Yet another variation on cascoded balanced line stage
Thanks Nelson Pass for the original design of the Zen balanced line stage.
This variation includes a cascode transistor on the gain device, and a cascoded current source to keep the bias current steady. The supply is +60/-20V. I could probably go all the way down to +60/-10V, and I already have the trafo for that negative rail :)
The bias LEDs in this design are amber Panasonic models that helpfully drop 2V when you put 10mA through them. The BJTs are Zetex E-Line packages that are compact and can dissipate useful power without heatsinks.
Q3 and Q4 are biased to 14V on the base. They hold the drain on Q1 and Q2 at 13V. I chose these bias levels to allow the output to swing down to +16V, given a 2Vrms input and a gain of 10 (20dB).
Q5/6 and Q7/8 form a steady 40mA current source, with the collector of Q7/8 held at -9.2V.
I'm a little skeptical of the role of Q3/4 in this circuit. By pegging Vd on Q1/2, it really changes the nature of the circuit. So, I am building the board with a socket for Q3/4. I will compare the sound with Q3/4 installed or jumpered.
Your thoughts and brutal flames are welcome.
Here is the final schematic. Please ignore the +/-24V supply markings: those are magic symbols in Eagle and I can't change their value easily. The real supplies are still +60V/-20V.
And here is the PCB layout I drew for it:
The bigger version of the PCB layout is here:
The board is 3.9"x3.2" with two layers. I had a problem because the Hovland MusiCap 7uF output coupling capacitors are massive: 3.25"x1". These would have occupied the entire board, so I moved them off to the sides, and their lead wires solder into the pads marked OUTCAP.
I should be getting these boards within a week. Praise be to $33/each PCB houses.
Final design changes
This design turned out to be unsatisfactory. After I built up both channels (but not on the breadboard), I measured a high amplitude 50MHz oscillation in the circuit. This was solved by inserting a 221Ω resistor between the base of Q5 and the top of the LED stack. For symmetry, I also made this change at the base of Q6. Q5 and Q6 are used to cascode and thereby improve the PSRR of the active current source. The insertion of the resistors should not impact this function, nor the sound of the circuit. Of course, by preventing oscillation, these two resistors will keep me out of trouble with the FCC :angel:
I also found the heatsink prescribed by Mr. Pass in "Balanced Zen Line Stage" for Q101 and Q102 insufficient. The heatsink -- Digi-Key part HS104-1 -- has thermal resistivity of 15.6°C/W. The TO-220 case of the IRF610 has thermal resistivity of 4.0°C/W from junction to sink. The IRF9610 is worse at 6.9°C/W. With my active current sources, I am drawing 200mA at idle, over a potential of 20V drain-to-source at Q101. This transistor needs to dissipate 4W. With this heatsink at standard temperature, the Tj will rise to 103°C. I found this to be slightly optimistic, since the temperature inside the chassis rises above standard temperature, and the IRF610 is rated to operate up to 150°C.
My solution was to use Digi-Key part HS104-2 with a thermal resistivity of 6.4°C/W, giving a nominal Tj of 67°C.
I also was not happy with the first revision of my circuit board. I made a second revision where the current flows straight across the board, and the signal lines don't cross or run parallel to the power. The modification for Q5 and Q6 is wired in point-to-point.
Cascoded current sources are cool, but may I suggest that you consider putting some more diodes to drop more voltage across the emitter sense as I would think "agility" of the current source was more a function of emitter resistor drop than having a high Vce
My intention was not to have a high Vce, but rather a stable Vce. The cascode holds Vc at 9.45V, Vb is 2V, and Ve is 1.45V. Current variation from Early effect should be nil, and I think 1.45V is enough across the emitter resistor to ensure that variations in Vbe are small by comparison.
What Ve would you recommend, and what effect would you expect? I could certainly see, for example, setting Vb to 6V for a 5.45V Ve and using a 110Ω 1% emitter resistor. But, I am indifferent about the effect this would have on the sound. I really don't care that much if the bias current drifts by a few mA over the course of an hour.
Is it really a good idea to cascode a fet and a bipolar? With a bipolar the drain current varies as the log of Vbe so when you cascode two bipolars you get a log / antilog function which comes out linear, but with a fet the drain current goes up as the square of Vgs. What square / antilog comes out to be is different to say the least. But in this balanced cct it may not matter because the nonlinearity cancels? I wonder how it would go with cascoded fets.
Also, is it really a good idea to have independent current sources down the bottom? A single one connected to *both* sources will definitely make the circuit balanced because it forces the total current to always be the same. When one side backs off the other side is made to turn on by an equal amount.
What is the reason for the cascode current source setup? I have never seen this before. Does it work over a wider range of voltages or have greater current constancy or something?
Congratulations on trying something new :)
Great circuit, I like the elegant use of the cascodes. I also liked the first pcb layout wrt the star grounding for the supply caps. In fact, I would have connected R16 & 17 directly to the caps as you did with R11 & 12.
To cut down on cross-pcb signal lines you could relocate the resistors for the outcap+ terminals to the right close to the supply caps.
I noted that in the second version you went away from that. In fact, the gnd for the V+ now is shared among several stages, which brings the danger of mutual influence with oscillations and/or higher distortion. That first layout was really, really good, in my view.
Cheers, Jan Didden
Thanks for the nice comments Jan. Did I ever mention how much fun I had building this preamp? Sounds great, if only I could make a nice chassis for it.
Here's one channel:
Plan view of same:
|All times are GMT. The time now is 10:36 PM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio