Another tube preamp/lineamp 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)
I had some rare lab time tonight, and I came up with another iteration of my ever-evolving/devolving tube preamp/lineamp circuit.
Schematic (9kB GIF)

This should please those of you who like to have a volume pot as opposed to stepped gain, or are prejudiced against global feedback loops :wink:

This could be used as-is for a lineamp or an instrument DI, or add a good 1:10 mic input transformer for a 0-60dB mic preamp.

It's fairly low distortion despite the lack of global feedback. The THD is overwhelmingly second-harmonic right up to the clip point (+28dBM).
 
NYDave:
Nice work, your ongoing lineamp. I had been looking for something for the 12BH7 when I first saw your ckt. Maybe someday I'll get to this one.
What are you recommending for the output trafo...anything in particular?

Peace!
Charlie
 
Well, on my breadboard, I used an Edcor WSM10k/600 and only connected half the primary. Performance was pretty good down to 40Hz; there's a rise in distortion down around 20Hz. A true 2:1 (2400:600) output transformer would probably be better, if you can find one.
 
Dave,

Nice to see an original design!

One question though - since the gain is an attenuator between stage #1 and #2 I am curious about how much input level stage #1 can take before starting to distort.

Have you looked into that?
 
Neat.

To be precise: Gain (in theory) goes to negative infinity dB. With a standard Audio pot, it will be twitchy below around 0dB gain, but you can go there. Of course the amp is not intended for low-gain duty.

At any pot position, the input stage will overload around 1V. When used with a 1:10 mike transformer, many high-output mikes on loud sources will clip the input even if you turn the pot down to a sane output level. Be prepared to use a lower tranny ratio or a pad.

At many gain settings, V1A may dominate THD. It would give lower THD if VR1 were higher. 500K could be used if the path to V1B Grid is very short. This can also economize C2 to say 0.22uFd.

C1 is barely big enough for 20Hz. (Remember you are bypassing 1/Gm, about 500 ohms, as well as R3.) For this and other reasons (electrolytic distortion), 100 uFd might be better. If you need a high-pass, I would do it in C2 (about the only other place you can).

With a really clean 12AV7, R2 dominates input noise. In lowest-noise application, lose it or use like 500 ohms.

The WCF can be sensitive to supply rail impedance, since it gets bottom-side drive from a small resistor to the supply rail. And after you followed my earlier advice, I now see this point injecting signal back into R7 and V1B, possibly complicating the analysis. Although with R13 not much bigger than R10, it may not make much difference, especially with the load-sensitivity of the WCF's balance. Consider trying C6 much larger, possibly more uFd than C3. The difference would be asymmetry and THD at low frequencies. It may be moot, but if you have a 200uFd 300V cap on the bench, worth an experiment.

> lack of global feedback.

Well: the WCF is two amplifiers locked in "Full Nelson", fighting each other, and re-distorting each other's distortion. It IS a 2-stage amp with overall feedback. The total distortion may be low, but I bet the high harmonics do not fall as fast as we might like. At 15mA and 2.5K load, you are running rich enough that at +4dBm it may not be a real issue.
 
is that output stage a 'totempole' like the LA2?

how is the rest of your mixer going?
 
[quote author="cuelist"]One question though - since the gain is an attenuator between stage #1 and #2 I am curious about how much input level stage #1 can take before starting to distort.

Have you looked into that?[/quote]

Yes. PRR's figure of 1V is pretty much on the money as far as what I found experimentally. I would quote the maximum average input level as 0dBU. Above this, you should pad the input.
 
[quote author="PRR"]At many gain settings, V1A may dominate THD. It would give lower THD if VR1 were higher. 500K could be used if the path to V1B Grid is very short. [/quote]

Well, one reason I chose the 12AV7--besides the fact that it's cheap and plentiful--is because of its low plate resistance, allowing the use of a lower value of gain pot. With Rp under 10K--in parallel with the 100K plate resistor--a 100K load is just about at the point of diminishing returns. Well, OK, it's actually about 90K due to shunting from the 1M fixed grid resistor. But your point is taken; a lighter load could still improve THD if only by a marginal amount.

As a very rough approximation--erring on the pessimistic side--the input capacitance of V1B is around 42pF. (Gain of 13, Cgk+strays about 3pF, Cgp+strays about 3pF). Let's add 11pF to represent 3" of typical shielded audio cable (or a foot of low-capacitance AES/EBU type). With a 100K pot, the rolloff corner frequency would be around 30kHz. Unless I'm way off the mark with my estimates, I can't go much bigger with the value of the gain pot without moving the corner freq. down into the audio band. But I will check this out experimentally next time I get a chance.

C1 is barely big enough for 20Hz. (Remember you are bypassing 1/Gm, about 500 ohms, as well as R3.)

OK, for 1800 and a 22uF cap, I come up with a corner frequency of 4Hz. The Xc of 22uF at 20Hz is 361 ohms. So, you're saying that the Xc should be much lower than 1/Gm? I never really thought of it that way--I always thought of it more in terms of Xc being selected to be much lower than the cathode resistor (Rk). But what you're saying does make sense, since--if I'm visualizing it correctly--the tube's cathode impedance is effectively in parallel with Rk and Xc.

With a really clean 12AV7, R2 dominates input noise. In lowest-noise application, lose it or use like 500 ohms.

Good point. The grid stopper is kind of a knee-jerk habit left over from my days of designing guitar amps, where it was often necessary. R2 could probably be eliminated. The 1K stoppers on the grids of the WCF are definitely needed, though; otherwise, the behavior gets pretty wacky just below the clip point.

Consider trying C6 much larger, possibly more uFd than C3.

It certainly couldn't hurt. As drawn, PS ripple is attenuated by about 20dB at the first pole in the filter, but I assumed a regulated supply. A 47uF cap would be even cleaner and wouldn't take up much more space. With the 22uF cap, the signal from V1B plate is down about 60dB, and the signal from V2A plate (which is already substantially lower than the input) is down 16dB--and they're in antiphase--so I don't believe coupling between the two stages is much of an issue there.
 
Looks like I'm talking to myself :wink:
I guess threads about theory and new designs will never attract as much interest as cloning.

Anyway, I drew some plate curves to compare the performance of V1A in my circuit running into a 100K load as opposed to the ideal "infinite load" condition.
12av7loadlines_sm.gif


What this seems to be telling me is that there's not a terribly huge difference in terms of gain and output swing for small signals below the point of grid conduction. The stage could be made more linear for larger signal excursions by altering its slope to fall across the more linear portions of the curve--moving away from the bottom of the graph, where the curves start to bunch up--but the penalty would be higher plate current and reduced voltage gain.

I've learned something valuable from this little enterprise. I've been reminded that drawing and interpreting load lines is fucking tedious :wink:
 
NewYorkDave

I would not assume that no replys means no one is reading. Sometimes people are learning IMO. This might be part of whyI like this forum people just don't jump in without knowing about something to add or ask about.

There have been times I have posted an idea or question with little feedback. I know if I don't have a good reply I often will not post. I think this is a "problem" with non face to face or phone interactions.
 
> Looks like I'm talking to myself.

No. the scroll-rate in this one section is way too fast. This thread must have fallen off the first page, which pretty much dooms it.

> I guess threads about theory and new designs will never attract as much interest as cloning.

I think the slower more-considered pace of theory/design threads would justify having a section just for them.

But we had that discussion already, and the consensus was against it.

> 12AV7... Rp under 10K

I figured 16K, but whatever.

> input capacitance of V1B is around 42pF. ... add 11pF

Granted.

> With a 100K pot, the rolloff corner frequency would be around 30kHz.

One of us is wrong.

The impedance at the wiper of a pot, with top and bottom at zero impedance to ground, varies from zero to R/4. The wiper of a 100K pot won't go over 25K. Taking Rp=15K, not over 115K/4= 29K. With ~50pFd capacitance, -3dB at about 110KHz.

Granted my 500K would be awful marginal.

> not a terribly huge difference in terms of gain and output swing for small signals below the point of grid conduction

Granted that I wasn't really thinking about plate resistance and signal level: a higher pot value won't significantly reduce distortion and will make layout critical.

> drawing and interpreting load lines is tedious

Which is why I wasn't going to do it...... and was not-right.

My next bad guess: bias V1A to put the plate voltage down around 60-70V. Gain may rise and noise (if all else is clean) may drop. Might need to go that way for ribbon mikes.

> Xc should be much lower than 1/Gm?

Lower than 1/Gm||Rk. This will typically be half to a quarter the cap you estimate from just looking at Rk.
 
[quote author="NewYorkDave"]
Schematic (9kB GIF)
[/quote]

What is function of R2????????? . Is it additional noise resistance?
With some tubes 1 k value of noise at input is dominant.
10 k is Hell.
Why we are doing transformers with good noise figure if
it can be destroyed by simple resistor?
Why you use mumetall transformer. You can use cheap ferrite to
obtain better results that you by mumetall and 10 k resistor.

If it is good for stability to avoid Kurz-Barkhausen s oscillations
and other hell, why you not bridge this resistor by some small
vf choke. This choke cuts resistor s noise at audio frequencies.
Or to use ferrite bead on grid input?
xvlk
 
> What is function of R2?????????

Asked and answered.

In guitar-amp design, it modifies grid-overload for gentle clipping with less overload recovery. It can also be handy on the test-bench, to let you know that you are overdriving the input with a very low impedance source and seeing results that won't happen in normal use.

Noise won't be awful with R2=10K. And if someone notices noise, it is trivial to short it out. (I think it would take awful layout to get MHz oscillations with this rig.)
 
[quote author="PRR"]> What is function of R2?????????

Asked and answered.

In guitar-amp design, it modifies grid-overload for gentle clipping with less overload recovery. It can also be handy on the test-bench, to let you know that you are overdriving the input with a very low impedance source and seeing results that won't happen in normal use.

Noise won't be awful with R2=10K. And if someone notices noise, it is trivial to short it out. (I think it would take awful layout to get MHz oscillations with this rig.)[/quote]

Yes, it is somethink about grid - leak limiter.
Then you can R2 can be bypassed by small capacitor.
Grid - leak limiting will be still in function and noise wanish.

But if you want 10 k at output, cheep ferite transformers have 10 k
naturally.

xvlk
 
[quote author="PRR"]
I think the slower more-considered pace of theory/design threads would justify having a section just for them.

But we had that discussion already, and the consensus was against it.
[/quote]

:sad:
Can we turn this idea over one more time ?
... cos I like the idea of a slower pace for these more thinking threads.

The long term worth of the LAB, I think, will be in this type of thread. They may bring new diy projects and solutions. The never ending newbie question, Meta threads and "look what I made" is fine, even though I think these could largley be replaced with FAQ and some more specific LAB Web pages.

A project and theory development section could be worth while and might even bring an EE or two that so far has not been tempted to joined us. You just never know what kind of George might turn up.

In a world of Geeks there is still some that might be more Geekier than most. :wink:
 
[quote author="PRR"]I think the slower more-considered pace of theory/design threads would justify having a section just for them.

But we had that discussion already, and the consensus was against it. [/quote]

I have mixed feelings. On the one hand, a separate forum for theory and original designs would certainly enhance my enjoyment of this place. But at the same time, it could make it altogether too easy for unscrupulous manufacturers or other ripoff artists to harvest other people's designs that were intended for noncommercial use by hobbyists. Of course, something as mundane as having to sort through a few pages of postings isn't going to be enough to dissuade a suitably motivated pirate. I guess I can see all sides of the issue, and therefore I am not qualified to run for President :wink:


12AV7... Rp under 10K

I figured 16K, but whatever.

No, you were right, I was wrong. I said "10K or less" based on a reckless assumption made after glancing at data on a "typical operation" table. But when I go to my loadlines and draw the tangent to the curve where the Q-point sits, I come up with 16.6k.

Granted my 500K would be awful marginal.

But still usable if you're careful with layout. The criteria I used was simply that the frequency where the reactance of Cin equalled the resistance of Rg should be above the audio band. That would ensure that the rolloff would be unobjectionable regardless of variations in the plate R of the tube, the setting of the pot, etc.

Simple passive circuits are one of the few instances where I'm willing to trust a simulator instead of using paper and a calculator or physical breadboarding. Here's how I draw the equivalent circuit of V1A coupled to the input of V1B:
rolloffequivckt.gif

Note that I did not include the fixed grid resistor in this analysis. That's because it's assumed to be either sufficiently large with regard to the pot, or simply not used at all. The pot is at 50% resistance to give the "worst case" source Z.

Here's the frequency response with a 100k pot:
rolloff100k.gif

-3dB at 100kHz relative to 8.5dB loss

with a 250k pot:
rolloff250k.gif

-3dB at 42kHz relative to 7.9dB loss

with a 500k pot:
rolloff500k.gif

-3dB at 24kHz relative to 7.6dB loss.

So a 500k pot is usable if you do need that extra dB and can tolerate a tiny amount of droop at the very top of the audio band. 100k gives the widest bandwidth with little loading of the tube, and 250k is a good compromise if you want even less loading but still insist on keeping the lowpass corner frequency well above audio.

drawing and interpreting load lines is tedious

Which is why I wasn't going to do it...... and was not-right.

I hope you don't think my aim was to prove you wrong. I like it when you and other sharp people criticize my designs; that causes me to consider my own reasoning more closely, keeps my analytical skills (such as they are) sharp, and ultimately leads to better circuits!

My next bad guess: bias V1A to put the plate voltage down around 60-70V. Gain may rise and noise (if all else is clean) may drop. Might need to go that way for ribbon mikes.

I understand this to mean increase the standing plate current and in turn increase the drop across the plate resistor. Or do you mean using a larger Rp and keeping all else the same?
 
> don't think my aim was to prove you wrong.

Please point out when I am wrong. I write so confidently that folks are afraid to call me on my slip-ups.

> I said "10K or less" based on a reckless assumption made after glancing at data on a "typical operation" table.

Those "typical" points are often not typical for voltage-amps. Rather they show-off the power of the tube.

Same on the plate curves. They take a few numbers, mostly at higher currents, and then use a French Curve to draw nice lines. Don't squint too far into the lower left corner-- it may be totally bogus.

I got my number from a plot of Gm, Mu, and Rp against current. This extended well below your operating current, with a plate voltage nearly what you used, so I assumed it had hard-data near your operating point and interpolation would give reasonably right answers.
 
[quote author="NewYorkDave"]
I have mixed feelings. On the one hand, a separate forum for theory and original designs would certainly enhance my enjoyment of this place.

But at the same time, it could make it altogether too easy for unscrupulous manufacturers or other ripoff artists to harvest other people's designs that were intended for noncommercial use by hobbyists.

I guess I can see all sides of the issue, and therefore I am not qualified to run for President :wink:
[/quote]

Dave has mixed feelings :wink: and yet that last post is a perfect example of what might end up in a forum like this.
GREAT POST
good words
pictures
well presented

one day in the distant future someone might enter a thread like this and suddenly ...
" I get it "
it makes it all worthwhile !

Often, I can't be bothered with all the deep theory that some of you guys come up with so a place where you can indulge might bring the rest of us with you. Some other guys might also open up and lay some sensational idea or application.

I think we need this.

As for the " unscrupulous manufacturers or other ripoff artists to harvest other people's designs that were intended for noncommercial use by hobbyists. "

BRING THEM ON !

If they can understand the idea and then bring it to production at a price that is fair ... I might even buy one myself.
On the other hand if a top flyte manufacturer comes to us and drops some good info on us then that's great. We have already been blessed with some great info from hard working commercial fellows.


[quote author="PRR"]Please point out when I am wrong. I write so confidently that folks are afraid to call me on my slip-ups. [/quote]
:green:
There's an invite ... so do it people !
No heros around here , JUST healthy debate and exchange of ideas.
I know PPR would be stoked to be shown wrong and prove that some on IS looking at his posts and taking it all in and THINKING about it.
:thumb:


whatever ... it's all good

just my 4 cents.


OH ! I almost forgot

... Dave for President ...
 
Back
Top