Help me pick an op amp to order

[N8Nagel]

So as you see here, over the weekend I found an issue with my TP-117 preamp and also found that the issue was a failed op amp on the input board, barely past the RCA jacks

http://audiokarma.org/forums/index.php?threads/tp-117-missing-channel-on-tape-1.813433/

So obviously the smart thing to do would be to order a couple NJM4558DDs the next time I have to order from Mouser or Digi-Key to replace the one I stole out of the busted unit and have a couple spares for the future. Or is it? A little research indicates that the 4558 isn't particularly well regarded, and here

http://nwavguy.blogspot.com/2011/08/op-amp-measurements.html

which I actually remember reading with interest previously, but didn't think it would be completely relevant to my life any time soon. However, it is - while being apparently a fairly serious comparison of op amps, the author puts the 4558 in a category titled "DON’T BOTHER: While the following parts had their glory days, they’re long gone. They offer no advantage over the above parts and are often much worse" and dismisses it out of hand without even bothering to measure a sample. Well, fair enough - HFE lists the TP-117 as having been made in 1989 and 1990 so there's that.

So this preamp is not a period correct restoration. It's a $20 thrift store find that happens to sound really good. If I can level up for less than $10 why shouldn't I? I guess the question is what would be the obvious choice, and do I need to worry about stability when increasing slew rate? Schematic is here:

https://www.hifiengine.com/manual_library/luxman/tp-117.shtml


it does appear that there are small capacitors (22uF/16V) between pins 4 and 8 (rails) and ground at a few points throughout the input board, but not on every op amp. Is that close enough or does that need to happen on every IC if I upgrade?

Oddly, some of the op amps in similar positions are M5238P not NJM4558D. Same question applies there I guess. Appears that CD (IC302), and V-DISC (IC306) are the M5238Ps if I am reading it correctly. Any comments on those "while I'm in there", ideas why those are different, and if there is an upgrade for those as well?

Sorry for the dumb questions, I'm a little bit out of my depth here, and thanks for any help. Who could think that one little component failure could push me down a rabbit hole like this...

 

[N8Nagel]

Looked up the data sheet for the M5238P

http://pdf.datasheetcatalog.com/datasheet/MitsubishiElectricCorporation/mXrxqrq.pdf

Looks like that is a much higher performance chip than the others, maybe Lux put them in those two positions because they were concentrating on high quality digital reproduction? but looking at the data sheet that guy is rated at a slew rate of 20 uV/s which is way higher than required, if those are stable in TP-117 pretty much anything should be yes?

 

[Retrovert]

OP: I was going to refer you to NwAvGuy. Congratulations, you've already found the expert who knows what he's talking about.

Looks like that is a much higher performance chip than the others, maybe Lux put them in those two positions because they were concentrating on high quality digital reproduction?

Many times opamps are selected based upon the lower-noise performance in higher-gain applications, and the higher-speed comes along for the ride. The circuit design must ensure the opamp is stable across the entire frequency range which may be presented, with filters to remove the unwanted HF components. I'd expect Luxman to get this right, but would verify it with a scope to be certain.

but looking at the data sheet that guy is rated at a slew rate of 20 uV/s which is way higher than required, if those are stable in TP-117 pretty much anything should be yes?

You hit the nail on the head by raising the stability issue.

Oscillation is a common problem with higher slew rate opamps, and dropping in faster parts often reveals weaknesses in the original circuit.

This is the big issue with random substitutions: supersonic oscillation, typically in the MHz range. Most never look for this and thus never see it. Stability is more than just randomly plugging in IC packages and seeing which "sounds" better. Many times that "improved" sound is merely bucket loads of oscillation and distortion.

TI somewhere has a big caveat about random opamp substitutions and states, for excellent reasons, that it won't assist audiophiles with such upgrades.

 

[N8Nagel]

So I guess the question is what would you use? Obviously NwAvGuy has some slightly different priorities than I do as I've got +/-15V rails and don't really care about power consumption so long as my PS doesn't blow up.

LM4562 looks plausible, but would really like an opinion from someone more knowledgeable as I have a little knowledge and a lot of hunch behind picking that part. Would that be a suitable chip to replace all of the op amps? There's a headphone driver one as well, that's a different part yet, and while I don't really see myself ever using this as a headamp, maybe I will someday?

 

[Bob]

couple of points. there were many, many designs based on the 4558. and these have
been around for decades. different internal circuits (and designs) can result in a better
opamp. but there are so many factors in their use.

unity gain stable/unstable means you'd have to look at phono preamps and the
EQ filters that lower gain down at high frequencies.

lots of preamps don't decouple the power supply lines like the mention above
of the 22uf so you might consider doing this (and possibly the individual
power lines).

a while back, there was chatter about running opamps in Class A. artificial
but since we're rolling opamps, why not?

I'd start with long pin sockets.(to allow wiring decoupling caps). then
before decoupling try various (lm4562, 2164, etc) opamps for the
sound you like then add decoupling in one channel to "hear"
any effects. of course, trim pins if they can short to chassis underneath.

you're somewhat lucky if the unit can supply enough power. I fixed/upgraded
a Dahlquist DQ-LP1 but could not change any of the opamps due to limited
power available due to limited space for a bigger power transformer
(search upgrading dahlquist dq-lp1)

I have several projects that I will be changing out opamps for better sound.

 

[dr*audio]

I've been using the LM4562 with excellent results for a few years now. It is unity gain stable so no worries about oscillation.

 

[N8Nagel]

I've been using the LM4562 with excellent results for a few years now. It is unity gain stable so no worries about oscillation.

Would you only replace the 4558s or consider replacing some of the other ones as well?

this really just started because I'm short one 4558, and after a little searching saw it wasn't particularly well regarded. If for a few cents more I can get an upgrade, however, I'm totally open to that.

 

[Retrovert]

NwAvGuy points out that most of the opamps are identical.

As you have found out, the difference between opamps comes down to noise, speed, bandwidth, sensitivity, and power. All are tradeoffs between the others. So you need moderate sensitivity, low bandwidth, low power output, low power consumption, low speed, and low noise.

The 4558 is slow, about 1 V/uSec with a few MHz bandwidth. Modern amplifiers are about twenty to fifty times faster with about ten to twenty times the bandwidth. That's a recipe for oscillation.

I can't give you a good substitute based on experience. This is a complicated area. A lot of what I know about these is for guitar effects where distortion is a beneficial side effect.

Many opamps, like the LM4562, a much faster device, will oscillate in older circuits. You'll need to add snubber capacitors and tinker with the feedback to prevent oscillation. A quick search for the LM4562 suggests it oscillates without capacitor compensation. Not a shock.

I can point you to diyAudio where an EE can give you an informed opinion for a suitable substitution.

Again, don't just plug in a faster opamp as it will oscillate at supersonic frequencies, typically in the MHz.

I would find an exact replacement and be done with it.

 

[Retrovert]

A lot of the opamp rolling is mindless substitution: this one fails, this one sounds bad, this one fails, this one fails, this one sort of works, this one fails, ahhhh, this one doesn't sound like crap so it's perfect!

You might want to look into the NE5532.

Any old, slow, low bandwidth opamp should be fine. Pokey is good. Noisy not so good, but one takes what one can get.

 

[N8Nagel]

A lot of the opamp rolling is mindless substitution: this one fails, this one sounds bad, this one fails, this one fails, this one sort of works, this one fails, ahhhh, this one doesn't sound like crap so it's perfect!

You might want to look into the NE5532.

Any old, slow, low bandwidth opamp should be fine. Pokey is good. Noisy not so good, but one takes what one can get.

Well, the *easy* answer is that a NJM4558DD is still available from Mouser for cheap. But I'm the guy that shoved a built 289 in a '55 Studebaker coupe with a 4-speed and a 3.73:1 LSD rear, so you see where I'm coming from here. Just my habit all my life; if something breaks, you replace it with the forged/high compression/larger diameter/mandrel bent/whatever part.

If I tried something "better" would a 100 MHz scope be good enough to tell if I were having issues? How would I check?

 

[Retrovert]

I hear you on the car upgrades, but the problem here is the analogy breaks down: engines don't oscillate the way opamps do.

I am a big fan of upgrades. I replace all of the resistors, capacitors, tube sockets, and wiring. Plus I polish the chassis.

In this case, however, upgrades are to be feared.

You can see the oscillation on a scope far out of your passband. Ringing, too. Look for overshoot/undershoot at the edges of the square wave. Spectrum analyzer is a lot easier. Another trick is to use a high-pass filter above 30 kHz and see what shows up. If you had no oscillation there would be no signal. Also the current usage will increase because the oscillation is sucking up huge amounts of power, and this increases the package temperature. I've touched oscillating opamps while using a scope and they're very hot. Another trick is to use an AM radio. This, BTW, is an easy way to hear diode Qrr noise.

Just don't use actual speakers, especially tweeters, until you've removed all of the HF oscillation.

The question is: are you prepared to hunt down the oscillation and rejigger the circuit to eliminate it?

 

[dr*audio]

You have to look at the circuit and the datasheet for the opamp. The LM4562 is unity gain stable and it is stable into a capacitive load up to 100pF. If you drive a higher capacitance than that you may need to install a resistor in series with the output. This is usually on the order of 100 Ohm to 1K. In the datasheet they show it used in a RIAA phono equalization preamp with a 470 Ohm resistor on the output. I would say you are safe in replacing any of the opamps in the unit with a LM4562, but do it one at a time and check the output of the opamp with a scope to make sure it isn't oscillating. I never had one do it, but it's best to be sure. Look at the schematic, there may already be a resistor in series with the opamp output in some circuits. If so, you're safe in installing the LM4562. Here is a link to the datasheet.
http://www.ti.com/lit/ds/symlink/lm4562.pdf
This isn't voodoo, just good engineering.

 

[Powertech]

If you are going to swap, I would second the NE5532 which is quiet and stable. You could use OPA2143 which has a JFET front end and is very quiet. Decouple each supply rail to ground with a 47 - 100uF electrolytic cap and then fit a 100nF ceramic cap directly across pins 4 & 8 as a supply shunt. The 5532 does exhibit a small DC offset at it's output and so should have a small (10uF) electro in series with it's output to stop volume pots from going scratchy. All final OP amps should have a small series cap (about 2nF) in series and a series resistor of about 100ohms to cope with connection cable capacitance. In terms of circuit stability, I would not go for unity gain. A small gain as a minimum on each stage tends to reduce the chance of unity gain oscillation.

Good luck.

 

[Retrovert]

I would say you are safe in replacing any of the opamps in the unit with a LM4562, but do it one at a time and check the output of the opamp with a scope to make sure it isn't oscillating. I never had one do it, but it's best to be sure.

Ummm, I don't think so. Those two parts are not equivalent. The LM4562 is widely reported to oscillate when used in place of the 4558 and for good reason. (Just search using Google if you don't believe me.)

Compare the specifications from the datasheets:

4558
Slew Rate: ±1,7 V/us
Gain Bandwidth Product: 3 MHz

LM4562
Slew Rate: ±20 V/us
Gain Bandwidth Product: 55 MHz​

Summary:

EIGHTEEN TIMES THE BANDWIDTH!
TWELVE TIMES THE SLEW RATE!​

Oscillation is therefore a major risk.

The original circuit was not designed for a fast part with high bandwidth. The behavior with such a part is therefore unknown. This is why the opamp rollers routinely destabilize circuits and blow up opamps.

The fact that oscillation or ringing were not seen does not mean it does not exist on a timescale not searched for. As I pointed out, the current draw is a good indicator of whether or not a circuit is stable, and the oscilloscope often looks clean when the part is not enough to burn a finger.

 

[Retrovert]

If you are going to swap, I would second the NE5532 which is quiet and stable.

Still a bit fast:

Slew Rate: ±9 V/us
Gain Bandwidth Product: 10 MHz​

But, nonetheless, far slower than the others.

You could use OPA2143 which has a JFET front end and is very quiet.

Yeah, I've seen that suggestion.

NwAvGuy doesn't like those JFET opamps and I'd trust his analysis. I wish he'd resume posting, his work was always spot on and highly educational.

Decouple each supply rail to ground with a 47 - 100uF electrolytic cap and then fit a 100nF ceramic cap directly across pins 4 & 8 as a supply shunt.

That is good practice for any opamp. Probably missing in the original. The feedback loop probably needs some HF compensation as well.

I have a small headphone amplifier I need to modify much like this. No decoupling, feedback is wrong, HF and anti-oscillation measures missing. Amazing the muntzing done to save money even when the result sounds terrible. Nobody notices, or so I conclude by the fact that the design has not changed in years.

All final OP amps should have a small series cap (about 2nF) in series and a series resistor of about 100ohms to cope with connection cable capacitance. In terms of circuit stability, I would not go for unity gain. A small gain as a minimum on each stage tends to reduce the chance of unity gain oscillation.

All good advice.

I wouldn't tinker with the original circuit until I understood it. That's the other side of opamp replacement.

 

[Retrovert]

The key to remember when performing substitutions with faster parts is that high-bandwidth amplifiers, be they solid state or vacuum tube, can amplify oscillators inadvertently created from delays in the feedback loop or from parasitics.

Consider the effects of tiny capacitance from wiring and leads combined with tiny inductance from board traces, leads, and wiring. This can add up to 5 pF, sometimes more. Who cares about that, right? Yeah, lots of designers. Talk to the designers of computer motherboards and cell phones.

In an audio circuit phase lag can create oscillators upwards of 10 MHz. You can't hear it, and the slow opamp can't amplify it, but that fast part? It's all over it. Yeah, baby, it's cold outside so stay here where the opamps are oscillating and it's toasty warm.

Even vacuum tubes—a very high bandwidth amplifier—suffer from this problem, where Miller capacitance can create oscillators in the AM band. The solution is adding stopper resistors which ruin the Q of the circuit. The same problem exists in all sorts of high-bandwidth circuits.

Just because one doesn't look for a problem doesn't mean it doesn't exist. Didn't you see that movie? You know the one I mean. The one where the oscillation monster is the basement? And it's awakened and stomps the amplifier on its way to Tokyo? Yeah, that movie.