New Jfet Emulation Idea - The Philbrick Op Amp

[Bill Mountain]

If someone could solder SMD it might be possible to fit this on a DIP socket.

I'm not going to do this myself but I was researching discrete tube op amps when I found this.  Just imagine.  Boutique "Discrete" epoxy coated op amps that you can drop into your favorite vintage distortion for that classic "warm" tone.

Thoughts?

[R.G.]

The Philbricks were the first opamps I ever used. Back in school, we had equipment donated to the school by the military, including a Philbrick analog computer.

It's certainly possible to do this, but like all JFET emulations of tube stuff, it's primarily a mental emulation that makes one feel good about somehow being connected to a tube circuit much more than it is any real emulation of sound or action - at least to the extent that the amplifying devices matter.

If you want a discrete, high performance opamp, look up the schemo for the JE-990. This is *still* preferred in some studios.

And like all opamps, the purpose of feedback is to hide the quirks and induviduality of what's inside the opamp.

On the Philbrick schematic, I suspect that the resistors are given in megohms, not ohms. You'd get precious little signal out of a 12AX7 with a 0.27 ohm plate resistor. I believe caps are in pF.

[Bill Mountain]

The Philbricks were the first opamps I ever used. Back in school, we had equipment donated to the school by the military, including a Philbrick analog computer.

It's certainly possible to do this, but like all JFET emulations of tube stuff, it's primarily a mental emulation that makes one feel good about somehow being connected to a tube circuit much more than it is any real emulation of sound or action - at least to the extent that the amplifying devices matter.

If you want a discrete, high performance opamp, look up the schemo for the JE-990. This is *still* preferred in some studios.

And like all opamps, the purpose of feedback is to hide the quirks and induviduality of what's inside the opamp.

On the Philbrick schematic, I suspect that the resistors are given in megohms, not ohms. You'd get precious little signal out of a 12AX7 with a 0.27 ohm plate resistor. I believe caps are in pF.

I was hoping you'd chime in.  I'm not so into discrete op amps as I amp into looking at other ways to control gain.  I'm back on a tube kick and I've seen simple tube circuits in the past that used feedback for gain control and I thought looking at tube opamps would clue me in.

I've heard of the JE-990 but never thought to apply it to what I use them for.

I guess I could look up some of those high voltage discrete opamp kits (for old school mic pre's) and adapt it to a dirt box.  I've spent so much time dealing with and suffering from the high frequency shrillness of regular opamps that I'm trying to find other (and sometimes super high headroom) options for gain stages.

Oh...and if I were to find a philbrick and an appropriate power supply would it be useful or would modern opamp be way better?

Also, what do you think that part label "8355037" in the top corner is?

[R.G.]

I'm not so into discrete op amps as I amp into looking at other ways to control gain.  I'm back on a tube kick and I've seen simple tube circuits in the past that used feedback for gain control and I thought looking at tube opamps would clue me in.

The basics are pretty simple. You get some gain, and then the gain goes down as you increase negative feedback. The nature of the internal circuit is hidden by a ratio about equal to the open-loop gain divided by the closed loop gain. When that ratio gets over maybe 100:1, you can start ignoring for simple purposes what is inside the amplifier. Super accuracy controlled by only the external components starts in as you run the open loop gain really high. Where the closed loop gain is only a bit larger than the open loop gain, the inside quirks show through to that extent.

Negative feedback always reduces the closed loop gain. It gets less ideal for cases where open loop and closed loop are close to each other. And "less ideal" is probably what you want if you still want tube character showing through.

I've heard of the JE-990 but never thought to apply it to what I use them for.

I guess I could look up some of those high voltage discrete opamp kits (for old school mic pre's) and adapt it to a dirt box.

I've spent so much time dealing with and suffering from the high frequency shrillness of regular opamps that I'm trying to find other (and sometimes super high headroom) options for gain stages.

Shrillness of regular opamps is usually from one of a few causes.
- running out of headroom on signal peaks, as you have devined.
- running out of open loop gain at high audio frequencies where the compensation for the opamp has cut gain so as to keep it stable. This lets the internal distortions show through the feedback
- the opamps are "too perfect" and reproduce highs more accurately, but "more accurately" not being what we expect.

Going to very high headroom, and perhaps discrete to avoid some of the compromises in stability compensation should help with the other causes. "Too perfect" is easy to head off with some judicious filtering in the opamp stages directly.

Oh...and if I were to find a philbrick and an appropriate power supply would it be useful or would modern opamp be way better?

I can't remember the high frequency response of the Philbricks. They may or may not be good enough for audio. They were often used for things which usually had responses under 100Hz, like aiming 16" naval guns.

Also, what do you think that part label "8355037" in the top corner is?

Hmmm. Good question. I don't know, can only guess. It's clearly a series element from the plate of the third triode gain stage to the grid of the final cathode follower. I'd try a resistor. It may be a special packaged circuit that does almost anything, but it could also be a special form of grid stopper. For a solid state emulation, I'd use a 1K and see what happened.

[teemuk]

This might shed some light.


You may also wish to view the entire tube opamp -related section at
http://www.philbrickarchive.org/


Basically, the neon bulbs act like voltage regulating zener diodes. Back in those days there were no zeners or if they were they were about hundred times more expensive than neon bulbs. Some iterations also used thyrites, which were early equivalent to varistors, for the same function.

For accurate "mojo" of the FET simulation you definitely must use neon bulbs with radioactive material, like in original Philbrick opamps.  

[Bill Mountain]

Suppose me (or someone with more free time) were to attempt this...realistically how many of the values could be maintained?

I would think to do it right I would have to read up on op amp design theory and work with the specs of the individual transistors.  But assuming it can be done as a simple Jfet emulation I wouldn't know how to bias the tranny's.

The negative input looks like a simple cathode follower and the positive input looks like a phase spliter.  V2 is where I get lost.  I think baising the second half of V2 (pins 6,7,8) would be the tricky part.  VR1 and 2 are new to me but I'm thinking it is some sort of voltage regulation.

What else would I need to figure out?

I just saw Teemuk's edits so, those must be the bulb things I see in Hifi stuff!

I was making a joke before but a drop in opamp with even worse specs than an LM308 could set the Rat clone world on fire!

[R.G.]

Basically, the neon bulbs act like voltage regulating zener diodes. Back in those days there were no zeners or if they were they were about hundred times more expensive than neon bulbs. Some iterations also used thyrites, which were early equivalent to varistors, for the same function.
[/quote]
Thanks. That helps. They're break-over clamps to keep the grid of the follower from being over-volted. Protection devices, no real function as long as things don't get too big or too fast, or both.

Suppose me (or someone with more free time) were to attempt this...realistically how many of the values could be maintained?

Good question. Unless you had high voltage JFETs or some composite with a JFET and a cascoded high voltage device to keep it from overvoltage, possibly none of them would be the same. At least you'd have to rethink them all. There do exist high voltage MOSFETs in TO-92 cases that could do the job, but then your worries about biasing would certainly be correct.

I would think to do it right I would have to read up on op amp design theory and work with the specs of the individual transistors.  But assuming it can be done as a simple Jfet emulation I wouldn't know how to bias the tranny's.

... as you note.

The negative input looks like a simple cathode follower and the positive input looks like a phase spliter.

The first duotriode is a differential amplifier, cathode-biased by R3 acting like a current source to some extent. The negative input causes the 1-2-3 section of the tube to steal current from the 6-7-8 side which is amplifying.

V2 is where I get lost.  I think baising the second half of V2 (pins 6,7, & would be the tricky part. 

I need to think about the second section of V2 a bit. Looks like a common-grid drive follower, which seems odd. It's not a source follower like I thought. Hmmm. Introspection needed.

[teemuk]

You are overthinking it. The second stage is a simple common cathode amp, followed by a direct coupled cathode follower in output. Both are wrapped inside a "local" feedback loop.

Think of this (naturally with tubes in place of transistors) put also add a single-ended follower output to it and the circuit is pretty much it.


Basically a fine way to spend four tube triode stages to make a 8-watt space heater that performs like a generic solid-state opamp from the weakest end of the scale.

[Bill Mountain]

You are overthinking it. The second stage is a simple common cathode amp, followed by a direct coupled cathode follower in output. Both are wrapped inside a "local" feedback loop.

Think of this (naturally with tubes in place of transistors) put also add a single-ended follower output to it and the circuit is pretty much it.


Basically a fine way to spend four tube triode stages to make a 8-watt space heater that performs like a generic solid-state opamp from the weakest end of the scale.

So is it "warm"!?!?!?!


So...if the stages themselves can be broken down so easily.  What makes it an opamp?  Just the way the follower, phase inverter, cathode gain stage, and follower are connected?

Would it be easier to use the discrete op amp stages in a Blues Driver and forget all of this mess?

Keep in mind this is more for learning as I have no need for this at the moment but it would be fun to work on for a future project.

[R.G.]

You are overthinking it. The second stage is a simple common cathode amp, followed by a direct coupled cathode follower in output. Both are wrapped inside a "local" feedback loop.

Well, actually, that's what my thoughts jumped to in the first place. It's dead simple, and what I expected.

But upon inspecting the schemo a bit, I found that there is no connection from the plate of the third triode to the fourth triode unless the neons are always fired. That seemed odd to me. Fired neons don't act like zeners, the forward voltage folds back. So you'd have a low voltage across the neons and a resistive divider of the 680K and the 4.7M to ground. That could work, but also struck me as odd. The neons can't really be a zener level shifter, so the resistors have to do the level shifting. That may be what happens, but it needs a little thought. So I decided to think.

Think of this (naturally with tubes in place of transistors) put also add a single-ended follower output to it and the circuit is pretty much it.

Yep. That's where I started, instantaneously. Then the shifting and neon issues occurred to me, and I decided to think. In any case, it can't operate like that because there are no PNP (or negative plate) tubes.

As I said, it needs a bit of thought. It may well contract down to the standard model of a three stage opamp, with diffamp, voltage amp, and follower, but - well, thought is almost never wasted.

[Bill Mountain]

No offense to Mr. Philbrick but could the lamps just be a poor design from the infancy of OP Amp design?

[amptramp]

General Electric put out a book showing circuit uses for neon bulbs.  Level shifting was one use, where the neons were used exactly as in the Philbrick op amp to maintain DC operation.

When power was applied before the tube heaters started operating, the voltage would rise to 90 across each bulb where the firing voltage was 90.  The stabilized voltage was 55 volts, so the two in series as shown here would be 180 volts to start, 110 volts in operation.  Similar circuitry was used in some oscilloscopes to maintain DC coupling without the need for a wide range of power supply voltages.  This allowed the gain to be maintained since the neons would be at a fixed voltage and not act as a voltage divider like a resistor would.  The circuitry shown with the resistor in series and the capacitor across the series connection of resistor and two lamps served to carry high frequencies without having the neons behave like relaxation oscillators as they would with capacitance directly across them.  It was also a cheap alternative to voltage regulator tubes for low-current uses such as screen supplies.

[R.G.]

OK. That makes sense. I know neons drop back from their breakover. If they stop at some significant voltage, that would account for the use.

In that case, with neons as a 110V level shifter, it does contract to a three-stage opamp, with a big power supply - +/-300V.

[PRR]

> there were no zeners or if they were....

They weren't.

What's odd about using gas discharge for voltage-drop? Tesla, Las Vegas neon, electric railroad thyratrons.... Zeners are very new and much less sexy.

Thanks for telling us about thyrites. The "surface tension" of silicon carbide (and similar) is fairly sharp and they can be "stacked" to any voltage. They were not used in my K2-W but it clears some fog.

Of course the real point (problem) is: the output of a tube is always Positive of its input. If the inputs are at zero volts (the "+" input generally went to a +/-1.3V trim, the "-" input is forced to follow), then the output can never go to or below ground. In analog computing, this is equivalent to having only positive numbers, not negative, which really pains the mathematician. Further since one tube stage won't come close to "infinite" gain, we need two stages, and now the second output won't even come near zero, only high positive voltages.

The Zener/neon/thyrite level-shifts from the +50V to +250V at plate down to -100V to +100V at output (actually CF grid). Without the losses of the 1Meg:2Meg divider which is acceptable between 1st and 2nd stages.

> Fired neons don't act like zeners, the forward voltage folds back.

Ah. Yes but. Dead up to 90V-130V, fold-back to 60V, then fairly constant from 0.5mA to 10mA (or bust). The exact voltages vary with gas, pressure, doping, and a pinch of radioactives. You can buy "good" regulators at 75V 105V and 150V (and rated 5mA-30mA). But in the day, NE-2s were kinda-consistently 60V when lit. (Apparently less at this very low current.) Since it is inside the loop, the exact drop is not-critical. Neons were cheap enough (and opamp markups so high) that it was NBD to weed-out the really-off ones. (However since the market was 110VAC indicators, the breakdown was always below 150V, and the holding was rarely much below 50V.)

> I can't remember the high frequency response of the Philbricks.

The slew-rate was really awful, even at +/-10V signals, nevermind the +/-100V levels they were optimized for.

Recall that output was often paper-recorder, or S/H (or just pause an integrator) to voltmeter. "Fast" was using 60Hz to sweep a function to a 'scope.

Note that there is Positive feedback between the cathodes of V2. The effective gain is much higher than you would count on your thumbs. 30*60 is around 2,000, but at DC the effective gain is over 5,000. Of course PFB affects distortion and speed, not usually for the better.

> No offense to Mr. Philbrick but could the lamps just be a poor design from the infancy of OP Amp design?

It's a brilliant design (and not due to Philbrick; he was a brilliant marketer). Previous opamps from respected designers were FAR bigger and not so good. The roots of the K2-W go to Loeb Julie, an undergrad student who butted heads with his professor, got out of amplifiers and became the guru of precision resistors and dividers.

Just put the K2-W aside. If an opamp is any good, and used *within design limits*, you can NOT tell what is inside of it. In most audio uses, we do NOT need full opamp function; we use the chips mostly because they are cheep and simple. If you need a "flavor", it is usually simpler to go for the flavor and forget all the other opamp virtues.