The LM4562's main foibles are rather high input current noise due to what is presumed to be a degenerated input stage (it takes only ~1.7 kOhms to equal voltage noise or about twice that to overtake a 5532) and a tendency to latch up on startup in single supply applications with high-impedance bias.
Because of the former, I would actually recommend against using one in an MM phono stage input if you are not going to give it an extra discrete input stage. I simulated things a while back (somewhere in a thread in Analogue Source), and high-frequency noise with a simulated cartridge source impedance actually comes out markedly worse than for an NE5534A, or LM833 or even NE5532. The best 2-channel opamp for this application might well be the trusty NJM2068, perhaps with a discrete output stage to offset its so-so output driving abilities.
I'll echo what people have said about "used correctly". I'm sure there are plenty of applications where the part's excellent transfer linearity, good common-mode linearity, good driving abilities and lowish voltage noise will come in handy.
Whenever I read "it sounds bad until you give it some gain", I really have to suspect oscillation being an issue, i.e. probably a layout problem.
Because of the former, I would actually recommend against using one in an MM phono stage input if you are not going to give it an extra discrete input stage. I simulated things a while back (somewhere in a thread in Analogue Source), and high-frequency noise with a simulated cartridge source impedance actually comes out markedly worse than for an NE5534A, or LM833 or even NE5532. The best 2-channel opamp for this application might well be the trusty NJM2068, perhaps with a discrete output stage to offset its so-so output driving abilities.
I'll echo what people have said about "used correctly". I'm sure there are plenty of applications where the part's excellent transfer linearity, good common-mode linearity, good driving abilities and lowish voltage noise will come in handy.
Whenever I read "it sounds bad until you give it some gain", I really have to suspect oscillation being an issue, i.e. probably a layout problem.
Uh, 36V *total*. So +/-18V. Same as a '741 or TL072.
Maybe you knew that and your fingers don't. But anybody uses this chip at +/-36V (72V total) will be disappointed.
That looks Xeroxed from the LM833/837(??) puff-sheet. Only the names changed to protect the innocent.
Last edited:
Hmmmm....the NJM2068 has noise specs that would virtually be the same as an LM4562 for phono use....but it is 3x slower and has 30 db more THD. so....???/
"Virtually the same"? Nope.
Practical NJM2068 voltage noise density is ca. 3.5 nV/√(Hz) or thereabouts, about 3 dB better than an NE5532 and about on par with an NE5534A.
NJM2068 input bias current is spec'd at 150 nA typ, so corresponding shot noise is likely to be at least on par with or lower than your average NE5532 (200 or 500 nA typ depending on manufacturer). And that sits at 0.7 pA/√(Hz). See Horowitz/Hill if in doubt.
So the NJM2068 has somewhat higher voltage noise but in all likelihood less than half the current noise, which is a nominal 1.6 pA/√(Hz) for the NJM4562.
Now in an MM application, the chip is faced with a source impedance ranging from below 1 kOhm in the bass to several tens of kOhms towards the upper end of the audible range. 10 kOhms turn 1.6 pA/√(Hz) into 16 nV/√(Hz), and that's an impedance you might see as low as 3-5 kHz - so up there current noise quickly dominates. Guess where noise bothers us the most? 6-10 kHz or thereabouts.
Hence why low(ish)-noise JFET inputs are so attractive in this application. If you want a combo MM/MC input that just needs switching of input impedance and gain setting resistor, you basically don't get around ultra-low-noise FETs, preferably a whole bunch. Otherwise an extra MC prepre is the wiser option.
Here's the noise simulation I mentioned earlier. NE5532 would be a hair noisier than LM833 at the bottom end.
The NJM2068's distortion spec is mainly limited by the relatively heavy output loading at relatively high gain and substantial output level (at least I interpret Vo = 5V as 5 Vrms). As NwAvGuy showed, the part clearly isn't entirely happy driving 1.75 kOhms at a gain of 7 - high-frequency distortion increases at 2 Vrms and maximum output voltage always stays a good bit behind NE5532 or OPA2134. As I said, this part would benefit from a discrete buffer. Let's not be too harsh on an inexpensive part that's been around since at least the early '90s. You can always use a less fussy 5532 if you can accept a bit more noise (still better than LM4562 overall).
Practical NJM2068 voltage noise density is ca. 3.5 nV/√(Hz) or thereabouts, about 3 dB better than an NE5532 and about on par with an NE5534A.
NJM2068 input bias current is spec'd at 150 nA typ, so corresponding shot noise is likely to be at least on par with or lower than your average NE5532 (200 or 500 nA typ depending on manufacturer). And that sits at 0.7 pA/√(Hz). See Horowitz/Hill if in doubt.
So the NJM2068 has somewhat higher voltage noise but in all likelihood less than half the current noise, which is a nominal 1.6 pA/√(Hz) for the NJM4562.
Now in an MM application, the chip is faced with a source impedance ranging from below 1 kOhm in the bass to several tens of kOhms towards the upper end of the audible range. 10 kOhms turn 1.6 pA/√(Hz) into 16 nV/√(Hz), and that's an impedance you might see as low as 3-5 kHz - so up there current noise quickly dominates. Guess where noise bothers us the most? 6-10 kHz or thereabouts.
Hence why low(ish)-noise JFET inputs are so attractive in this application. If you want a combo MM/MC input that just needs switching of input impedance and gain setting resistor, you basically don't get around ultra-low-noise FETs, preferably a whole bunch. Otherwise an extra MC prepre is the wiser option.
Here's the noise simulation I mentioned earlier. NE5532 would be a hair noisier than LM833 at the bottom end.
The NJM2068's distortion spec is mainly limited by the relatively heavy output loading at relatively high gain and substantial output level (at least I interpret Vo = 5V as 5 Vrms). As NwAvGuy showed, the part clearly isn't entirely happy driving 1.75 kOhms at a gain of 7 - high-frequency distortion increases at 2 Vrms and maximum output voltage always stays a good bit behind NE5532 or OPA2134. As I said, this part would benefit from a discrete buffer. Let's not be too harsh on an inexpensive part that's been around since at least the early '90s. You can always use a less fussy 5532 if you can accept a bit more noise (still better than LM4562 overall).
Last edited:
Some months ago I took some time and did - out of plain curiocity - a listening session with all the different opamps in lying around my "drawers":
JRC 2114D
JRC 4560
JRC 4558DD
LM 833N
NE5532P
LT1364
TL072IP
TL082CP
LM4562NA
OPA2604AP
OPA2134PA
OP275
OPA2228P
LME49720HA .
Ad a "test rig" served rather "stock" Marantz CD-67 (only opamp had been fitted into a socket), audio output (fed into an integrated amplifier) was taken from before the HDAM (i.e. straight after the opamp) and I listened via headphones.
What can I say - there were certainly audible differences (for me), the best three would be JRC 4558DD, TL082CP and OPA2134PA.
JRC 2114D
JRC 4560
JRC 4558DD
LM 833N
NE5532P
LT1364
TL072IP
TL082CP
LM4562NA
OPA2604AP
OPA2134PA
OP275
OPA2228P
LME49720HA .
Ad a "test rig" served rather "stock" Marantz CD-67 (only opamp had been fitted into a socket), audio output (fed into an integrated amplifier) was taken from before the HDAM (i.e. straight after the opamp) and I listened via headphones.
What can I say - there were certainly audible differences (for me), the best three would be JRC 4558DD, TL082CP and OPA2134PA.
A friend tell me, that the AD825 in a phono preamp with passive RIAA network so as as front end in a power amp is much better.
he received the recommendation during a call with LC Audio in Denmark - go to
AD825 module from LC Audio
Because this is a single OP AMP, adapter PCB is necessary:
L C Audio Technology/AD825
Maybe this is true, maybe not. Are there listening tests between this both OP-Amps ?
Unfortunately there is no schematic of the internals from LM4562:
https://www.ti.com/lit/ds/symlink/lm4562.pdf
but schematic of AD825 is available (basically only one voltage gain stage by use of a folded cascode):
https://www.analog.com/media/en/technical-documentation/data-sheets/AD825.pdf
he received the recommendation during a call with LC Audio in Denmark - go to
AD825 module from LC Audio
Because this is a single OP AMP, adapter PCB is necessary:
L C Audio Technology/AD825
Maybe this is true, maybe not. Are there listening tests between this both OP-Amps ?
Unfortunately there is no schematic of the internals from LM4562:
https://www.ti.com/lit/ds/symlink/lm4562.pdf
but schematic of AD825 is available (basically only one voltage gain stage by use of a folded cascode):
https://www.analog.com/media/en/technical-documentation/data-sheets/AD825.pdf
Last edited:
I think the 2156 is very similar to the 1656 but has a different front end with two more ntype mosfets so that the input range is rail-to-rail. Otherwise the specs are quite the same. I used the 2156 recently when I rolled a headphone driver that used one half in a 2X gain shunt mode and the other half in voltage follower mode to double the current output. Here I felt the follower needed to be able to tolerate R-R input range. Sounds good to me with 33 Ohm Grados.
That is an inherent property of all OP Amps and discrete amps with only one voltage gain stage in the NFB loop.
In order of best possible sonic quality this isn't a disadvantage - so I think.
Let me introduce you to the AD797. It has one gain stage (folded cascode BJT) and 146 dB open loop gain: link to manufacturer's datasheet
_
I think somewhat more price-accessible is the AD8597, at 1.1 nV/√(Hz), and $5.75 a unit at Mouser. $2.50 per, qty–1000 (one reel). Surface mount, tho. No DIP type packaging. But hey, its also 2020 AD, and the dual-in-line package is pretty much a thing of the past. Even at my advanced age, I've learned to solder surface-mount without making a hideous mess. Well, not all the time. LOL…
⋅-=≡ GoatGuy ✓ ≡=-⋅
MY comment was directed toward comparison of an LM4562 with a NJM2068---NOT a NE5532.As far as I can see, the 4562 has a voltage noise of 2.7nV√Hz @ 1 Khz and current noise of 1.6pA√Hz. The specs for an NJM2068 are harder to interpolate, as they only give an "Equivalent Input Noise Voltage" of 0.44µV (FLAT+JIS A, RS=300Ω ); the 4562 specifies 0.34µV (20-20KHz). So where are you getting the conclusion that the 2068 is quieter?
The NJM2068 noise values are A-weighted, so a lot less than reality (the cuve showing noise against source impedance cannot be flat 20kHz noise as the values are less than the Johnson noise). From that curve its likely the NJM2068 has a lot lower current noise, its noise values aren't starting to jump rapidly until 10's of kohms. My guess is its about 6nV and 0.5pA, giving 0.86µVrms flat (20kHz BW) at low impedance, 2.9µVrms at 20k source, which A-weighted appear as 0.44 and 1.6uV
I never said I thought it was better, just that I have a somewhat irrational liking for it. It's ability to tolerate pretty high voltage rails is useful at times. In my eyes most modern op-amps are good enough that I can get excellent audio performance with almost all of them. I bought my stock of OPA604s a while back to fix something (probably an MC^2 amplifier module) and discovered that they're rated for up to +/- 24V rails. This is fairly useful and prompted me to buy some extra.
A lot of people seem to think that buying expensive parts is a substitute for good engineering. You can throw AD797s and API2520s at a crappy circuit all day long and at the end of the day, it's still a crappy circuit. This is not to say that these chips don't have their place (they're very useful in low-noise circuits), but spending $100 on AD797s will not make a mediocre line stage design any better.
Quite honestly, the LM4562 is pretty hard to best. If you don't like how it sounds, there four possibilities that come to my mind.
1) It is a poor choice for the circuit for some reason or another. Input impedance on the 4562 is pretty low, I know of a few circuits that flat out won't work with it.
2) Flawed circuit design that stinks regardless of the chip used.
3) The extremely low noise and distortion performance of this chip is making flaws in other parts of the circuit (or another piece of equipment) apparent
4) You find the clean, near-zero distortion to be unpleasing and boring. There's no shame in this, the point of Hi-Fi listening is personal enjoyment. If this is the case, recognize it, since you won't like chips like the AD797 either.
That would make the part have more voltage noise than the NE5532, whereas NwAvGuy found 3 dB less (O2 gain stage, gain = 7x, sum of external impedances presumably ~= 490 ohms). Mind you, I kind of suspect that an NE5532 might be closer to 5.5 nV these days. (Then his OPA2134 would be at about 8 nV...)
The datasheet curve suggests noise being up by 3 dB at ~1.2 kOhms or so, which would be in the 4.0-4.5 nV vicinity. Given that it probably hasn't been updated since way back when but manufacturing may have improved, seeing 3.5 nV these days does not seem entirely unlikely.
It is looking like at least NE5534 non-A level noise performance (4 nV / 0.6 pA), which is not at all outlandish actually. I think the best combo I've seen in Horowitz/Hill for a bipolar input opamp was something like 2.5 nV / 0.4 pA, wasn't a particularly fast part though... In any case, getting to NE5534 noise levels in a cheap dual opamp is not too shabby. The somewhat lame output stage is, unfortunately, a trait shared with other '80s low-noise opamps. Designers were expecting these to be primarily used for high-gain tape head or mic preamps and the like, or a phonopre (preferably with "Neumann pole" for reduced loading).
I'm using the AD797 in line stages (Av = 16 dB) - superb part with great output drive and ultra low noise for source impedances up to c. 2k. This makes them great for use with a 10k Log pot where the noise is optimal in the mid setting from 12 o'clock through to about 4 o'clock position. The input bias current is a bit high, but you should always use DC blocking when a pot is involved. I don't normally comment on 'sound' here on diyaudio but the AD797 sounds good - it is a great part. You do need to follow the data sheet recommendations and it has to be decoupled properly - it will oscillate if not or if the layout is sloppy.
The best part for MM is still the NE5534 - you need to go to a very good JFET input opamp if you want to best it in that application and they will be a lot more expensive for the same distortion and output drive capability. The reason for the NE5534's great noise performance in MM applications is the input noise current is only 0.6pA/rtHz - half of the AD797 and only about 40% of the LM4562 (1.6pA/rt Hz max). So a 5534 is actually quieter than an AD797 or an LM4562 when used in a phono amp - but of course distortion is not as low.
I've used the LM4562 as a line stage amplifier (Av = 14 dB) and as unity gain buffers and never had a problem - its quiet in that application, has very good drive capability into low loads (600 Ohms) and ultra-low distortion. I published two preamps here and here that uses it as the line stage amp along with measurements on the one ('X-ultra mini one').
Some folks have had problems with the LM4562 - apparently they can easily get damaged if overdriven and once that happens, the distortion is much higher permanently (see diyaudio member PMA's comments on this). Personally, I never ever had a problem with the LM4562 - its a great part in my book for line stages, buffers and as a general purpose building block for things like headphone amps (with the appropriate power buffer stage), tone controls and DAC I/V applications.
I would not break into too much of a sweat over these things - I'm using all three devices in my products as well. Just make sure if you are doing a phono amp, you use a 5534 or a JEFT opamp with the same or lower noise voltage (the JFET opamp noise current will be negligible).
The good thing about all three of these devices of course is that they are all still available in DIP8 - none of the good newer devices are unfortunately.
(BTW, for general noise comparisons (NOT for phono) in straight amplifier stages, you can find a noise comparison tool here. This was developed from Steve Hagman's spread sheet and allows you to compare up to 7 opamps for noise and plot some graphs as well)
The best part for MM is still the NE5534 - you need to go to a very good JFET input opamp if you want to best it in that application and they will be a lot more expensive for the same distortion and output drive capability. The reason for the NE5534's great noise performance in MM applications is the input noise current is only 0.6pA/rtHz - half of the AD797 and only about 40% of the LM4562 (1.6pA/rt Hz max). So a 5534 is actually quieter than an AD797 or an LM4562 when used in a phono amp - but of course distortion is not as low.
I've used the LM4562 as a line stage amplifier (Av = 14 dB) and as unity gain buffers and never had a problem - its quiet in that application, has very good drive capability into low loads (600 Ohms) and ultra-low distortion. I published two preamps here and here that uses it as the line stage amp along with measurements on the one ('X-ultra mini one').
Some folks have had problems with the LM4562 - apparently they can easily get damaged if overdriven and once that happens, the distortion is much higher permanently (see diyaudio member PMA's comments on this). Personally, I never ever had a problem with the LM4562 - its a great part in my book for line stages, buffers and as a general purpose building block for things like headphone amps (with the appropriate power buffer stage), tone controls and DAC I/V applications.
I would not break into too much of a sweat over these things - I'm using all three devices in my products as well. Just make sure if you are doing a phono amp, you use a 5534 or a JEFT opamp with the same or lower noise voltage (the JFET opamp noise current will be negligible).
The good thing about all three of these devices of course is that they are all still available in DIP8 - none of the good newer devices are unfortunately.
(BTW, for general noise comparisons (NOT for phono) in straight amplifier stages, you can find a noise comparison tool here. This was developed from Steve Hagman's spread sheet and allows you to compare up to 7 opamps for noise and plot some graphs as well)
Last edited:
JCX brought it up ages ago and I think it's still a great idea for MM: gang all four channels of a OPA1644 together and enjoy ~3 nV/rtHz JFET input opamp (with the ~2x E_noise reduction and still negligible I_noise). There's a couple newer high-performance CMOS opamps now that might even be better all said and done. Haven't examined them carefully, as I'm all digital anyhow.