Tube line amp circuit

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

NewYorkDave

Well-known member
Joined
Jun 4, 2004
Messages
4,378
Location
New York (Hudson Valley)
My on-again, off-again attempt to build a tube mixer continues to yield some useful (to me, anyway) bits. I had some rare free time the other day, so I breadboarded a line amp circuit.

schematic (10kB GIF)

This can put out a pretty hefty signal, and can even drive 600 ohms directly with reasonably good performance. (Here's a good use for all those UTC 600:600 +30dBM transformers you have laying around :wink: ).

Into 600 ohms, distortion is pretty low right up to maximum output, at which point the output stage goes into cutoff and the THD soars. Into a higher-impedance "bridging" load, it can swing up to +35dBU before the output even starts to flatten. Through a 2:1 stepdown transformer, it can pump +27dBM into 600 before clipping.

Open-loop gain is about 40dB and there's 20dB of negative feedback as shown, giving a closed-loop gain of 20dB. The 22k feedback resistor could be replaced with 56k to give a loop gain of 26dB, but the THD figure suffers somewhat. I would not recommend using less feedback than this.

The THD at all levels is mostly second harmonic. When working into a bridging load, at +30dBU, there were no measurable harmonics past 5th. I'm defining "measurable" as being above the noise floor, which on this breadboard was about -88dB (unshielded wiring floating all over the place). Higher-order harmonics appear at measurable levels when working near +25dBM into 600 ohms. Interestingly, when I added the grid stopper to the lower half of the 12BH7, the maximum output increased, but so did the proportion of odd harmonics.

These measurements were made with resistive loads. Through a transformer, the percentage of odd harmonics increased somewhat (probably because I was using a cheap transformer). The figures include whatever noise and distortion my old HP 200 tube oscillator was contributing. Individual harmonics were measured with an HP wave analyzer, from the fundamental all the way up to the highest harmonic that would produce any sort of reading above the noise floor.
 
Whoops... I just noticed a little oversight on the schematic. Contrary to the text, C2 and C3 do NOT have to be 250V or greater :wink:. They can be 25V 'lytics. Also, it would be good to use 400V types for C1 and C6, since these points will rise to full supply voltage if the tubes are unplugged. I'll update the schematic when I get a chance.
 
Thanks for one more very usefull circuit!

Would it be possible to make the gain variable by means of variable feedback?

Also, what pot value would you use to attenuate the signal before the 12bh7 tube?

You know, I would like to try this circuit myself, but would be nice to have gain and output control, if it´s possible.

Also, let us know how you are going with your tube mixer. Is there any metalwork already done?
 
The notes on the schematic are now corrected.

[quote author="rafafredd"]
Would it be possible to make the gain variable by means of variable feedback?[/quote]

Yes, within a limited range. Too much feedback, and the circuit (like most others) becomes unstable. Also, if the feedback resistor is reduced too much, it loads down the output tube. If there's too little feedback, generation of upper harmonics becomes objectionable. Just as a guess, I believe the circuit should work well with feedback resistor values from 6.8K (for about 9dB of gain) to 56K (about 26dB gain). I don't recommend using it outside of this range.

Also, what pot value would you use to attenuate the signal before the 12bh7 tube?

None. It's not a good idea to put a volume pot inside a feedback loop. Besides, even if there were no feedback loop, the plate of the second triode is direct-coupled to the grid of the 12BH7, and provides the bias for it, so that part of the circuit would have to be totally redesigned for AC-coupling.

Since this is a line amplifier, I would use a 100K pot before the grid of the input tube as a volume control. This isn't a good idea for mic preamps (since the signal level is so low at that point), but it's fine for line-level signals.

This can, however, be used as a mic preamp (of up to 46dB gain) if you put a suitable 1:10 mic transformer in front of it. In that case, of course, you would not use a pot on the input.
 
I found some more "lab time" today, and I've cooked up an improved version of the circuit with a wider gain range and lower distortion. I'll post more info later after I have a chance to make a drawing.
 
Here's the fruit of today's labors:
9kb GIF

Please note that, among the other changes, the input tube is now 12AV7. This tube is kind of like a 12AU7 on steroids. It has more gain, higher transconductance and lower plate resistance; and like a 12AU7, it can handle large voltage swings without clipping. I didn't use the usual 12AX7 here because I wanted the amp to handle input signals up to line level without clipping, and the 'AX7 overloads so easily.

With higher open-loop gain available, the closed-loop gain can now be increased to as much as 40dB while still maintaining a reasonable feedback factor. This means that the amp can serve as either a line amp or a preamp. With a 1:10 input transformer, gains of up to 60dB are possible. When set for lower gains, such as 20dB, it can drive a 600 ohm load to almost half a watt with low distortion.

At lower gains, what little THD there is, is almost all second harmonic. For instance, at 20dB gain and +25dBM output, second harmonic is 0.15%, third is 0.03%, 4th is 0.02%, 5th and higher are buried in the noise floor, and THD is 0.2%.

At higher gains and high output levels, some higher harmonics start to creep in, but second harmonic still predominates. At 40dB gain and +25dBM output, second harmonic is 0.89%, third is 0.17%, 4th is 0.01%, 5th is 0.02%, 6th is 0.03%, 7th and 8th are at .01%, 9th and above are down in the noise floor, and THD is 1.1%. The rise in distortion at higher gain is not hard to understand, since at 40dB of closed-loop gain we're only about 14 dB below our open-loop gain. That's about the limit of what I consider acceptable--I don't want the THD to be higher than 1% at peak operating level, worst case--and that's why I specify a maximum gain of 40dB for this circuit.

My sloppy breadboard, with its wires flying all over the place and its haphazard grounding and decoupling, was not stable below 14dB of gain--so I put that down as the minimum gain figure on the schematic. This could probably be improved in a properly-built prototype, or with some tweaking of the open-loop frequency response.

C5 is shown as a 22uF/250V cap; but if one could find a quality cap of a higher value in the same voltage rating, that would be even better.
 
are these gonna be the make up amplifiers? did you say its all passive apart from that?
 
Nice to see a circuit using the 12AV7. (since I have a box of RCA black plates... :wink: and since I haven't had time to sit with the Sylvania manual I have that has the RC amplifier data in it...)
 
are these gonna be the make up amplifiers? did you say its all passive apart from that?

Yes and yes. It's a line mixer, so there are no preamps. It's just a passive mix network followed by a preamp, a volume control and a line amp. This is duplicated six times for left and right program, reverb sends 1 and 2, and PFL/monitor outputs left and right.

I'm fond of the notion of a "universal" line amp/preamp because I like the idea of building a console using the same amp in all positions. That's what Langevin had in mind when they brought out their 5116B in the early '60s, so it's not a new idea.

Nice to see a circuit using the 12AV7. (since I have a box of RCA black plates... and since I haven't had time to sit with the Sylvania manual I have that has the RC amplifier data in it...)

If you don't have your Sylvania manual handy, I found a couple of scans posted on the web:
Data page
RC-coupled amplifier data

I normally avoid designing circuits with out-of-production tubes, but I couldn't resist this time because the 12AV7 was just perfect for my application. Plus, there's still plenty of them around.
 
If it were to be used as a pre amp, what method would be the best for balanced out? I'm thinking just a transformer right?
 
Yep, a good quality 1:1 transformer should work fine. Maximum output might be a little lower than what I measured into a resistive load since the reactive load of a transformer is harder to drive. You could also use a 2:1 output transformer and adjust the circuit for 6dB more gain to compensate for the stepdown of the transformer, as long as the gain of the circuit is not adjusted above 40dB.
 
hey dave,

this is why your console will sound better than mine. as you only have a handfull of active stages you can choose your components carefully without two much hassel sourceing them all. i have to use whatever i can buy cheaply in multipuls of 16.
 
C5 needs to be rated full supply voltage. It will probably go there during power-up.

> I wanted the amp to handle input signals up to line level without clipping, and the 'AX7 overloads so easily.

One of the big Macintoshes has a 12AX7 fed 200V input signals.

And I know I have used a 12AX7-type as input and feedback point with 8V peak input signals. The grid bops 8V, the cathode bops 7.7V, the grid-cathode only sees the 0.3V difference.

If feedback is "effective", then the input signal level does not matter.

> not stable below 14dB of gain

Did it oscillate at 1 Hz or 1 MHz?

At the top: those grid-stopper resistors seem excessively high. None of these tubes are liable to take off in 10 MHz oscillations, and 10K is liable to add to your 1 MHz phase shift.

But your bass coupling sure looks unstable. Three caps: this is a phase shift oscillator. Stagger the poles as you have done and it isn't too bad, but C5 is load-sensitive so you don't know where it will end up in actual use.

Didn't I post a fewer-pole variant? Maybe not. revised schematic

This has only one dominant bass-pole inside the feedback loop. (The cathode caps cause phase shift only over a narrow band and never reach large phase shift.)

The output cap does not have feedback around it, but it should be large anyway, and you can budget the lost feedback cap to make C5 47uFd instead of 22uFd.

You need an input cap now, but for general use you need one anyway. (Who knows what is on the other end of the wire?) BTW, because of bootstrapping, the input cap can be VERY small, even 0.001uFd.

This should be unconditionally bass-stable at any gain. The limit is now how much DC upset the feedback network causes. At 8K, it may be considerable. You could increase the feedback loop impedance: while this reduces the gain of V1a, it also linearizes it, so THD may not change much. Or you could even return R3 through a large cap to ground. Now V1a's cathode will be up over half supply voltage, but you already have V2b's cathode at that height. V1a will work with less voltage, but V1a does not have to make much signal to slap V2b into clipping. And you could make R13 fixed and vary R3 to set gain.

If working only at low gain, you might stick with your plan but make C1 MUCH smaller. Like 0.001. The 170Hz roll-off will vanish in the feedback.

The power decoupling is in the wrong place, maybe. The output stage has about the same PSRR as V1b, so they can eat the same power. V1a runs at lower level and might profitably be decoupled. However I suspect that all three stages could run on the same decoupling network and not be unstable. You can't put three high-gain inverting voltage amps on one decoupler, but the WCF has no gain and no phase inversion.
 
Right. I should have said that the 12AX7 biases optimally at a lower grid voltage, and therefore overloads more quickly with high grid to cathode voltage swings. I did once consider "tapped biasing" as you show in your mod, so that G and K would swing together with a fairly constant differential between them, but I wasn't hot on the idea of putting yet another coupling cap in the path, as well as limiting the maximum voltage swing of that stage, which is what fixing the whole deal more volts above ground will do. Perhaps I should rethink that. The grid-to-cathode voltage of the second stage still presents a limiting factor, though, which means that the 12AX7 might still not be suitable for this circuit if I want it to handle big swings.

[quote author="PRR"]Did it oscillate at 1 Hz or 1 MHz? [/quote]

Certainly closer to 1MHz than 1Hz, although I didn't bother to measure the frequency of oscillation. I just saw that telltale thickening of the trace on the 'scope when I increased the FB beyond a certain critical point.

At the top: those grid-stopper resistors seem excessively high. None of these tubes are liable to take off in 10 MHz oscillations, and 10K is liable to add to your 1 MHz phase shift.

Actually, those are there just as much to prevent grid blocking as to prevent HF oscillation. Without the stoppers on the WCF, the clipping behavior is asymmetrical. The values are arbitrary, though, so I'll try lower values and see if they work satisfactorily.

But your bass coupling sure looks unstable. Three caps: this is a phase shift oscillator. Stagger the poles as you have done and it isn't too bad, but C5 is load-sensitive so you don't know where it will end up in actual use.

I actually outsmarted myself with that one, since I was closer to getting it right the first time around. (See the first schematic in this thread). On the second iteration I thought, perhaps naively, that tapping off the FB after the coupling cap could provide some automatic compensation for load reactance as long as the output cap was suitably large to avoid excessive phase shift with any sane value of load impedance.

Your modification makes more sense in many ways, but I was dead set against having high DC across the feedback resistor because I wanted the circuit to lend itself to variable gain if desired, e.g., switched feedback resistors. Perhaps tapping off the feedback as I did in the first design, along with coupling to the cathode as shown in your mod, would be the best compromise. Or, R13 and R3 could be fixed at whatever values correspond to the lowest gain desired, and R3 could be shunted by a variable or switchable resistance (coupled to ground through a big cap) to increase gain as desired.

You can't put three high-gain inverting voltage amps on one decoupler, but the WCF has no gain and no phase inversion.

Actually, the upper half of the WCF is a common-cathode amplifier and is thus inverting, but it certainly doesn't have any gain. In fact, it has loss from grid to plate.

Thanks for your constructive criticisms. I'm planning on a little lab time tomorrow (actually later today, since it's now 4:15AM) so I will try out some of the ideas you've suggested.
 
I did get a little lab time yesterday, but not enough to try out everything I wanted to try.

At any rate, here's a schematic of yesterday's iteration. I rearranged the PS decoupling for better PSRR. I also eliminated a pole in the feedback network, adjusted the DC conditions in the 12AV7 stages, and increased the size of the coupling cap that feeds the lower grid of the WCF. I don't have my notes with me at the moment, but I'll write in the voltages later if I get a chance.

Maximum output is a couple of dB lower than the previous design, due to reduced plate voltage on the output stage, but is still plenty for most purposes. The amp is stable down to 10dB gain. Next time, I'll play around with compensating the open-loop high frequency response to roll off below the frequency where the phase shift is greater than 180.
 

Latest posts

Back
Top