OK audio gurus: What are the highly regarded FET-input op amps these days?

[basinstreetdesign]

I have been using LM4562 for a while now with no complaints.  They are VERY linear, quoting an amazing 0.00003 % typ distortion, 55 MHz GBW, unity gain stable and with Ibias of only 72 nA.  They are in standard pinout in soic-8 and stk at DK at USF$2.79 for 1.

https://www.digikey.ca/products/en/integrated-circuits-ics/linear-amplifiers-instrumentation-op-amps-buffer-amps/687?FV=1f140000&k=lm4562&pkeyword=lm4562&stock=1&ColumnSort=1000011&cur=USD&lang=en

They are characterised for distortion, CMRR, PSRR, etc up the yin yang.

[Marco]

There have been audio op-amp distortion measurements and comparisons with double-blind listening tests.

Increasing the distortion by 80 dB to do a "listening" test ... is that to emulate the audio chains you will find everywhere with 200 opamp circuits?

[GregDunn]

I have several audio boxes which use either the TL07x or NE553x op amps; some of them are just low or unity gain amplifiers, others are single or double pole filters.  I've driven them with millivolt levels (phono preamp) up to 5-10V levels and they behave just fine.  All are suitably low noise for audio and transparent.  We actually did a double-blind test against a very high end tube preamp about 30 years ago and couldn't tell them apart.  Pretty much any op amp in the same category will be more than good enough if you pay attention to the spec sheet when picking them.  I think the NE5532 is unity gain stable whereas the NE5534 requires compensation?

[Mechatrommer]

I guess in audiophewlery, first there's blind test or post. And then there's double blind post, and then there's showing real circuit. I see a lot of work was done on filtering, those tantalum and ceramics caps?  and then there's discreete power stage AB topology with careful biasings. And then there's speaker box resonant calculator or whatnot. Is it worth  arguing something that contributes less than 1% from the whole system? And then there's tube transistor that sounds white, man.. i cant hear 1MHz anyway..

[Zero999]

I have several audio boxes which use either the TL07x or NE553x op amps; some of them are just low or unity gain amplifiers, others are single or double pole filters.  I've driven them with millivolt levels (phono preamp) up to 5-10V levels and they behave just fine.  All are suitably low noise for audio and transparent.  We actually did a double-blind test against a very high end tube preamp about 30 years ago and couldn't tell them apart.  Pretty much any op amp in the same category will be more than good enough if you pay attention to the spec sheet when picking them.  I think the NE5532 is unity gain stable whereas the NE5534 requires compensation?

The NE5534 is stable in gains of 5 or less, but is faster than the NE5532.

Yes, it's all a load of bollocks, the distortion produced by modern op-amps is negligible, compared to speakers and the room acoustics.

I have been using LM4562 for a while now with no complaints.  They are VERY linear, quoting an amazing 0.00003 % typ distortion, 55 MHz GBW, unity gain stable and with Ibias of only 72 nA.  They are in standard pinout in soic-8 and stk at DK at USF$2.79 for 1.

https://www.digikey.ca/products/en/integrated-circuits-ics/linear-amplifiers-instrumentation-op-amps-buffer-amps/687?FV=1f140000&k=lm4562&pkeyword=lm4562&stock=1&ColumnSort=1000011&cur=USD&lang=en

They are characterised for distortion, CMRR, PSRR, etc up the yin yang.

The LM4562 is a completely unsuitable replacement for a J-FET op-amp. It's a BJT input op-amp with a high bias current of 72nA, compared to, 65pA typical, 20nA worst case, for a J-FET op-amp such the old TL072 and a 55MHz of GBWP could be far too fast for a circuit and layout designed for one with a GBWP of 3MHz. The input current noise for the LM4562 is 1.2pA/√Hz which will make it vary noisy when the inputs are connected to a high impedance. The old TL072, with an input current noise of just 0.01pA/√2 will be less noisy in applications when the input is connected to a higher impedance than 15k.

Note that it's only specified for that very low level of distortion when in unity gain configuration and with an output voltage of 3VRMS, not that you'll notice the increased distortion at lower and higher output levels. It's just a case of the figures being manipulated to make it look good.

[CodyDowling]

I just designed in the OPAx140 from Texas Instruments. (I'm using the OPA2140, for dual op-amps, but they also have a single and a quad version, IIRC). The specs that I cared about looked good, but without knowing what you're using for it's a little hard to say if it's a good fit. I was worried about relatively low noise, relatively good accuracy, and ridiculously high input impedance with a reasonably low input capacitance. It fits nicely for that application (monitoring audio signal to feed into a DC servo), but it will depend on your application specifics.

[cvanc]

Thanks for all the comments.  A couple of you asked what the specific repair was.  It was an old piece of stereo equipment, a "Quad 34" preamplifier.  I'm guessing here, but it's maybe early 80's vintage?

The preamp is op-amp based and is populated with TLE2071 and TLE2072 chips (http://www.ti.com/lit/ds/symlink/tle2071a.pdf).  This is a part I had not encountered before, and as I said at the beginning you can still get them but they're kind of expensive.  The board has many high impedance nodes and a fairly spread out layout - not something to blindly drop a super fast modern chip into.  It would probably oscillate.

It wasn't really about improving the performance after repair.  It was about finding a replacement part that was more common and hopefully less costly.

Again, thanks for the advice and opinions.

[RandallMcRee]

Just for grins, here is the bible on opamps for low distortion applications. The 4562 *is* very good.

http://www.nanovolt.ch/resources/ic_opamps/pdf/opamp_distortion.pdf

There are some applications where these specs do matter. Samuel Groner, for example, designs oscillators with vanishingly low distortion. That is necessary when using said oscillator to measure other equipment.

[David Hess]

Consider what the LF411 and LF412 are.  This is not quite straightforward because the two parts are not directly related.  The LF411 is a single version of the LF353 dual and the LF412 is a dual version of the LF351 single.  So:

LF351(S)/LF412(D)   4MHz 13V/us 25nV/SqrtHz@1kHz
LF411(S)/LF353(D)   3MHz 15V/us 18nV/SqrtHz@1kHz 1.8mA/ch

The only thing which makes these parts suitable for audio is the low transconductance of their JFET differential inputs which allows higher bandwidth and slew rate while maintaining unity gain stability.  So replacements will need JFET inputs with class-AB outputs.

Modern jelly bean replacements are the TL051/TL071/TL081 and TL052/TL072/TL082.  The TL05x series has better precision.  The TL07x series used to be the noise tested versions of the TL08x series but today they are the same.

JFET Singles:
LF351 SO only from STM, Eww
LF356 $0.83 5MHz 12V/us
LF411 $1.29 3MHz 13V/us
TL051 $0.60 3MHz 20V/us 18nV/SqrtHz@1kHz   Improved TL071/TL081  Higher Precision
TL071 $0.55 3MHz 13V/us 18nV/SqrtHz@1kHz   Noise Tested TL081
TL081 $0.52 3MHz 13V/us 18nV/SqrtHz@1kHz   

JFET Duals:
LF353 $0.60 3MHz 13V/us
LF412 Discontinued
TL052 $0.70 3MHz 20V/us 18nV/SqrtHz@1kHz   Improved TL072/TL082   Higher Precision
TL072 $0.68 3MHz 13V/us 18nV/SqrtHz@1kHz   Noise Tested TL082
TL082 $0.75 3MHz 13V/us 18nV/SqrtHz@1kHz   

JFET Quads:
TL054 $2.62 3MHz 20V/us 18nV/SqrtHz@1kHz   Improved TL074/TL084   Higher Precision
TL074 $0.64 3MHz 13V/us 18nV/SqrtHz@1kHz   Noise Tested TL084
TL084 $0.60 3MHz 13V/us 18nV/SqrtHz@1kHz   

Somewhat improved replacements for these:
AD711/AD712/AD713
LT1792/LT1793   Single only but lower noise, lower distortion, and lower input capacitance.

In his book, Douglas Self discusses a bunch of more modern JFET parts:

OPA134/OPA2134  Not much of an improvement.
OPA604/OPA2604   Not much of an improvement.
OP627   Great improvement.

[Marco]

A great improvement towards what exactly? When even the venerable AES needs to do something extremely silly like making the circuit amplify the distortion by 80 dB to do a listening test you can easily see HiFi kinda lost the plot.

[David Hess]

A great improvement towards what exactly? When even the venerable AES needs to do something extremely silly like making the circuit amplify the distortion by 80 dB to do a listening test you can easily see HiFi kinda lost the plot.

Study the datasheets and find out.  The OP627 has lower distortion, lower noise, and higher common mode rejection then the jelly bean parts it can replace.

There are some really improved JFET parts for audio applications but I limited myself to suitable replacements for the LF411 and LF412 and parts discussed elsewhere.  The LT1113 and LT1169 come to mind.

As a practical matter, I would stick with the cheap TL07x/TL08x series unless lower noise or lower distortion in a non-inverting configuration is required and instead use better circuit topologies including unloading the output of the operational amplifier.

Linear Technology had a great discussion about non-linear input capacitance in JFET operational amplifiers causing distortion in non-inverting configurations but Analog Devices has apparently deleted it.

[Audioguru]

i cant hear 1MHz anyway..

The audiophools in the double blind listening tests might be old like me (73) with normal for their age high frequency hearing loss. Then they cannot hear missing high frequencies and cannot hear distortion and hiss anyway.

Old geezers without hearing aids cannot hear the distortion and hiss of a lousy old 4558 dual opamp. The datasheet does not say a single word about how much distortion it has and lists "typical" noise but lists no maximum amount of noise like audio opamps do.

My hearing aids boost the highs to normal without any distortion or hiss. In a noisy mall, street or restaurant their automatic noise reduction mode works very well. 

[Mechatrommer]

i'm not entirely sure what you mean, good or bad. anyway what took you so long? the OP was calling for you (look at the topic!)

[cvanc]

... anyway what took you so long? the OP was calling for you (look at the topic!)

 

[Marco]

The OP627 has lower distortion, lower noise, and higher common mode rejection then the jelly bean parts it can replace.

Do you think there is anyone in the world who could even hear the difference compared to an OPA140 in an A/B test for a realistic circuit? (ie. not the silly AES one.) Let alone a quality test.

[David Hess]

The OP627 has lower distortion, lower noise, and higher common mode rejection then the jelly bean parts it can replace.

Do you think there is anyone in the world who could even hear the difference compared to an OPA140 in an A/B test for a realistic circuit? (ie. not the silly AES one.) Let alone a quality test.

In the right application, sure it will be audible.  The OP627 will support a higher closed loop gain at the same distortion level as the TL071/TL081; raise the closed loop gain high enough and there will be an audible difference.  In high sensitivity applications, the difference in input noise will be audible.  The OP627 also maintains lower distortion with lower impedance loads but this should be solved with better circuit design; do not drive low impedance loads with operational amplifiers when distortion or precision matters.

[Marco]

Compared to an OPA140 ...

[David Hess]

The OPA140 series should be lower distortion but it is not available in DIP packages.  However I doubt the TI datasheet specifications represent reality; it looks like they used a very low source impedance for the distortion tests which is unrealistic for a JFET part.  The included schematic does not show an input cascode so distortion from non-linear input capacitance is likely but they do not discuss this.  Maybe their JFET capable complementary bipolar process solves this problem but if so, they are not saying.

Build a test circuit and do comparisons yourself.