Being discrete

May 28, 2018
 by Paul McGowan

In yesterday's post, I described the differences we heard between op-amps. The TL072 with the FET input softened the sound at the expense of detail, forcing us back to the original 709C. It was a bit more aggressive than we liked but it didn't lose anything.

We wondered if it wouldn't be possible to have the best of both worlds. Or, maybe even something altogether better, and that's when we rolled our sleeves up and started to really understand the op-amp. And what we learned changed everything.

We use the term "op-amp" as a generic reference to an integrated circuit linear amplifier. But, that is not correct. In fact, the term op-amp is shorthand for a specific type of circuit, the Operational Amplifier. From Wikipedia:

An operational amplifier (often op-amp or opamp) is a DC-coupled high-gain electronic voltage amplifier with a differential input and, usually, a single-ended output. In this configuration, an op-amp produces an output potential (relative to circuit ground) that is typically hundreds of thousands of times larger than the potential difference between its input terminals. Operational amplifiers had their origins in analog computers, where they were used to perform mathematical operations in many linear, non-linear, and frequency-dependent circuits.

The popularity of the op-amp as a building block in analog circuits is due to its versatility. By using negative feedback, the characteristics of an op-amp circuit, its gain, input and output impedance, bandwidth etc. are determined by external components and have little dependence on temperature coefficients or manufacturing variations in the op-amp itself.

What's all this gobbledygook mean? Well, to understand that (as Stan and I had to so many decades ago) it's necessary to wrap our arms around the op-amp's basic circuit. That circuit is relatively simple. We can make an op-amp with as few as three active devices: a pair of transistors (or tubes) for the input differential pair, and a single transistor for the output gain stage. That's it. If we want to further fancy it up we can add an output pair acting as a buffer to drive cables properly.

Here's a picture of a simple, 5-transistor discrete op-amp.

What makes this a discrete op-amp vs. an IC op-amp? Nothing, when it comes to the circuit. They are identical from a circuit standpoint. What makes this a discrete op-amp vs. a chip op-amp is simply packaging. Discrete means separate bits to make the circuit, integrated means the bits are integrated into one packaged circuit (integrated. Get it?).

Tomorrow I'll help you understand what's going on with this circuit.

Subscribe to Paul's Posts

14 comments on “Being discrete”

  1. In ancient times (the '60's), I was a young EEngr and did many a simulation of things physical on an analog computer, e.g. heat flow studies, mechanical. and hydraulic systems. Capacitor feedback turns an op-amp into an integrator, and two in series represents a 2nd-order differential equation, thereby simulating most physical processes. One very interesting simulation involved control of an electrical power generating dam where turbine speed was mechanically measured and controlled with a two-stage hydraulic valve. After analyzing the system in a general way, all of the physical constants had to be determined by having the computer simulation match chart recordings taken at the actual site. The first computer was tubed (100v), then solid-state (10v). Function generation (squaring, non-linear) was always crude, but at the end of my brief career (1968), I worked on a hybrid system where a digital computer did the function generation.
    Now of course, digital computers are fast enough to do the calculus and everything else in real time.

    Mannie Smith
    Norfolk, VA

  2. This brings back a flood of old memories. I studied electronics in the days of analog computers. Digital counterparts especially in real time were decades away. I remember designing circuits that would perform functions in calculus, specifically integration and differentiation. One of the most important analog computer simulations was aircraft performance for the military. Analog computers were acres of discrete electronics and patchbays. Later I got to play with Square D logic blocks. Anyone remember those? You could make all kinds of neat circuits like JK flip flops and timers. I saw my first PLL on my first job. It was built from discrete parts by GE for a laser measuring instrument that worked on the principle of two laser beams from the same laser split and converging on moving steel at two angles, one where the steel was moving towards the beam and the other away from the beam. The difference in the dopplar upshift and downshift was a velocimeter in the 10 mhz range at about 3000 fpm. The signal was sent to the PLL and a pulse counter integrated the output to get length.

    Soon afterwards the first PLL IC was introduced. When I saw my first PLC I knew instinctively what it was, what it could do. PLCs changed everything and got cheaper and cheaper as they got increasingly more powerful. If there aren't PLCs in most electronically controlled things you buy there will be and they will integrate seamlessly into the internet of things, the coming 5G network. PLCs changed everything. Control went from hardware based to software based. Designing around them is one PITA because you have to specify every nut and bolt including all interfaces with the outside world. Programming is also a PITA for me. When I'm working on a project that uses them, I specify what they must do and let other people handle the details. I limit my control system design mostly to ladder logic. I suppose there are some people who enjoy it. At least now there are standard protocols like Backnet and Modbus plus. In the old days every manufacturer had his own protocol and they were and still are incompatible with each other without protocol converters. Those do exist or the internet of things wouldn't be possible. What a change I've seen in my lifetime from discrete analog computers for simulation to the internet of things using AI, quantum computers, and every means of communications known including light itself.

    As the systems become increasingly complex, human understanding of them will become harder and harder and predicting everything they might do could become impossible. When machines start designing other machines it's over for the human race. We'll be obsolete. And it won't be all that far off.

    1. Machines designing machines was essential for the microprocessor revolution. This was called a "Silicon Compiler" and it was the brainchild of Stanford professor Carver Mead.

      The good news is in my correspondence with some leading-edge AI researchers. They assert the equations that drive AI predict they will not equal and replace human thought processes, and they won't take over like evil geniuses. People who are familiar with silicon fabrication further agree that Moore's Law was the Singularity and it is nearly over.

      Stephen Hawking and Elon Musk, like most experts commenting outside their areas of expertise, are simply wrong. Ray Kurzweil is hopelessly over-optimistic, as he was most of his career.

      OTOH, Norbert Wiener, the father of cybernetics, predicted the effectiveness of Internet marketing, social media and the political consequences of Internet trolls in the 1950 book "The Human Use of Human Beings". We have met the enemy, and they are us.

      1. I remember reading that book in the mid-'60s in college. Didn't recall that it was written in 1950, 16-17 years earlier. Also studied Thomas S. Kuhn's "The Structure of Scientific Revolutions" 1962, and was surprised to learn that it went into a fourth edition in 2012. Looks like I'll have to go back to college and get my degree refurbished...if only I could afford it. 😎

  3. In 1975 I was starting to design an amp which could be used for guitar and mobile disco purposes, and I had the happy idea of using a 741 op-amp for the input stage. Although they must have existed by that time I had never seen a power amp circuit using it. The output stage was a conventional complementary class AB running on +-30/40v rails, the input had its own +-15v supply, and there was a voltage amplification stage to mediate between the different voltages and ensure that the 741 output never would exceed 4v peak to peak and thus, since the amp was direct coupled, allowed a 0-20khz frequency response. It was a simple circuit which worked very well, and was soon taken up by others I knew and rippled outwards from them. I think the reason for its 'success' was its simplicity, plus the fact that it was more powerful than most amps at the time, being capable of 300w into 8 ohms in bridged mode. I kept an example of it in regular use for nearly 20 years until my electric guitar and disco days were behind me. The 741 had poor bandwidth and slew rate, and was noisy, but once better integrated op-amps became available the simplicity of circuits using them meant that their incorporation into consumer and professional audio products was a bit of a no-brainer.

    I do not think this is the case for high-end gear, I have an anecdote to illustrate that but, not wishing to jump the gun on Paul's future posts, I shall hold off with it for a while.

  4. What about the companies that are making discrete op-amps that plug into chip op-amp sockets?

    And, there is also a camp that believes the op-amps sound better when soldered instead of simply inserted?

    1. I'll pass on the pin compatible discrete op-amps, because that would be gun-jumping. As to soldering sounding better than socket insertion, if both pins and socket were gold-plated I would think it unlikely. If both are just tinned then there is a, probably very small, risk of contact problems.

      1. Hmm, good thought. Paul, when you do your annual cleaning disconnecting everything and putting it back together, do you take your tubes out and clean the pins or just replace the tubes? How do you clean the tube sockets? Inquiring sound minds want to know.

    2. All good electrical connections require melting metal to bridge the gap. The electrons have to be in one big cloud to flow. For pressure contacts like switches, relays and sockets, this comes from micro-welds that can be formed from friction when the switch contacts or socket pins wipe into place; or from localized current hot-spots that arc over and melt some metal together.

      For long term contact like sockets and connectors, there is a minimum current rating for this reason. Gold to gold contacts typically have a minimum current of 10uA, which is enough to create and maintain the micro-welds. Since audio circuits go to zero current for long periods of time, it is important to plug and unplug periodically for a fresh wipe. Soldering obviates this maintenance.

  5. Hi Paul, happy memorial day to you and yours! So, not being an engineer just being around audio engineers and manufacturers and owning some of them based gear this is what I have come to learn. Number one, from what I was told one of the advantages of op-amps Is that they are very low in distortion and very neutral with respect to their sonic characteristics. In the equipment I have owned using them I have found that To be true to my ears.Real smooth, transparent and neutral sounding.Everything good about what you would Like to be hearing solid-state. one downside I’ve noticed about op-amp’s Based amplifiers is that the manufactures seem to need to have a large amount of capacitance in the power supply reason being
    ( from what I’m told ) is that the The downside of them is that they can’t move a lot of currentFor any length of time so they need extra capacitance for power reserve.Do you agree?

    1. I have never heard that, Andy, so I don't know if it's true. It doesn't make a lot of sense. Op-amps are DC amplifiers that are typically quite power supply insensitive because they also have a great deal of power supply rejection characteristics depending on how they were designed.

Leave a Reply

© 2022 PS Audio, Inc.

linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram