My CCS-loaded ECC99/5687 amp for HD580/600/650
May 25, 2006 at 3:54 AM Thread Starter Post #1 of 6

Rescue Toaster

100+ Head-Fier
Joined
Sep 30, 2005
Posts
132
Likes
11
Greetings! Here is the design I'm building. The last of the parts should be arriving tomorrow. The design is a parafeed (CCS-loaded) single tube stage with autotransformer output for 300 ohm headphones.

In this circuit there's really only 4 parts that carry AC signal, the tube, the parafeed capacitor, the output transformer, and the cathode resistor. The cathode resistor only carries input signal current though, not output. With good constant current sources above the tubes, those 4 parts *alone* completely determine the performance of the amplifier. There is zero power supply interaction.


The amp:

Schematic

22ma constant current sources load the tubes. The output transformers are any suitable 10K:600 wired as autoformer for 5:1 stepdown, providing a 7500 ohm load to the tube. The bottom of the transformer returns to the cathode instead of ground (neccesitating 4-pin balanced headphones) to eliminate degenerative feedback in the cathode resistor. With the CCS on top, the output current loop is entirely contained and the tube operates with fixed bias. The input volume control can of course be eliminated and replaced with whatever. The two gas tube shunt regulator (with teflon cap bypass) provides further power supply isolation and sets the voltage. The parafeed capacitor has to be tuned to the output transformer chosen.

The power supply:

Schematic

A custom transformer from electraprint w/ electrostatic screen feeds two 6AX4-GTB damper diodes in full wave. They provide a very slow turn-on to ensure the ECC99 is fully heated. The 47 ohm resistors are really just to drop some extra voltage to lower dissipation in the supply CCS. Small film caps can be used because the CCS-fed shunt (the supply CCS & gas reg tubes in the amp) provide excellent isolation. Even so ripple is only 45mA before the regulator. The power transformer's electrostatic screen runs with the heater supply to the ECC99.

I'm using Gary Pimm's cascoded constant current sources, which are probably overkill. You could easily use something like the C4S from bottlehead for each tube, and a simple DN2540 CCS in the power supply.

The basic structure of the amp could be easily setup to run any high-GM triode (ECC99, 5687, probably 6N6P though I haven't done the math, please no 6C45's, too much gain and they won't like the fixed bias). Essentially anything with a plate resistance around 2-2.5K that operates in the 120-180v range and ~20ma could be used. My suspicion is the 5687 is the best choice, but I'm gonna try out the ECC99's first, since I have them.

For transformers I'm starting with some real cheapo Edcor 10K:600's (4:1, wired as autoformer so 5:1). They are nothing special ($12 ea!) but don't really do anything objectionable. The best transformers are probably Lundahl 1674 amorphous cores ($100 ea) wired as 4:1 and then as 5:1 autoformer. If you absolutely must keep your headphones as 3 pin, find an existing ~5:1 transformer, I'd go with the Magnequest B7 15K:500 ($100 ea). Still return the primary to the cathode, but run both secondaries to ground. Another excellent choice is the Magnequest TL-404 which is already an autotransformer, and has many taps for lower impedance phones, but the amp won't have enough power for low impedance, low efficiency phones.

Basically I wanted the ultimate platform for any tube as a single-stage amp. I wanted to be able to use the variety of awesome 10K:600 transformers out there, almost none of which can handle DC current, thus CCS parafeed. It might be beatable with some *extremely* high end SE transformers made to handle the current instead of the CCS's, but you lose a lot of flexibility and they'd be extremely expensive to have them custom made. No-one sells the right transformers out of stock, believe me I've looked.
 
May 27, 2006 at 12:04 AM Post #2 of 6
Neat!

I was always wondering anyone would start using ECC99. I think it is still in production.

But I have a question, does this work with Un-Balanced headphones? I guess I feel happier with the tranny tied to groound. Does it work though? I am not certain.

T
 
May 27, 2006 at 5:14 AM Post #3 of 6
Tomo, the bottom of the transformers should *always* be returned to the cathode, not ground. If you use a transformer and not an autoformer like me (the Magnequest B7 15K:500 is best bet, or have electraprint wind something), then you can tie the bottom of both secondaries to ground.

Alternately, if the two halves of your tube bias up closely, you can still use an autoformer. Connect the two cathodes together (there is no signal current in the cathode resistor, so no crosstalk), and then the cathodes become output 'ground'. The 'ground' if your 3-pin output will be at ~4 volts, so don't treat is as a line out, but you can use normal headphones then.

Some pics of construction underway!

High end copper & teflon top plate.

Mounting power supply & CCS underneath

PS wired up.

Everything still works (I have tested the PS before, but not the CCS). Gary's constant current source worked first time, came right up to 62ma (down from 66 in the schematic, I'm only going to run 20ma per channel). The heatsink is a normal flat plate/fan, but I used one of those threaded aluminum block guys instead of a nut. I used insulated 820B mosfets and arctic silver thermal compound.

Tomorrow, I hope to build each channel's CCS & get the VR tubes in, start to lay out the input ground.
 
May 27, 2006 at 7:37 AM Post #4 of 6
Yeah,

That was the answer I wanted. It seemed weird to connect two cathodes together as common-ground of a sort. But I am glad to hear that if the tubes are matched it is OK.

Do you know how much gain you can get from this?

T
 
May 27, 2006 at 7:53 AM Post #5 of 6
Using a 5:1 output, gain is only about 3.5. Probably not suitable with passive-iv no-output-stage DACs, but fine for any DAC with a real output stage or any ~2v rms out preamp. I don't like having to attenuate excessive gain. A 5687 would probably have even less gain, but you can run it hotter (25-30ma) and that gets Rp down plenty so you can do 4:1 output to get it back.

So that might be a good angle to go if you don't want to do balanced phones. 5687 running 28ma each side, about 4v bias or so.. 140v. Use normal 10K:600 transformers for 5K:300, return the primary of the transformers to the cathode, but both secondaries you can ground. The ECC99's plate resistance is high enough I'm not entirely satisfied doing 5K:300 (thus 5:1 autoformer), but it should *work*, might be a little heavy in the 2nd harmonic.
 
May 30, 2006 at 4:28 AM Post #6 of 6
Did a runup today w/ both channel CCS's and the ECC99 installed. I had to replace the CCS current set resistors because the mosfet's aren't quite like the spice models (surprise), but everything is great now.

The main CCS in the power supply doesn't have enough heatsinking so I backed the current off a bit, to 56ma in the supply, and 19ma each channel, 18ma through the gas tubes.

Amazingly, the two halves of my ECC99 came up within a VOLT of each other! 154 & 155v. You can't buy tubes matched that well if you tried.

CCS's, gas regs, and ECC99 wired.

Only the parafeed caps, transformers, and input attenuator still to go. Also gotta put a 4-pin connector on my HD600's before I can listen.
 

Users who are viewing this thread

Back
Top