Variation on the JC-2 preamplifier

[jonusgrumby]

How is 50 ohms over the top?

The device curves show minimal gain increase at operation over 3mA. A current source should represent an infinite impedance, so 20 additional ohms is not a problem.

What am I missing?

[john curl]

6ma would be better. Take it from me, I do this for a living. Of course, I designed in a pot, and YOU took it out.

[GRollins]

Seems likely that the phrase "increased negative feedback" is lurking just beneath the surface here.

Grey

[jonusgrumby]

I do this for a living too, although my professional products are for space and aircraft applications, not consumer electronics. Working in that environment has meant rigorous worst case analysis on every design. I tend to carry that over in my audio products. For whatever reason, it seems audio designers fall into a line of thinking that if 3 mA is good, 6 mA must be better. But better why? You certainly don’t need it do drive the next stage. The voltage swing is miniscule and the small miller compensation capacitors require next to nothing to drive even well beyond audio frequencies.

To anyone that might build this variation, I encourage you to use the parts I specified. If you do, you will be assured of duplicating the results I achieved. Additionally, if you use the tubeshunter kit, these parts will fit on that board. There is no room on the board for a bias pot.

If you wish to experiment with these values, have fun. That’s what it’s all about.

[jonusgrumby]

Actually, the feedback network I use (R10 and R9) increases the closed loop gain and decreases the amount of neagtive feedback.

[john curl]

It seems to me that you 1/2'd the slew rate and increased the input stage distortion 4 times, but I could be mistaken. Please don't second guess the original designer.
This is like some guy buying a Porsche 911 and modifying it seriously by changing the fuel mixture, or timing and saying that they have 'improved' the performance, without offering proof. This same person designs engines for a tractor factory. (nothing personal implied)

[jonusgrumby]

I mean nothing personal to you Mr. Curl. Not sure the same applies. I assure you, I hardly fit the tractor engine analogy.

I have not demeaned your original design. My application was different and my parts selection reflect that. Your design has proven itself by standing the test of time.

Ultimately the slew rate will be determined by the load capacitance, not the front end stage current (unless the output is unloaded). The front end voltage swing is so tiny (only a few millivolts) and the load capacitance so small that 3 mA is still well in excess of 30V/uS worst case.

The input stage distortion on my version will run these parts in a very linear part of their operating curve. The lower values may, or may not and likely not as linear as 3 mA will. Again, because the swing is so tiny, the distortion difference is unmeasurable. I didn't concentrate on that, it was setting the bipolar output current that concerned me. I have previously stated my reasoning on that.

My suggestions for improvement were aimed at the tubeshunter kit. I think they are worth considering for anyone interesting in building this kit. I didn't build an original JC-2 nor do I have one for reference. I have auditioned Mark Levinson products in the past, but that is as close as I got. I did build my version of the tubeshunter kit. It performs flawlessly in my intended application and the bias measurements were exactly as calculated. The second channel performance matched the first within a few millivolts. Distortion was well below .01% at audio frequencies on both channels. I'm not sure what aditional proof one need supply.

[john curl]

The JC-2 line amp design was first designed 35 years ago for professional amplications. Originally it used larger To-5 output devices, with heatsinks, and operated at about 50ma. Later, when it was adopted to the Levinson JC-2 we experimented with smaller TO-92 parts, such as the 2N4401/4403. However, we ALWAYS used thermal epoxy around to parts to dissipate the heat.
I doubt that anyone lately is using heatsinks on the TO-92 devices, so they are deliberately 'starving' the output devices, in order to keep them from overheating.
The 'changes' brought out here, will only compromise the design, because you will have increased the input distortion, lowered the slew rate, and the lower open loop bandwidth (very important in the original design).
Please don't try to 'improve' the design until you understand it well.
By the way, the original design had a slew rate of 100V/us.

[jonusgrumby]

Ouch.

30 milliamps is hardly starved operation.

I do understand the design, it's elegant, but not that complicated.

I have provided sound engineering rationale for every decision I made. My circuit has a higher open loop gain than your original. Your input stage may have had a tad over unity gain, but that's it. With a normal distribution of parts, it would have likely been closer to a gain of .75 (at best). The increase in the drain resitor in my approach increases the input stage gain to 7.5 or more.

It is debatable if your design with the new SK and SJ devices is higher, since the operating parameters you suggest are undefinded. With some parts it will be slightly higher, with others it will be slightly lower. As for what I suggest, the decreased current (and therefore gain) in the bipolars is compenstated by the increased gain provided by the input stage. So much for open loop gain.

Again, the output load sets the slew rate. Connect a 1 kHz squarewave to your amplifier circuit and adjust output for 5 V p-p. Use an oscilloscope to observe the output waveform first unloaded and then with a 1000 pF capacitor. Assuming yours remains stable, and it should, you'll see the change in slew rate. Crank on your cherished pot all you like and it won't change the output slew.

Do you really think I don't understand how it works?

[john curl]

No, the slew rate is determined by the first and second stage idle current and the paralleled combination of Cob and compensation capacitance.