Let's just start with the fact that the LM4562 is a brilliant op amp, so I am not dissing it here. Nonetheless I have never used it for any length of time and have generally thought it to be precise and detailed, but utterly uninvolving. I am now aiming for a proper low distortion solution, and of course it hoves into view.
First of all, is there a better amplifier out there than that LME series? I also want to perhaps use it in an error correcting role (as well as in a conventional feedback role). On the error correction I am a touch worried and think that a Barrie Gilbert op amp might be better (Jan, do chime in here, because I know you have done this.), but they are a bit noisy. The even trickier part is that I'd like to pre-emphasise the HF part - and do all that amplification in the 4562 - and then keep the loop gain open for higher frequencies in some chip amp (which may then go back to be error corrected).
Is the 4562 the right amp to do this with, is the first question, and the second is whether I can stabilise an output amp like a 3886 when the pre-emphasis has to come down. Not just below 10 but below unity. A fat capacitor in the feedback loop of an amplifier isn't quite what they were designed for. The aim of this ridiculous dancing is to not have a distortion response rising with frequency (something we had pretty much accepted as a de facto result of even the best work.
Your views?
First of all, is there a better amplifier out there than that LME series? I also want to perhaps use it in an error correcting role (as well as in a conventional feedback role). On the error correction I am a touch worried and think that a Barrie Gilbert op amp might be better (Jan, do chime in here, because I know you have done this.), but they are a bit noisy. The even trickier part is that I'd like to pre-emphasise the HF part - and do all that amplification in the 4562 - and then keep the loop gain open for higher frequencies in some chip amp (which may then go back to be error corrected).
Is the 4562 the right amp to do this with, is the first question, and the second is whether I can stabilise an output amp like a 3886 when the pre-emphasis has to come down. Not just below 10 but below unity. A fat capacitor in the feedback loop of an amplifier isn't quite what they were designed for. The aim of this ridiculous dancing is to not have a distortion response rising with frequency (something we had pretty much accepted as a de facto result of even the best work.
Your views?
Thanks for that recommendation. I like the low noise aspect of it, though it's pretty expensive. Has anyone actually listened to it and what are the general opinions? Personally I loathe the sound of the LM4562 but I haven't really given it a full chance, tailored to its characteristics. All I have done is an A/B with my favourite op amp, the LM6172, and it didn't come off that well, though you could certainly tell the hard work that had gone into it. Very detailed but just not really very engaging - and not really the direction I wanted to go in.
While looking at this one, I also came across the OPA1604 which seems to be pretty much the same thing except with 10dB less OLG and therefore about 3x the distortion. It's a nicer price, at around £1 per amp, so has anyone had a chance to try it out?
If the 4562 sounds poor then you need quantify it and find out what the problem is.
I would do some frequency response tests and see what results you get.
Sometimes something that sounds better is actually distorting the signal.
Classic example is guitar valve amps.
Flat sounds boring but lots of harmonics make the guitar sound better.
Marshall made a fortune from that fact.
I would do some frequency response tests and see what results you get.
Sometimes something that sounds better is actually distorting the signal.
Classic example is guitar valve amps.
Flat sounds boring but lots of harmonics make the guitar sound better.
Marshall made a fortune from that fact.
The only way you're going to audibly hear a difference in these unobtanium-low distortion opamps is if they're wildly misbehaving, which speaks to implementation. We're talking <-120 dB distortion levels here under line-level.
The 1612 is also internally 2-pole compensated (at least that's what it looks like) and has a bit of a kink in it's loop gain phase plot. Which is going to up the ante a bit in a EC/composite role, as you're suggesting. But certainly a high performance part.
As far as the 3886, you can push it into G > 10 noise gain compensation above audio bandwidth and probably have it stable in unity/sub-unity forward gain. Section 7 here will be most helpful: http://web.mit.edu/6.101/www/reference/op_amps_everyone.pdf
The 1612 is also internally 2-pole compensated (at least that's what it looks like) and has a bit of a kink in it's loop gain phase plot. Which is going to up the ante a bit in a EC/composite role, as you're suggesting. But certainly a high performance part.
As far as the 3886, you can push it into G > 10 noise gain compensation above audio bandwidth and probably have it stable in unity/sub-unity forward gain. Section 7 here will be most helpful: http://web.mit.edu/6.101/www/reference/op_amps_everyone.pdf
composite amps with another op amp wrapping a audio power chip amp don't have much scope for improvement given the low and uncertain gain intercept, minimum stable local gain of most power chip amps but some improvement is possible
I have made headphone amps with good 'audio' FET input op amps wrapping 100 MHz DSL driver cfa op amps
the composite multiloop feedback linear stability is much easier to design with the output much faster than the input op amp
I've used AD8610, OPA627 but don't really have much concern for/belief in particular op amp's 'sound' once you get to a decent quality level
today I would look at OPA827 and the more reasonably priced OPA164x parts for line level input, possible volume pot feed
I have simmed a Black's Error Feedforward (Quad 'Current Dumping' amplifier in the audio world)
I have made headphone amps with good 'audio' FET input op amps wrapping 100 MHz DSL driver cfa op amps
the composite multiloop feedback linear stability is much easier to design with the output much faster than the input op amp
I've used AD8610, OPA627 but don't really have much concern for/belief in particular op amp's 'sound' once you get to a decent quality level
today I would look at OPA827 and the more reasonably priced OPA164x parts for line level input, possible volume pot feed
I have simmed a Black's Error Feedforward (Quad 'Current Dumping' amplifier in the audio world)
Given that that highest performing 3886 composite currently loose in the wild uses the LME49710 (half a 4562) I would suggest that is up to the task. distortion does rise a bit to 20kHz, but 10ppm @20KHz is not going to keep me awake at night.
Non-trivial though...
Non-trivial though...
Attachments
maybe modern production 3886 are all above average?
the designs seem to take the 'typical' GBW spec as the effective minimum
My guesstimate puts the safe with production variation margin local closed loop gain corner as low as 100kHz for LM3886 with a greater than 4:1 range
designing for stable extra loop gain from the nice input op amp really doesn't give much 20 kHz added safe composite loop gain
adding a zero in the loop to stretch the LM3886 BW becomes very problematic if you have to allow for low GBW devices
the designs seem to take the 'typical' GBW spec as the effective minimum
My guesstimate puts the safe with production variation margin local closed loop gain corner as low as 100kHz for LM3886 with a greater than 4:1 range
designing for stable extra loop gain from the nice input op amp really doesn't give much 20 kHz added safe composite loop gain
adding a zero in the loop to stretch the LM3886 BW becomes very problematic if you have to allow for low GBW devices
Last edited:
I can only assume that, as he worked for the company at the time he had access to some statistics us mere mortals don't have. There is a plot of closed loop gain to 10MHz with a sample of one* available here http://www.diyaudio.com/forums/vend...er-achieving-0-0004-thd-n-55.html#post4742560 .
I know I couldn't do better so I bought the amps. One day might actually hook them up!
I should also note that I am not convinced I would be able to hear the difference between a super composite and a standard gainclone.
*well aware a sample of one doesn't challenge your observation in the slightest, but one measurement is hopefully better than none
.
I know I couldn't do better so I bought the amps. One day might actually hook them up!
I should also note that I am not convinced I would be able to hear the difference between a super composite and a standard gainclone.
*well aware a sample of one doesn't challenge your observation in the slightest, but one measurement is hopefully better than none
.
Distortion will strongly depend on load it drives. On a light or virtually no load 161x is extremely linear. Once it gets under a couple of kOms it starts to give up the leaders position to 4562. 4562 remains linear down to sub kiloOhm load.
MAX9632 is an interesting competitor - has LME49990 like linearity and driving capability (tested myself) but it is a single opamp part.
MAX9633 is a dual one, but... it has extremely high current noise. I even though that this was a typo in DS, but not, I tested it, really matches the spec.
Last edited:
Why is that? LM4562 has -36db less THD spec, -4.5db less voltage noise, and -20 db less CMRR.
You have to understand the circuit to figure out which opamp properties are important - basically all of these good audio opamps, if _used correctly_, are completely indistinguishable from each other (by ear) apart from noise level (if in a preamp dealing with low-level signals).
However the "used correctly" is quite a mine-field. Some circuits will cause one opamp to oscillate and another not. Some will overload some opamps output sections, but not others, some will expect very low input bias currents, other circuits don't care. Some opamps are much more sensitive to RFI than others and require suitable RFI filtering on the inputs.
However the "used correctly" is quite a mine-field. Some circuits will cause one opamp to oscillate and another not. Some will overload some opamps output sections, but not others, some will expect very low input bias currents, other circuits don't care. Some opamps are much more sensitive to RFI than others and require suitable RFI filtering on the inputs.
I hate to say it, but there is very little recorded music that hasn't already passed through a bunch of NE5532s and TL072s. By most standards the LM4562 is a superior op-amp to the NE5532.
I'm still (perhaps irrationally) fond of the OPA604, maybe because some equipment I really like used it. That's a FET input op-amp though. It can take up to +/- 24V rails which is quite nice at times.
I'm still highly skeptical that anyone can hear the difference between two op-amps when they are used correctly. Of course, some people don't necessarily understand that they can't just swap these chips in a circuit willy-nilly and expect them to work properly.
I'm still (perhaps irrationally) fond of the OPA604, maybe because some equipment I really like used it. That's a FET input op-amp though. It can take up to +/- 24V rails which is quite nice at times.
I'm still highly skeptical that anyone can hear the difference between two op-amps when they are used correctly. Of course, some people don't necessarily understand that they can't just swap these chips in a circuit willy-nilly and expect them to work properly.
OK, I'm really late to the show, but from my experience the LM4562 is just extraordinary. A flatness of response which is incredible, an equivalent input noise figure which (so far) none of the alternatives above come close to besting, a generous +-36 V power supply tolerance, 120+ dB of power-supply-ripple-rejection and 110+ dB of common-mode-rejection as well. Almost unmeasurable opamp A vs B (they're dual units) cross-talk and intermodulation distortion.
And $1.62 unit price from Mouser.
What a deal!
If the basic specs aren't enough, the Texas Instruments datasheet is full of juicy bits you can do with the things, and they're quite capable of driving a 600 Ω 'conventional low-impedance' microphone-input load.
The only “bettering” I could do would be to gin up a custom JFET pre-opamp gain stage for a very, very low output transducer such as a MM cartridge. With tantalum film resistors, well chosen JFETs, tight circuit board design and ground-planes absolutely everywhere, (i.e. between traces!), you might get down to 0.7 to 0.8 nV/√Hz for a front end. Not so well on a hot day! (recalling that the additive effect of thermal noise is … thermal).
Anyway, NE5532's have double-to-triple the input equivalent noise.
The OPA604 is topping 10 nV/√Hz.
But the take-away from the above mini-discussion is that you can employ a custom, linear, no-feedback, ultra-low-noise JFET gain stage in front of whatever you hope to achieve with the op amps. And get your noise down to a respectable sub–1.5 nV/√Hz figure of merit.
Just saying,
⋅-=≡ GoatGuy ✓ ≡=-⋅
And $1.62 unit price from Mouser.
What a deal!
If the basic specs aren't enough, the Texas Instruments datasheet is full of juicy bits you can do with the things, and they're quite capable of driving a 600 Ω 'conventional low-impedance' microphone-input load.
The only “bettering” I could do would be to gin up a custom JFET pre-opamp gain stage for a very, very low output transducer such as a MM cartridge. With tantalum film resistors, well chosen JFETs, tight circuit board design and ground-planes absolutely everywhere, (i.e. between traces!), you might get down to 0.7 to 0.8 nV/√Hz for a front end. Not so well on a hot day! (recalling that the additive effect of thermal noise is … thermal).
Anyway, NE5532's have double-to-triple the input equivalent noise.
The OPA604 is topping 10 nV/√Hz.
But the take-away from the above mini-discussion is that you can employ a custom, linear, no-feedback, ultra-low-noise JFET gain stage in front of whatever you hope to achieve with the op amps. And get your noise down to a respectable sub–1.5 nV/√Hz figure of merit.
Just saying,
⋅-=≡ GoatGuy ✓ ≡=-⋅