Re: posts about electronics
Some of us are still designing analog audio circuitry, and we still need testers such as the QuanTech to sort out the good from the marginal.
It is unfortunate that QuanTech went out of business, many years ago, but this one instrument gave me the ability to make breakthroughs in low noise design, starting 45 years ago. Without it, I would never have discovered the low noise properties of the 2N4403, 2N4405, and even the 2N4401, in 1968.
The 2N4401-4403 push pull parallel combination was the basis of the Levinson JC-1 that we introduced at the NY AES, 40 years ago having a noise spec of 0.4nV/rt Hz without an input transformer. Some people did not believe it was possible at the time.
Today, with selected modern devices, it is easier to do, but back then, it seemed almost impossible, at room temperature.
The QuanTech noise analyzer allowed me to get real feedback as to what the real tradeoffs were in bipolar transistors, and also jfets, especially as they got better.
I will always be grateful for being assigned to it, back in 1967, when I was with the Ampex Instrumentation Recorder group, and given this assignment to find the lowest noise bipolars in our inventory, because it could be tedious and I was the new kid in the group. It just so happened that I liked comparing transistor noise, so when I moved to the Audio Division, 6mo later, I got samples from every manufacturer I could and I tested them on the QuanTech, finding surprising breakthroughs in both bipolars and jfets over the years, AND eliminating expensive, but noisy jfets, that were batch sensitive.
Hopefully, today's up-in-coming audio designers will make their own 'discoveries' that will allow 'breakthroughs' based on their open mindedness and passion for the subject.
The test equipment will, of course, be different, but if one is fortunate to have access to something that gives a real 'breakthrough'.
For example, in low temperature physics, the guy who used lasers to 'cool' almost to absolute Zero comes to mind. Anybody see the program on low temp physics showing this a few years ago?
Previously, I put up 4 separate schematic diagrams of different approaches to circuit topology, that were available in 1968, or 45 years ago.
Now I could like to briefly comment on each of them as to their specific characteristics.
The first (A) is just a simple transistor. It could be NPN or PNP, but I drew it with PNP to emphasize that PNP's are quieter for voltage noise, (as shown with the QuanTech) than an NPN. You just have to think 'upside down' and you automatically get a noise improvement. This is the most basic active circuit possible, it can either be common base or common emitter, and it has fairly low noise, IF the device used is the RIGHT ONE. The weaknesses of this topology is high distortion and the need for both emitter and collector cap coupling (at a minimum). Local resistive feedback will increase the input noise significantly, if used. Overall feedback is difficult, because this topology has so little gain in a realistic design.
B is a paralleled combination of input transistors of A. It is 6dB QUIETER, but the base current would be 4 times more, and this could be a problem.
C is a simplified example of what one might do, IF they want to get the lowest noise, yet use an OP AMP to supply more gain, voltage swing, and feedback.
This sort of topology is even used today, perhaps with paralleled N channel jfets, but it could have been implemented 45 years ago, as well.
It is a pretty good approach, if you can get the biasing right, but it is a relatively expensive solution, especially 45 years or so, ago. A number of ultra low noise lab preamps are made with this approach. Generally the input noise is the noise of the input transistor or jfet, and its usually necessary feedback resistor.
Now here is where it gets interesting. How do you get 10 ohm equivalent noise IF you already have a 10 ohm resistor that is in series with the input device(s)? This is a dilemma!
I would like to further comment on what is necessary to make an 'exceptional' audio design. As I said before, knowing your 'building blocks' can be very useful, even giving someone an engineering 'edge', over the competition, until they can find out what you know and do the same. And 'Thus it is always so!' '-)
However, I achieved 0.4nV/rt Hz capability in 1968, but the JC-1 did not come out until 5 years later. Much of the difference between early discovery, and proper implementation came before finding an OPTIMUM TOPOLOGY to utilize. This took me several more years to develop.
Now what is the problem? Well, it is a matter of distortion, cost, and low noise, and these requirements can be contradictory at first application.
For example: One could just use a single transistor, or even better, parallel a bunch of them, and then use an op amp for rest of the gain. This could have been done with four paralleled 2n4405 devices and a uA709, back in 1968 and we would have had ultra low noise and low distortion too! Yet, what are the 'drawbacks'? Cost might be one, because a low noise +/- 15V power supply could be expensive, and any IC op amp was not 'dirt cheap' in those days. Another design decision was whether to cap couple or not. The input bias current could be somewhat high, was it too much for a MC cartridge to bear?
If so, what relatively high value of input C would be necessary to not add excessive 1/f noise to the equation?
What about the optimum input impedance range? I would think 100 ohms or less would be best in this design. That sort of leaves out a making general purpose test amp this way. In other words, to get the BEST low noise performance, you have to direct couple the input, if possible, and keep the useful impedance range limited to perhaps 100 ohms, to get the lowest noise with this method.
Now, should the discrete transistor array be cap coupled to the IC? Here, there is a trade-off between size, voltage rating, and value of the cap. Cap quality might be important too. Of course, even in 1968, we could DIRECT COUPLE the discrete input stage to the op amp, BUT we would have to offset voltage in both op amp input terminals. This could be done, even then. So then what? We make something that meets a noise and a general distortion spec. for a MC input, but is it worth doing? We have replaced a very cheap and compromised input transformer, with a solid state gain stage, but will it be really better, and worth it? Why not just use a better input transformer? (more later)
PNP transistors are always quieter than NPN transistors. This is because the BASE of the PNP is N material. The Rbb' of the PNP is usually about 1/2 the value compared to NPN transistors with similar construction.
Demian, please look up the part numbers that I spoke about. 2 types are the 2N4403 and the 2N4405.
Finally, D shows what most people would opt for, based on what they are used to, a differential pair. It is simple, lower distortion than a single device, easier to bias, BUT it is 3 db NOISIER than a single device.
The fundamental limitation of this input stage, EVEN with virtually perfect devices, makes it 2.5 times or more, noisier than what we could do 45 years ago. And that is why this is a good, but not an optimum approach with MC cartridges.
However, you just made the same fundamental mistake that Dick Burwen did at the time. You presumed that all transistors with similar ratings are made essentially the same way. The 2N4403 is the TRUE complement to the 2N4401 and it is quieter than the 2N4401. (At least it used to be)
The 2N2907 has the 2N2222 as a complement. The Rbb' of these devices is greater than the 2N4401-2N4403 devices. The QuanTech sorted this out for me in 1968.
In discrete bipolar transistors, the PNP complement usually has about 1/2 the Rbb' or intrinsic base resistivity. IC's have had problems with making PNP transistors.
The 2N4401-2N4403 are the complements that really work better than almost all preceding part pairs for low voltage noise, AND they were NOT represented as low noise devices at the time.
Now, if I can, I would like to give an alternative approach to the low noise problem.
This is not necessarily a universal solution, and I personally haven't used it for the last 30 years, BUT this does give low noise without an input transformer.
The advantages of this design is that it can run on a couple of D cells, and this lowers the price and gives a relatively quiet power supply. It is direct coupled both input and output, BUT it does not amplify DC. Even 45 years ago, (if I had thought it up at the time) it could achieve the equivalent noise of a 10 ohm resistor,