Differential Amp W/ Differential Power Supply: Theory Question(s)

[RicharD]

I'm playing with the idea of making a tube circuit that is completely differential from the input through to the output.  It stands to reason that a differential power supply would be the way to go, ala "world of sand designs."  We do differential output stages with tubes and unipolar power supplies all day long every day.  My main question is, if I build a differential supply, should I keep my potential difference the same?  ie. If I'm designing for a B+ of 350V, do I make my bipolar supply +/- 175V or +/- 350V?  I'm thinking I need the +/-350V to keep the potential the same across each tube section.  Would it be a waste of time and money to build a bipolar supply? 

Just trying to think outside than amp cabinet.   

[eleventeen]

Non-theoretical answer:

As I understand diff circuits, their primary attraction is their incredible sensitivity to very small voltage changes; which concurently imposes a stingent requirement for extreme stability on the part(s) of the surrounding and setup components/circuitry. The ultimate expression of a diff amplifier is, at least in my mind, a comparator, which is usually designed to take a small change in input voltage and cause the output to snap to the + rail whenever the + input rises above the - input, or snap to the - rail whenever the - input falls below the + input.

Now such a structure would be fabuloso in terms of designing a precision power supply, but one has to ask, fo' what? Don't we like voltage sag?

And likewise, I don't exactly understand why you would vote to employ such a "tool" to accept a guitar input and pass through nuance and "touch" to the output stages.

[RicharD]

Short answer = not a geetar amp.   

[eleventeen]

Yeah, for building such a thing out of discrete components, I think you'd have to temperature compensate the bloody daylights out of the circuitry; employ 1% resistors all over the place, and super high quality everything or the circuit would just blast all the way in one direction or the other. And not give you much oppty to check out why or how. Tube aging would be another contributor to extreme irritation.

When you look at older hp (for example) test-gear circuitry built from discrete parts pre-LM324 you will see pairs of resistors and transistors "clamped" together with "S" spring-metal deals to keep their temperatures matched. And, maybe I'm crazy, but perhaps you could develop a line of 12AX7 heat sinks, LOL.

Now in the "I'm willing to be corrected" department, with tube circuitry you are not going have 100,000 or even 1,000,000 gain factors like are common with chipskys. But in general, for every stage, I think you'd have to build about 3x the circuitry to get the job done: 2x everything because it's differential, and 1x pure temp compensation. 

[sluckey]

One big advantage of differential amp technology is wasted if you don't feed it with a balanced/differential input. Common mode rejection is one of the basic advantages of using bal/diff inputs. Not many bal/diff guitar pickups out there. My Taylor 614CE has a differential output. I play it thru a Fender Acoustisonic SFX amp which also has a differential balanced input (balanced XLR). The system really cuts down on external noise interference. And the output stage is also differential, allowing for direct speaker connection, like most modern ss amps.

Sounds like you need to supplement the amp hobby with another hobby. Wanna feed some baby squirrels?  

[RicharD]

>One big advantage of differential amp technology is wasted if you don't feed it with a balanced/differential input.

Right now the experiment is a mic preamp.  Other possibilities are compressors, mixers, monitor amps, etc.  I'm fixing to answer my own question out in the lab.

>Sounds like you need to supplement the amp hobby with another hobby.

Understatement!  I have over 100 drawers of parts storage but only 1 is marked transistors.

>Wanna feed some baby squirrels?

We fostered a pair of kitten this summer.  My eldest daughter decided to adopt them.  Now her landlord say no pets.  Guess who still has kitties?  We have 1 nasty mean squirrel living in our back yard.  His ears look like they've been clipped, like the white trash do to their pit bulls.  He's about the biggest squirrel I've seen and he totally pesters our pets.  My feral cat and Border Collie are out to get him.

[sluckey]

His ears look like they've been clipped

Haha! He's a veteran, or stupid (Yo, Adrien)! I've got a lot of those too. Mostly horny males that are too eager with females that are not quite in love yet!   
Or a slow squirrel that got too close to a mother's nest. 

[DummyLoad]

be aware of V_h-k if V- exceeds 100V

[RicharD]

I'm a dum bass.  1st off, there are no negative complements to the 12A*7 family.  So I had to change my line of thinking.  A quick pencil sketch and I fingered out that I'd have cathodes sitting close to -350V.  Nope.... not gonna happen for the reason  ISO pointed out, unless I build 2 filament supplies.  Balancing the 2 halves has proven to be tedious at best.... and this is only a single stage.

[DummyLoad]

meh! you're giving up too easy... :p

[RicharD]

3 of them caps are gonna explode.

[DummyLoad]

whooops!     yup they is... just imagine them reversed... mkay!   

[PRR]

The "point" of +/- supplies is that you can set the output at Zero Volts DC and avoid an output cap.

Except in tube circuits, you only have N-type devices. And the output is always more-Positive than the input.

So to get from a plate at say +100V to a next-grid at -2V (or -102V so that its plate may be near zero volts DC) you have to do a trick. And all the tricks suck.

With a -200V supply, first plate at +100V, next grid at -100V, you can use a 3:1 divider from first plate to next grid to -200V and come out at the right place. But this 3:1 loss of gain hurts.

You can do something with Neon bulbs. Philbrick sold a few of these. Noisy. Zeners are a bit better.

You almost always end up with a cap. While you might accept the loss of gain at DC, the lure of getting back some AC gain (and losing noise) is just too great.

You do all this for DC amps. Aside from VTVMs (which work different), there are "Analog Computers" which solve differential equations and plot the solution. Because of the small gain-bandwidth of tubes (especially with lossy level-shifting between stages), they generally work slowly; and it is useful to be able to "pause" which amounts to holding a DC level.

> a mic preamp

Must it pass DC? (Is it a calibrated barometer?)

If not, then +/- supplies are really more distraction and expense than good audio engineering.

But go ahead and do it. The pointlessness isn't really clear until you mess with it.

[eleventeen]

The pointlessness isn't really clear until you mess with it.

Tagline material, that.

[RicharD]

>Except in tube circuits, you only have N-type devices.

I actually wired all the way to the cathode of the neg 1/2 and the bypass cap is what unjammed the rusty cogs in my head.  I may have actually drooled.  When I realized I'd need a cathode down at -350ish, I knew there was a problem.  My initial thought was to use a set of bottles for the positive 1/2 with a typically grounded 12VDC filament supply and another set of bottles for the neg 1/2 with a separate 12VDC filament supply bonded to the neg 350V.  At this point I had messed with it long enough for the pointlessness to become really clear.  Iso's circuit looks good on paper but what happens at the next stage?  The power supply just got complicated.

>Balancing the 2 halves has proven to be tedious

Operator error.  1 scope channel was dialed out of callibration.  I was totally banging my head against the wall trying different tubes etc. even though the cathodes were sitting at the same voltage.

[HotBluePlates]

I have 2 observations that are not "answers" but "food for thought"...

... a tube circuit that is completely differential from the input through to the output.  It stands to reason that a differential power supply would be the way to go ...

Why?

Thought #1:
It seems to me that if you had a true differential source, it would be one were the "+" and the "-" signals are equal and opposite, and each gets its own amplification. If that's true, the signals are positive and negative with respect to each other, not to anything in the circuit itself. So the circuit components could all be between ground and B+. Further, since each individual signal tends to pull current in anti-phase to the other signal, the whole circuit tends towards minimal current variation when in use. What I mean is that ideally, the push-signal and the pull-signal induce equal and opposite currents.

*If* this is the case, giving each side its own power supply actually costs twice as much and puts a bigger burden on the supply, while making filtering more difficult. So if the plan is a differential signal path along the lines of a long-tail pair, a single supply polarity makes the most sense.

Thought #2:
Let's say you do have a solid reason for needing an output that bobbles about ground potential (like a solid state complementary pair output). You decide a bipolar supply is the way to go. You can always hook up a tube with the plate more positive of the cathode, and the cathode a little more positive than the grid. It doesn't matter if the plate is at 300v, cathode at 2v and grid at 0v, or if you have the plate at 100v, cathode at -198v and grid at -200v; both situations feel exactly the same to the tube.

Yes, supplies and reference voltages for the heaters become tricky. look at any schematic for a tube regulated power supply or the power supply of a tube o'scope to see examples of giving tubes the operating voltages they need(and the reference voltages they need) while only being concerned with "ground" at the output. Obviously, there are times when -150v or +150v is the "ground of a subsection of the whole circuit.

[Shrapnel]

Thought #1:
It seems to me that if you had a true differential source, it would be one were the "+" and the "-" signals are equal and opposite, and each gets its own amplification. If that's true, the signals are positive and negative with respect to each other, not to anything in the circuit itself. So the circuit components could all be between ground and B+. Further, since each individual signal tends to pull current in anti-phase to the other signal, the whole circuit tends towards minimal current variation when in use. What I mean is that ideally, the push-signal and the pull-signal induce equal and opposite currents.

Thx HPB, this just got me thinking....

Pure differential input... 2 non-differential channels side by side... mixed in a P-P fashion at the "output." If we know we are going to get a non-differential signal in on the input... ground the "PI" input on the opposite side of the signal, or not, as "PI" operates ala 2-ch Marshall "1974" 18W mixing. Single power rail would work, nodes broken down as usual.

IF we are looking for line level out only, what would need to be done to a "firefly" to get it right? Load resistor and voltage divider?
Otherwise, what would the difference be driving a more powerful power amp section?

[RicharD]

Thanks HBP.  That's good stuff.  As pointed out, my thinking was completely flawed since all tubes are N type devices.  The bipolar power supply has been tossed out the window.

My reasoning for wanting to build this circuit fully differential is to get away from costly input and output transformers.  They're roughly $50.00 each.  That's $200.00 for a stereo pair. 

Here's the next question:

If I have a differential pair of cathode followers each with a Zout of 440 ohms, is my total Zout 220, 440, or 880 ohms?  Seems like they're in parallel so my Zout should be 220 ohms.  220 ohms would be excellent and certainly would get me out of the output transformer.  440 ohms would be acceptable but 880 ohms is probably too high.

[HotBluePlates]

Pure differential input... 2 non-differential channels side by side... mixed in a P-P fashion at the "output." ... ground the "PI" input on the opposite side of the signal ...

A sorta-definition you may already know about:
"Differential" amplifiers have a pair of inputs and amplify the difference between the 2 inputs. One of those inputs could be driven with a signal and the other grounded, and the diff-amp simply amplifies the difference between the incoming signal and ground. This is how a typical long-tail pair works, although the "grounded input" is often attached to a resistor between the tail resistor and ground (as part of the feedback circuit).

But it sounds like you're describing a concept a lot like this Vacuum State preamp. Ignore the topology for each amplifier segment; the amplifier accepts a differential input and has a differential output by having 2 amplifier paths, one for push and one for pull (or + and - if you like).

My reasoning for wanting to build this circuit fully differential is to get away from costly input and output transformers.

Well, I don't know that your thinking was completely flawed; positive and negative supplies can be used, but the approach and goal are different from solid state practice.

But input and output transformers also provide isolation. What happens when you have a condenser needing phantom power? Will your input stage be happy with 48vdc on the grid?

If I have a differential pair of cathode followers each with a Zout of 440 ohms, is my total Zout 220, 440, or 880 ohms?

I haven't messed much with studio electronics, but my gut instinct is that a balanced output is rated and "balanced from each side to ground. So on that basis, I'd think with each leg being 440 ohms from ground that the output impedance is 440 ohms. There are a few books I could check for an answer, or PRR will pop along shortly.

[PRR]

> fully differential is to get away from costly input and output transformers.

A microphone hisses like a 200 ohm resistor.

A tube hisses like a 1K resistor which is also at 1000 degrees C.

The hiss voltage of the tube is about 6 times higher than the hiss of the mike.

That's why a tube mike amp "always" has a 1:7 to 1:15 input transformer.

Alternatively you can live with the hiss. On a rock stage it may go unnoticed. With hot condenser mikes the higher mike-hiss (and higher mike output) may make tube hiss unimportant.

You can reduce the tube hiss by working at a higher transconductance. If the cathode were cold we want Gm to be less than 1/200 ohms. since it is three times hotter than the mike (~~300C) we need GM 9 times better, 1/22 ohms. This is 45,000 microMhos. If you have two input devices, it wants to be 90,000uMho each.

8417 will exceed 20,000uMho if you run well over 100mA. There is an exotic microwave tube does 45,000 at only about 45mA. So ten 8417 at 1A ~~300V, or four exotics at 180mA 200V..... the cost is in the area of fine transformers. (Won't be cold in your studio!)

The output nominally drives 600 ohms at >+20dBm. This is 18mA peak. Few tubes get below 1K plate-cathode impedance so direct drive is inefficient. And for balanced output you need two tubes each driving 300 ohms with 18mA peak. This is far beyond 12AU7 turf. A couple 6V6 will dump 18mA easy. The gain into 300 ohms is about unity plate-loaded or about as half cathode-follower (and a heavily-loaded CF is not dead-clean).

With 4:1 iron the peak current is just 5mA and the load toward the tube is 10K. 12AU7 can pull this fairly well, 6V6 can loaf.

Yah know: Hammond 125ESE wired for the highest tap and driven with triode-wired 6V6 at 350V 35mA will make the 7.75Vrms on the highest secondary tap and pull to 40Hz. Source impedance is 10 or 5 ohms presented to the 600 ohm port: no sag even wih many loads. Gain from grid to secondary is about 0.3; this is typical of solid line-amps. You need 25Vrms from the driver; standard Champ design. You want another gain of 50; direct-to grid won't give the gain or will have excess hiss, any 1:5+ input iron into any normal triode volt-amp will do fine.

Radio shack sells a 1:7 transformer with XLR for like $17. It aint the fattest chunk in the bog, but I've done good work with it. It would plug straight into a Champ. Hack the 6V6 to triode and disconnect exverything off the OT 16r secondary.

Note the transformer isolation.

> this Vacuum State preamp

The input device is the FET. If you cheat a little, why not cheat a lot? Use the fine $10 chip, capture your sound, THEN (another day when your head is clear) route it through any handy tube-amp.

> instinct is that a balanced output is rated and "balanced from each side to ground.

What ground? The core concept is "floating". Ever see a military field telephone? Two wood boxes and a reel of 2-cond wire. As far as the system knows, there is no ground, just the load bridged across the far end of the circuit.

Two 440r CFs push-pull is a 880 ohm output. I also wonder if 440 ohm small-signal can make the required dBm power. In small studio with only >10K loads, 880r source is tolerable.

Note that "Vacuum State" has two 1uFd caps going out. 0.5u effective. If loaded in 600R the bass falls at 530Hz. Even at 10K load it is 32 Hz. This is not a Serious Pro Output (though it may be sweet in the small modern studio).

[RicharD]

Sing with me now!

There's a hole in the bucket dear Liza dear Liza.........
I guess back to transformers it is.  I hadn't made it to the point of actually plugging in a microphone or putting a real load on it.  The good news is it'll easily convert to a stereo pair single ended using transformers.  The bad news is I probably have excessive gain, more NFB.  I'm doing +41dB w/o transformers.  IIRC, the CineMags I have will bump me another 20dB.


>If loaded in 600R the bass falls at 530Hz. Even at 10K load it is 32 Hz.

I know this isn't really directed at my circuit, but I did notice a serious bass response issue late last night.  I'm flat from 60 - 20kHz.  It's starts rolling off FAST by 55Hz.  By the time I reached 20Hz, the waveform was garbage.  My function generator's output is not floating so I isolated it with a cheap Rat Shack 1:1 audio transformer.  I fingered this is what was causing the roll off, and at best must be contributing to the distorted waveform.

[RicharD]

Tonight's experiment had less than promising results.  I was able to determine my roll off problem was entirely the fault of the low brow Rat Shack 1:1 isolation transformer.  I only had a 100k load on my output so I knew that wasn't the problem.

Basically my circuit is a stereo pair connected differentially so getting back to single ended was as simple as swapping a few wires.  I added a 1:15 mic to grid transformer and that's where thangs headed south.  The transformer has so much gain that the input of the first stage was severely clipped.  I could only apply < 100mV.  0dBU is completely out of the question.  Attached is the 1st stage EF86 circuit I copied from the Philips data sheet. I measured bias at 2.3VDC so counting on thumbs I assume the input threshold is about 2.3Vp-p or 1.6Vrms.  Removing the cathode bypass cap helped and got me into the 200mV range.  Making it a plate follower with a FB ratio of 20:1 got me to the 500mV range.  A 1:7 input transformer would probably get me to 0dBu, but I don't have one nor can I throw money at one right now.  My original concept was to not have an input transformer and simply sticking one on the front ain't gonna work.

Looking back at some old Ampex schematics, the listed inputs aren't anywhere near 0dBU.  The 351 shows -77 and the MX-10 shows -60.  The MX-10 uses a triode connected EF86 with a gain of roughly 25.  That's about where I was when the circuit was looking it's best. 

Is shooting for 0dBu input expecting too much?  I wouldn't think so.  Regardless, it's time to start over from scratch.  No loss though whereas I learned quite a bit the past few days.

[HotBluePlates]

Right now the experiment is a mic preamp. ...

I added a 1:15 mic to grid transformer and that's where thangs headed south.  The transformer has so much gain that the input of the first stage was severely clipped.  I could only apply < 100mV.  ... Attached is the 1st stage EF86 circuit ...

Looking back at some old Ampex schematics, the listed inputs aren't anywhere near 0dBU. ... Is shooting for 0dBu input expecting too much?

You do say this started as a mic preamp. Mics don't output 0dBu, and you already know that is line-level. So the simple answer is to look at the difference in mic and line amps. If line amps have gain, it's ususally because they are anticipating a loss later in the signal chain, like faders or mixing circuits.

And without checking, I'd think line amps would have a very modest step-up, if any at all.

I would say the mic preamps I have can swallow line-level signals (I've fed CD's through them on the way to a power amp and speakers), but there are 2 levels of gain on that unit. Switching from High Gain to Low Gain removes a stage of amplification (maybe 2; there are 4 triodes total per channel), and might alter the feedback circuit as well.

And a circuit that can amplify a very small signal will probably run out of steam with a big signal. Not necessarily a problem, but you have to design for the big signal if you want to handle them.

I measured bias at 2.3VDC so counting on thumbs I assume the input threshold is about 2.3Vp-p or 1.6Vrms.

Stuff like this is never an issue in guitar amps, so it never comes up. But if you read any of the old tube books, the writers always are hesitant to apply a signal that drives the grid more positive than -1v. That appears to be the point where grid current effects start in most tubes, which also means distortion. So you'll need to leave at least 1v more bias than the signal you plan to apply, and you really ought to think about the peaks not represented in that RMS figure if you want pristine clean.

Meandering story that applies:
I was once in a introductory engineering class where we had a project to build a beam out of balsa wood. We studied the various configurations a beam could take and the calculations for how much load each would withstand. The project was to make a beam 12 inches long that could hold up to 250lbs, and have the lowest cost to make. Each glue joint had a cost assigned, and various thickenesses of wood had different costs.

We calculated to the nth degree. We investigated 4 or 5 different configurations to make the beam. We didn't/couldn't test the plan ahead of time, as the teacher had a special test rig to use. Out of 8 or 9 groups, each with their own beam, all failed the load test except 1 team. Where we screwed up was using the thinnest, smallest stuff that we calculated that would do the job. We didn't account for unknown factors that we didn't know how to calculate, or for weaknesses in the wood. We didn't allow a safety margin.

Anyway, I'm saying if you want to apply a big music signal that has occasional peaks, you'll want to allow for the safety margin and stay well away from 0v on the grid.

[PRR]

> engineering class where we had a project to build a beam
....We didn't allow a safety margin.

My dad, an EE with solid general-engineering basics, once observed that in Electronics we often use safety factors which would be grossly reckless in other engineering.

> build a beam

Beams are often used to support people in ways that would be very bad if the beam failed.

The floor-load in my new room is two people, two dogs, a couch and some books. Not even 1,000 pounds on 12'*30', under 3 pounds per square foot. Code requires design for 40 psf. If you work this out as 180 pound people it is 80 bodies. Will they even fit? 360sf/80= 4.5 square feet each, a 25 inch square for each person. This is a VERY crowded but not impossible cocktail party. OK.

Having figured my 40psf, I buy wood. On test, most of it breaks at 3,000psi tension, but I must figure it as 1,000psi. If I use steel, while most is good for over 30,000psi, some codes force calculation on 16,000psi.

With the whole family in the room, my beams are nominally 40 times stronger than I "need". If everybody on my road jammed into my one room, I'd still have nominally 3 times more strength than I "need".

And there have been cases of residences with higher loads, and falling floor beams. Newspaper-hoarders run over 25 pounds per foot of stack. When you fill a room 2 feet high you exceed nominal design strength. 5 feet exceeds average actual strength. Indeed hoarders are sometimes discovered by floor trouble as the piles get over 4 or 6 feet.

Not just beams. Weakest link in the chain. There was the old house with a party upstrairs, and then the party was downstairs.... the whole floor dropped. It had 16 foot studs up two floors, with a ribband to support the upper floor. The beams were fine, and the studs, but the ribband was not sufficiently notched (perhaps only nailed). Good enough for a century of immigrants with few possessions and no large parties, not for a college kegger with wall-to-wall bodies.

OTOH, in electronics our cost-critical safety factors are rarely over 2:1 and sometimes like 1.2:1

[PRR]

Mike input levels run from 0.1mV (ribbons on harpsicord) to 2V (hot condenser too-near drums or Fender Twin).

Obviously a 2V signal can go straight to the recorder, You probably don't have to balance such a hot line, though you may want to balance for the Phantom Power. If this is all you ever do, the mike input design is trivial.

Building a GOOD 0.1mV input is tough. That's what separates a gitar-amp from a fine mike amp.

A Langevin I met had two options: 30mV input for general use (it would take more but the bass could be uglier), and 100mV. That was the general range before the recent (last 5 decades) fad for too-close miking.

> The 351 shows -77 and the MX-10 shows -60.

Those are full-up levels. At this gain, mike+amp hiss exceeds tape hiss. In normal use you will run the gain 10dB-30dB lower. If you have to lose 40dB, it probably clips before the knob and you should use a pad ahead of the mike-amp. Real-world levels on dynamics are rarely below 2mV=0dBfs nor much above 50mV=0dBfs for "sane" sources. Large condensers have 10dB pads for a reason.

> Is shooting for 0dBu input expecting too much?

Yes. It can happen in modern too-loud studios, but it is stupidly high.

If stupidly-hot is ALL you do, use a line-input (not mike) transformer and low-gain tube, or even go straight to a diff-pair's grids. Say 0.775Vrms peaks reference, 7 microvolts noise in two tubes, that's 100dB S/N which is essentially not reproducable in most listening rooms.

And you can pad a too-hot input down. 2V mike signals are usually padded 20dB, to 0.2V; even that is hot for classic mike inputs.

Take 50mV maximum input (that's peak; if you still own a VU meter and calibrated amplifier it will read 10dB-20dB lower). 1:14 step-up is 0.7V at first grid. (However at gain=188 this is 121V at plate.... huge!) You see another reason why classic tube inputs don't accept over 100mV (and newer ones often use 1:7 iron for modern hot studios). While you can accept "any" input with cathode degeneration or overall NFB, the cathode resistor degrades the noise-figure you want for the harpsicord.

BTW, pentodes have far more hiss than a triode, altho with 1:14 iron you may not be in real trouble.

Define your real needs. Flea-squeaks? Heavy steel percussion? With fairly good gear, are you running into hiss or overload?

[DummyLoad]

I assume the input threshold is about 2.3Vp-p or 1.6Vrms.

i don't think that assumption is correct, but i'll deffer to those in the know.

the way i see it, the charts give a clue as to what the input threshold would be - the parameters you highlighted in the charts indicate an AV of 188 with Vo of 54Vrms at 5% distortion.

 
given that AV is Vout/Vin, and you've been given AV at max Vout;  then Vin*AV = Vout at an acceptable distortion level. so solving for Vin; Vout/AV = Vin; in this case about 287mVrms or 406mVpk @ 5% output distortion.

input signals larger than 287mVrms simply increase distortion and AV decreases.




  

[Merlin]

You might like to look at some tube phono amps for examples of low noise, low level design. You can feed a magnetic cartridge (5mVrms max) directly into a good input tube (triode only, no pentodes), and the tube won't be the major noise source. A dynamic mic up against a speaker will give you 1-2mV, so it is possible to do without an input transformer in that case.

Signal transformers are no longer used in professional audio gear anyway; they are too expensive and non-linear. It's cheaper and very nearly as quiet to do the job with a transistor front end. Unless you're a purist, the logical approach would be a hybrid design; this would also make phantom power easier to include.

Heck, if you want to make life real easy you could use one of those new balanced mic-input ICs, and then use tubes after that.

[PRR]

> i don't think that assumption is correct

Often good-enuff for simple sane triode amps.

Pentode offers more ways to skew things; this one seems to set the plate voltage quite low. Perhaps this gives reasonable output swing with high gain.

> Vout/AV = Vin; in this case about 287mVrms or 406mVpk @ 5% output distortion.

My thumbs say almost 200 divided by about 50 is 1/4 Volt. With 1:15 input iron the max input is like 30mV. That's very reasonable for far-mike work. Which is where you NEED the free-gain of a 1:15 input transformer. If he's got a 414 with cymbal scars on it, he'll need to re-think the input levels.

[PRR]

> look at some tube phono amps for examples of low noise, low level design

Good for study, bad for plagiarism.

> tube won't be the major noise source

Phono pickup is a 5K-47K noise source. Over 10K in the "hiss" octaves. A bit higher than a typical tube. In fact it was designed/evolved so-that tube hiss was overwhelmed.

Dynamic mike line is a 200 ohm noise source. This was picked for line-loss: higher would lose treble (phono lines are 3 feet while mike lines are longer), and lower (than around 20 ohms; some old-old mikes ran this low) gets into copper loss.

Then there are "hi-Z" dynamic mikes. Now rare, once common. The easiest dynamic capsule is 4 to 20 ohms, low for a line, far too low for good noise tube amplification. Anyway these replaced crystal mikes, which have ample voltage level into open grid. You need a transformer, you don't want to buy all four at once when you start with one or two mikes. Mikes with built-in transformers were the best low-investment PA rig for decades. Lines over 50 feet may be problematic, but start-up PA rigs were often that short.

With 200 ohm mikes, phono inputs might be a low-hiss plan _when_ interfaced at 5K-50K impedance. With a transformer.

Without step-up, mike self-hiss voltage is about 5 times lower than phono pickup self-hiss, lower than tube self-hiss, and could be annoying.

> A dynamic mic up against a speaker will give you 1-2mV

No; somewhat more.

And a MAJOR difference from phono design: the output from a record is a compromise against recording time. Classic LPs will rarely peak over 50mV. There are some 12" Singles with higher recorded level but these slam the limits of available cutter-heads. You won't find 200mV peaks from a phono cartridge. 50mV (re:1KHz) input overloads are 99.9% good enough.

The Live Performer has no such limit. Long-term average of horn and wind is limited, but one-note levels of trumpet and trained singer may be VERY loud. Electric musicians have the option to use bigger amps.

Speaker blurbs show 95-99dB sensitivity but this is far-field beaming. The base efficiency is closer to 85dB/W/meter hi-fi, 92dB/W/meter for serious guitar cones. Take 90dB SPL at 4 feet for simplicity.

Power of gitar amp runs 5W-100W. Richard has been seen with a 25W, use +14dB power factor.

In the far-field, SPL rises 6dB for each halving of distance. When you get within one radius, maybe one diameter, of the source, you are near-field and SPL is near constant. Say 12" cone. "up against" is roughly like 1 foot and therefore +12dB factor for distance.

90dB+14dB+12dB= 116dB SPL in near-field.

Sensitivity of SM58 or EV635A is near 0.2mV at 74dB SPL. (Newer sheets reference 94dB SPL, but I'm old-skooled.)

116dB SPL - 74dB SPL is 42dB up. Voltage is 125 times higher.

0.2mV * 125 = 25mV.

That's perhaps low. A Champ may be 10mV, but a Twin with D-120s may be 70mV.

And these are sine-equivalent numbers. The peak is 1.4 times higher. A 25W amp may play at "40W" overdrive, which adds slightly to speaker peak voltage but cone phase-shift may pervert transients 3dB-10dB higher.

If clean reproduction is wanted (it may not be), then 100mV-200mV is a target. That's at the jack; with a step-up transformer more like 500mV-3V(!) at the grid on loud sources.

And modern $99 condensers have 3X to 10X the output, putting you near Volt at the jack and several volts at the grid (but noise become moot).

Dynamics or padded condensers in 4th row of full orchestra may never exceed 50mV.

Mikes in modern loud close studio practice can deliver 100mV-1,000mV. However they may not need a 15dB SPL noise floor (50dB noise floor goes unnoticed when a Twin is at full roar) so non-optimal noise design (transformerless, pads) is an option.

> transformers are no longer used in professional audio gear anyway

Most workhorse gear lost the iron in the 1980s. And of course lost the tubes 20 years before that. 

Everything old is new again. There is a fad for tubes an transformers; not across a 64-in console but a few for specific flavors.

I like the MIDIman AudioBuddy as a low-fault $99 stereo mike preamp. It's sure better than some of the vacuum and iron crap I have used, though not as good as the very best (which was always rare). There's heaps more on the market in every price range. I assume Richard has considered such a path.

> get away from costly

Your budget may be wrong for your aspirations. Or maybe not. Yes, a useful tube preamp cab be cobbled from spare parts and spare time. If you have studio clients with money, it would be wiser to save time and use known-good costly parts, or even buy the whole box from a sexy maker. OTOH I suspect you don't have paying clients, are probably not the next Rikki Martin, and your day-job may be slack (my contractor pals are between jobs a lot), and you have piles of stuff. Go for it.

[RicharD]

The old circuit is in the recycle bin.  Far too many flaws in my original idea.  A lot could be said  such as ISOTone's input threshold analysis was very much in line with lab results, etc etc.  Bottom line is, the circuit wasn't right.

>pentodes have far more hiss than a triode

Wired in triode mode, the EF86 is an entirely different beast.  Not a gain monster like when in pentode mode.  From the curve chart I was expecting a gain of about 25 so I designed just a touch of feedback making it a plate follower with ~ a 20:1 ratio.  My mileage varied considerably whereas I only got a gain of 10.  I was only getting a gain of a little more than 10 out of a transformer I know for certain is 1:15.  Something is a little wonky and I did check my probe callibration whereas that drove me bonkers the other day.  Regardless, this circuit is working much better.  Tied into the rest of the circuit, I'm getting a total gain of 400 or 52dB.   0.3Vp-p was just shy of my max signal in to get an undistorted  sine wave out (120Vp-p) with the entire circuit sitting wide open.  Standing alone (volume pot turned down), I can apply 2Vp-p and maintain a clean output.  This circuit seems more on track.

[Voxbox]

This is a really interesting thread, so I thought I'd throw in my tuppence worth.
Speaking as a Broadcast (hardware) Audio Electronics Engineer we used balanced (differential) inputs for everything because of the distances between equipment and potentially hostile electromagnetic environments.

Someone said earlier that the guitar pickup isnt balanced, well actually it is, we just choose to ground one end of it.
Unfortunately it's a high-impedance low-output device so is not so good for driving long lines (guitar cables). Even some microphones can be capable of driving very long cables (hundreds of yards or more, I'm thinking of effects mics on golf courses), but they have impedance-reducing transformers in them to convert the high-impedance transducer into a low-impedance balanced output that can drive the line better with less losses and interference as their output is floating and balanced. Floating means that neither leg of the output is referenced to ground. I dont beleive there is such a thing as an electronic (active) floating input. The interference that can get into the cable is also reduced greatly due to the fact that it is a floating balanced circuit. If transformer balanced, and therefore floating, then this common-mode rejection (reduction in interference that commonly gets into both wires in the balanced circuit) is likely to be higher than that of an electronically balanced circuit, however circuitry may have improved since I was dealing with these things on a daily basis in the 80s and 90s.

So, why not add a balancing, high to low impedance transformer for a guitar pickup? You could then drive your balanced, differential input of your amp and reap the benefits of balanced circuitry that Broadcast installations have been using since the Dawn of Broadcast Time. The internals of the amp dont need to be balanced at all, no Broadcast equipment is balanced internally, its just the inputs and outputs to the outside world that need to be balanced.

I recall that Alembic used to produce a bass with a balanced output on an XLR. It could be done on a guitar as well. But its going to sound different! I tried an electronic balanced output on my PRS Custom once; the single coil sounds were great, strong and clear, but the humbucker sounds I didnt like so much. Its just a question of taste however.

The thing is, for "historical" reasons, guitarists seem to be comfortable in using basic unbalanced circuitry and tend to accept the attendant losses in performance and increase in circuit noise. Manufacturers wont advise us otherwise as it will cost them more to make instruments with balanced outputs and amps with balanced inputs. We accept what we have because we have always had it because we dont know differently.
So it's great to see this subject being aired here amongst some of the most knowledgeable folks in the amp electronics community.

Check out these guys for good quality transformers to experiment with -
http://www.lundahl.se/index.html

Cheers, Colin

[RicharD]

Lundahl makes very nice stuff although I have no experience with them from this side of the pond.  I've been using CineMag for my microphone thangs.

[PRR]

> Something is a little wonky

Yes, and some of it is not seeing the WHOLE circuit.

Say you had a 200 ohm source. The 1:15 ratio reflects 45K. You have 22K more resistance. Assuming infinite tube gain, you have made a 45K+22K voltage divider. You can't actually read the 1:15 up inside the transformer because 45K internal resistance is against the 22K which your meter can reach.

If you find transformer gain of 10, then evidently your reflcted impedance is more like 11K, and your signal source reflects as 49 ohms. Which is not realistic for dynamic mikes.

Seen another way, if tube gain is infinite then the 22K reflects out as 98 ohms, which is mighty low. If tube gain is 25, then the 470K reflects as 19K, and the input on the left side of the 22K is 41K, 180 ohms to the mike.

Note also that even with 49 ohm source you are losing 1/3rd of your signal through excess loading. You have kissed-off 3.5dB of noise figure before you find a grid, whereas the transformer should deliver to grid a 1dB noise figure.

Also, the 200 ohm source reflects as 45K which is a noise resistor but also a signal. Your 22K adds another resistor which generates noise but does not generate signal. That's another hit on the noise figure.

You want NO resistors loading either grid or cathode nodes. Transformer to grid, cathode heavily bypassed.

For signals up to a fraction of cathode voltage, for non-precision audio, let the tube work open-loop.

For larger signals or when modules must be interchanged without gain error, use a small resistor under the cathode return and bring NFB around. The source of the NFB needs excess voltage gain and excess current to drive the small cathode resistor.

While I'm picking on you:

You may run the first tube at fairly good current, keep the plate load nearer 100K even 50K. That keeps Gm up and noise down. Depending on your application, this may be a pointless frill.

The 1Meg volume control is dubious. Assume it is half-up, 250K wiper impedance, and assume 100pFd into the next grid. This is -3dB at 6.7KHz! Broadband audio usually can't use pots over 250K unless you can defy universal capacitance.

You also have to consider the noise power in a 1Meg pot, though with gain of 25 you are probably OK. There's ~~6uV of hiss at that half-up wiper. Tube gain 25 pot gain 0.5, this is 2uV at grid, 0.13uV at mike. That's cutting it close. 250K pot gets the noise down.

[PRR]

> I dont beleive there is such a thing as an electronic (active) floating input.

The standard transformerless "balanced" mike preamps ARE (nearly) "floating" (up to +/-5V or so; fine for studio but sometimes insufficient in "real work"). You may ground either input and lose your CMRR but still get full signal.

The do have from 5K to 100K each side to ground, which I suppose is not fully floating and sure will degrade CMRR from an unbalanced source. (What saves us is that microphones are usually true floating.)

The winding technique on gitar pickups would not balance well. That could be done, and we could easily wind to a lower more long-line-friendly impedance (50-500 ohms). But how far do you want to be from your speaker? And will you pay for a transformer to un-balance and raise impedance for a better match to the tube? "Recording" guitars exist; Les Paul liked the idea. They never caught on.

> add a balancing, high to low impedance transformer for a guitar pickup?

The pickup and the cable have capacitance. Each transformer adds capacitance. Transformer action makes cable capacitance effectively lower. Is the two tranny plan lower-C than just going straight? If we used 100 foot lines, yes. If we use 6 foot lines, no. Somewhere around 30 feet (where too many guitarists work) you have to ask how much you can pay for low-C transformers... and musicians can't spread equipment costs over 16 minutes of ads an hour amortized over a decade.

> accept the attendant losses in performance

It works out, with Alnico pickups and #42 wire, that we can reach 5KHz with an output level that just-almost strains a 12AX7 grid and overwhelms tube noise(>80dB signal/hiss ratio). We do NOT want >5KHz off a plucked steel string (inharmonicity).

Hum is dominated by having a wide-open magnetic pickup in a room with AC power. (Humbucker is better but reduces zing.)

Buzz.... is a problem, but typically musicians tolerate FAR worse than it needs to be. Tarnished plugs, broken cavity shielding, cheezy cable shielding, running cables over wall-warts... I've worked a lot with long unbalanced lines and if you are careful (broadcast engineers can't always take the time) and the room is not hostile (some broadcast facilities are nasty with buzz-fields), it can work well at lower levels than guitar.

All of these points may be argued. Balanced guitars exist. Wide-range pickups exist, and you can set the inharmnicity low-pass further along the chain. Tolerance for buzz may be less in recording than in Mannys Den. But I don't feel the "primitive" state of guitar interfacing is mere stupidity or cheapness.

[RicharD]

>You want NO resistors loading either grid or cathode nodes. Transformer to grid, cathode heavily bypassed.

Ah, so my NFB was loading the secondary of the input transformer.  I can totally see that ..... now.  I was running under the assumption that plate followers = good thangs, lower distortion & improved input headroom.  I assume a 12AU7 plate follower would be acceptable as a 2nd stage? 

>You may run the first tube at fairly good current, keep the plate load nearer 100K even 50K.

In my previous circuit, I was trying to maximize gain by setting my load line perpendicular to the tube curves.  Old habits die hard and completely unnecessary in this case.  In my new scenario, I went with a 47k plate resistor which is 1/5th of the 250k pot.  That's an acceptable ratio.  I ended up running the cathode voltage a little higher than I intended to due standard resistor values.  I figured it's better to increase than to decrease.  I have the cathode bypassed with 100uF which is double the value shown on the Philips data sheet.  I don't know if that constitutes "heavily bypassed" but my output at 20Hz was 33.5Vrms.  200 & 2k = 35Vrms.  20k = 36.5Vrms.  (yes my meter is cool with frequencies besides 60Hz).  Lab results were just about dead on with load line predictions.  Mr. transformer is 1:15 again like he's supposed to be.  Gain for the stage measure 23 which is negligibly less than 25 as shown on the load line. 

>While I'm picking on you:
I love it!  I'm learning a lot.  This little lesson on mic input transformers has been a good one.  I am extremely appreciative of your posts.  Thank you so very much!

-Richard
 

[RicharD]

>You want NO resistors loading either grid or cathode nodes. Transformer to grid, cathode heavily bypassed.

What about using a grid stopper between the input transformer and the grid input?  That's series resistance so it doesn't seem (to me) to be a load.  I've been reading about transient distortion and a grid stopper seems to help.  There appears to be a compromise between hi frequency cutoff and and eliminating transient distortion.  Funny thang is, with nothing but a scope connected to my square wave generator,  I'm seeing a slight inadequate treble response.  As soon as I connect to a mic input transformer,  I'm getting a high frequency transient distortion.  A 68k grid stopper seems to reduce the transient about 50% and is bringing me down about -1dB at 20kHz with the roll off starting at 10kHz.  150k eliminates the transient and kills (and I do mean kills) the hi frequency (-3dB at 5kHz).  I may be splitting hairs whereas I believe the transient is well above 20kHz.

Attachment snipped form EMRR's required reading:
http://ia331414.us.archive.org/2/items/High-fidelityTechniques/Langham1950High-fidelityTechniques_text.pdf

[RicharD]

The latest experiment has very promising results.  An obstacle I've run into is V1 tends to over-drive V2.  Eliminating V2 leaves me with less than adequate gain. I've been through 1/2 a dozen V1 tube stages from EF86, 12AX7, 12AU7, 12AT7, 6FQ7, & 6DJ8.  I've visited V1 currents from 1mA to 12mA.  V1 either has too much gain and not enough input sensitivity, or acceptable input sensitivity but not enough gain.  I got close with the 6DJ8 but what a current hog.  The limited operating voltage was not cooperating with the rest of my circuit + the frequency response was surprisingly poor.   ISOtone suggested the often over-looked, much under appreciated 6BQ7A.  I think this tube is the bee's knees.  It has a nice middle of the road mu, very linear curves, and a nice low plate impedance.  My lab results have considerably less gain than calculated, but my currents are dead on.  I can account for the lower gain of V1 beacuse of the 2nd stage loading.  I don't know why V2 is so much less.  V3 is pretty much right on.  TubeCad say my cathode follower has a Zout of 154 ohms.     I grabbed a better (CineMag) input transformer that I had left at ISOtone's lab which allows me to switch between 1:17 & 1:9.  It's response is much much flatter than the old UTC iron I was using.  That UTC transformer may have been a huge contributing factor to my frequency response issues with the 6DJ8 circuit.  I didn't discover it's funky response curve until a later experiment.  I do have a 33k grid stopper between the input transformer and V1 to curb some transient distortion.  The value was selected by trial and error weighing a balance between frequency response and square wave spike reduction.  The schematic and lab results are attached.  I need to drag the Zilla board over to ISO's to measure A weighted harmonic distortion.  A distortion analyzer is at the top of my wish list. 

[PRR]

> either has too much gain and not enough input sensitivity, or acceptable input sensitivity but not enough gain.

I don't understand.

"Sensitivity" is required output level divided by gain.

> What about using a grid stopper

No. Don't. But its noise-effect and transient effects are unimportant until you work-out a gain stucture. And a test set.

> I've been through 1/2 a dozen V1 tube stages

DEFINE YOUR SPECS!

Is this for harpsichord, Fender Twin? Ribbon? Condenser? Is low-hiss a Goal? (In home studio with high background noise, or with large/hot condenser, maybe not.)

You mentioned zero dBu somewhere. Take this as 0.775V in 600 ohms.

You have a 1:15 input transformer. 0dBu becomes 0.775V*15= 11.6Vrms, 16V peak. This will drive a 6V6.... why do you need a preamp?

OTOH, dynamics on harpsichord give under 1mV. Times 1:15 is 15mV. First-stage tubes with gain from 15 to 150 (12AU7 to pentode) will bring this to 0.225V to 2.25V out of the tube. This is reasonable. In fact there is no strong reason to have BIG gain in the first stage. You "need" enough gain so that your second stage does not have to be designed for lowest-hiss. Gain of 5 or 10 is often fine. Maybe more if you will have a loss-network (gain pot) between first and second stage. Of course if that loss is larger than first stage gain, now the second stage is the dominant hiss source.

You will usually need a second stage of gain. You will end with a "power" stage to drive modern 10K loads. If you won't use an output transformer, this suggests a low-current high-gain stage followed by a high-current cathode follower. (It isn't just the small-signal output impedance. 10K driven to +20dBu is 1.1mA peak. The tube idle current  must be much larger than 1.1mA, like 3mA-5mA.)

Mike signals will rarely be 0dBu. When they are, you use a 20dB pad in front of the input transformer (so it does not have to be sized 10 times bigger).

So you need a range of 1mV to 70mV from the mike or pad. That's a larger range than a guitar. If you scale for 1mV in makes 1V out, overall gain of 1,000, then a 70mV input makes 70V out which is frightening.

The classic mix-desk approach was to scale preamps for gain of 35dB, about 1:50. Then a fader. Then a mix-network with about 15dB loss, to another 35dB lo-hiss amp. This gave low line level suitable for in-room use. For outside lines there would be master fader and a beefy line-driver.

* * * * ... I started this 8 hours ago and lost any train-of-thought I had.

[RicharD]

>I don't understand.

Once again my poor writing skills shine through.  “Sensitivity” is not the right word.  “Input threshold” is what I meant.  I've obviously learned that there is a trade off between gain and input threshold.  I'm realizing that I may have set the bar way too high.

DEFINE YOUR SPECS!

I guess that would be prudent.  
6P's  "Prior planning prevents piss poor performance."
Basically I want an all tube signal path which can handle a large variety of condenser microphones.  
Output level +6dBu continuous, +18dB peaks.
The unit should be able to stand alone between a microphone and a tape recorder (or DAW).
For lack of a better term, call it studio grade.
Input transformer:
CineMag CM3440-A  Mic to grid.  
600 ohm in = 1:9
150 ohm in = 1:17

1ST Stage:
>No. Don't
>there is no strong reason to have BIG gain in the first stage
>if you will have a loss-network (gain pot) between first and second stage

I went ahead and lost the grid stopper.  TubeCad does not come with the 6BQ7A factory loaded.  I had to key in the parameters myself.  I musta done something wrong whereas TubeCad says I should get an AV of 31, but my load line plot shows an AV of 25.  I am measuring 22 (your mileage may vary).   I don’t really consider this BIG gain.  I noticed that when I’d dime VR1 (Rl of V1), my gain out of V1 would reduce.  I suspect my NFB of the following stage was improperly loading the 1st stage.  I dialed the pot back until this stopped, pulled it, measured it, and stuck that value resistor (150k) above the pot.  This gives me a loss of about AV=0.1, roughly the inverse gain of the 2nd stage.


2nd Stage
> Of course if that loss is larger than first stage gain, now the second stage is the dominant hiss source.
>You will usually need a second stage of gain.

It’s seeming like I don’t really need a 2nd stage since my preceding divider is just about in balance with it’s gain (AV=10).  The question is:  Does the plate follower reduce or add hiss?  That’ll have to be tonight’s experiment whereas I need to retrieve my lab amp and speakers from ISO’s lab.   Thus far, everything has been done with a function generator, scope, and DMM.

3rd Stage
> You will end with a "power" stage to drive modern 10K loads.
>If you won't use an output transformer [/i]
Previous experiments have been done sans an OPT.  I now have a 2:1 Edcor in the circuit with a 10k dummy load and a TL082 based meter driver circuit (that needs work).  The meters slam when powered off….. I digress….  The Edcor OPT is seriously messing with my frequency response.  I will be changing this out in lieu of CineMag CMOT-2H which is also a 2:1 output transformer.  Assuming most of my parameters for the 6BQ7A are correct, TubeCad says I have a Zout of 154 ohms.  With a 2:1 OPT, I should be able to drive a 10k load w/o any problems.  I don’t see any need to go to a power pentode CF, but apparently I might be riding in the wrong boat.  It seems like my entire circuit is missing the target.  I’ve lived a lifetime of learning from my mistakes.  Why should this be any different?   : )

> Mike signals will rarely be 0dBu. When they are, you use a 20dB pad in front of the input transformer
I happen to have ½ a dozen Daven  500 ohm T-Pads.  I know the norm is 600 ohm.  Will a 500 ohm be ok?  Does that put too much load on the microphone?

> The classic mix-desk approach was to scale preamps for gain of 35dB, about 1:50.
This appears to be my target.  You asked for my specs.  Remember who you’re talking to…. Clueless in Texas.  : P  Remember this thang started out with me trying to wire tubes as P type devices and no signal transformers.   I’d like to think I’m finally on the right path.  The 6BQ7A sure does have very linear curves.  You can smack it with pretty good levels, it doesn’t have excessive gain, nice output impedance as a CF, and they’re $1.55 each.  I must be missing some shortcoming.

> I started this 8 hours ago and lost any train-of-thought I had.
I’m guessing you survived the storm.  It’s been raining all day here.  That’s not normal for Sept.

-Richard

[RicharD]

>And a test set.

Until I can get down to the studio and borrow a pair of good mics, I gotta work with what I got.  Hopefully my daughter will visit and I can capture some piano playing.  So for the moment, I have a crappy mic, in front of a crappy speaker, tuned into a crappy radio station and the volume is set low because people are sleeping.

Obviously the mic feeds straight into the protoboard.  Up at the top right is the 12VDC regulated supply for filaments and control power.  Next to is the 48VDC regulated phantom power supply.  Top left is the 350VDC B+ supply.  I like to build the power supplies in the back to keep the majority of dangerous voltages clear and away.  Left to right across the main rail is the top channel input transformer.  I'm currently not jacking with the top channel.  I'm experimenting on the lower channel only for the time being.  The other channel is live so that my power supply is properly loaded.  I'll go true stereo soon enough.  Next to the octal plug transformer are the 48V phantom supply relays.  The tubes left to right are 6BQ7 12AU7 & 6BQ7.  I only have one output transformer.

To the right of the main circuit is a normal breadboard with a TL084 meter driver circuit.  Next to it are a pair of vintage VU meters meters mounted in a cigar box.  You ain't really trying until you build something into a cigar box.

My load is a simply awful Alesis power amp.  The speculation sheet says it's input is 10k.  I'm driving both channels from a single source so I assume my load is now 5k.  The speakers are NHT book shelf speakers.  It's a less than perfect test set, but it's what I have to work with.

So....  With this all set up, I got busy.  1st thangs first, crank it up wide open with no source.  Hum is better than I expected.  I've done enough protoboarding to know to expect some noise.  My biggest noise source is all the unshielded wires flying off the input transformer.  As soon as a mic is connected, this noise source vanishes and I can turn my bench light back on.  I compared the 3 stage with the 150k dropping resistor above the gain pot to a 2 stage eliminating the 2nd stage plate follower.  I couldn't really tell a difference.  I also tried it as a 3 stage w/o the 150k attenuation resistor.  The added gain might be nice in some situations.  I have the VU calibrated to 0dBu.  With my radio source set back ground level (quiet enough to not disturb sleepers), I could make 0dBu with the unit wide open, 150k in and input set 1:9.  That seems pretty good.  With the 150k removed, 0dB was easy to get to.  I turned the radio off and cranked up the power amp.  I was able to hear the rain outside, the dog walking around, and the air conditioner turn on.  This is with a lousy Audix F15 discontinued cheapo cymbal mic.  I'm anxious to get the other transformers mounted up and some decent microphones in place.  Thus far, thangs seem right.

[PRR]

> a large variety of condenser microphones.

Recording WHAT?

Far-miked harpsichord in large room at 74dB SPL?

Close-mike trumpet, drums, or Fender Twin at 124+dB SPL?

Large/hot condensers, generally, don't demand lowest-hiss performance. They gotta have an amp in them anyway. Whenever convenient, the designer sets the gain/output so that the mike's own hiss is higher than any sorta-low-hiss thing you put beyond it. If you run AKG 414, you barely need an input transformer to get though a tube. A 150:600 1:1 is handy to to overwhelm tube hiss and mainly as a place to insert the Phantom power.

Also if you close-mike big percussion or guitar amps, hiss just is not a problem. A good mike and preamp have room-referenced hiss near 15dB SPL; bad rig may be 25dB SPL. When sounds reach 120dB SPL, and you scale for this to touch 0 dBfs, the hiss is down in your 16th bit and will play-back far below the ambient sound of most listening rooms.

And indeed you may not need gain. I did a BIG percussion riot 10 feet out with 414s and I figured my board was set at a net gain of +6dB. That left some headroom, because I did not know the piece and this guy had energy to spare. Of course equipment and room hiss was completely out of the picture. Even the few rests were so short your ear could not recover 96dB before the next bang (heck I think some of that racket lingered in the room for an hour).

First stage gain of ~20 is probably fine. If first stage is lowest-hiss and second stage optimized for gain has twice or even 5 times (pentode) as much hiss, 1uV times 20 plus 5uV is far less than a dB increase of noise figure.

You probably don't need +6dBu/+18dBu. The input hiss of a 16-bit A/D converter must be OTOO 30uV. This means any gain over 30 on large condenser or 150 on dynamic is ample to make the mike hiss overwhelm the converter hiss. Once you get there, you can do a final gain trim digitally. Digital gain is prone to expose low resolution; but a "24-bit" converter fixes that and needs even less gain to get in its happy zone.

I think the 6BQ7 is under-appreciated. But it would not be my go-to for a high-level stage with high-ratio input iron. You could possibly revert to the transformerless balanced input plan and get good hiss with condensers and 6BQ7. But I've had a "harrowing" day and can't think.

[PRR]

.

[RicharD]

Talk about swinging a thread 180!  Nice circuit.  Back to the original topic: diff amp, diff supply, and no transformers.  I'll protoboard it as soon as I have enough board space.  Right now I'm a little too invested in the 6BQ7A project to tear it down.

>I think the 6BQ7 is under-appreciated. But it would not be my go-to for a high-level stage with high-ratio input iron.
Apparently it's not anybody's "go to" tube except for HP.  (about 1/4 of my pulls have HP stamped on them)  It sure does seem right for my purpose.  Maybe there's something sucky about it I don't know.  It certainly isn't as sexy as the EF86 or 6DJ8.  I'm not quite ready to throw in the towel just yet.  I should have good mics tonight as well as good output transformers.  I wanna run a few more tests at home and then drag it down to Yellow Dog to run it in a "real world" setting.  Oh yeah, I've also added input pads.  I've been weighing building NY Dave T-Pads vs. just buying the Clairostat pads.  1 is cheap but labor intensive, the other is KISS but $40.00 ea shipped.  I'm also awaiting the arrival of some LM3916 LED VU drivers.  Analog meters = schweet but expensive.  LED meters are easily 10x less expensive. 

[PRR]

> Nice circuit.

Stupidly expensive. And for your whopping 25 Watts, the performance is not that great.

But it is what you were dreaming-of in the first post.

It could be tweaked better. Not wonderfully better.

As biased now, the negative rail is mostly waste, just long-tail "constant current" bias. A -6V rail and some NPNs would give better performance at lower cost and heat.

It IS one of the few contraptions which will direct-connect to a mike with or without Phantom Power, no input transformer or caps. The lower pair of 6.8K are a mid-lane condenser DC load, but these could be zero (grounded-center ribbon) or infinity (floating dynamic) with only slight effect.

Equivalent noise resistance is about 1K. So it degrades the self-hiss of a typical dynamic. Hotter condensers have higher self-hiss (and output) voltage; this would not degrade an AKG 414's hiss. Nor many of the large-capsule "newmann-like" $99 products.

Maybe the worst fault: without iron, run balanced and well within ratings, it may have no "tube sound" at all. It may sound like the fault-free $9 chip, with a bit more hiss, and a LOT more cost and heat. How cold is it there this week? (Up here I could already use another 25 Watts in the bedroom.)

[RicharD]

We don't know the meaning of cold in Austin, although this summer compared to last summer has been mild.  Only a 7 day streak of triple digits, not 30 something like last year.  Hurricane weather never makes it this far inland, but for some reason this recent tropical depression dumped buckets here.  8" in 3 days.  I think today's high was 92.  My truck is loaded with construction trash and it's all soaking wet.   YUK!

I didn't get the mics today, but arranged for them for Saturday.  Transformers are here.

[RicharD]

Just when you thought it was safely on it's way to page 2, Richard downloads the 30 day trial of DSSF3.  Stayed up way too late last night reading the help files (very broken English) and calibrating my sound card to the software.  I've never hooked a DMM up to my computer.  At $80.00, I think I'm gonna purchase a license.  Cool package.  http://www.ymec.com/hp/signal2/thd.htm

So the 1st attachment, DSSF3.gif are the THD and response graphs for my sound card.  I hard patched outputs 3 & 4 into inputs 1 & 2.  Obviously my unit under test cannot have greater resolution than the sound card it's connected to.  (BTW:  My sound card is an old 8 channel Terratec MT88D.)  It's important to know this plus it's important to know where zero dB is and calibrate the software accordingly.  There's nothing more confusing in electronics than erroneous telemetry.

Today I drug my big ol protoboard into the bedroom and patched into my PC. I ran 9 different scenarios.  The 2nd attachment is the best result.  If anyone cares, they can all be seen here:  http://www.sotxampco.com/Temp-Telemetry/

So.... it really is all about the pad.  I was surprised to see distortion increase as I turned the gain and/or master levels down.  I was expecting the opposite.  Padding the input seemed to have little or no effect on THD or frequency response.  Switching the input transformer to 1:17 from 1:9 had dramatic effect on high frequency response.  I was a little surprised by this given the "spendy boy" nature of CineMag transformers, but I guess it compares to geetar amps where one can trade off low frequency response for raw power.  Adding a 150k resistor above the gain pot seemed to make the circuit sit in the pocket.  Response was flattest and THD was most consistent. 

I did get my paws on some good mics.  I'll record some sound bits soon.

[DummyLoad]

sine waves are nice and orderly and things... how does it like a square wave?   

[RicharD]

see attached

[RicharD]

I'm sold.  I'm gonna buy a license.  The next attachments are waterfall spectrum analysis.  The first one is no real signal in with everything dimed.  I think this is my worst case noise floor.  It get a whole lot prettier with the pads set to infinity.  The 2nd one shows an A major chord.  It's like a function generator with Hammond organ draw bars.

[RicharD]

http://www.sotxampco.com/MP3/Mad-Rush.mp3

Above is a sound recording I just made.  Remember this circuit is built on the breadboard.   The mics are Previs.  They are wooden body condenser mics.  The mic pre was set 1:9 w/o any pads.  Master @ 100%, gain at 25%.  The instrument is a Young Chang upright piano.  The musician is my daughter playing from memory.  The song is Mad Rush by Phillip Glass.

DoH....

The DAW is Samplitude software.  The hardware is a Terratec EWS88MT.  Mics placed close behind the sound board.  Panned hard left and right.  No EQ, compression, or anything else added to the recording.