Low-voltage Tubebuffer by David Lin

[Nisbeth]

This circuit (http://www.diyzone.net/diy/reference...j8buffer_2.htm - bottom schematic) has been discussed a few times before. I would like to know if it matters if the supply is +/- 24V or single 48V as the buffer already has caps on both input and outputs?

Also, someone mentioned in another thread that the circuit would benefit from a higher supply voltage. Which resistor values should be modified if I change the supply voltage?

TIA!

/U.

EDIT: What's the minimum load impedance that this circuit can comfortably drive?

[SY]

I would be suspicious of the distortion performance- you've got a bad combination of low cathode load and low operating voltages. On top of that, the circuit attempts to drive roughly 20 mA through the tube at these low voltages, which is probably futile. Grid current and concommittant distortion will be the inevitable result.

[Nisbeth]

So you're saying: "Don't bother at all" ?


/U.

[SY]

If you want to use low voltages, it's best to use a tube designed for them (e.g., 8056). If you want to use a common, high-quality tube, be prepared to feed it proper voltages. ECC88 works best with more than 50 volts from plate to cathode, preferably 80-120V.

[moamps]

Originally posted by Nisbeth
I would like to know if it matters if the supply is +/- 24V or single 48V as the buffer already has caps on both input and outputs?

No.

... Which resistor values should be modified if I change the supply voltage?...

Chatode resistor(s).

....What's the minimum load impedance that this circuit can comfortably drive? ..

IMO, 10k or more.

Originally posted by SY
..I would be suspicious of the distortion performance- you've got a bad combination of low cathode load and low operating voltages. On top of that, the circuit attempts to drive roughly 20 mA through the tube at these low voltages, which is probably futile. Grid current and concommittant distortion will be the inevitable result.

The tube in this design is set to (IMO)

Va=40-45V
Ia=6-7mA
Vg=-0.6-0.7V

which is a good setting.

Regards,
Milan

[SY]

Milan, drop a 1K load line (close enough approximation) on the tube curves. Look at the distortion, even divided by mu. Then check out the grid current at a 0.6V Vk-g. Ouch.

[Morgan L]

The design is flawed due to the low value of R4. And 24V is not much for a 6DJ8 to work with. As noted, grid current is a likely problem. The ECC86/6GM8, OTOH, was made for this type of application and doesn't suffer from grid current. And it is probably easier to find than an 8056 nuvistor. It has a lower mu than 6DJ8, but that doesn't matter here, and the pinout and heater are the same. Use 180 ohm for R3 and 5600 ohm for R4, for a quiescent current of about 4 mA. I have used ECC86 in CDPs on the internal +/15V rails (quiescent ca 2mA) with good results, so I'm confident about this tube.

Good luck with your project!

Morgan

[PRR]

> if it matters if the supply is +/- 24V or single 48V

No. In fact I don't see why it eats +/-24V.... maybe another stage in the same box already needed it.

> the circuit would benefit from a higher supply voltage.

Tubes generally do. It isn't easy to make electrons cross empty vacuum, you need to suck hard.

However, the 6DJ8 is not in terrible distress here.

> Which resistor values should be modified if I change the supply voltage?

Triode circuits scale with voltage. You don't have to change anything until the plates glow, or the supply drain becomes a problem.

At over 100V, a 6DJ8 will be running too hot for long life. And passing 16mA, which really is not necessary in Hi-Fi design. Sometime before that point you should change the 820 ohms that SY is picking on to a higher value, 3K to 6K or maybe more.

Lookit. Build it but leave R4 where you can change it easy. 820 ohms, especially at low voltage, will give significant 2nd-order distortion. That may be why you play with tubes. 3K will give enough current and lower THD, which could be good or bad. At higher supply voltages, the THD will be low but the heat and power will be annoying, use a higher R4.

With hi-Z load, and 820 ohms on the bottom, clipping will happen about 4V peak. Just below clipping the THD will be about 0.5% mostly-2nd. At typical power-amp input level of 1.4VRMS, THD will be like 0.2%-0.3%. If your system is so clean it is "sterile", this is a lovely dollop of sweetness. If your system is colored already, this won't do much.

> What's the minimum load impedance that this circuit can comfortably drive?

It is already (as shown) driving ~1K. An added load higher than 1K won't hurt it much. For lowest THD you want a bigger R4. With high supply voltage R4 can even be bigger than load resistance, but with the fat 6DJ8 you may not want to go past 10K.

As drawn, the 1uFd output cap gives -3dB at 17Hz (-1dB at 34Hz) with a 10K load, so you really don't want to go even that low unless your woofers don't go below 80Hz. Keep the load well above 10K or put more uFd and money in C6.

> bottom schematic

Oh, crud!!!! I was looking at the top one.

morgan> the low value of R4. And 24V is not much for a 6DJ8

I think Morgan is also looking at the top plan, because R4 in the bottom plan does something else and certainly is not "low". Also I think Milan is right that the DJ runs with much more than 24V across it (in either plan): this is not the simple "split" that you expect when you see bipolar supplies.

As for the bottom plan: I'd say you could kill the 2SK170 that way even at +/-24V or +48V..... no, you have to go close to 100V total to fry the '170.

I'm not sure why both the 6DJ8 and the '170 are self-biased: seems like the hard way to set an operating point.

With the '170 trimmed to some "good" operating point for the 6DJ8, THD in hi-Z loads will nearly vanish. At 10K load it will be under 0.05%, so I'm not sure why you would do it this way instead of (heresey!) ditching the 6DJ8 and using the '170 as a simple cathode follower. Either would be approximately equally transparent, but you avoid 3 watts of ugly heat.

[SY]

Geez, PRR, I was looking at the top one, too! Replacing the bottom resistor with a current source will be very helpful- with the resistor in that spot, the downswing looks terribly different than the upswing. The CCS (well, sort of constant, anyway), can help that for the first volt or so, but that particular implementation has a limited compliance and output resistance. The latter is not nearly as much of an issue as the former.

Using FETs as CCS is what put me off of using CCS for decades before I was persuaded to try them again. Using stiffer circuits (like the cascode bipolar CCS in my preamp) turned me around.

If I had to guess about the use of a +/- rail instead of a straight 48V, I might indeed suspect that the bipolar supply was already there. The other possibility is that the designer wanted to use 35V instead of 63V capacitors in the power supply and decoupling.

As a side note, I can't seem to read any of the text on the linked page. Is that my browser, or does that page use some kind of weird character set?

[Brian Beck]

With the '170 trimmed to some "good" operating point for the 6DJ8, THD in hi-Z loads will nearly vanish. At 10K load it will be under 0.05%, so I'm not sure why you would do it this way instead of (heresey!) ditching the 6DJ8 and using the '170 as a simple cathode follower. Either would be approximately equally transparent, but you avoid 3 watts of ugly heat.

Heresy indeed! Non-linear input capacitance in the ‘170 is the reason to prefer the 6DJ8 in the prime role. I seem to be the only one carrying this torch here these days (John Atwood on another forum does too). Nasty PIM results, especially with higher gate circuit impedances such as volume pots, although in small doses this can add a zip or false “air” that some find desirable. The “liquidity” of tubes should be interpreted as relative absence of PIM. Worth 3 watts to this old boy.