It's powered by 2 AAA cells, so I suppose my answer is yes.
You've gotten me more interested in creating a schematic of the thing. It is, BTW, a Transcend brand 5-band EQ HPA from the '80's.Why? I said I wouldn't use an amp intended for general purpose instrumentation for audio. I don't even care for the TL072 for audio.
The LM358/324 app notes tell you specifically to use a pull down resistor from the output to Vcc- to get rid of the crossover distortion. Without it, it is intolerable. With it, it's not a low power op amp anymore.
The ONE GOOD THING a 358 has going for it is the buffered PNP input stage. Inputs will common mode to the negative rail (or GND with a single supply). That feature is needed way too often in peripheral support circuits, and real audio grade op amps just won't do it.
The ONE GOOD THING a 358 has going for it is the buffered PNP input stage. Inputs will common mode to the negative rail (or GND with a single supply). That feature is needed way too often in peripheral support circuits, and real audio grade op amps just won't do it.
Given the title of the thread, which did not specify any application niches, I think that the LM324/358 is still a contender for worst opamp on the planet. Given the price, it's probably also the most commonly used.
The fact that it includes V-/gnd in its common mode input range is most likely a reason why the LM324/358 so widely used/copied. I use it all the time to roll circuits for SMPS CVCC controllers in my day job. I would not let it near any of my usual home projects. Other responders to this thread have echoed that sentiment more or less vehemently (though not necessarily in agreement with my basic premise). Other builders of audio circuits have used it (I give them the benefit of the doubt) without considering its weak points (many!!!!).
I actually found it somewhat amusing that one could slather some lipstick on this notoriously bristly pig by throwing a bias resistor between output and gnd/V- to introduce the output stage to the unprecedented condition of Class AB bias. I'll have to search out the particular app note where that dodge is mentioned. I've only seen that option applied to far, far, better opamps.
The fact that it includes V-/gnd in its common mode input range is most likely a reason why the LM324/358 so widely used/copied. I use it all the time to roll circuits for SMPS CVCC controllers in my day job. I would not let it near any of my usual home projects. Other responders to this thread have echoed that sentiment more or less vehemently (though not necessarily in agreement with my basic premise). Other builders of audio circuits have used it (I give them the benefit of the doubt) without considering its weak points (many!!!!).
I actually found it somewhat amusing that one could slather some lipstick on this notoriously bristly pig by throwing a bias resistor between output and gnd/V- to introduce the output stage to the unprecedented condition of Class AB bias. I'll have to search out the particular app note where that dodge is mentioned. I've only seen that option applied to far, far, better opamps.
The improvement is much greater on the 358 than it is with audiophile op amps. With these better op amps, any differences show up in the 4th decimal place if at all. With the 358 it shows up in the 1st decimal place. It's like somebody fixed it.
4558's don't let me sense power line zero crossings without a negative supply. The 358 does. That makes the 4558 "worse". The 358 with a bias resistor is on par with a 4558 without when it comes to sound. And why would I even use up my stash of LF412's in the EQ section of a DJ light controller, when 358's do the job just as well Triacs aren't going to be upset by a little cossover distortion, or going into slew limit on the highs.
4558's don't let me sense power line zero crossings without a negative supply. The 358 does. That makes the 4558 "worse". The 358 with a bias resistor is on par with a 4558 without when it comes to sound. And why would I even use up my stash of LF412's in the EQ section of a DJ light controller, when 358's do the job just as well
Triacs aren't going to be upset by a little cossover distortion, or going into slew limit on the highs.
I'd guess, you using it wrong. Did you ever compared your believings with proper measurements?
I think the same... until I listen a YBA CD Special. The opamps are the 4558, and the circuit are cheap!, the only components that YBA have improved are the output capacitors.
All other components seems of low quality, but the sound is glorious!. Listen and see this CD have that I think that is equal, or more important, the elimination of the electro-mechanical noise than the circuit and the quality of the components; in the inside the work are made for reduce vibrations with felts and a mastic, and the use of good cabling.
If the sound is so good, good, good and the opamps are the 4558, the problem can't be the opamps; other problems originate the bad sound. Listen a CD Special and think for yourself.
Regards
All other components seems of low quality, but the sound is glorious!. Listen and see this CD have that I think that is equal, or more important, the elimination of the electro-mechanical noise than the circuit and the quality of the components; in the inside the work are made for reduce vibrations with felts and a mastic, and the use of good cabling.
If the sound is so good, good, good and the opamps are the 4558, the problem can't be the opamps; other problems originate the bad sound. Listen a CD Special and think for yourself.
Regards
Much of the problem with op-amps is advertising/datasheet claims that it works for Most applications, including many it isn't good at.
Another problem is any pre-conceived notion of exactly how you'll set the gain is guaranteed to cause sub-optimal performance; because, for a given application, one particular model of op-amp will only deliver optimal performance at its gain sweet spot, which depends on its stability needs and that figure will never match the figure that you actually wanted.
To say that simply:
Forcing it to do what you want, instead of what it needs, guarantees bad performance.
In that case, swapping the op-amp is a good idea, because there is probably a different model op-amp that has a performance sweet spot much closer to the needs of your particular application.
The 4558 can perform well, the 4580 can perform fantastically, but neither do a good job outside of a very narrow range of parameters. That is true of all amplifiers. Fortunately, they are all different, so it is possible to select a different one that comes closer to doing the job you wanted done.
Now, let's see a few problems:
Screeching tone? The gain is too low--replace with something compensated for lower gain.
Wonky response that goes to crap at high output? Linearity is insufficient, so either use it as a parallel amp or replace with something that has higher current capacity.
Dull as can be so that imaging/soundstage is bad although tone is good? The gain is too high, so replace with a faster or more nervous op-amp that actually requires that much gain.
The above problem-resolution comment assumes a well working power circuit with clean and sufficient voltage (which may not be the case).
P.S. Instead of replacing the op-amp, it is also possible to modify the support circuit to be suitable to the op-amp you already have. For some of the more difficult op-amps, the prospect is doable but overly time consuming, in which case, choosing a more suitable op-amp could be a more expedient solution.
P.P.S. In cases like these, where a given model of op-amp is often used wrongly, I think that we should suspect a very bad inaccurate datasheet.
Another problem is any pre-conceived notion of exactly how you'll set the gain is guaranteed to cause sub-optimal performance; because, for a given application, one particular model of op-amp will only deliver optimal performance at its gain sweet spot, which depends on its stability needs and that figure will never match the figure that you actually wanted.
To say that simply:
Forcing it to do what you want, instead of what it needs, guarantees bad performance.
In that case, swapping the op-amp is a good idea, because there is probably a different model op-amp that has a performance sweet spot much closer to the needs of your particular application.
The 4558 can perform well, the 4580 can perform fantastically, but neither do a good job outside of a very narrow range of parameters. That is true of all amplifiers. Fortunately, they are all different, so it is possible to select a different one that comes closer to doing the job you wanted done.
Now, let's see a few problems:
Screeching tone? The gain is too low--replace with something compensated for lower gain.
Wonky response that goes to crap at high output? Linearity is insufficient, so either use it as a parallel amp or replace with something that has higher current capacity.
Dull as can be so that imaging/soundstage is bad although tone is good? The gain is too high, so replace with a faster or more nervous op-amp that actually requires that much gain.
The above problem-resolution comment assumes a well working power circuit with clean and sufficient voltage (which may not be the case).
P.S. Instead of replacing the op-amp, it is also possible to modify the support circuit to be suitable to the op-amp you already have. For some of the more difficult op-amps, the prospect is doable but overly time consuming, in which case, choosing a more suitable op-amp could be a more expedient solution.
P.P.S. In cases like these, where a given model of op-amp is often used wrongly, I think that we should suspect a very bad inaccurate datasheet.
Interesting info.
How do you determine the gain 'sweet spot' for particular opamps ?.
Dan.
Look the Bode plot on the datasheet.
Don't these 4558 make good unity gain buffers? Include it in a nested feedback scheme, with a "better" op amp in the front end, and then the circuit doesn't "sound" like a 4558 any more. That's a popular DIY headphone amplifier.
Well, except that a '4558 isn't a particularly good load driver (all of them, though some are worse than others). That's why people eventually released 4560 and 4565 types, which are usually preferred unless you have a very sloppy circuit layout indeed. Beyond these, low noise was all the rage, so low-Ib inputs got dropped in even beefier successor parts. Extra-high supply voltage versions rated for up to +/-22 V were only available for the NJM4558/4560 (DX version), and anecdotal evidence suggests that these ratings were a bit of a stretch.
That being said, NJM4558s have very commonly been used for line in/out buffers indeed. My old '97 vintage Onkyo ProLogic HTR is stuffed with NJM4558Ls... guess SIP package parts were cheap because noone wanted 'em anymore. One of these days I should upgrade the post-volume pot (50k) 16.5 dB amp. There's an NJM2068 available in SIP but I have no idea how these fare for input impedance distortion; otherwise the NJM4565L may be the best bet.
Re: LM324 replacement, I'd want to reverse-engineer the circuit first. A quad suggests that maybe we're looking at a balanced input stage plus an output amp with discrete buffer? Resistor values and layout would be interesting, also whether the buffer is Class A, AB or (gasp!) B. Decent quads are few and far between, but I might look at the trusty MC33079. Low gain tends to keep the distortion of its mimotic output stage in check pretty well and ought to suffice when a decent buffer stage is present; some Class A bias may well be worth a shot if needed. Resistor values may need to be lowered as well. Otherwise it's generally considered a well-behaved part, unlike the temperamental LM837 which is best avoided. There's an LME49740 as well (4562 family quad), which would be a nice modern part but quite possibly too "hot" for a layout from way back when.
That being said, NJM4558s have very commonly been used for line in/out buffers indeed. My old '97 vintage Onkyo ProLogic HTR is stuffed with NJM4558Ls... guess SIP package parts were cheap because noone wanted 'em anymore. One of these days I should upgrade the post-volume pot (50k) 16.5 dB amp. There's an NJM2068 available in SIP but I have no idea how these fare for input impedance distortion; otherwise the NJM4565L may be the best bet.
Re: LM324 replacement, I'd want to reverse-engineer the circuit first. A quad suggests that maybe we're looking at a balanced input stage plus an output amp with discrete buffer? Resistor values and layout would be interesting, also whether the buffer is Class A, AB or (gasp!) B. Decent quads are few and far between, but I might look at the trusty MC33079. Low gain tends to keep the distortion of its mimotic output stage in check pretty well and ought to suffice when a decent buffer stage is present; some Class A bias may well be worth a shot if needed. Resistor values may need to be lowered as well. Otherwise it's generally considered a well-behaved part, unlike the temperamental LM837 which is best avoided. There's an LME49740 as well (4562 family quad), which would be a nice modern part but quite possibly too "hot" for a layout from way back when.
The broadly favorable M-audio Revo 5.1 sound card had Two 4580's set up as parallel amplifiers, to drive loads better, resulting in literally award winning sound quality. That is also possible with the 4558.
That does actually double the cost for resistors and chips, because there are twice as many of each. Its even kind of funny because in that case, the difference between junk output or hi-fi output, is about 15 cents.
The majority of chip amplifiers of all sizes are current limited, so this news just happens to apply to most chips.
With better-performing, more powerful types being readily available, I don't know why one would go to the trouble of paralleling 4558s though. One NJM4556A would have to be as powerful as, uh, at least 4 of them?
But speaking of parallel 4580s, who can spot the goofs in this headphone driver?
But speaking of parallel 4580s, who can spot the goofs in this headphone driver?
Last edited:
I think you're confusing the 4558 with the 4556, which is commonly used as a unity gain buffer in headphone amplifiers due to its 70mA output current.
I think the 4580 can sound quite good in the right application(s), but I've never cared for the sound of the 4556.
I see people here all the time paralleling 4556's to drive a certain brand/model of headphones.
They can parallel all the 4556's they want, but the sound quality still won't be in the same class of chips like the TPA6120.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.