The SSTART Preamplifier
For those familiar with tube-based high end gear, the Conrad Johnson ART preamp and its descendants surely rank among the more fascinating components ever made. Why? Because the ART is comprised of a single gain stage — a common cathode topology using paralleled triode sections... and that's all. So why on earth would anyone do such a thing? Doesn't it have a really high output impedance? Surely it can't provide much current. And just the one triode? Wouldn't that invert the signal? As it turns out, only the last objection has any validity, and that's easily taken care of by reversing the plus and minus connections at the speaker. The other points can be dealt with if you approach them correctly.
There's a perfectly valid reason for using such a simple circuit. It's the absolute, bare minimum number of gain stages and that's an irresistible lure for those who have discovered that cleaner, simpler circuits generally lead to better sound. But how to get beyond the undeniable limitations?
Most any tube can provide sufficient gain for a line stage. The problem is that the output impedance — for all intents and purposes, the impedance of the load resistor — will be 5k to 10k, minimum, and that just won't do. That high an output impedance will interact with interconnects and amplifiers, causing all sorts of mischief. The usual answer is to append a cathode follower (for those who are only fluent in solid-state, that's the tube equivalent of a source or emitter follower) and let that define the output impedance. Unfortunately, that adds another stage to the circuit, and that's precisely what we're trying to avoid doing.
All is not lost, however. There is another way. Suppose you want to swing 10V against the load resistor. If you've got one tube, the current available from that tube defines what the load resistance needs to be in order to produce the desired 10V. Unfortunately, the output impedance will be much higher than you want. But if you take a second tube and parallel it with the first one, supplying the same amount of current, then the load resistor (hence the output impedance) can be cut in half. And if you double that current, the load resistance can be halved yet again. As a fringe benefit, all that cumulative current means that we will be able to drive any reasonable load and quite a few unreasonable ones.
Better still, paralleling gain devices reduces noise.
Ah, ha! Now, we're getting somewhere!
I had thought about doing something like an ART clone, but to be honest, although I love the sound of tubes, having to replace them annoys me. Worse yet, the quality of the tubes you can buy these days is, shall we say, uneven. And they're expensive. And they're hot. Ugh. Okay, so what about solid-state?
Much has been made of the "tube-like" qualities of FETs. Well, it depends on which qualities you're looking at. Without going through a lengthy comparison of the characteristics of tubes versus FETs, let's acknowledge the elephant in the room: FETs have more in common with pentodes than with triodes such as the ones used in the ART preamp. All gain devices distort, but triodes' gain characteristics cause them to distort differently than pentodes. FETs, unfortunately, just won't mimic the triode sound.
Still, what might a solid-state analog of the ART topology look like?
MOSFETs might appeal to some, but it seemed more logical to look to JFETs.
You can find people who will champion any given part ever made, and they will inevitably want to know why you didn't use their favorite piece, since it's obviously superior to whatever you chose to use. When it comes to JFETs, there are several candidates that spring immediately to mind, among them the Toshiba 2SK389. 2SK109, 2SK170, and 2SJ74. The fly in the ointment being that they've all been discontinued (the '170 just this spring). Before you argue with me, please check the Toshiba general catalog. There's a difference between "But my friend Ralph just bought some a month ago," and full production. Yes, there are still some in the retail pipeline, but you'll have to accept the fact that the parts are formally extinct. And if people can't get their hands on the parts they can't build the circuit. For that matter, even the Toshiba JFETs that are in production aren't easily found here in the United States, so I try to use them sparingly.
One part that is reasonably priced and available from Fairchild is the J310. It has the benefits of being able to pass a fair amount of current, low noise, and low capacitance. If you're willing to hunt them down and pay the price asked, the Toshiba 2SK246 makes a nifty alternative. One problem that crops up immediately is that the J310 can only handle 25 volts. And if you intend to pass much current, you'll need to limit the voltage anyway because the part is only rated for 350mW of heat dissipation. It's possible to build a preamp using a 25V rail, but given the interest in so-called "hot followers" and other low gain amplifiers it would be better if the preamp could swing more voltage at the output.
Between the availability criterion and the need for a JFET that can take a bit of voltage (the majority can't), the project was almost stillborn. With that in mind, I decided to allow one compromise and use a cascode. A cascode is a gain device that sits atop another one. In doing so, it takes the voltage load off of the one underneath. It also alleviates some of the capacitance…icing on the cake in this case. The part I chose for the cascode device is also relatively inexpensive and widely available: the IRF610. Many people will already have some on hand.
The circuit — which I call the Solid STate ART, or SSTART for short — is simplicity itself. Like the ART preamp, the gain stage consists of paralleled devices operating in common Source mode, meaning that they are hooked up such that the signal goes in the Gate and out the Drain. I tried different numbers of parallel J310s, but settled on four. If you feel like experimenting, you can use more or less, depending on your goals. Note that I show each JFET having an independent Gate stopper (the 100 Ohm resistor right before the Gate). You'll find it easier to experiment if you use this simple modular trick to keep things from breaking into oscillation.
Each JFET is biased independently by the 51.1 Ohm resistor at its Source. This value can be increased if you like, but be careful about decreasing it because a strong signal can clip the input.
The Drains of the four parallel JFETs are gathered together and run into the Source of the IRF610 cascode. The voltage divider that feeds the Gate of the IRF610 holds it at about 15.5V. Given an average voltage drop between the Gate and the Source of the IRF610, the Source will settle somewhere around 11 to 12V. This sets the voltage at the Drains of the J310s, thereby limiting the heat dissipation. Use a heat sink on the IRF610, as it will get warm. Something like the Aavid Thermalloy 530613B00000G is perfect for the job.
The load resistance, the 270 Ohm 5W resistor, splits the difference between the rail voltage and the cascode's Drain. This determines the available output voltage swing, which is on the order of 35Vp-p. This can be increased, should you feel the need, by increasing the rail voltage and adjusting the load resistance and the cascode's voltage divider, but it will suffice for most applications as it is.
The 1000pF capacitor at the input limits the bandwidth to around 100 kHz. Use a silver-mica or polystyrene cap for best sound quality. Or you can do without. The circuit has terrifyingly wide bandwidth, which I happen to like, but it can get you in trouble if you live near a radio transmitter or if the Martians happen to choose your backyard for a landing site on their next visit. My advice is: If in doubt, leave it in. If you use a lower resistance volume control, increase the cap accordingly.
Inevitably, some folks will grumble about the DC blocking cap at the output. It's necessary unless you want to start adding more transistors. Before anyone works themselves into a fluster, please devote a moment to the thought that if you trace out the entire current flow of a supposedly capacitorless solid state piece, you'll find that the power supply caps are very much in the signal path. Another point to ponder is that there are a number of high end solid state designs using DC blocking caps that are flagrantly in the signal path. One last thing…given that, even after all these years of effort on the part of solid state designers, tube designs are still widely regarded as being the touchstone for sound quality. (When's the last time you heard someone tout a tube preamp's "solid-state-like" imaging or dynamics?) Open up a tube circuit and what do you find? DC blocking caps. Lots of them. Just use a good quality film cap and things will be fine.
One thing the DC blocking cap will not do is save you from turn-on thumps. Fear not, you have options. One is to always remember to turn the preamp on first, then the amp. A couple of seconds will suffice. When you're done listening, turn the amp off first, then the preamp. I've been doing this for years, so it's second nature to me. Another possibility is a manual mute switch that shorts the output to ground. The elegant solution is a relay that does the same thing at turn-on and turn-off. Or you could just leave the silly thing on all the time. A channel draws less current than an average night light.
And now a word about parts matching. For JFETs, the matching parameter of choice is Idss. This is the current you get if you tie both the Gate and Source to ground and measure the current that flows through the device via a resistor between the Drain and the rail voltage. In the case of the J310, the rail voltage for matching should be on the order of 10V. Use a 100 Ohm resistor between the Drain and the rail. Measure the voltage across the 100 Ohm resistor and use parts that have about the same voltage reading. For this circuit I used J310s from the lower end of the part's current range (25mA to 30mA), as they gave greater flexibility. If you find that you have higher Idss devices, set the voltage divider for the cascode device to about 8V by replacing the 22.1k resistor with 51.1k. You might also consider increasing the value of the JFET biasing resistors to reduce the idle current through the JFETs.
The IRF610 doesn't have anything to be matched with, unless you want to match the left and right channels. In this case, use Vgs matching. Nelson Pass has written more on MOSFET matching than anyone I know and it is all available at passdiy.com. Or you can just use any old pair you happen to lay hands on. You can't design a much simpler signal path for a solid-state preamp, it's inexpensive to build, and the parts aren't hard to obtain. Not a bad experiment, overall. Just don't forget to swap the speaker leads due to phase inversion.
Does it sound like the Conrad Johnson ART preamp? Well, no, but it's a SSTART...