A question about the dreaded LM358

[Jmariner]

I understand what crossover distortion is, but came across some bench tests that I found a bit confusing because the chip is still loathed in guitar use, and potentially music equipment in general.
Apart from the dead spot due to the crossover distortion, is it unrealistic to say the spikes in harmonic content can be heftily exploited and clipped off for highly rich, even order harmonic content? The asymmetry is even there from step 1.



In stock schematics provided by mfr., it's not uncommon to see a load resistor across the output so the device functions in Class-A, and this greatly cleans up the major harmonic defect, and ultimately, there are still far superior and cheap choices for general audio needs. So, is there any justifiable potential in regard to only involving it as a clipping stage, and working off exploiting those huge peaks with an abundance of even order harmonics? I assume there are better ways to accomplish this, but it does seem like a convenient tactic to shrink down a high gain, hard clipped stage.

I'm not looking for "lazy" topologies to avoid digging into the science, and grasp the challenges I have placed on myself. That is a subject I have had the bad habit of expressing here, but it is a predicament fairly embarrassing to me lol. Ultimately, simple or complex in architecture, would it be the definition of polishing a turd? lol

[ElectricDruid]

I understand what crossover distortion is, but came across some bench tests that I found a bit confusing because the chip is still loathed in guitar use, and potentially music equipment in general.

For me, the most useful charateristic of this op-amp is that it can drive its output very close to ground (definitely within 20mV, typically within 5mV). This is often useful for stuff like LFOs or envelope followers, and that's mostly where I'd use it. For audio, there are better options *if you want hi-fi*.

Apart from the dead spot due to the crossover distortion, is it unrealistic to say the spikes in harmonic content can be heftily exploited and clipped off for highly rich, even order harmonic content? The asymmetry is even there from step 1.

I'm not seeing anything "even order" there. That's a full spectrum of harmonics. How could you "clip off a spike in harmonic content"? I'm not understanding you, sorry.

In stock schematics provided by mfr., it's not uncommon to see a load resistor across the output so the device functions in Class-A, and this greatly cleans up the major harmonic defect, and ultimately, there are still far superior and cheap choices for general audio needs. So, is there any justifiable potential in regard to only involving it as a clipping stage, and working off exploiting those huge peaks with an abundance of even order harmonics? I assume there are better ways to accomplish this, but it does seem like a convenient tactic to shrink down a high gain, hard clipped stage.

It also has a dismal-by-modern-standards slew rate, which is one of the principle characteristics of the LM308 which everyone raves about because of the RAT distortion.  So it might be great or awful, depending what you want.

I'm not looking for "lazy" topologies to avoid digging into the science, and grasp the challenges I have placed on myself. That is a subject I have had the bad habit of expressing here, but it is a predicament fairly embarrassing to me lol. Ultimately, simple or complex in architecture, would it be the definition of polishing a turd? lol

No, there's no such thing as a turd in this context. People have made some cool stuff using those "terrible" voice-changer chips from kid's toys, for example. The audio quality is shocking, but that's the point - it takes a perfectly reasonable signal and adds a ton of lo-fi character and distortion.

The real questions are "Does it do what I want?" and "Does it sound good, according to my own personal sonic aesthetic?". Those questions can't be answered by either datasheets or anyone else. You just have to get stuck in and try it.

[Jmariner]

You are correct, there's both even and odd order harmonic peaks, but I simply mean clipping it to reduce the extreme magnitude of the peaks. Without hearing it, I'm guessing there would be something abnormal about the sound and texture. The dB gain in those peaks is insane. In that test and others I have reviewed, it showed a load resistor to drive the chip into class-A operation reduces distortion gradually as the load resistance is decreased, but the current it sees by the time most would find it acceptable for general audio use is quite ridiculous at just over 5mA lol.

My thoughts were as you increase gain through the feedback loop, it will probably become quite unpleasant quickly without soft clipping it in the feedback loop, or at least something in the output scheme like a voltage divider that diodes tie into. My favorite op amp based distortion I came up with many years ago runs in Class-A, and I've never bothered to put it on my scope to see if I'm making a compromise to it's performance for the bruit gain it produced, and huge increase in bass. I have stuck with running an NE5532 in that build, but a couple of the more expensive chips sound a tiny bit better in it. With this being the case, I thought getting some of these to tinker with was worthwhile.

The chips initial appeal to me was it is cheap, single supply and can be run on an incredibly low D.C. as you also mentioned. In general, it sure looks like it will be a lot of fun to experiment with.
I went ahead and ordered some so soon enough, I will hear one to know if it's an obvious amount of distortion content, or if it's more so obvious on test equipment. 50 for about $3 on the slow boat from China was too tempting. One thing I remember about sampling op-amps in a circuit you use often is how awful some turn out to be. It's probably been 20 years since I heard one, but I still remember the god-awful hum of the TL082 in a couple of my commercial half-stack heads. I haven't seen them around in my parts for well over a decade, so I very well may have pitched them lol.

[duck_arse]

50 for 3$? from china? can't you find them somewhere closer, more reliable? also, why would TL0xx's hum? the 08x was the better specced part, as well.

[ElectricDruid]

50 for 3$? from china? can't you find them somewhere closer, more reliable? also, why would TL0xx's hum? the 08x was the better specced part, as well.

Wasn't the 08x series the "normal" ones, with the 07x the "low noise" ones, and 06x the "low current" ones?

It's been so long, I don't remember any more. I just use 072 for everything unless there's a good reason not to (sometimes there is!).

But yeah, totally agree - TL082 shouldn't hum. Something else was up.

[amptramp]

I keep reading that the TL07X series is a low-noise version of the TL08X but the noise specs are identical - 18 nV/(SQRT Hz) and 0.01 pA/(SQRT Hz).  The TL08X series has slightly higher worst case input offset voltage, lower typical but higher maximum input bias current, slightly worse CMRR, slightly higher supply current, lower typical open loop gain, faster rise time,

The mystery to me is why TI produced two separate series with such closely matching characteristics.  Maybe one foundry couldn't quite get the specs of the other, but I see no difference in noise performance between them.

[duck_arse]

well, now I don't know if I was told it wrong, or I was remembering it wrong.

The JFET-input operational amplifiers in the TL07xx series are similar to the TL08x series, with low input bias and offset currents, and a fast slew rate. The low harmonic distortion and low noise make the TL07xx series ideally suited for high-fidelity and audio preamplifier applications.

The TL06x (TL061, TL062, and TL064) family of industry-standard operational amplifiers (op amps) mirror the TL07x and TL08x family of op amps with lower power consumption. These devices provide outstanding value for cost-sensitive applications, featuring high input impedance, wide bandwidth, high slew rate, and low input offset and input bias currents.

then:

The TL081, TL081A and TL081B are high speed J–FET input single operational amplifiers incorporating well matched, high voltage J–FET and bipolar transistors in a monolithic integrated circuit.

now:

The TL08xH family (TL081H, TL082H, and TL084H) is the next-generation family of the industry standard TL08x (TL081, TL082, and TL084) high-voltage general purpose amplifiers. These devices provide outstanding value for cost-sensitive applications requiring high slew rate with high voltage signals, such as motor drive and inverter systems.

[Kipper4]

Well I love em.
Horses for courses.
Led drivers
LFO
Envelope detectors.

Wouldn't be my first choice for audio unless it's all I had.
I bought a bucket load when they were a couple of bucks per 100 units.

[PRR]

The mystery to me is how anybody can "dread" a 19 cent chip.

[R.G.]

You are correct, there's both even and odd order harmonic peaks, but I simply mean clipping it to reduce the extreme magnitude of the peaks.

I may be misunderstanding your statement, but no, you can't clip harmonic peaks. That plot seems to be a spectrum of harmonics produced when (notionally at least) feeding the thing a low distortion sine. That is, a pure(ish) sine wave would appear as one peak at the fundamental. Distortion products show up at multiples of the sine frequency. The peaks in the spectrum can't be "clipped" as they are frequency domain, not time domain. Clipping the input in time domain will introduce even higher peaks in the frequency domain plot.
Time domain response is fundamentally different from frequency/spectrum domain.

Without hearing it, I'm guessing there would be something abnormal about the sound and texture.

Crossover distortion is unpleasant sounding, all right. But at low levels it's more of a non-specific unpleasantness than its own sound.

The dB gain in those peaks is insane.

If I'm understanding you correctly, you may be mistaken about what that plot means. I know that I can't tell what it means. It looks like a plot of a fundamental at 1kHz and its harmonics; the fundamental seems to be about 105db(ish), with a second harmonic down by about 50db from the fundamental and the third and many others down by about 40db from the fundamental. I can't tell much more about it without more information, like what the input level is, what the reference 0db level is, and so on. But I don't think the harmonic peaks can be thought of as gain peaks.
"db" isn't a gain, it's a ratio; more properly, ten or twenty times the logarithm of the ratio of two things. A voltage ratio of ten times (as in, gain of ten) could be described as 20db, for instance. A voltage divider causing a voltage drop of a factor of two (that is, 1/2) can be expressed as -6db.

My thoughts were as you increase gain through the feedback loop, it will probably become quite unpleasant quickly without soft clipping it in the feedback loop, or at least something in the output scheme like a voltage divider that diodes tie into.

Again, you have to watch time domain versus frequency domain. Soft clipping can only be done in time domain. It will only increase distortion products, never decrease the distortion that the device itself does.
There is a competing effect - the opamp's reduction of its own open-loop distortion by feedback. At low frequencies, the total distortion is effectively reduced by negative feedback. The distortion gets quite low below 1K, for instance. But at higher frequencies, the opamp's open loop gain is lowered to keep it stable, so there is less gain available to reduce distortion. So crossover (and other) distortions are not as effectively reduced at higher frequencies. But clipping in the feedback loop won't reduce distortion, only increase it.