Nice JFET op amps to replace TL072

[Kesh]

I'm moving from pedal builds to slightly more hi-fi gear, mic preamps and other home studio kit and I'm looking for a replacement for the TL07x family op amps. And in a DIP package too (sorry Analog Devices).

Bipolar op amps were dominated by the NE5532 and it's a great chip, and the LM4562/LME49720/LME49860 an obvious upgrade for that niche.

I really just want the good, but workhorse, jfet op amp that's a replacement to the TL072. A lot of op amp talk is full of audiophile nonsense that I am sceptical about. DIYSB seems less full of that guff.

[psychedelicfish]

The OPA2134 family are FET input available in DIP. The noise and distortion figures from them are ridiculously small

[merlinb]

The official upgrade to the TL072 is the TLE2072.
www.ti.com/lit/ds/symlink/tle2071a.pdf

Also, analog Devices do plent of DIP opamps, although I don't know how many are FET...

[Paul Marossy]

Also, analog Devices do plent of DIP opamps, although I don't know how many are FET...

I like their OP275. It's supposed to have a "Butler front end" and is a combo of BJT & FET. I found a page once on the net that point out that they are outperformed by the NE5532 and a couple others I can't remember right now.

[merlinb]

I like their OP275. It's supposed to have a "Butler front end" and is a combo of BJT & FET. I found a page once on the net that point out that they are outperformed by the NE5532 and a couple others I can't remember right now.

The OP275 has worse noise, distortion and drive capability than the 5532, objectively. But a lot of people do like it, presumably because of its pleasant distortion. It has a BJT input, however, so is not a suitable replacement for the TL072.

[Kesh]

The TLE-2072 doesn't look great. I just want a quality jelly bean like the 5532 but jfet.

Of course the worse voltage noise is not an issue for high impedance inputs like a 10:1 trafo on a mic pre, as current noise dominates.

[merlinb]

The TLE-2072 doesn't look great. I just want a quality jelly bean like the 5532 but jfet.

What do you dislike about the 2072? You may struggle; the TL072 is the jelly bean. Anything else is inevitably going to be less common.
OPA2134 maybe?

[wavley]

I kinda like the OP275, it's good sounding in some circuits I have, it's stock in the Boomerang+ Phrase Sampler.  Maybe I do like the sound of it's distortion, I kinda approach everything as an "Everything can be made to clip, it's about clipping gracefully" attitude. 

I have some mic pres with OPA2134 chips in them, I think they're pretty good sounding.  Of course a lot of audiophools dislike the 5532, but it's an honestly good op amp.  It always makes me giggle when I think of the hundred or so 5532's a recording passed through before making it to any format and then someone says they sound like crap in this last piece of equipment on which they are listening.

[Paul Marossy]

Of course a lot of audiophools dislike the 5532, but it's an honestly good op amp.  It always makes me giggle when I think of the hundred or so 5532's a recording passed through before making it to any format and then someone says they sound like crap in this last piece of equipment on which they are listening.

Yeah, that is funny when you think about it. That NE5532 is in a lot of pieces of equipment out there. And it's not a bad opamp AFAIAC.

[Kesh]

What do you dislike about the 2072?

voltage noise mostly. the datasheet doesn't seem to quote it at 1K, but 48 nV per root Hz typical at 10 and 12 typical at 10K, potentially more, is too much for mic preamps

to everyone else, agreed the 5532/4 is great. where is the jfet equivalent of a great, cheap and universally used op amp?

[12Bass]

ADA4627-1 is my current favorite sounding JFET op-amp, though it is single, not inexpensive, and would require a DIP adapter.  OPA1642 also sounds decent, is reasonably priced, and would also require a DIP adapter.  For an improved, general TL072 replacement, I would suggest the OPA1642.  ADA4627-1 would be more for cases where you are seeking the highest fidelity.   

[PRR]

>> dislike about the 2072?
> voltage noise mostly. ... ... too much for mic preamps

Last point first. 200 ohm sources need LARGE input devices. Large BJTs, or even larger FETs (since FETs aint as good as junction devices). While it is just-possible to get mike-level hiss with big FETs, the demand doesn't justify making chips that way. In any other industry, we'd look to change the transducer impedance; alas we have a closet full of 100r-300r microphones we must use.

Or a transformer (which has other advantages; also disadvantages). With iron, even the horrid noise-voltage of a 12AX7 can be made to vanish below mike self-noise.

Fight or don't fight. There are several chips MADE FOR transformerless mike preamps. They have the large input devices which are not needed in hardly any other application, plus the monkey-motion to extract the differential and deliver an output. Because the FAT input devices dominate a high-cost IC process, there are a zillion mike preamp plans using jellybean chips with mass-produced fat switch BJTs as inputs. IMHO the chip is best for getting-on with the business of making music, the paste-up plan is best if you make a million preamps and order two million 2N4401 low-bidder.

>> dislike about the 2072?
> voltage noise mostly. the datasheet doesn't seem to quote it at 1K, but 48 nV per root Hz typical at 10 and 12 typical at 10K, potentially more

Noise versus frequency follows some basic trends.

Hiss per root-Hz tends to be flat from somewhere in the audio band to somewhere in the radio band.

> datasheet doesn't seem to quote it at 1K

What is 1KHz hiss? In most audio stages, "all the hiss" is the top octave or so. 10KHz IS the hiss-zone you will hear most. (Phono/Tape preamps are different.)

At radio freqs the hiss may rise or fall depending on circuit; IAC, we audio-heads do not hear radio.

At low freqs hiss ALWAYS rises. It is a fundamental of the universe, often called 1/f noise. It sounds like random rumble. As implied by "1/f", it rises 6dB per octave lower. The transition from flat broadband hiss to rising 1/f random rumble may be modeled by a Bode plot. Therefore a spec of 48 at some low freq and a spec of 12 at some high freq imply a noise-corner at 48/12 = 4:1 higher than the low freq spec. In this case, 40Hz. In Bode it is 12 from 40Hz to above the audio range. Being Bode, it is really 18 at 40Hz, 14 at 80Hz, asymptotic to 12.

40Hz is a very good 1/f corner. Old Ge could be over 1KHz. Good modern clean Silicon is often 100Hz-200Hz. (I suspect all recent production is better than this but they never update the numbers on old datasheets when process improves.)

12nV/rtHz is the hiss of a 14K resistor. Yes, if you use all 10K pots and 2K feedback resistors, the jellybean FET-input opamps will stand out hissy. (Also there is no urgency about using zero input current devices.) If you use 100K pots and 50K NFB, at most settings the various resistor hisses will drown the FET hiss.

[merlinb]

voltage noise mostly. the datasheet doesn't seem to quote it at 1K, but 48 nV per root Hz typical at 10 and 12 typical at 10K, potentially more, is too much for mic preamps

It's shown in fig. 41, about 13nV/rtHz. It's close to 4dB quieter than the TL072 across the board.

http://www.uni-kl.de/elektronik-lager/416511

For a one-chip 200 ohm mic preamp solution, maybe the INA103?

http://www.ti.com/lit/ds/symlink/ina103.pdf

[R.G.]

Can you hear the difference between the TL07x and the new Belchfire HyperAmp in actual equipment and conditions?

Or does some other issue overwhelm the differences?

[Kesh]

At the moment I'm actually playing with transformer mic preamps, my intention, in my own experimental way, to come up with my own "colour" mic pre, so I am looking, at first, at the colour from the transformers themselves. High ratio trafos have more of this supposed magic colour mojo, so using the voltage noise/current noise rule of thumb has led me to jfet op amps, probably to switch to tubes or discrete jfets at a later stage of experimentation. Again, for more magic mojo colour, this time class A. But for now I want to see if I can actually hear the transformer colouration I keep reading about, so am wanting as pure a gain stage as possible.

Eventually I am looking to (hopefully) top out at over 60db of gain (my dynamics are gain hungry), so am whittling away noise at every stage. And as many classic mic pres have step down output transformers, and apparently this is the main source of colour, and necessary if I switch to tubes, I may have to produce even more gain, but we'll see what can be done. So maybe 40 to 50dB of gain from non-iron (or mu-metal) stages.

At +60dB on single chip instrumentation amp type mic pres, with higher impedance mics, I have heard significant noise below 10K: shshshsh, rather than ssssssss (and even lower, for which I can't find the onomatopoeia, maybe khkhkhkh?) and obviously this part of the spectrum is much more important in music, so I don't think it should be discounted.

So this whole thing led me to asking: where is the standard issue, not too pricey, jfet audio op amp of a decent quality, akin to the bipolar 5532, or even the 4562 that costs 70p if I get the NJM version.

Can you hear the difference between the TL07x and the new Belchfire HyperAmp in actual equipment and conditions?

Or does some other issue overwhelm the differences?

Possibly at +60dB, but of course room noise, even in a treated room, makes the whole noise thing fairly pointless, but I'm having fun.

[amptramp]

From back in my days designing low-noise preamps for infrared scanners:

With photoconductive detectors of about 50 ohms impedance, we designed an amplifier stage using differential 2N4405 PNP transistors (PNP are usually lower noise and higher voltage than the corresponding design of NPN transistors) operated from a current source and feeding an LM318 op amp.  We determined the LM318 had an input noise of 11 nV/root Hz (and this was before the manufacturer actually came out with the same number), so it was suitable as the output stage since the differential stage gain would allow the LM318 input noise to be overwhelmed by signal.  We used a 2N4405 as the current source transistor for the differential stage.  The biggest problem with the design was the compensation network since there was an additional phase shift in the feedback network.  The final design including the resistive bias source for the photoconductor came out to 0.64 nV/root Hz.  We took similar care with the voltage regulator that powered all the stages to get to a power supply noise of 33 nV/root Hz.

The 2N4405 is a plain ordinary metal-can PNP switching transistor with nothing special about it other than low noise at low input resistances.  It is optimum for about 50 ohms but remains usable with 100 - 300 ohm inputs.

National later came out with a low-noise NPN differential pair rated around 1 nV/root Hz which would have made it all much simpler.  With the LM318 (and the NE5534), the input stage collectors are taken to the offset adjust pins.  You can simply take both inputs to V-, shutting off the input stage of the op amp and connect the NPN collectors to the offset pins, meaning you can replace their input stage with your low-noise one, and since you have the same number of stages, you don't have a compensation problem.

Low-noise design isn't what I would call fun, but when you do it right, it is satisfying.

[amptramp]

BTW where the National LF356 spec sheet says it gets 16 nV/root Hz, don't believe it.  The number we measured was 20 nV/root Hz, meaning it is slightly worse than a TL071.  We did our own tests on that one.

[R.G.]

I don't know if  they still make the LM381,  but it was a no-holds-barred low noise preamp design from way back. There may be lower noise special cases, but it was truly remarkable back when. There are probably  better one solutions now, but for a part with history and lots of apps, it might be a solution to look at.

[PRR]

> the LM381,  but it was a no-holds-barred low noise

It's long-gone. TI got the design but marks it OBSOLETE and further has mis-filed it under "battery fuel gauge".

It's hard to find. It was aimed at HIGH-volume uses: cassette decks. When was the last time you bought a cassette player? (I know when I did: a month before the WTC fell I bought a new car, one of the last with a cass player, and I bet it warnt no LM381.)

LM381 is about 1uV total hiss in 20KHz bandwidth.... fine for several-Kohm uses but far from the lowest voltage-noise (OR current-noise) part around.

I keep looking at it though. It was good but far from fabulous. Mostly an opamp with a pair of 2N3904 in front. For lowest-hiss you hadda skip one of the pair. The internal bias ran it at a good op-point for cassette-head hiss. There was a trick to shift that about 4:1, so you could nail a narrow range of medium impedances. You can't run it leaner for hi-Z, and the devices are not fat enough for low-Z. (NatSemi did publish a transformerless mike amp with several LM381 but their audio suggestions were often clueless.) Noise corner is specced way up at 1KHz, which is poor for tape/phono work. Output stage is pretty lame, you need something more to get out of the box. Gain change can't be too extreme because it goes unstable below Gv=10, and because the input has teeny common-mode range.

It IS a consumer tape-head amp. For rational head design, when tape is moving the oxide hiss will overwhelm amplifier hiss, so it doesn't have to be "perfect" to be "flawless".

At the same time the LM394 super-transistor was a MUCH better start toward a low-hiss input. at similar current the '394 was about 3dB less hiss, and you could bias ANY way you needed for low hiss from above 100K to well below 1K source. Yes, it cost more, and yes you need a lot more for a preamp. But it was good enough to inspire the Jensen 990. (This could be what you were thinking of?)

> PNP are usually lower noise ...than the corresponding design of NPN transistors

True in theory. Rarely seen in practice. Perhaps the slim advantage in theory is overwhelmed by processing variations. Many (maybe most) successful low-hiss designs start NPN. It seems to be a non-factor in audio.

[Kesh]

But it was good enough to inspire the Jensen 990.

and graeme cohen's preamp. which was my last build.