CELLINI-2 a hybrid

[Bigun]

Having built SS amps and recently my first SET (http://www.diyaudio.com/forums/tubes-valves/167872-my-cellini-triode-amp.html) I'm interested in bringing these families together to explore a hybrid that might offer some of the sonic benefits of the triode but with greater current drive and much less weight in iron. Well actually I'm just curious.

I've picked out what I feel are two proven building blocks. I'm not sure what kind of PSU I want yet, the circuit allows for different voltages for the tube and the buffer if desired.

The front end is a 6N23P triode that I believe to be flexible enough to operate at lower B+ and yet deliver good results. If I allow higher B+ I might also be interested in using something I already have on the shelf such as a 6SN7 or triode wired 6E5P.

The back end is the Krill diamond buffer invented by Steven Dunlap. It has been used in his all-SS amplifiers for a number of years with reports of good sonic results.

So the question is - what do I need to know and consider here having never made a hybrid before ?


p.s. I figure the tube/valve forum is a good place for this because you guys aren't afraid of glass or sand !

 

[Arnulf]

Keep the offset potentiometer value large enough to keep the tube load reasonable, add a grid stopper and consider adding some capacitance to compliment grid stopper and Cmiller. 6N23P (~= ECC88) is a VHF/UHF tube, you can never be too careful

Otherwise looks very straightforward, build it and report how it sounds !

 

[Bigun]

Keep the offset potentiometer value large enough to keep the tube load reasonable, add a grid stopper and consider adding some capacitance to compliment grid stopper and Cmiller. 6N23P (~= ECC88) is a VHF/UHF tube, you can never be too careful

The offset potentiometer recommended for Krill is around 1Meg, although I suspect there is some tradeoff with noise ?

I've never worked with a VHF tube before, I'll follow your advice to avoid parasitic oscillations, I think I've read similar suggestions for use with high mu tubes.

 

[pedroskova]

You could probably use a voltage doubler / tripler off of the SS PSU to provide for an ECC88 / 6dj8 type of tube for 100 or so volts. That would just leave you with a small filament transformer.

 

[Arnulf]

The offset potentiometer recommended for Krill is around 1Meg, although I suspect there is some tradeoff with noise ?

Yes but I don't think you went for CCS just to load it down with say 2.5K or so (two halves of 10K in parallel) You trade distortion for noise and vice versa and the general consensus established, uhh, well long time ago, seems to be that Rload should be at least 2-3x Rinternal for bearable amount of distortion, and more if possible.

 

[Bigun]

[...]seems to be that Rload should be at least 2-3x Rinternal for bearable amount of distortion, and more if possible.

...maybe the 'load' resistor gives me a means to tweak the sound of this thing ?

Anyhow, the datasheet I have for the 6N23P is one of those short-form bilingual jobs, it's not that clear to me what the internal resistance of this tube is - does anyone have a better datasheet perhaps ?

Pedroskova, a doubler's a good suggestion, because the CCS plate load should have good PSRR since a doubler will otherwise increase B+ noise I think ?

 

[pedroskova]

Pedroskova, a doubler's a good suggestion, because the CCS plate load should have good PSRR since a doubler will otherwise increase B+ noise I think ?

Yeah, they can be noisy, but maybe some of the design solutions from Cavalli Audio might work.

Caveat emptor...I'm no EE

 

[SemperFi]

Nice. That output buffer looks like one of those buffer chips that pinout like opamps but are only current boosters.
Why will a doubler be more noisy? U can make fullwave doublers too.
Will a unipolar positive voltage doubler create some asymmetry in the Vcc+/Vee- which inturn might give u a DC offset problem?
Then again, a doubler is probably more of a hastle than its worth. The input cap is a good idea anyways. I think ur schem is pretty good to go. But hey, we must discuss all options here!

 

[SemperFi]

...but really. Shouldnt the ouput stage have its own separate supply anyways? I'm nitpicking maybe. Since u think hybrid u ought to try a mirror load for the tube and simply DC couple the whole way.

 

[Bigun]

There are two reasons I thought a doubler could add to noise is simply that we have more diodes creating switching noise inside the box (I'm thinking of a single chasis). But it's probably not an issue to worry about. And a separate supply for the front end is also probably unnecessary with proper decoupling and the PSRR benefits of a CSS (or a choke).

With a thermionic valve plus a Krill buffer I should be calling this amplifier the "THRILL"

 

[anatech]

Hi Gareth,
You will need a regulated supply for the tube B+, omit this at your peril!

Why? Because you have a feedback network that exists through the power supply - even with a separate winding for HT. This will also greatly reduce / eliminate any hum issues as well as cross channel cross-talk. A reg for each channel would ensure great separation between channels.

The DC offset shouldn't be very much here. Why not first try a 100K resistor to signal ground first? From there you might have some more knowledge as to the magnitude of the possible DC offsets. When using straight Diamond buffers, the offsets I was seeing where on the order of 35 mV maximum. Not worth correcting if you ask me. These were also connected to a tube front end in a hybrid. If you find that you do get a DC offset that is stable, try fine tuning one of the current sources first. That way you avoid touching the audio path as much as possible. Of course, it would also be easy enough to use a pair of zener diodes to fix a lower voltage, adding capacitors to filter the DC. From each end you can connect your control, then take a resistor in series to your audio input to the buffer. No problems loading your audio down.

As for generating switching noise, that's easy to avoid. Don't use stupidly high values of capacitance in your filter. You should probably be able to get away with 47 uF or lower along with some series resistance and an R-C-R-C type filter before your regulator(s). Bias your heaters up to about + 30 ~ 50 VDC DC while retaining the AC feed (Don't use DC heater supplies!). The regulators may be a simple zener - pass transistor type. If you want one that performs really well, use a current source for the base-zener feed instead of the more common resistor type. That will create a nicely engineered, quiet power supply solution for you.

Your plans for a CCS load will work great. You can adjust that current for minimum distortion, and it will sound better. Keep in mind that you will retain the basic non-linearity of the tube, it will just be reduced a fair amount. Good choice.

-Chris

 

[Arnulf]

...maybe the 'load' resistor gives me a means to tweak the sound of this thing ?

Anyhow, the datasheet I have for the 6N23P is one of those short-form bilingual jobs, it's not that clear to me what the internal resistance of this tube is - does anyone have a better datasheet perhaps ?

Sorry, I have only now noticed this post.

mu = gm * Ri -> Ri = mu / gm

('S' is the other name for transconductance, 'gm')

I'm pretty sure your datasheet contains the two values needed, no matter how short it is

 

[Bigun]

Ri = mu / gm

good point and thanks !

Well, in this case Ri = 34/12.7 k = 2k7 which will be roughly the output impedance of my common cathode voltage amplifier stage for most frequencies of interest.

Now, the input impedance of the diamond buffer is pretty high, at least 1M as I have 3 cascaded emitter followers in it. I assume this is good enough ? And potentially it's worth designing things so that this Zin is dominated by the dc bias resistor so that Zin is less signal dependent = more linear ?

 

[Arnulf]

Now, the input impedance of the diamond buffer is pretty high, at least 1M as I have 3 cascaded emitter followers in it. I assume this is good enough ?

Definitely, I only mentioned this because of your biasing potentiometer arrangement the value of which wasn't marked. It would be pointless to ruin all your hard work invested in bringing load impedance up by overlooking that potentiometer or going for a "safe" (low-Z) value while hoping for lower noise (which you mentioned earlier).

 

[Bigun]

Thanks Arnulf, good advice. I actually don't have a high value pot, so I'll go with Anatech's suggestion of a simple resistor to ground at first. Actually, I think I can make better progress if I just build the output stage and use my Cellini original as the tube frontend.

It'll be a week or so as I have some other business to take care of first.

Anybody any strong feelings on choice of current source ? I was wondering about my options and there are quite a few, including bootstrapped resistors driven off the output.

 

[jarkaa]

Regarding PSU

Hello Bigun,

I am following your progress with great interest because I have similar thoughts of using the Krill OPS with tube front end since I got a bagfull of NOS tesla e88cc tubes.

If you take a look at Dunlap´s power supply schematic for the Krill there is a voltage doubler (tripler) already in it. It is regulated down with zeners and silicon.

My current Krill PSU measures +-95 V before the regulators with 36Vac ct trafo.
My initial idea was to modify the regulators with 82v zener diodes to obtain higher output voltage for the tube or replace Dunlap´s regulators with Maida´s.

 

[anatech]

Hi Gareth,
Keep your CCS simple and cheap. A cheap (not high brightness) red LED as a voltage reference, emitter resistor and a transistor is all you need. If you're dealing with very high or variable voltages, cascode the current source with a power transistor an dyou'll have close to the best there is.

Now for the best part. Mount the LED in thermal contact with the current setting transistor and the CCS is now temperature compensated.

I'm not crazy about bootstrapped type circuits. It's those capacitors that are now in the circuit.

Hi jarkaa,

My initial idea was to modify the regulators with 82v zener diodes to obtain higher output voltage for the tube or replace Dunlap´s regulators with Maida´s.

Complete overkill. Stay simple to start, then once you have everything working you can play around. You can also now compare the performance differences to see if it's worthwhile, but at least you can design the Maida with known voltages and currents as a design target. So make it go first, then add complexity.

-Chris

 

[Bigun]

I'm back from business trip so I can start to work on this again - although another business trip is lurking for start of November already.

Anyhow, I'm getting quite excited about this project so feel free to pressure me!

I think I need to have something to listen to as quickly as possible so my plan of attack is to start searching my parts bin for one channel and figure out how to kluge something together. Then I can play with it, you know, the usual thing, keep fixin' it 'til it breaks

It will mean using parts left over from my TGM amplifer days, 2SC5200/2SA1943 output devices and BD139/140 pairs everywhere else. As for LEDs, I have a bag of nearly 100 of these strange yellow-orange coloured one's to use up, they light up like really bright tubes !

 

[Bigun]

Hello Bigun,

I am following your progress with great interest because I have similar thoughts of using the Krill OPS with tube front end since I got a bagfull of NOS tesla e88cc tubes.

That's great, always more fun when you're doing something like this to have some others on the same trail. How do we persuade you to start building ?

 

[jarkaa]

Some progress

Hello,

I am drawing a pcb for the Krill follower at the moment.

I already have 2*35Vac trafo, tubes and a proto housing reserved for the project.

I also built a test psu on breadboard according to the Krill´s schematic but without the regulators in place. It seems to give +-49V directly from the bridge rectifier and +-94V after the voltage doublers.

VGS topology gives me trouble. A Common cathode amplifier like this: http://www.tubecad.com/2005/April/blog0042.htm seems tempting but I have very little experience with tube circuits.