As we continue our blaze down the path of discovery, we’re frequently tasked with the laborious effort of comparing and contrasting spec sheets and running tests for various components in applied circuits, in search of that hidden “Holy Grail” component that outperforms all the rest. Then, when our eyes are glowing two shades of blood-shot red from near endless nights of reading notes and spec sheets and staring without blinking at a computer screen as the Audio Precision test equipment spits out it’s report, we’re often left with a pile of data that still means relatively little to us until we ultimately plug the gear in and listen to it. Occasionally, we’ll find something that truly stands out above the rest, with an heir of “almost too good to be true” clouding our minds with suspicion. It behooves us to dig deeper for truth, even as we case the joint for any Nazis that may have snuck in to try and rob us of our hard-fought quarry (did I mention we recently moved into an abandoned Aztec temple teaming with booby traps). So we tend to keep things under lock and key for some time until a suitable amount of trial runs and beta-testing clear our conscience enough to venture releasing such sacred and forbidden knowledge into the world.
One thing that has blown our head-gaskets of late, is the elusive topic slew rate and our studies into its impact on audio signal processing. Slew rate is defined as “the change of voltage or current, or any other electrical quantity, per unit of time.” In other words, it’s the measure, in volts per second, of how quickly an electronic component can react to and re-produce a shift in voltage for an audio signal that’s being passed through it. Typically, a faster slew-rate represents a higher fidelity and a more accurate recreation of the original source audio that’s being processed through the component or circuit. However, in application, slew-rate isn’t always maintained with results you would expect. And in general, the spec is widely misunderstood by the consumer market. Thus, we felt it might be time to visit this topic for the sake of steering customers in the proper path regarding modifications.
For the sake of this blog post, we’ll draw our attention to the TL072, which is a low-noise dual JFET-input operational amplifier available from Texas Instruments. This op-amp has been a work-horse in the world of solid-state signal processing since the late 70s. It’s been used in many applications from mic pres to summing mixers to guitar pedals to amplifiers to you name it. It boasts, a slew-rate of 13 V/μs (volts per microsecond), meaning it can effectively recreate a shift from 0 volts to up to 13 volts (on average) per microsecond (1 one-millionth of a second). The human ear has been documented to possess a hearing range of approximately 20Hz to 20kHz, which means it can react to and discern sound waves that measure 20 cycles or oscillations per second to up to 20,000 cycles per second. More realistically, the spectrum is slightly narrower in range for some people only going up to between 15kHz and 18kHz. The range tends to meet its upper limit sooner as the aging process and hearing loss from overexposure to high sound-pressure levels (SPLs) reduces our hearing sensitivity over time (but all those KISS concerts and excessive jamming in the garage with amps dialed to “11” were probably worth it... not to mention one time you took your dad’s .44 magum out for target practice and forgot to wear hearing protection). So, for every 13 volts of swing the TL072 is capable of reproducing within the interval of a millionth of a second (1,000,000 Hz or 1MHz), the human ear is only really able to discern a small portion of the chip’s processing capability (roughly 1/55th based on a 15kHz to 18kHz average range). It’s largely subjective as to how much we perceive otherwise as our body and central nervous system may perceive on a broader spectrum that’s not entirely understood. Not to mention, it’s also likely the amplifier and loudspeakers or headphones being used to reference said audio are limiting the picture way before that even...
Realistically, the rest of the circuit really can’t be forgotten, as we find more often than not, an op-amp that clocks a faster slew rate will often introduce high-frequency oscillations, often ultrasonic or above the normal human hearing range. These oscillations may also be termed Radio Frequency (RF) oscillations as they occupy a range occupied by radio wave communications (20kHz - 300 GHz). Regardless of what you call these oscillations, they not only pollute the signal pathway with noise but eat up precious headroom that lowers your mix ceiling and brings unwanted color from distortion into the equation. In fact, most circuit designs with op amps that create issues like this, must feature what’s known as compensation capacitors in the feedback loop between the op amp output stage and one of its differential inputs. These compensation caps, running in parallel with a feedback resistor in the loop, create what’s known as an resistor/capacitor low-pass filter (R/C LPF).
As most of you may know, a low-pass filter (LPF) effectively filters or suppresses frequencies above a certain threshold, while allowing all other frequencies below the threshold to pass by unhindered. This is much like a high-cut (low-pass) filter in an equalizer circuit or plug-in. The actual frequency threshold in the LPF is largely determined by the value of the capacitor, as capacitors introduce capacitive reactance into the equation. Capacitive reactance is the measure of a capacitors opposition to alternating-current (AC) which comprises the audio signal. This reactance is also a function of frequency, meaning that depending upon the capacitive value (in Farads), the relationship with allowed frequencies and opposed frequencies may vary. In other words, compensation caps are meant to tame the oscillating beast and make it more presentable at the “dinner table.” They filter and suppress unwanted oscillations like those caused by a faster slew rate. Usually, the smaller the cap value, the higher the frequencies it reacts to and opposes. In the example pictured, they’re using a 100 nano-Farad cap in parallel with the 100kohm feedback resistor. Unfortunately, in the process of taming the beast, they also tend to negate most of the benefits of a faster slew rate as the high-frequency content the chip may be capable of reproducing due to fast slew rate, can never go higher than what the R/C LPF is filtering off. Not to mention, these R/C LPFs also negatively affect things like phase response. Phase response, which is a function of frequency, largely has to do with how accurately an audio processor can reproduce signals from input to output without letting one range of frequencies slip out-of-phase and end up sounding too “loose” or “flubby” (these are scientific terms… at least I didn’t say “too warm”). This has a big impact on our perception of transients. So you start trying to tie bows around things to make it sound prettier, you may end up strangling something else. All us first-time parents and/or dog-owners have to learn that lesson the hard way, but our later children and pets will thank us for it (I’m joking, of course). Luckily, the one thing adding a compensation cap or R/C LPF will improve is your headroom, because those high-frequencies, whether audio or not, can still consume precious current that depletes your amp circuit of headroom, and distortion may not be too far off.
All that being said, a fast slew rate may sound ideal, but it’s not always going to make a world of difference you think it will, and it may even hurt your overall specs unless you know how to balance things out by making additional changes. In general, we’ve found that unless a circuit is specifically designed around a fast-slew rate chip, with the intent of reducing the RF oscillations without tossing the benefits out the window, it’s hard to warrant using a faster chip that’s going to break your bank to get there. I mean, it’s kind of like building a rocket car capable of breaking the sound-barrier, but then putting a governor on the engine that limits its speed to 55 mph, and then restricting it to public roads where the speeds are even more strictly enforced. Now you almost know how Batman feels, sans the emotional anguish of losing your parents to cold-blooded murder amidst addrenaline fueled attemps to curb those mood swings through masked, viligante justice. But, hey! I’m more of a Spider-man guy myself...
There’s lots to consider… so you must carefully consider these options before making your choice in designs and/or modifications, otherwise the aging knight in the temple may say of you:
He probably couldn’t hear too well through that chain-mail cowl anyway, but it’s always best to defer to the judgement of elders when forging a path of discover through a forbidden temple.
In strictly summing conditions, there are other considerations to make for slew rate. You may be summing 8 to 48 channels or more through a single summing node (ie a TL072) and the shifting voltages of multiple channels may elicit a need for a faster slew rate so as to register and recreate the unique sub-divisions between signal sources on the individual channels without blurring the picture and producing a muffled, increasingly incoherent output signal that sounds like garbage. But once again, the circuits inability to maintain that rate or reproduce due to LPFs and what not, as well as the human ear’s inability to discern much more than the 18kHz really leaves one wondering what are we really gaining. Consider the numerous time-honored applications of the TL072 would provide that it’s slew rate really hasn’t been brought much into question with regards to amplification and/or summing applications. That being said, we have a question:
Why would you need a faster slew-rate?
Or more importantly...
Just how slow of a slew-rate can you get away with before it starts to obscure your overall sonic picture?
We’ve even discussed this in depth with the engineers at Texas Instruments, and they confirmed our suspicions. You don’t really need to go faster in most audio applications. But the advances to design, engineering, and manufacturing of op amps since their early introduction in the late 60s produce better specs all around, so we just deal with it by building or changing the circuit to compensate. As is the case with the Burr Brown OPA2134, available from Texas Instruments, which boasts a slew rate of 20 volts per microsecond, it often needs additional compensation caps and filtering, which is okay because the THD+N, noisefloor, headroom, and crosstalk specs all see improvements with relatively few (if any) unmanageable side-effects… but there are still better options (more below).
As for the slowest rate you can maintain, it’s hard to say. Most audio op amps don’t go that slow and hover around 15V/us, and the ones that do run slower don’t sound bad. In fact, the highly sought-after discrete opamps like the API 2520 produced by API for use in their coveted circuits, only produce 1.5 volts per microsecond. That’s quite a bit less than even our example in the TL072. Yet, people pay $60-80 a pop to drop these into their circuits. And I don’t think I’ve heard any complaints yet with API as they’re on the up-and-up last time I checked. In fact, if you look at the maths (plural), you’ll find that most RF filtering in circuits applied in the form of compensation caps, often reduces the slew rate down to as little as 1 volt per microsecond anyway.
So does slew rate matter much? Not much, from what we’re finding. Of course, we may eventually find a slice of humble pie waiting for us in the back of the fridge once we devour the other leftovers, but the results as of late are kind of stacked against slew rate. And one thing seems apparent: if your audiophile bias for specs leads you down a trail of skepticism away from op amps that appear compromised in the slew rate arena, you’re possibly throwing out the baby with the bathwater. In fact, if you can manage a slower slew rate op amp, you can often remove compensation in the form of a R/C LPF from the circuit and restore a little sweetness to the once bitter waters. We even found a new chip from Texas Instruments called the OPA2209 that has a slew rate of 6.4 volts per microsecond, almost half of the TL072, yet it trounces the TL072 and
just about every other chip in its price-range with regards to THD+N, noise, crosstalk, etc. It’s even a slight improvement over the OPA2134. It’s got all that without creating some of the same unwanted RF oscillations the others are responsible for, which allows you to remove some of the compensation caps to improve your phase response. In fact, it’s even got a relatively low “DC-offset.” Op amps are DC-coupled, which means that in addition to passing the alternating current the audio signal is on, they can also pass direct-current (used to power the active components like opamps and transistors), which can cause unwanted shifts in the signal between power rails and ground and unwanted noise and artifacts through your speakers. Circuit designers combat DC-offset following the outputs of op amps by installing coupling caps that allow the audio’s AC signal to pass while blocking any DC-offset remaining on the op amps output. A lower DC-offset may allow for the occasional removal of AC-coupling caps after the op amp in the circuit path, which will have an even better impact on your overall phase response if you do this at the appropriate times. There are lots of things to consider when doing this, and it does require some fine-tuning. However, the benefits of the chip can’t be denied.
The only catch with the OPA2209 or it’s single-sided counterpart, the OPA209, are surface mount chips. Meaning, unless you get creative with install, you won’t be able to mount those in a footprint designed for through-hole components like most vintage analog mixers afford. Luckily, we came up with our own solution: a through-hole DIP-8 surface mount chip adapter we are calling the Rock-it Socket.
I’d also like to make another note about modifications, specifically with regards to decoupling caps. Decoupling caps are meant to do one things: to eliminate the possibility of AC-noise generated in the power supply or cross-talked from other signal pathways, from ever reaching the IC or op amps power inputs. Decoupling caps allow a safe pathway for AC-noise to be shunted directly to ground, and are usually installed between a power rail connection such as the power pins on an IC and a dedicated ground connection. We’ve found that applying additional decoupling caps or even compensation caps to errant IC chips and op amps, can often be a delicate process, as you’re essentially installing caps on the solder-side of the board where there was never any footprint laid on in the circuit for them originally, and you’re tethering them between two points with about a million exposed contacts around them. These caps, if not properly insulated and tacked down, can short out against the legs and terminals for other components or signal pathways, or be snagged and knocked off the circuit entirely, which means fried “doo-doo” in your mix, amidst the rising smell of burnt components. Additionally, it often takes a lot of time to dress these caps and solder them in place, getting things nice and neat so there’s no problems down the road. Imagine applying two or more caps per chip on a console teaming with over three-hundred opamps! It’s a daunting task.
Luckily, the Rock-it socket we designed has surface mount caps for decoupling already in place and all you have to do is apply a solder dot to the proper terminal or install a quick wire jumper, and the Rock-it Socket will tie a surface mount decoupling cap between the power pins of the IC and a dedicated ground connection.
We’ve found the OPA2209s (adapted and decoupled with our Rock-it Sockets) outperform just about every other chip we’ve tested or used before, regardless of which position you place it in. So whether it’s a mic preamp or a line driver or an input buffer or a pan or fader buffer, or you name it, the OPA2209 creates better specs in the circuits it’s used in 95% of the time. Just goes to show that slower isn’t necessarily worse. We’re happy to offer the Rock-it Socketed OPA2209 in newer revisions of our console modifications, as well as via a stand-alone product if you’d like to perform your own upgrades.
So just to recap, faster slew rates look nice, but often don’t add up to much. The process for building components has just become more sophisticated with components being produced with tighter-tolerances, allowing for an ever-increasing ceiling for specs as well as a fixed replication of specs from one component to the next. We can all agree that “Better is BETTER,” but just because it looks good on paper, doesn’t mean it’s always going to be superior, especially if the results are only noticeable to the ears of a gnat, let alone high-dollar test equipment that measures on a spectrum far above what humans can discern… but if you’re mixing for a porpoise (which can hear as high as 150kHz) and referencing through special underwater speakers that can reproduce above 100kHz, that faster slew rate might be nice (he said facetiously). Otherwise the slew rate specs don’t always mean too much to us, kind of like the credentials of revered automotive technician and decorated Marine, Gomer Pyle. Mayberry still loves you Gomer, but if I were working HR, you wouldn’t have made it past the first interview... Also, Goobers need not apply.