My thoughts on that are that 300mv output is still huge. That would give around 4.5 watts RMS when fed into a (typical) power amp and 8 ohm.
What about at 20 or 30 mv output which might realistically equate to around 50mw into 8 ohm when fed into a typical power amp. Also at even lower levels still. 20vrms in and say 1 or 2 mv output is a totally valid condition too.
I don't know the answers to all that as I haven't the means to make such measurements but it would be interesting.
The class ab output stage of an opamp generates its distortion when supplying current to a load.
If the level of signal voltage (opamp output) is very small, the opamps "distortion" is still at the same level as it was for a much higher voltage output. The opamp supplies the same current into the feedback network regardless of the feedback resistor value. As the feedback resistor value goes down the voltage goes down keeping the current the same.
Noise (another analogy, think wanted signal vs noise (opamp distortion))... its like saying that 1 mv of noise riding on 100 volts of signal isn't very much. Attenuate the signal with a normal pot and the noise and signal go down together. Attenuate with a set up that keeps the noise (read distortion) at 1mv and only reduce the signal and then as signal goes down the "noise" (distortion) becomes more dominant.
What do others think to that ?
Because the conventional resistive divider type of volume control attenuates signal and noise and distortion and everything else progresively down to zero.
The active gain stage does not achieve the same overall result.
Mooly
Because the conventional resistive divider type of volume control attenuates signal and noise and distortion and everything else progresively down to zero.
The active gain stage does not achieve the same overall result.
In digital format, percentages of distorsion increase with lowering levels.
Hower absolute values of distorsion do not, just like the Baxandall active gain control, I think.
Because the conventional resistive divider type of volume control attenuates signal and noise and distortion and everything else progresively down to zero.
The active gain stage does not achieve the same overall result.
In digital format, percentages of distorsion increase with lowering levels.
Hower absolute values of distorsion do not, just like the Baxandall active gain control, I think.
Correct, the noise and gain of the first stage determines the noise factor of the entire system. If the first stage gain is zero and the stage is purely resistive the noise can be approximated by KTBR, (K= Boltzmann Constant; T = Temperature; B- Bandwidth; R is resistance. This is common knowledge to the communication/RF engineers.
I believe I understand the proposed mechanism, but have not Mr Self's measurements proven that even at high levels of attenuation the distortion is buried in the noise? Is anyone suggesting this phenomenom would not show up under these test conditions? Or that distortion at a level below the noise floor could be audible?
I think there is something different about the way this configuration sounds compared with a conventional passive volume control. But seems as though it must be something else. I have no idea what though
I think there is something different about the way this configuration sounds compared with a conventional passive volume control. But seems as though it must be something else. I have no idea what though
I see the stark fist of removal has been applied, I can't believe it was not obvious from the start.
how is distortion attenuated with a passive control? its level remains the same, as a percentage of the signal being attenuated does it not? that seems the only useful way to represent distortion to me, as a percentage of the signal, not vs fullscale. noise? perhaps, noise in the signal anyway, but the noise created by the resistors, lower CMRR by resistor mismatch in balanced systems, non-linearity from contacts etc etc, i'll take a constant vanishingly low noise and perfectly matched levels, with cheap and handy means to control it perfectly over these any day.
besides, modern digital attenuation as part of software, DSP or the dacs themselves, is operating at 40 and 64bits float, I havent studied how this relates to the interpolation/dither of the signal though, anyone got any info or links on that?
besides, modern digital attenuation as part of software, DSP or the dacs themselves, is operating at 40 and 64bits float, I havent studied how this relates to the interpolation/dither of the signal though, anyone got any info or links on that?
Last edited:
I did a quick LT spice simulation on the active volume control. I used a discrete class A buffer in the op-amps feedback loop (thats just my thing), and not the straight op-amp approach used by Douglas Self in his Elektor design.
Indeed, as expected with the 1k volume control pot, the current into the summing junction from the driving source is high at high gain settings (I set Rin to 0.1 Ohms and Rfeedback to 999.9 ohms). However, at normal output levels of 1V, its around 1mA and a 5534 or 4562 will easily drive that type of load to single digit ppm levels. At 10V out of course you are asking for 10mA but you are never going to see that in a practical situation. On the discrete class A buffer version, I get 14ppm at 20KHz at 10V out. At 1V out I get 2ppm. Of course, the driving source needs to be up to the job, but if these are decent opamps then no issue (and at single digit ppm distortion levels, class AB distortion can hardly be an issue). You could of course raise the value of the pot from 1 k to 5k or 10k, but the result would be increased noise.
Final Analysis:- It does what it says on the can - very neat circuit technique.
Indeed, as expected with the 1k volume control pot, the current into the summing junction from the driving source is high at high gain settings (I set Rin to 0.1 Ohms and Rfeedback to 999.9 ohms). However, at normal output levels of 1V, its around 1mA and a 5534 or 4562 will easily drive that type of load to single digit ppm levels. At 10V out of course you are asking for 10mA but you are never going to see that in a practical situation. On the discrete class A buffer version, I get 14ppm at 20KHz at 10V out. At 1V out I get 2ppm. Of course, the driving source needs to be up to the job, but if these are decent opamps then no issue (and at single digit ppm distortion levels, class AB distortion can hardly be an issue). You could of course raise the value of the pot from 1 k to 5k or 10k, but the result would be increased noise.
Final Analysis:- It does what it says on the can - very neat circuit technique.
Last edited:
Your simulation at high gain has shown that the 2012 preamp with 4 x paralleled 5532 halves driving the 1k pot should have no problems at normal signal levels. We know this already though from the published measurements. 1 x 5532 half driving a 10k pot as in the '96 preamp I built should have even less trouble, albeit with higher noise.
The previous discussion was over what happens at high attenuation or low gain settings. But Mr Self has kindly taken some measurements that show that there really shouldn't be an audible problem at low gain either.
I'm interested in why it seems to sound different though, and I wonder if it has anything to do with enclosing the unity gain buffer (buffering the pot wiper) in the feedback loop that sets the gain. Could this make the opamp more prone to instability that doesn't easily show up? If so it could help explain why I was hearing differences between brands of 5532? Does anyone find this plausible? Interesting also that you're suggesting the use of a buffer at the output as well - this would make 3 opamps enclosed in the loop...
The previous discussion was over what happens at high attenuation or low gain settings. But Mr Self has kindly taken some measurements that show that there really shouldn't be an audible problem at low gain either.
I'm interested in why it seems to sound different though, and I wonder if it has anything to do with enclosing the unity gain buffer (buffering the pot wiper) in the feedback loop that sets the gain. Could this make the opamp more prone to instability that doesn't easily show up? If so it could help explain why I was hearing differences between brands of 5532? Does anyone find this plausible? Interesting also that you're suggesting the use of a buffer at the output as well - this would make 3 opamps enclosed in the loop...
The op amp driving the pot has a 20 ohm resistor in series with it - as does the tone control output op-amp. 20 Ohms should be enough to isolate the load and prevent any oscillation with 5532/34/LM4562 devices.
Since the signal level into the active gain stage is not going to be greater than a few hundred mV, the op-amps can easily provide enough drive. The assumption I am making here is that the pre-amp output at 1V will drive any AMP to full power. If of course, you are looking at the pre-amp performance at full output swing, then you need to be able to drive up to 10mA - quite high, but the above op-amps can do it with very low distortion.
BTW, did you do your listening tests with headphones or speakers?
Since the signal level into the active gain stage is not going to be greater than a few hundred mV, the op-amps can easily provide enough drive. The assumption I am making here is that the pre-amp output at 1V will drive any AMP to full power. If of course, you are looking at the pre-amp performance at full output swing, then you need to be able to drive up to 10mA - quite high, but the above op-amps can do it with very low distortion.
BTW, did you do your listening tests with headphones or speakers?
" . . . and I wonder if it has anything to do with enclosing the unity gain buffer (buffering the pot wiper) in the feedback loop that sets the gain. Could this make the opamp more prone to instability that doesn't easily show up?"
Each of the output op-amps in the active gain stage is compensated (C10-C13) and the input opamp are running at unity gain, so there should not be any oscillation. Do you have access to a scope?
Each of the output op-amps in the active gain stage is compensated (C10-C13) and the input opamp are running at unity gain, so there should not be any oscillation. Do you have access to a scope?
I did my listening with both 'phones and speakers - the board supplied by the Signal Transfer Co. includes a headphone amp based on the high performance heaphone amp in Self's Small Signal Audio book. I noticed the "character" using both types of transducer.
I have a 'scope and did check it with sine and square wave signals after completing the build - there was no sign of instability with the Fairchild 5532s I used initially. But this is not a comprehensive check and no guarantee that it is always stable. I guess I would need access to test gear that can measure THD at the expected levels to be sure.
Isn't the unity gain configuration with 100% feedback likely to be the least stable configuration? And then there is another loop around it.
I recall reading C. Bateman's capacitor distortion series and in the part where he discusses buffer amplifiers for driving large signals into caps, he tried a few unusual configurations with opamps and found that some devices weren't stable with particular configurations that on paper should have been ok. I suppose this is what led me to thinking along these lines.
I have a 'scope and did check it with sine and square wave signals after completing the build - there was no sign of instability with the Fairchild 5532s I used initially. But this is not a comprehensive check and no guarantee that it is always stable. I guess I would need access to test gear that can measure THD at the expected levels to be sure.
Isn't the unity gain configuration with 100% feedback likely to be the least stable configuration? And then there is another loop around it.
I recall reading C. Bateman's capacitor distortion series and in the part where he discusses buffer amplifiers for driving large signals into caps, he tried a few unusual configurations with opamps and found that some devices weren't stable with particular configurations that on paper should have been ok. I suppose this is what led me to thinking along these lines.
The 5532's are very definitely unity gain stable, and all of the devices in the signal chain have some series resistance where it matters so this will isolate any load capacitance. I would be very surprised if you had oscillation - but you should check inside the loop as well, not just on the output.
I know it sounds crazy, but you may want to put some damper resistance in series with your headphones - if you drive good phones from a low Z source, they can sometimes sound a bit brutal (i.e. 'surgical'). 15-30 ohms on 32 ohm phones helps them loosen up a bit soundwise.
Re differences in sound, if you look over on the e-Amp thread, you will see I did a comparison between my older Ovation 250 amp and the new e-Amp. They sound markedly different, and its not just me - the guy I listened with also remarked on it. The older amp has a darker, more rounded and laid back presentation, while the newer one is more forward. Soundstaging on both is very good. So, I am not surprised that you are hearing differences between the two pre-amp designs. As to what it is, at this stage we can only speculate. Give your ears time to tune in and you may prefer the sound of the latest design more - thats audio for you!
I know it sounds crazy, but you may want to put some damper resistance in series with your headphones - if you drive good phones from a low Z source, they can sometimes sound a bit brutal (i.e. 'surgical'). 15-30 ohms on 32 ohm phones helps them loosen up a bit soundwise.
Re differences in sound, if you look over on the e-Amp thread, you will see I did a comparison between my older Ovation 250 amp and the new e-Amp. They sound markedly different, and its not just me - the guy I listened with also remarked on it. The older amp has a darker, more rounded and laid back presentation, while the newer one is more forward. Soundstaging on both is very good. So, I am not surprised that you are hearing differences between the two pre-amp designs. As to what it is, at this stage we can only speculate. Give your ears time to tune in and you may prefer the sound of the latest design more - thats audio for you!
That does indeed sound crazy. Is this about the headphones or the stability of the driving amps? Are headphones really designed for such high source impedances? From reading NwAvGuy (aka RocketScientist) I'd say that the answer is a 'no' (he has a thing about low Z drive) but then he's not an infallible authority
The heaphone amp already has a 30R in series with the output, and while different series R may alter the sound from headphones due to damping effects this is missing the point really. Beside the headphone amp is a seperate amp.
The character I'm hearing is also there when using my power amp and speakers, and in my case that's far more revealing as I only have cheap Senheiser 'phones but my speakers (B&W801) are very neutral and revealing of amplifier differences (or to put it another way: absolute b#stards to drive Not that this should matter to the preamp...)
The 5532 should be unity gain stable by itself given proper decoupling etc, but what I'm speculating is whether this is still guaranteed when you stick it in the feedback loop of another amplifier. In theory I guess it should be, but perhaps there's some sort of interaction going on that doesn't show up under normal test conditions and yet gives a certain "sound". It is driving a fairly heavy load of 680R also (but now 1k in the 2012 preamp) due to the virtual gnd at the -ve opamp input, which could be having an effect.
An alternative speculation might be that because the ouput opamp in the Baxandall gain control is already somewhat taxed just driving the pot, it has a harder time driving the power amp as well. But you would expect this to show up in the THD measurements (I guess we can assume they were made with a "standard" load on the output).
The character I'm hearing is also there when using my power amp and speakers, and in my case that's far more revealing as I only have cheap Senheiser 'phones but my speakers (B&W801) are very neutral and revealing of amplifier differences (or to put it another way: absolute b#stards to drive
Not that this should matter to the preamp...)The 5532 should be unity gain stable by itself given proper decoupling etc, but what I'm speculating is whether this is still guaranteed when you stick it in the feedback loop of another amplifier. In theory I guess it should be, but perhaps there's some sort of interaction going on that doesn't show up under normal test conditions and yet gives a certain "sound". It is driving a fairly heavy load of 680R also (but now 1k in the 2012 preamp) due to the virtual gnd at the -ve opamp input, which could be having an effect.
An alternative speculation might be that because the ouput opamp in the Baxandall gain control is already somewhat taxed just driving the pot, it has a harder time driving the power amp as well. But you would expect this to show up in the THD measurements (I guess we can assume they were made with a "standard" load on the output).