opamps are just like power amps. They draw current from the supply rails. As output current increases the supply rail currents change as well.
Only if the opamp stays in ClassA do the supply rail currents follow the output currents. If the opamp transitions into ClassAB then the supply rail currents do not follow the output. They become these half wave rectified things that contain harmonics way up into the MHz. That is the VHF that must be decoupled locally.
This is not oscillation, this is normal amplifier operation.
I don't think the laws of physics have changed all that much in that last 20 years.
Those are a little on the low side indeed. In order to equal 5534 e_n levels, Rf||Rg would have to be no higher than .75k, so half that should be OK in practice. For IC3, either 560R and 1k or 470R and 820R easily fulfil this condition. IC2's noise usually is quite irrelevant but it sees fairly high levels, so I'd prefer 1k + 1k, 1k5 + 1k5 or 2k2 + 2k2 there.
Other than that, it seems like a very reasonable construction.
Well, it does require the amps to be unity-gain stable, which means you're giving away some GBW. Besides, you obviously can't get lower than unity in a non-inverting circuit. It's reducing high-frequency noise though.
I think the approach kinda makes sense for the gain stage, but not for the pre-gain stage (IC2). It would be easy to include some passive filtering after that one instead.
Incidentally, concepts like these with redistributed gain are not exactly new. They go back to the late '70s at least, as you'll find when inspecting, say, Grundig mass-market amps and receivers like the R/Vx000 series. Not with 10k pots then, of course, but with a respectably low noise floor in spite of 100k pots (75 dB(A) re: 50 mW / 4R). (The most attention to noise in mass-market amps today seems to be paid by NAD.)
Even those were just a poor man's version of what the fancy preamps of the time did (e.g. Yamaha C-1a, C2) - two-stage volume control with a 4-gang pot. While you have to pay some attention to tracking error, that really gets things sorted noise-wise. No worries even with Klipschorns, assuming the power amp is good enough. When the technology trickled down to integrated amps around 1990, units like the Technics SU-VX800 could boast SNRs of 86 dB(A) / 50 mW / 4R.
Now 4-gang pots are about as common as hen's teeth these days (they vanished from mass-market amps in the early-mid '90s), but you can equally combine a normal pot and a stepped attenuator (down to a 2- or 3-position switch). It's just not quite as comfortable.
PS: The whole 2-stage volume control thing works so well because of a fundamental disparity in required instantaneous dynamic range vs. total dynamic range.
Let me illustrate this.
Required instantaneous dynamic range is defined on the sonic side of things. Noise level needs to be low enough to be inaudible, and you have a certain desired listening volume plus headroom. This is essentially limited by human hearing and the surroundings (and recordings, of course). In practice, required DR is like 80, maybe 100 dB tops.
However, electroacoustic gain from preamp output to ears is subject to variations:
Speaker sensitivity varies, from less than 85 dB SPL per 2.83 V @1m in small speakers to over 105 dB SPL in big horns.
Power amp voltage gain typically ranges from 26 to 32 dB.
Listening distances and rooms vary (some interaction with speaker sensitivity here).
The picture is a little clearer when looking at the world of headphones, headphone amps being close relatives of preamps. Transducer sensitivities range from little over 90 to more than 130 dB SPL per 1 Vrms here (assuming a zero-ohm output). As you may imagine, those at the low end commonly don't go quite as loud as desired, while those at the other end tend to be plagued by hiss.
Required total dynamic range basically is required instantaneous dynamic range plus electroacoustic gain variation.
IOW, if you want to have 80 dB of instantaneous dynamic range for any kind of headphones, required total DR will be 120+ dB.
Now you can either try and "brute force" it by building a gain stage that achieves a 120+ dB SNR (which still is doable), or include some attenuation after the gain stage in order to keep noise levels in check when needed. The latter tends to get the volume pot into a range with better tracking, too.
Therefore, building something like a super-low-noise tone control may be a noble endeavour, but I don't think it is necessary. You only need to get impedances low enough for distortion to be minimized. Of course it still depends on where in the circuit the tone controls are located, but even then I don't think they have to be super-ultra-low-noise as long as they're in any kind of sane place.
Let me illustrate this.
Required instantaneous dynamic range is defined on the sonic side of things. Noise level needs to be low enough to be inaudible, and you have a certain desired listening volume plus headroom. This is essentially limited by human hearing and the surroundings (and recordings, of course). In practice, required DR is like 80, maybe 100 dB tops.
However, electroacoustic gain from preamp output to ears is subject to variations:
Speaker sensitivity varies, from less than 85 dB SPL per 2.83 V @1m in small speakers to over 105 dB SPL in big horns.
Power amp voltage gain typically ranges from 26 to 32 dB.
Listening distances and rooms vary (some interaction with speaker sensitivity here).
The picture is a little clearer when looking at the world of headphones, headphone amps being close relatives of preamps. Transducer sensitivities range from little over 90 to more than 130 dB SPL per 1 Vrms here (assuming a zero-ohm output). As you may imagine, those at the low end commonly don't go quite as loud as desired, while those at the other end tend to be plagued by hiss.
Required total dynamic range basically is required instantaneous dynamic range plus electroacoustic gain variation.
IOW, if you want to have 80 dB of instantaneous dynamic range for any kind of headphones, required total DR will be 120+ dB.
Now you can either try and "brute force" it by building a gain stage that achieves a 120+ dB SNR (which still is doable), or include some attenuation after the gain stage in order to keep noise levels in check when needed. The latter tends to get the volume pot into a range with better tracking, too.
Therefore, building something like a super-low-noise tone control may be a noble endeavour, but I don't think it is necessary. You only need to get impedances low enough for distortion to be minimized. Of course it still depends on where in the circuit the tone controls are located, but even then I don't think they have to be super-ultra-low-noise as long as they're in any kind of sane place.
If low noise is what you want you can't beat a transformer/transformer feedback. <.1nV has been achieved and can not be equalled by any reasonable paralleling of active devices.
Thanks Bonsai, nice summary of the pros of using an RC filter and I believe it also partly answers my question in that any additional signal imposed by the large R due to the opamp output current draw is pushed down to frequencies where the opamp PSRR is highest thanks to the filter time constant. I suppose another advantage might be that by using seperate RC filters for each opamp the channel or circuit stage seperation could be improved.
How? I don't get it. Can you explain this?
Owdeo
Hi Owdeo
That kind of thinking was taken to task as far back as November 1987 by respected U.K. technical writer Ben Duncan, in an article from HiFiNews and Record Review about the flaws in the original Marantz/Philips CD players.
" Also in this 2nd generation machine, a valiant attempt effort has been made toi improve PSRR by inserting 47 ohm resistors (R9,R10) nin line with the supply..In conjunction with C7 and C8, the outcome is mild filtration of the supply, at the same time creating an irregular supply impedance over the audio band.
Since I.C.s 1 and 2 are intimitately combined in a dual package with common supply pins, imperfections in the decoupling forces them to interfere with one another, whenever I.C1s current draw acts to modulate the supply terminal, I.C.2 suffers error feedthrough."
Kind Regards
Alex
Hi Owdeo
That kind of thinking was taken to task as far back as November 1987 by respected U.K. technical writer Ben Duncan, in an article from HiFiNews and Record Review about the flaws in the original Marantz/Philips CD players.
" Also in this 2nd generation machine, a valiant attempt effort has been made toi improve PSRR by inserting 47 ohm resistors (R9,R10) nin line with the supply..In conjunction with C7 and C8, the outcome is mild filtration of the supply, at the same time creating an irregular supply impedance over the audio band.
Since I.C.s 1 and 2 are intimitately combined in a dual package with common supply pins, imperfections in the decoupling forces them to interfere with one another, whenever I.C1s current draw acts to modulate the supply terminal, I.C.2 suffers error feedthrough."
Kind Regards
Alex
Thanks Bob, appreciate the advice. There seem to be quite a few proponents of the use of RC filtering here, but since neither yourself or Douglas Self (at least going by his published designs) use it we should all be able to conclude that it isn't required for achieving good measured performance or stable operation in this application.
Seems as though there's still a question mark over whether it has the potential to affect sound quality in so far as it alters the way the supply currents flow around the opamp and the way it interacts with the supplies, which perhaps would not really show up in THD measurements unless it makes them worse. Or do you think this is just woolly thinking and that if it really is benficial it would result in reduced THD measurements?
Cheers,
owdeo
Thanks sgrossklass, nice analysis. I also came to similar conclusions so it is very welcome to have them confirmed
I think the coupling caps are all too small also, there would certainly be unnecessary capacitor distortion as a result. Still I recall it sounding very good at the time it was popular, very dynamic and musically involving, if a tad "grainy". That was a long time ago though...
Hi Alex,
Very interesting, thanks for posting this. I wonder what the RC filtering proponents have to say to it?
Having said that, I recall some letters from Ben Duncan published in EW&WW in the '90s criticising Self's work and an article proposing an explanation for cables sounding different due to "micro diodes" formed in copper. I recall these were all thouroughly addressed and disected by Self beyond any doubt at the time, and it made for some enjoyable reading
So perhaps he is not such a respected technical writer... also Hifi News is the most evil of all the subjectivist mags don't you know Cheers,
owdeo
PS Good to hear from you, will send you an email
I say that 47R is way too high if the opamps are being asked to deliver significant output currents. I use RC filtering but go for resistors below 10R even when not delivering line driving current levels. If the chosen cap has ESR of 0.2ohm then even a 2R2 gives a worthwhile 20dB filtration at higher frequencies.
Ben Duncan's articles in HFN&RR I followed carefully when they came out and even built my own version of his AMP-02, using Linear Technology opamps as recommended. These days I'd not use devices like LT1037 for audio though
His arguments quoted above don't seem to pay much attention to the fact that opamps PSRRs decline -20dB/decade and usually begin falling within the audio band.Indeed, but didn't they start the whole subjectivist bandwagon by first suggesting that measurements were meaningless and allowing "faith-based audio" to flourish?
Now the common man with no electronics knowledge who goes into a hifi shop to buy a low-end system is almost certainly going to be sold a bunch of fancy cables that cost as much as one of the components as the salesman will assure him that they make all the difference. Even the mains IEC cable is apparently vitally important! Worse still you see frequently in these magazine reviews the statement that the item under review didn't sound too good at first listen, but then improved miraculously to the best thing they've ever heard after it "burnt in". Heaven forbid it might be that their hearing had adjusted to the way it sounds...
In my view to some extent it's a grey area like everything in life, but while there's perhaps much more to be done to correlate how measurements relate to our perception of sound, this "mystcising" of audio into some kind of black art has stopped the technology from advancing anywhere near as much as it should have in the last 30 years. Eg why aren't all high-end speakers active with DSP correction by now?
And please don't say because it doesn't sound as good
It must be interesting for manufacturers these days as it appears that to succeed in the high-end area their product has to receive favourable reviews in these magazines, and since the reviews are always subjective without any attempt by the reviewer to validate his assessment, they have no way of ensuring success just by designing good products...
Now the common man with no electronics knowledge who goes into a hifi shop to buy a low-end system is almost certainly going to be sold a bunch of fancy cables that cost as much as one of the components as the salesman will assure him that they make all the difference. Even the mains IEC cable is apparently vitally important! Worse still you see frequently in these magazine reviews the statement that the item under review didn't sound too good at first listen, but then improved miraculously to the best thing they've ever heard after it "burnt in". Heaven forbid it might be that their hearing had adjusted to the way it sounds...
In my view to some extent it's a grey area like everything in life, but while there's perhaps much more to be done to correlate how measurements relate to our perception of sound, this "mystcising" of audio into some kind of black art has stopped the technology from advancing anywhere near as much as it should have in the last 30 years. Eg why aren't all high-end speakers active with DSP correction by now?
And please don't say because it doesn't sound as good It must be interesting for manufacturers these days as it appears that to succeed in the high-end area their product has to receive favourable reviews in these magazines, and since the reviews are always subjective without any attempt by the reviewer to validate his assessment, they have no way of ensuring success just by designing good products...
Hi Richard
You mustn't have been long out of nappies back then ?
After doing a bit of googling, I found I made an error - it was the AMP-01 I based my pre on, not the -02 which came out later ('89-'90). So I must have been still in the womb
My CD player at the time was a Philips CD-160.
Owdeo
Owdeo
I guess we will have to compare notes on that lot with actual listening?
I don't go along with fancy interconnects and mains cables either. I do however try to obtain the highest possible technical performance by blueprinting and matching devices, then far more attention to the PSU area.
Kind Regards
Alex
Owdeo
I guess we will have to compare notes on that lot with actual listening?
I don't go along with fancy interconnects and mains cables either. I do however try to obtain the highest possible technical performance by blueprinting and matching devices, then far more attention to the PSU area.
Kind Regards
Alex
In my view to some extent it's a grey area like everything in life, but while there's perhaps much more to be done to correlate how measurements relate to our perception of sound, this "mystcising" of audio into some kind of black art has stopped the technology from advancing anywhere near as much as it should have in the last 30 years. Eg why aren't all high-end speakers active with DSP correction by now?
The only time I've heard a Meridian system playing was back at the old Heathrow Penta show and that indeed was my impression - the sound made my ears bleed so I left that room jolly fast
That's not to say digital sucks - its all in the implementation.
The only time I've heard a Meridian system playing was back at the old Heathrow Penta show and that indeed was my impression - the sound made my ears bleed so I left that room jolly fast
That's not to say digital sucks - its all in the implementation.
I wasn't necessarily referring to that brand, but I do believe that if something about the sound was that offensive it must have been a measureable problem and was not the result of using digital filters per se. The problem is working out what and how to measure it...
Having worked for a company that produced a digital speaker correction system (though not as a designer), I can vouch that pollution of the analogue stages by all the digital switching noise from the DSPs can be a real problem that is very hard to prevent, so I agree with you Alex.