Inigo Phono Stage 1.0
Over the years there have been numerous formats for recorded music, each with its associated playback hardware. Some, such as the player piano, have fallen by the wayside, perhaps undeservedly. Others, like the 8-track tape, died a well-deserved death. Some persist on the fringes, reel-to-reel tapes being a good example; arguably the finest quality music storage medium ever available to the consumer, but the fiddle factor involved in threading a tape, cleaning tape heads, and so forth tended to limit its acceptance. When the CD was introduced in the United States, it was promoted as "pure, perfect sound, forever." It was intended to supplant the LP record and for a while it looked as though it would succeed, but every time the pundits declared vinyl finished as a format, it would lurch to its feet, like the battered swordsman from The Princess Bride:
"My name is Inigo Montoya. You killed my father. Prepare to die!"
And we all know how that ended.
But vinyl playback requires more than just a turntable, it also requires a dedicated preamplifier stage — a phono stage. Whether phono stage design is a labor of love or a trial by ordeal depends on your point of view. From the start, you're confronted by conflicting requirements and plagued by parts availability problems. To begin with, you need a minimum 40dB of gain. If you intend to use a moving coil cartridge, you'll need more. You need to achieve this amplification with an absolute minimum of noise. And you have to add a three-part filter network to meet the requirements of the RIAA playback specification.
And because I'm a masochist, I decided that it also had to be balanced, use passive EQ, and no feedback. And while I was at it, I wanted it to make my afternoon tea, as well. It took a while, but I achieved most of those goals. Sadly, I had to give up on the afternoon tea. The Inigo's input is comprised of a complementary differential using low noise JFETs (Q3, Q4, Q9, Q10, Q11, Q12, Q17, Q18) in parallel pairs to lower their noise contribution. The JFETs are, in turn, cascoded by bipolars (Q7, Q8, Q13, Q14), which lower their effective capacitance and relieve them of the necessity of dealing with the full rail voltage. For this iteration of the circuit, I chose to load the front end with current mirrors (Q1, Q2, Q5, Q6, Q15, Q16, Q19, Q20), which provide the lion's share of the gain.
Most phono preamps put all the RIAA equalization in one place. I've split it in two, which affords more flexibility in choosing capacitor values. The first EQ is the 2122 Hz, also known as 75uS (R24, R25, C1). The configuration I'm using is sometimes called a "flying" EQ because it uses one filter network connecting the opposing sides of the circuit together. You could easily arrange separate EQ for each side, but this way uses fewer parts. I've labeled the filter inputs and outputs to match the output of the first gain stage and the input of the second gain stage. Laying the circuit out this way makes it easier to see how each block functions.
The second gain stage is identical to the first.
The second RIAA EQ (R49, R50, R51, C4) incorporates both the 50 (aka 3180uS) and 500Hz (318uS) EQ elements into one compact unit using a single capacitor. Like the 2122Hz, this is a flying EQ.
And finally, to buffer the second RIAA EQ from the outside world and provide a low output impedance, I've included a JFET buffer (Q41, Q42). Sounds simple, doesn't it? Like so many other things, the devil is in the details.
For one thing, phono stages require a power supply that is extremely quiet. (Now you know why I covered regulated power supplies last time out... sneaky, eh?) The semiconductors, likewise, must be of the highest quality. If you use ordinary parts, you'll discover that the circuit works just fine, but it will be annoyingly noisy. The general topology affords plenty of lattitude, and I've got several variations on this same general theme, some with resistive loads instead of current mirrors, some with cascoded outputs, etc. One interesting variant uses the THAT 300 and 320 for the current mirrors. I may revisit this circuit in the future with some of the other options.
How much gain to provide is a thorny topic. Phono cartridges vary widely in output, so there's no such thing as a one size fits all solution. If you use a conventional moving magnet cartridge with a phono stage providing 70 or 80dB of gain, you'll find the combination difficult to live with. The slightest turn of the volume knob will blow you out of the room. At the other extreme, if you use a low output moving coil with a standard 40dB phono stage, you may not achieve decent playback volume, even with the system running wide open.
In this case I settled on 50dB of gain. This is sufficient for moving magnet and moving iron cartridges, and will also do a good job on mid- to high-output moving coils. There are a number of ways to jimmy the gain either higher or lower, depending on your needs. One easy way to increase gain is to lower the value of the resistors connecting the two halves of the bias for the input JFETs (R12, R37). But beware — everything is interrelated. If you increase gain, bandwidth decreases. Check the RIAA EQ once you settle on a gain.
Which brings us to the accuracy of the RIAA equalization. There's always someone who complains that you should be able to just throw random parts into a circuit and achieve perfection without effort. For brute force applications like basic power amplifiers, that sort of approach is semi-feasible. This circuit will not come unstuck if you make reasonable changes to the parts values, but if you stray too far afield, you'll need to fiddle a bit to get maximum performance. Why? Because things like the capacitance of the transistors will affect the accuracy of the RIAA playback characteristic.
Don't fret. Here's how to fine tune the EQ.
Start with the resistors that set the 50Hz level (R49, R50). Compare the level at 50Hz with the output at 1kHz. It should be 16.94dB higher. If it's too high, decrease both resistors. If it's too low, increase them. These two resistors must be the same value.
Subject to the tolerances of R51 and C4, the 500Hz level should be fine. Should the capacitor be off a little bit, you can adjust R51 to compensate. You're looking for the signal at 500Hz to be 2.65dB higher than at 1kHz.
Finally, the 2122Hz level is adjusted by changing R24 and R25. In this case the signal at 2100Hz (it's easier to repeat a round number on most function generators than it is to dial in the last 22Hz) should be down 2.82dB relative to the 1kHz signal. If it needs to be more negative, then increase the value of the resistors. Conversely, if it's too low, decrease the value. Again, both resistors need to be the same value.
As a practical matter it's easiest to use pots to try different resistances. Once you get it right, measure the pot, then select fixed resistors that match that value. You can usually get a good match using two 1% resistors in series. For instance, the 3180Ω value for the 500Hz/318uS resistor (R51) can be created using 3.16k and 20Ω in series. The tighter tolerance your capacitors, the easier your life will be. It's getting easier to find 2% film caps these days — I recommend polystyrene or polypropylene. If you've recently robbed a bank, you can use Teflon, but I fear that the price is beyond anything I can justify.
Treat the RIAA equalization as an iterative process. After you change one part, go back and recheck the others. Everything you do will affect everything else. It takes time and effort to get it right.
Some folks might be puzzled by the apparently small value of the input resistors. It's easily explained — each half of the circuit presents half the usual load to the cartridge. If you assume a normal 47k input impedance and divide that in half, you'll end up with 23.5k for each leg. As of this writing, Mouser shows both 23.2k and 23.7k in stock. Either will do. It's not necessary to obsess over the input resistance; many famous preamps have used values such as 47.5k or 49.9k and no one has raised an eyebrow. You can hit 47k precisely by paralleling two 47k resistors per side. The circuit will not complain if you want to use 100Ω or even 10Ω cartridge loads at the input. Just split the value and put half at each input. If your cartridge needs a little capacitance, put double the capacitance to ground at each input.
At the minimum, the input JFETs should be matched for Idss. The bipolar transistors in the current mirrors should be matched for gain. If you have access to test equipment, then buy a handful of transistors and have at it. If not, Nelson Pass has published simple circuits to match JFETs and transistors at www.passdiy.com. Matching the transistors in the cascodes is optional, but certainly won't hurt because the current flowing from one base to the other across the midpoint of the voltage divider will sum to pretty near zero. If the transistors are unmatched, they will modulate each other a bit. This is not something to lose sleep over, just one of those things that appeals to people who like to get finicky about details.
Try to match the total current drawn per quarter when matching the input JFETs. Mix and match pairs until you get the upper left, upper right, lower left, and lower right pairs of JFETs as close as possible. They set the operating current for the entire circuit. The cascodes pass whatever current is presented at their emitters, and the current mirrors, as you might guess from the name, only reflect the current they are fed. It matters because the DC balance at the outputs will be influenced by the opposing currents. If the upper right quadrant, which pushes against the lower right quadrant, is higher in current, then the midpoint between those two transistors will shift upwards. This will lead to asymmetrical waveforms if the stage is driven hard.
And that, in turn, leads us to the coupling caps between stages. They block DC, thus allowing us to dispense with servos, which means that the circuit is free of feedback. They also act as a subsonic filter to reduce the effect of warped records. I've set them relatively low — roughly 5Hz taken together — but you can adjust them quite easily by changing the values of R28, R46 and R52 (plus its mate in the other output buffer — not shown). Incidentally, I did test a rather unconventional servo; one that might even meet some peoples' definition of ‘out of the signal path,' but I'd rather use caps than a servo. If you feel differently, there are any number of potential tie-in points that you might consider. The inputs are an obvious choice, but you might also consider dividing the resistor that couples the two halves of the circuit in two, then using the midpoint to receive the servo's correction signal. The servo I experimented with used an optocoupler to vary the emitter resistance for the output devices. It works perfectly if you're into that sort of thing.
On to the specifications. These were taken from my prototype and represent realistic values that you can hit with a little attention to detail.
Gain = 50.45dB
RIAA curve accurate to less than 0.1dB from 20Hz to 21kHz
Input overload >30mVrms
THD = 0.65%
My usual disclaimer about THD applies here. In fact, even more so than usual. My bench is right under the main circuit breaker box for the house and there's enough EMF in the air to light the city of Paris. I've looked at the output nasties and it's comprised of — no surprise — almost pure 60Hz hum. It's annoying enough when you're testing an amplifier, but as anyone who has ever built a phono stage can tell you, they're highly sensitive to interference. This design is no exception. The take-away message from this is, 1) don't be put off by the THD figure, it's inflated by our microwave oven running whilst I'm trying to measure distortion, 2) use an external power supply, 3) use a very quiet power supply, and 4) try to cobble together a well-shielded box for the circuit. Your ears will thank you for it.