Yet Another JFET Phono Preamp

[wrenchone]

Attached is a schematic for a passive equalization RIAA preamp using a buffered cascoded common source input (like a Pacific with a couple of bows and ribbons), and a complementary feedback pair for the second stage after the EQ network.

I've been working over the past three years on various schemes for the second stage(after the EQ network) to replace the common source stage used by the Pacific, which has relatively high distortion when forced to pump out a volt or two. Some of the things I've tried are a heavily biased small signal MOSFET (see the "All American" thread), and using optimally loaded JFET SRPP stages. The complementary feedback pair I'm trying here simulates very well, and since the two halves of the pair are not direct coupled, the JFETs are relatively easy to select. BTW, lest you complain about the interstage coupling cap, it's there so that both stages can be optimally biased for gain and voltage swing.

The amp is in the advanced prototyping stage right now. I'm very curious as to how It'll sound. This will also be a test for the new fancy-pants shunt regulator shown in the schematic. It simulates as having ~3 milliohms static output impedance, and recovery time is extremely fast (1-2 usec) for a 100 mA load step with 500 nsec edges.

[Salas]

Thorough engineering as always from you Wrenchone. Curious to know about your soon to be made subjective tests too. The shunt part is nice, I know it will do well, because a very recent one that I posted for the Pass B1 looks topologically similar enough, and a DIY audio member had excellent practical results with a quick mock up, used for the almost no PSRR Pass BOZ preamp. And yours is faster.

[SY]

Dumb question, especially because it's hard to read the parts values. Assuming that R105 is 3k3 and R109 is 22k6 (that's what it looks like at this resolution), what does the buffer following the input amp bring to the table?

[wrenchone]

More gain, more RIAA accuracy. Without the buffer, the impedance driving the EQ network is the drain load resistor in parallel with the drain impedance, and you need to adjust the first resistor in the EQ train to account for this. The cascoding raises the drain impedance, but it ain't infinite. With the buffer there is no guesswork, and more consistency unit to unit. In a pinch, you could get rid of the buffer - it's more useful if the input stage is not cascoded.

[SY]

So if RIAA conformance is the issue, then the 22k6 can be tweaked?

What's the gain difference? Gain is about gmRload, and Rload is just changed from 3k3 to 2k88. That calculates to 1.2dB, and you have to add four active devices in the signal path. Do you lose as much on noise as you gain on gain? I don't know the noise specs of the follower devices.

[wrenchone]

In a word, yes. You'll lose a little gain due to loading, but not all that much. My simulations don't show much of an effect on distortion magnitude or distribution by including the buffer. I haven't tried listening tests with and without. Without the buffer, I'd just use a reverse RIAA network and tweak for flat response.

The PN4393s are nominally chopper FETs, but they make pretty good low noise amps for a person on a budget they aren't anywhere near as quiet as the 2SK170, but they are adequate for MM applications, say in the second stage of a MM preamp.

[wrenchone]

Attached is a simulation file for the new shunt. I used a 27V reference zener as it was the highest value 1W zener I could find in a hurry. Results would be similar with a 30V unit, just higher voltage.

[wrenchone]

Here is the transient response of the circuit. It takes good low impedance output capacitors to get optimum response. The value shown for output filter (660uF, 11 mohm), would be the value you get from two parallel Panasonic FM series caps, 330uF, 50V. The extra PNP gain stage is what really peps up this regulator.

[wrenchone]

Here is a somewhat larger schematic of one side of the preamp.

[Salas]

Excellent circuitry!