Hi,
According to me they use transformers.
However, semiconductor alternatives have existed for a long time for both inputs and outputs that perform quite credibly.
Ciao T
According to me they use transformers.
However, semiconductor alternatives have existed for a long time for both inputs and outputs that perform quite credibly.
Ciao T
Hi ThorstenL
What does mean "to perform credibly" ? Is there any reason why the simple balanced circuit presented by Self should not "perform credibly" because of its lack of sophistication ? The simplicity of the circuit may be explained by the fact that his project seems to target hi-fi rather than professional installations.
With a little help from a negative resistance ?
They may not be so easily available than the ICs used here.
What does mean "to perform credibly" ? Is there any reason why the simple balanced circuit presented by Self should not "perform credibly" because of its lack of sophistication ? The simplicity of the circuit may be explained by the fact that his project seems to target hi-fi rather than professional installations.
Hi,
Not necessary, but can be useful. Depends on the transofrmer
They are made from Op-Amp's. So they can easily be made using parts that are readily available.
Measure CMRR at 20KHz for a good quality transformer. And then for the alternatives. I object not to the lack of sophistication, true balanced circuits are hardly more complex, I object to the lack of putting to use long known and widely documented designs to use in what is claimed to be state of the art...
I hope you will not perpetuate the demeaning "no need to put in serious design work, it's only HiFi" found among some pro designers?
Of course, just like THD and SNR, so CMRR is merely a number and has no real reliable correlation with sound quality. Designs should be "balanced" (pun intended), yet the benefits of correctly designed balanced I/O's are not small.
If we are fitting balanced in's and out's at all, why not make sure we use state of the art circuitry? Why stick to stuff that most pro designers would not have used in the 1980's? It actually will not increase circuit complexity much, if any.
Ciao T
Not necessary, but can be useful. Depends on the transofrmer
They are made from Op-Amp's. So they can easily be made using parts that are readily available.
Measure CMRR at 20KHz for a good quality transformer. And then for the alternatives. I object not to the lack of sophistication, true balanced circuits are hardly more complex, I object to the lack of putting to use long known and widely documented designs to use in what is claimed to be state of the art...
I hope you will not perpetuate the demeaning "no need to put in serious design work, it's only HiFi" found among some pro designers?
Of course, just like THD and SNR, so CMRR is merely a number and has no real reliable correlation with sound quality. Designs should be "balanced" (pun intended), yet the benefits of correctly designed balanced I/O's are not small.
If we are fitting balanced in's and out's at all, why not make sure we use state of the art circuitry? Why stick to stuff that most pro designers would not have used in the 1980's? It actually will not increase circuit complexity much, if any.
Ciao T
Got my copy of Elektor today and read Doug's article. I'm glad to see material like this in Elektor. As always, I enjoyed reading Doug's article and look forward to the next installment. He does his homework and is not afraid to present some unconventional circuits. His articles always make you think, even if you don't agree with some of the philosophy.
The circuits are clever, but the added complexity is a bit off-putting. I don't think working so hard to get the noise down in a line stage is worth it, but that is just a philosophical difference opinion.
Personally, I prefer to see pot wipers looking into the inputs of JFET op amps when possible - no wiper current, either AC or DC.
When Doug discussed the pot getting warm I did smile and think back to the Lirpa days of Audio Magazine, noting that this is the April issue of Elektor.
I do have to take issue with Doug's description of the noise reduction mechanism at work when op amps are paralleled. There is no "partial cancellation" of uncorrelated noise sources.
Think about what happens when we parallel two transistors to gain a noise advantage. If they are voltage driven and each have the same transconductance, the input signals add on a voltage basis at the output, rising 6dB into a given load resistance. But the uncorrelated noises of the two transistors' currents only adds on a power basis, going up by 3 dB. Signal goes up 6dB, noise goes up 3dB; voila! A 3dB net noise advantage.
Back to the op amps in parallel. The noise from a given op amp goes down by 6dB because of the resistive divider formed by the output current-sharing resistors (remember, to the noise from op amp A, the output impedance of op amp B looks like a short to ground). This effectively happens to each of the noise sources of op amps A and B. Each noise component suffers a 6dB attenuation to the summing junction. But then the two noise voltages at the summing junction add on a power basis and the total noise goes up 3dB, now lying net 3dB lower than the noise that was individually present at the output of each op amp. The signal is not attenuated at all in the summing node because the same signal is at the output of each op amp. No "cancellation" is taking place.
Cheers,
Bob
The circuits are clever, but the added complexity is a bit off-putting. I don't think working so hard to get the noise down in a line stage is worth it, but that is just a philosophical difference opinion.
Personally, I prefer to see pot wipers looking into the inputs of JFET op amps when possible - no wiper current, either AC or DC.
When Doug discussed the pot getting warm I did smile and think back to the Lirpa days of Audio Magazine, noting that this is the April issue of Elektor.
I do have to take issue with Doug's description of the noise reduction mechanism at work when op amps are paralleled. There is no "partial cancellation" of uncorrelated noise sources.
Think about what happens when we parallel two transistors to gain a noise advantage. If they are voltage driven and each have the same transconductance, the input signals add on a voltage basis at the output, rising 6dB into a given load resistance. But the uncorrelated noises of the two transistors' currents only adds on a power basis, going up by 3 dB. Signal goes up 6dB, noise goes up 3dB; voila! A 3dB net noise advantage.
Back to the op amps in parallel. The noise from a given op amp goes down by 6dB because of the resistive divider formed by the output current-sharing resistors (remember, to the noise from op amp A, the output impedance of op amp B looks like a short to ground). This effectively happens to each of the noise sources of op amps A and B. Each noise component suffers a 6dB attenuation to the summing junction. But then the two noise voltages at the summing junction add on a power basis and the total noise goes up 3dB, now lying net 3dB lower than the noise that was individually present at the output of each op amp. The signal is not attenuated at all in the summing node because the same signal is at the output of each op amp. No "cancellation" is taking place.
Cheers,
Bob
This isn't just about opinions, philosophical or other, its about sub-optimal design. Any time one metric in a design solution is increased beyond what's required then at least one other is bound to suffer for it. For a treatment of design (in general, not just preamp design) as an optimisation process I recommend Brook's book 'The Design of Design'
Amazon.com: The Design of Design: Essays from a Computer Scientist (9780201362985): Frederick P. Brooks: Books
I'm with Doug on this - the way that uncorrelated noise sources add by power and not amplitude is indeed because there's partial cancellation. Meaning that at times the two uncorrelated noise sources are out of phase, at other times they're in phase. In the (50% of the time) they're out of phase there's cancellation because one's positive and the other's negative. Which is why noise adds only on a power basis, not an amplitude basis. The 3dB difference is the result of cancellation or, if it were two light sources adding, we'd say 'interference'.
Well there's a lot of commentary here about extreme designs.
I am afraid that's exactly what the high end is all about, be it cars, planes or amplifiers. If we were all talking about mid fi then the watchword would be 'balance'.
As to the noise of parallel opamps: just measure it with a sound card or an AP with a bunch in parallel and then a single device. Case closed.
I am afraid that's exactly what the high end is all about, be it cars, planes or amplifiers. If we were all talking about mid fi then the watchword would be 'balance'.
As to the noise of parallel opamps: just measure it with a sound card or an AP with a bunch in parallel and then a single device. Case closed.
Yep - except I'm not going to agree that the existing state of high end in audio is a pity, rather its a market that's ripe for disruptive innovation. I don't agree that mid-fi is all about 'balance' either (I note you're not explicitly saying that it is) - its overly purchase price-focussed (and to a lesser extent fashion-influenced) in my estimation. Optimising on a purchase price metric means longer term issues like serviceabilty (which high end generally does better than mid-fi) are sacrificed.
The light analogy is not appropriate. I don't see why all these explainations are needed, the same logic applies to the expected value of the sum of two random variables. It's just math.
BTW I forgot "Stacked Amplifiers Lower Noise" EDN October 1982, I got an extra check for winning best design idea. Stacked DAC's work too.
Attachments
Last edited:
AD524/624 since they lend themselves to adding the outputs with no additional components. Unfortunately EDN does not archive all the old stuff.
Last edited:
On reflection about the Baxandall volume control in Doug's design I wonder if there might be a concern with it, especially in the low-impedance design context. First of all, I think the Baxandall volume control is a wonderfully clever design. It does a lot for you in efficient ways, and the fact that it is able to get an approximation to a log characteristic with a linear pot means that you can take advantage of the usually-better tracking of linear pots.
The way that it obtains high attenuation is at the root of my concern. In the high-attenuation state, the wiper, which is connected to the inverting op amp input, is close to, or at, the end connected to the output of the op amp. In the extreme, at full attenuation and with a 1k pot, you have the input signal running throught the 1k of the pot into the output of the op amp, which is a virtual ground by virtue of the feedback. Ideally, therefore, you get no output.
At the same time, the op amp output stage is called upon to deliver substantial amounts of current to its output node to sink the input signal current, which can be on the order of several mA. Indeed, this current is on the same order of magnitude that would be required to deliver the input signal voltage into a 1k load. One will thus get the current-dependent output stage distortion from the op amp (e.g., crossover distortion), but without any signal voltage at the output.
This means that the THD% due to this distortion source will go up as the attenuation goes up. When the attenuation is at 20dB, the effective multiplication of the distortion is 10X. Higher % distortion at a lower level is usually not a good thing sonically. At VERY low volume, you hear nothing but the output stage crossover distortion.
Although I am not sure about the magnitude of the effect, one also needs to consider the small-signal finite impedance of the virtual ground at the output of the op amp. With a 1k pot and the control set to 40dB attenuation, that virtual ground impedance wants to be much lower than 10 ohms. Moreover, will that virtual ground impedance rise much with frequency at the upper end of the audio band due to the falling loop gain of the op amp as frequency rises? What will be the resulting frequency response of the volume control due to this effect when set at -40dB?
Just some food for thought.
Cheers,
Bob
The way that it obtains high attenuation is at the root of my concern. In the high-attenuation state, the wiper, which is connected to the inverting op amp input, is close to, or at, the end connected to the output of the op amp. In the extreme, at full attenuation and with a 1k pot, you have the input signal running throught the 1k of the pot into the output of the op amp, which is a virtual ground by virtue of the feedback. Ideally, therefore, you get no output.
At the same time, the op amp output stage is called upon to deliver substantial amounts of current to its output node to sink the input signal current, which can be on the order of several mA. Indeed, this current is on the same order of magnitude that would be required to deliver the input signal voltage into a 1k load. One will thus get the current-dependent output stage distortion from the op amp (e.g., crossover distortion), but without any signal voltage at the output.
This means that the THD% due to this distortion source will go up as the attenuation goes up. When the attenuation is at 20dB, the effective multiplication of the distortion is 10X. Higher % distortion at a lower level is usually not a good thing sonically. At VERY low volume, you hear nothing but the output stage crossover distortion.
Although I am not sure about the magnitude of the effect, one also needs to consider the small-signal finite impedance of the virtual ground at the output of the op amp. With a 1k pot and the control set to 40dB attenuation, that virtual ground impedance wants to be much lower than 10 ohms. Moreover, will that virtual ground impedance rise much with frequency at the upper end of the audio band due to the falling loop gain of the op amp as frequency rises? What will be the resulting frequency response of the volume control due to this effect when set at -40dB?
Just some food for thought.
Cheers,
Bob
Hi,
The same note applies to the Volume control used for the 2012 Preamp, BTB.
Incidentally, while looking for material for the High End Tone Control thread I came across the Urei 537 Schematic dated June 1978... It showed buffer Op-Amp's for the pots of the tone controls... The tone control is not per se Baxandall but operates on a similar principle.
http://www.waltzingbear.com/Schematics/Urei/537.JPG
Ciao T
The same note applies to the Volume control used for the 2012 Preamp, BTB.
Incidentally, while looking for material for the High End Tone Control thread I came across the Urei 537 Schematic dated June 1978... It showed buffer Op-Amp's for the pots of the tone controls... The tone control is not per se Baxandall but operates on a similar principle.
http://www.waltzingbear.com/Schematics/Urei/537.JPG
Ciao T
Has anyone put together a BOM for Farnell? I've found most of the components at Mouser.com, but the Vishay pots seem to be only available at Farnel, unless there are others that can be substituted. Plus the Vishay are fairly expensive as pots go. If I could order everything from Farnel it may be more expensive as I'm in the US, but getting everything under one roof does appeal to me. I hate trying to find parts from multiple vendors. 
BTW, I will be using the Elektor PCB. Thanks in advance.
BTW, I will be using the Elektor PCB. Thanks in advance.
I would not use 1k pots in a Baxandall. 10k seems to me to be the best compromise. However, in he quest for ever lower noise . . .
Cap selection gets more difficult as well with low value pots.
As for opamp loading when driving the feedback network te solution is simple: just buffer it. 3 transistors and three resistors gets you doesn't to c 5ppm 10Volts into 600 ohms at 20KHz.
Cap selection gets more difficult as well with low value pots.
As for opamp loading when driving the feedback network te solution is simple: just buffer it. 3 transistors and three resistors gets you doesn't to c 5ppm 10Volts into 600 ohms at 20KHz.
The Baxandall tone control is used widely everywhere because of its simplicity. Lots of similar looking circuits have same kind of approach just difference in component values and resultant impedance network differs. There is another way of doing a tone control which IMHO is better than Baxandall in many aspects such as slope order, low noise and zero interference at flat position.
Dennis Bohan of RANE corp. designed and patented one such tone control. It features the slopes of higher order ranging from 12dB/oct to above which are not possible in Baxandall which is limited to 6dB/oct unfortunately. Higher order slopes lets you to vary bass and treble without altering midband frequencies.Its known as "Accelerated Slope Tone control"
One can read the white paper on the below mentioned link.
www.rane.com/pdf/acceler.pdf
Dennis Bohan of RANE corp. designed and patented one such tone control. It features the slopes of higher order ranging from 12dB/oct to above which are not possible in Baxandall which is limited to 6dB/oct unfortunately. Higher order slopes lets you to vary bass and treble without altering midband frequencies.Its known as "Accelerated Slope Tone control"
One can read the white paper on the below mentioned link.
www.rane.com/pdf/acceler.pdf
Last edited:
Very thought provoking is all that when you start to think about it.
It would be interesting to take a similar volume stage in isolation, perhaps with fixed resistors in place of the pot to give some value of high(ish) attenuation and then follow that with a gain stage to bring the overall level of the two stages back to unity.
Then do an A-B test of some kind by being able to feed either the Baxandall stage or direct feed into a reference set up for evaluation.
Could any effect be measured for real I wonder with suitable test gear ?
Crossover distortion?
I guess this is worth a look:
How about that thing? I couldn't find it for NE5532. In the schematic is actually NE5534 also using a slight gain, but it is biased towards class A, tested "known good" and it looks both accessible and readily doable.
P.S.
There are other op-amps, such as JRC4560L and many more, very pleasant run parallel, and so much less straining noise when driving that 1k load. But I thought that NE5532 Parallel could drive a 1k load okay? Doug's preamplifier seems to belong with a large scale, very low gain, power amplifier so that noise output occurs when output doesn't.
Edit: I guess that my question for you is, should we swap or should we bias the op-amp?
Crossover distortion?
I guess this is worth a look:
How about that thing? I couldn't find it for NE5532. In the schematic is actually NE5534 also using a slight gain, but it is biased towards class A, tested "known good" and it looks both accessible and readily doable.
P.S.
There are other op-amps, such as JRC4560L and many more, very pleasant run parallel, and so much less straining noise when driving that 1k load. But I thought that NE5532 Parallel could drive a 1k load okay? Doug's preamplifier seems to belong with a large scale, very low gain, power amplifier so that noise output occurs when output doesn't.
Edit: I guess that my question for you is, should we swap or should we bias the op-amp?
All op amps make crossover distortion and other output-current-dependent distortions - its just a matter of degree. So the concern about the Baxandall volume control would apply in all low-z configurations. Also, remember, most preamp output stages are not called upon to drive a low impedance, because most power amplifiers have a decent impedance for an input greater than 10k. In the design being discussed here, however, the op amp in the Baxandall tone control is ALWAYS being required to drive a relatively low output impedance.
Cheers,
Bob