Incidentally while you lower the impedance to minimize circuit noise, power supply noise increases due to more current being drawn by the opamps. So then you need pay more attention to maintaining low PSU impedance and having a decoupling topology that doesn't inject that PSU noise into your GND.
Incidentally while you lower the impedance to minimize circuit noise, power supply noise increases due to more current being drawn by the opamps. So then you need pay more attention to maintaining low PSU impedance and having a decoupling topology that doesn't inject that PSU noise into your GND.
The transformer symbol is used to represent amplifier ( class d) in Skogs's schematic ( not the usual triangle symbol - same as opamp- we usually see, it's a convention to use this Skogs 😉 - and why i was lost at first ).
Of all the headaches this hobby can provide, simple line level attenuation has been the absolute worst for me😛 I just cannot find an elegant solution. I'll probably add another opamp dedicated to attenuation.
I appreciate the help!
I appreciate the help!
Well i don't see this as headhache rather options with pro and cons. To make the choice between option can indeed introduce headache thoug 🙂
I know two things to lower level ot pain: simulation and prototyping.
Both will help to make educated choice.
In other word try to simulate the different idea, then see the one which satisfy your mind or give best results and prototype it. Or prototype and measure each idea. Only things to be gained that way imho.
I forgot a third anti headache: asking things here. But then you have to make choice again as there is many ways to skin an attenuator ...which induce mild headache 😀
I know two things to lower level ot pain: simulation and prototyping.
Both will help to make educated choice.
In other word try to simulate the different idea, then see the one which satisfy your mind or give best results and prototype it. Or prototype and measure each idea. Only things to be gained that way imho.
I forgot a third anti headache: asking things here. But then you have to make choice again as there is many ways to skin an attenuator ...which induce mild headache 😀
I agree its a royal pain. But if you get a winding machine, wire and some ferrite cores then it won't much bother you again as you can then create your own autoformers for any ratio you need.
Here's what my autoformers look like, attenuating the output of my latest DAC to improve the SNR. They're at the back with the yellow tape wrapped around, RM12 ferrite cores.
Here's what my autoformers look like, attenuating the output of my latest DAC to improve the SNR. They're at the back with the yellow tape wrapped around, RM12 ferrite cores.
Last edited: