Minus 300 volts on the cathodes (and, presumably, somewhere around zero volts common-mode on the grids) until the 12AX7 was up to full temperature?
Did these 12AX7s have extraordinarily robust heater-cathode insulation? The 1962 RCA datasheet shows a maximum heater-cathode voltage of +/- 200 volts.
-Gnobuddy
Perhaps it might be useful to give an audio summary of op amps and their place in hi fi reproduction.
First, we did NOT use op amps, because DC coupling was very difficult, and inefficient. Yes, op amps were sometimes necessary for instrumentation and analog computers, but not really for audio. We just used cap or transformer coupled stages, for the first 50 years of audio reproduction. For example, the Marantz 7,8,9,10 tube products did NOT use op amps. Just cap coupled or transformer coupled stages.
Op amps were derived from a separate need in industry, that first started with tubes (like Bear put up), then solid state discrete (like Dick Burwen of Analog Devices designed in the '60' s, and finally the first IC op amps developed by Bob Widlar at Fairchild in the mid 60's. At first, they seemed like a miracle! The negative feedback reduced the distortion to almost unmeasurable, they were tiny, DC coupled, and flexible. One of my first jobs in 1966, was to fully characterize a group of uA709 op amps for a military project. Within their limits, they were pretty darn good, and we thought, that once they were affordable (they could cost up to $100ea at the time) they 'might' do audio pretty well. A few years later, in 1968, the uA741 came out, and it became a standard building block for servos, etc at Ampex where I worked. It was simple, convenient, overload proof, and SLOW! It would just barely pass an audio signal, but this did not stop people from using it, because it saved time, space and money. (IC's had gotten down to a few dollars each by now)
Slew rate was measurable, but at the time its problem (TIM) was underestimated, and people actually made tape recorder and studio mixer electronics with uA741 op amps. This was partially due to the test equipment used, usually SMPTE IM, at the time. It simple did NOT show the problems with slew rate limiting. Later, when the HP 339, and the ST1700 THD analyzers were made available, TIM (SID) was more easily measurable, and the uA741 and its equivalents were shown to be embarrassingly full of distortion. Now, there were other slightly better IC's available. One that we used since 1970, was the Harris 911, with 2.5V/us slew rate. This was an approved IC by Dick Burwen, to replace discrete modules, but it was expensive, about $5.00 each--the equivalent of 1 hour of an engineer's salary, at the time.
Now, were we now in op amp heaven? Not really, but enough for now.
First, we did NOT use op amps, because DC coupling was very difficult, and inefficient. Yes, op amps were sometimes necessary for instrumentation and analog computers, but not really for audio. We just used cap or transformer coupled stages, for the first 50 years of audio reproduction. For example, the Marantz 7,8,9,10 tube products did NOT use op amps. Just cap coupled or transformer coupled stages.
Op amps were derived from a separate need in industry, that first started with tubes (like Bear put up), then solid state discrete (like Dick Burwen of Analog Devices designed in the '60' s, and finally the first IC op amps developed by Bob Widlar at Fairchild in the mid 60's. At first, they seemed like a miracle! The negative feedback reduced the distortion to almost unmeasurable, they were tiny, DC coupled, and flexible. One of my first jobs in 1966, was to fully characterize a group of uA709 op amps for a military project. Within their limits, they were pretty darn good, and we thought, that once they were affordable (they could cost up to $100ea at the time) they 'might' do audio pretty well. A few years later, in 1968, the uA741 came out, and it became a standard building block for servos, etc at Ampex where I worked. It was simple, convenient, overload proof, and SLOW! It would just barely pass an audio signal, but this did not stop people from using it, because it saved time, space and money. (IC's had gotten down to a few dollars each by now)
Slew rate was measurable, but at the time its problem (TIM) was underestimated, and people actually made tape recorder and studio mixer electronics with uA741 op amps. This was partially due to the test equipment used, usually SMPTE IM, at the time. It simple did NOT show the problems with slew rate limiting. Later, when the HP 339, and the ST1700 THD analyzers were made available, TIM (SID) was more easily measurable, and the uA741 and its equivalents were shown to be embarrassingly full of distortion. Now, there were other slightly better IC's available. One that we used since 1970, was the Harris 911, with 2.5V/us slew rate. This was an approved IC by Dick Burwen, to replace discrete modules, but it was expensive, about $5.00 each--the equivalent of 1 hour of an engineer's salary, at the time.
Now, were we now in op amp heaven? Not really, but enough for now.
Part 2. Back in 1970 or so, both Dick Burwen and I independently started to use HA-911 IC op amps for professional audio products. Dick designed the first audio products (mixing board and pro preamp) for Mark Levinson, while I attempted to use the same basic devices in a mixing board for the Grateful Dead. My attempt failed after a tryout with the band, but Dick Burwen's designs were more successful. Today, I attribute that to Burwen's use of BETTER PASSIVE PARTS, and selecting the 911's to remove those with excess x-over distortion (there were a large percentage).
By the time that I re-met Mark Levinson after a 5 year hiatus from first meeting him, he had these Dick Burwen designed products for sale. I, however, after failing to make the grade in 1971, thinking that the IC was the major problem, had returned to making fast discrete designs, (both OP-AMP and Transamp topologies) following the guidelines that Matti Otala had put forth since 1970.
This forced me to make simpler, faster designs, with a high open loop bandwidth, rather than concentrate on open loop gain. Of course, the low open loop gain made distortion more difficult to eliminate, so I used the most elaborate topologies like complementary differential jfet input, to get the lowest open loop distortion that was practical. I worried about every active part and its inherent linearity, switching to jfets when I could. The result was OK distortion, very low noise, and very high slew rate, e.g. 100V/us for the line amp.
Originally, Mark Levinson just made these line modules of my design for the Grateful Dead, but then he plugged them into his preamp gain stages and realized that they did sound better than even the best selected IC's at the time. That is when he stopped using the Burwen designed and built modules with the selected 911 IC inside, and went to my class A, very fast discrete designs, and the Levinson JC-2 preamp was born. While the line amp for the JC-2 was a trans-amp, the phono gain stage was a true discrete OP AMP. This preamp really took on Audio Research, the best tube equipment being made at the time, giving people a solid state choice, without having to keep their old Marantz or Mac tube preamps, or buy the Audio Research. This product really went far, but it has been bettered over the decades. Enough for now.
By the time that I re-met Mark Levinson after a 5 year hiatus from first meeting him, he had these Dick Burwen designed products for sale. I, however, after failing to make the grade in 1971, thinking that the IC was the major problem, had returned to making fast discrete designs, (both OP-AMP and Transamp topologies) following the guidelines that Matti Otala had put forth since 1970.
This forced me to make simpler, faster designs, with a high open loop bandwidth, rather than concentrate on open loop gain. Of course, the low open loop gain made distortion more difficult to eliminate, so I used the most elaborate topologies like complementary differential jfet input, to get the lowest open loop distortion that was practical. I worried about every active part and its inherent linearity, switching to jfets when I could. The result was OK distortion, very low noise, and very high slew rate, e.g. 100V/us for the line amp.
Originally, Mark Levinson just made these line modules of my design for the Grateful Dead, but then he plugged them into his preamp gain stages and realized that they did sound better than even the best selected IC's at the time. That is when he stopped using the Burwen designed and built modules with the selected 911 IC inside, and went to my class A, very fast discrete designs, and the Levinson JC-2 preamp was born. While the line amp for the JC-2 was a trans-amp, the phono gain stage was a true discrete OP AMP. This preamp really took on Audio Research, the best tube equipment being made at the time, giving people a solid state choice, without having to keep their old Marantz or Mac tube preamps, or buy the Audio Research. This product really went far, but it has been bettered over the decades. Enough for now.
Part 3. Now, what seems 'wrong' with OP AMPS, either discrete or IC?
First, the op amp topology, generally has one dominant internal rolloff (pole) and the open loop bandwidth can be less than 1Hz to over 10KHz, depending on the design. In general, all op amps get better and faster with increased unity gain bandwidth. For example, the first IC's might have had a 1MHz unity gain bandwidth, yet today's roughly equivalent IC's (for the task) might have a gain bandwidth of 30MHz or more. All else being equal, this increased the effective slew rate proportionally, so 10-20V/us is relatively easy and cheap to achieve, compared to 1/2 to 2V/us of the older devices. One confounding problem is that the effective slew rate is effected by both the gain bandwidth AND the input stage transconductance. Yet, when you reduce the input stage transconductance, the input noise goes up. A great example of an optimum design is the AD797, designed by Scott Wurcer. It has the lowest noise that is practical, yet 20V/us slew rate. By the standard of slew rate, this should be enough, so why isn't everything made with this formula? Well, one thing that is left out, is the OPEN LOOP BANDWIDTH.
This is the heart of the controversy. IS open loop bandwidth really important for audio quality? Only the ear can really decide.
What you need is two IC's: one like the AD797, with a low open loop bandwidth, and a part like the AD825, that has a similar overall topology, but has a HIGH open loop bandwidth. Is there a subjective difference? Some think that there is. Only open listening tests will tell the difference. Now what about discrete designs, well, we can use alternate topologies that might be even more linear, and even operate open loop if necessary. This is what I chose for my best designs. Discrete designs can use fully complementary jfet devices virtually everywhere, and this makes the difference, along with true class A output. What is the trouble with op amps? Low open loop bandwidth, it would appear.
First, the op amp topology, generally has one dominant internal rolloff (pole) and the open loop bandwidth can be less than 1Hz to over 10KHz, depending on the design. In general, all op amps get better and faster with increased unity gain bandwidth. For example, the first IC's might have had a 1MHz unity gain bandwidth, yet today's roughly equivalent IC's (for the task) might have a gain bandwidth of 30MHz or more. All else being equal, this increased the effective slew rate proportionally, so 10-20V/us is relatively easy and cheap to achieve, compared to 1/2 to 2V/us of the older devices. One confounding problem is that the effective slew rate is effected by both the gain bandwidth AND the input stage transconductance. Yet, when you reduce the input stage transconductance, the input noise goes up. A great example of an optimum design is the AD797, designed by Scott Wurcer. It has the lowest noise that is practical, yet 20V/us slew rate. By the standard of slew rate, this should be enough, so why isn't everything made with this formula? Well, one thing that is left out, is the OPEN LOOP BANDWIDTH.
This is the heart of the controversy. IS open loop bandwidth really important for audio quality? Only the ear can really decide.
What you need is two IC's: one like the AD797, with a low open loop bandwidth, and a part like the AD825, that has a similar overall topology, but has a HIGH open loop bandwidth. Is there a subjective difference? Some think that there is. Only open listening tests will tell the difference. Now what about discrete designs, well, we can use alternate topologies that might be even more linear, and even operate open loop if necessary. This is what I chose for my best designs. Discrete designs can use fully complementary jfet devices virtually everywhere, and this makes the difference, along with true class A output. What is the trouble with op amps? Low open loop bandwidth, it would appear.
Realistically, the most demanding audio signal an op-amp can be called upon to output (assuming conventional +/- 15 volt rails, or thereabouts) is a 30 volt peak-to-peak sine wave at a frequency of 20 kHz.
Simple calculus shows that the corresponding required maximum slew-rate is 1.88 volts per microsecond. Tops.
So that old Harris 911, if it met its slew-rate spec, was actually capable of handling full-power, full-audio-bandwidth signals, with a little safety margin, as far as slew-rate goes. I'm not familiar with this device, and don't know about its other audio performance problems (distortion, noise, etc), if any.
Since op-amps, like other semiconductors, come with rather loose parameter tolerances, it makes sense to specify a device with a little more slew-rate safety margin, if it doesn't cost too much.
Well, by the late 1970s, IIRC, you could buy a TLO71/ 72/ 74 with a specified slew rate of 13 volts per microsecond. That was six times the highest slew rate you would ever need for audio use. A whopping 600% safety margin - you would never need more than that. The TLO series also addressed other audio requirements, such as reduced crossover distortion.
The TLO series op-amps were, to all intents, perfect for most line-level audio use.
There was still a (very) little room left for improvement. It was possible to do a little better in the noise department than the TLO series when dealing with very small signals, such as the output of a cassette tape head.
Signetics launched the NE5534 / 5532 opamps, and that was pretty much it. They had the same 13 V/uS slew rate as the TLO7x series opamps, but a few decibels (about 13 dB) lower input noise. These were, for all practical purposes, perfect for any normal audio use.
There have been spec improvements since then, but they are irrelevant for general audio - you cannot hear the improvements, because the 5532/5534 is already better than the human ear. Keep the input signal in the audio band (meaning, if there is garbage up at 50 kHz in the signal, filter it out!), and the 5532 cannot be audibly improved upon.
This is a tough pill for most of us to swallow, especially because of all the "New And Improved!" advertising we're constantly subjected to. We believe everything can be improved upon, and newer is usually better.
But some things really are already so perfect they can't, in any practical sense, be improved upon. The chambered Nautilus is so perfectly suited to it's way of life that it has remained unchanged for at least 400 million years, and maybe 500 million. Many faster, sleeker, newer life forms have come and gone since the Nautilus began slowly propelling itself through primordial oceans.
The NE5534 / 5532 is the Nautilus of the audio op-amp world; so perfect for normal audio use that they can't be audibly improved upon. Bean-counters will find ways to make them cheaper, chemists will find ways to make them use less toxic materials et cetera. But audibly, in any normal home audio use, they can't be improved upon. (If you work for NASA and are trying to pick up the incredibly weak signals from Voyager, then the 5532 may not be what you want.)
And even the TLO7x series is still perfectly usable whenever you don't need the very best noise performance.
What is wrong with opamps? For audio use, nothing; culturally, the only thing wrong, is that nowadays many of us look these incredible gift-horses in the mouth, lacking an appreciation of how perfect a solution they are.
-Gnobuddy
Here is what Douglas Self had to say in an EE Times article:
-Gnobuddy
Here's a link to the full article: Op amps in small-signal audio design - Part 1: Op amp history, properties | EE Times
-Gnobuddy
Gnobuddy, you really think that it is that easy to decide on an audio op amp? Sorry, but we have decades of experience that show that the 5534, TLO-72, etc, may be good, but not GREAT audio gain stages. You have ignored the research done by Walt Jung, Matti Otala and many others, who spent years measuring different op amps. However, you have to LISTEN for the differences, rather than assume that all is well.
As far as I can tell, Otala was a twit who apparently didn't understand that if you feed an amp signals that are too fast for it, it will fail to reproduce them.
He also apparently failed to understand the limits of the human ear, or of audio reproduction equipment. They don't put out 100kHz square waves, and if they do, they are defective.
He had his fifteen minutes of fame with engineers, followed by the widespread realization that a simple lowpass filter on the input of any reasonably well-designed audio amp would put an end to TID. After that, Otala's TID seems to have been banished to the usual place where bad engineering or scientific ideas slip away to die quietly in the dark.
Jung? I definitely didn't ignore him. I remember reading his "Op-Amp Cookbook" decades ago, and good reading it was, too. It's been a long time, and I've read a lot of other books on op-amps before and after that, but I seem to remember seeing lots of interesting and elegant circuits in Jung's book.
-Gnobuddy
What's truly amazing is the human mind's ability to conjure up ghosts, goblins, and other imaginary things, and then to believe in them despite all evidence to the contrary.
"Mediocre" essentially means "average"; and that is exactly what the normal, average, human mind does. We've been doing it for a couple of hundred thousand years, too.
We can't escape the limitations of our human brains; but our greatest geniuses have devised a way to tell when we are fooling ourselves, and when we are not.
That way is the scientific method. Abandon it, and you are back to casting entrails, reading tea-leaves, and imagining that op-amps have invisible, unmeasurable defects that only your own golden ears can detect.
It is subjectivism and the abandonment of the scientific method that leads to mediocrity, my friend. There are thousands of years of human history to prove it. And no amount of indignant scorn from you can change that. 🙂
And now I wish you all a very fine evening, while I wander off to admire the little collection of wonderfully virtually-perfect 5532 op-amps in my parts-box. 😀
-Gnobuddy
As far as I can tell, Otala was a twit who apparently didn't understand that if you feed an amp signals that are too fast for it, it will fail to reproduce them.
He also apparently failed to understand the limits of the human ear, or of audio reproduction equipment. They don't put out 100kHz square waves, and if they do, they are defective.
He had his fifteen minutes of fame with engineers, followed by the widespread realization that a simple lowpass filter on the input of any reasonably well-designed audio amp would put an end to TID. After that, Otala's TID seems to have been banished to the usual place where bad engineering or scientific ideas slip away to die quietly in the dark.
Jung? I definitely didn't ignore him. I remember reading his "Op-Amp Cookbook" decades ago, and good reading it was, too. It's been a long time, and I've read a lot of other books on op-amps before and after that, but I seem to remember seeing lots of interesting and elegant circuits in Jung's book.
-Gnobuddy
Not to affect your rant, but degenerating the input pair does wonders for eliminating TID. In that sense, it was a useful identification. But it surely has taken on a life of its own thereafter.
And one nice thing about all these new fanci-pants opamps is, provided proper decoupling, they tend to enjoy even greater stability than their predecessors. And (generally) need less juice to do their job. Who doesn't like a bit more stability!? (Yes, I'm sure someone will come along with a poorly chosen opamp for its respective application to argue a point contrary)
I guess that math only works for rockets, lasers, medical devices, etc.
What's truly amazing is the human capacity for self-delusion.
Here we go again! '-) It is a real shame that someone like Dr. Matti Otala, who put so much time into research into audio quality can be considered a 'twit'. Flawed maybe, but not a twit.
If I listened to the likes of you "scientists" I would have just hung it up and not did Lounge at all. I truly understand how Jim Williams of Audio Upgrades became so bitter.
The real shame is that criticizing the use of 741's for audio had a place 40yr. ago and some folks can't let go of it. The world has moved on, get over it. The open-loop BW stuff has no basis in reality.
Robert, why the bitterness yourself? Those very "scientist" you're kicking to the curb are some of the very same individuals that enabled your very career. Or, for that, matter, my own. Perhaps some gratitude of how we got to where we are?
At the same time, acknowledging the demands of your input/output and designing accordingly is hardly mediocre. Throwing 1000x the needed specification necessary to pass an undistorted signal doesn't make it any more linear. At a certain point, we're hard constrained by thermal noise (both in terms of electrons bouncing around randomly and discrete collisions of atoms against our microphone diaphragms) in terms of recording and playback. But almost infinitely more so by our ears and brains. The electronics side of the equation can become so much better than our brain and ears that there's nothing left to gain. Marginal or otherwise. Yes, there's plenty of engineering required to make sure gremlins do not constrain one's ability to linearly pass the audio signal.
Never mind starting with a highly compromised medium such as vinyl, even acknowledging its aesthetic appeal.
A photon travelling ~3E8 m/s is not mediocre. It's hard capped by reality.
None of this is to say your pursuit is unwarranted. But also no need to be nasty. Sheesh.
At the same time, acknowledging the demands of your input/output and designing accordingly is hardly mediocre. Throwing 1000x the needed specification necessary to pass an undistorted signal doesn't make it any more linear. At a certain point, we're hard constrained by thermal noise (both in terms of electrons bouncing around randomly and discrete collisions of atoms against our microphone diaphragms) in terms of recording and playback. But almost infinitely more so by our ears and brains. The electronics side of the equation can become so much better than our brain and ears that there's nothing left to gain. Marginal or otherwise. Yes, there's plenty of engineering required to make sure gremlins do not constrain one's ability to linearly pass the audio signal.
Never mind starting with a highly compromised medium such as vinyl, even acknowledging its aesthetic appeal.
A photon travelling ~3E8 m/s is not mediocre. It's hard capped by reality.
None of this is to say your pursuit is unwarranted. But also no need to be nasty. Sheesh.
Last edited:
- Status
- Not open for further replies.