Transformer coupled preamp

[WaveshapeIllusions]

In my various attempts to male a tubey preamp without using tubes, I have come across a few things. Currently I'm using a Fetzer Valve stage, and it sounds pretty nice.

I was looking at a basic Class A tube amp schematic, and decided that I'd like to transfer it for use with JFETs. I've often heard that the transformer is part of the tube amp tone, so I picked up a small 1:1 audio transformer.

Looking at the output stage specifically, the B+ is ran from one end of the primary to the plate at the other. There is no load resistor.

I've not seen this done with transistors before, so I was wondering if it was even possible to just transfer it like that, replacing triode with JFET. Obviously there will be DC in the primary which makes saturation much more likely,
but that's part of the tone to me. The main concern to me though, is that gain is related to the ratio between source and drain resistances (I think). Since the primary resistance is rather low, this may make for a rather low gain circuit. I'd of course bypass the source resistor with a cap, but I'm not sure how much that will help the gain with such a small drain load. Should I add resistance in series with the drain? Is this nothing to worry about? Thanks in advance.

[WaveshapeIllusions]

So I guess it's not a good idea?

[R.G.]

This is not intended as criticism, just information. What you're doing in trying to learn these things is good.

I was looking at a basic Class A tube amp schematic, and decided that I'd like to transfer it for use with JFETs.

You are aware that transferring tube circuits to JFETs is maybe a fun idea, but it's not really "tube-y" in any way, right? The old saw that "JFETs are just like tubes, kinda, somehow" is pretty much wrong.

I've often heard that the transformer is part of the tube amp tone so I picked up a small 1:1 audio transformer.

There is no "the" tube amp tone. Tube amp tone is a continuum, and quite wide. Transformers, with their inherent frequency and linearity restrictions when used as required for the output stage of a tube amp, do add filtering and some kinds of distortion. However there is no guarantee that any old transformer, most especially a 1:1 audio transformer with wildly different design requirements will sound anything like a tube amp in general or one tube amp in particular. The conclusion does not follow from the diffuse generic rumor.

Looking at the output stage specifically, the B+ is ran from one end of the primary to the plate at the other. There is no load resistor.

There is a load resistor, you just can't see it. This is the classical single ended transformer coupled setup, where the load on the plate/drain is the transformer primary inductance in parallel with the transformed "resistance" of any secondary loading. It's an AC only resistor.

I've not seen this done with transistors before, so I was wondering if it was even possible to just transfer it like that, replacing triode with JFET.

In general, yes, inductive and transformer loading on JFETs is possible. It may well and probably is not possible to just plop in a JFET where there was a tube with random transformer, and get any particular tone.

Obviously there will be DC in the primary which makes saturation much more likely,  but that's part of the tone to me.

Not knowing the transformer or the DC currents means you can make no estimate other than some guess that the transformer will even pass audio. I have found that most people who like the sound of saturation have wildly incorrect ideas about what saturation does to sound.

The main concern to me though, is that gain is related to the ratio between source and drain resistances (I think)

This is true for resistor-capacitor circuits, not inductive or transformer loading; at least, it's much more complicated for the last cases.

Since the primary resistance is rather low, this may make for a rather low gain circuit. I'd of course bypass the source resistor with a cap, but I'm not sure how much that will help the gain with such a small drain load. Should I add resistance in series with the drain? Is this nothing to worry about?

For frequencies where the transformer will load the device properly, the effective drain load is the transformer impedance ratio (1:1 in this case) times the load on the transformer secondary. This is true for gain calculations too. If you're designing, trying to hit some target, yes, it's something to worry about. If you're experimenting to see what you get when you do ... this... then, don't sweat it, try it out, and see what you get.

Remember that the drain will DC bias to the power supply because of the low DC resistance, so power dissipation in the JFET is set by the gate-source voltage setting current in the channel. You can estimate this by measuring the DC voltage across the source resistor and applying Ohm's law.

[WaveshapeIllusions]

Thanks RG. Always glad to learn new things about it. Being that this will be in a guitar, the load would then be the input stage of (in my case) the amplifier, right?

I agree that the tube amp tone is rather variable. In this case I'm aiming for warm with primarily second harmonic distortion.

I tested the transformer with alligator clips just to make sure it passed audio. There was definitely lower signal level and some low end loss. C1 was about 6 dB lower relative to the second harmonic than without the transformer. There was definitely some fuzziness on the notes, I thought it sounded OK. However, this was with the Fetzer Valve which has a load resistor and coupling cap. It sound different with the transformer direct to drain I'm sure. Perhaps it would be easier to use a standard load resistor and just put the output through the transformer?

Thanks for explaining things to me. I like to expirement with my setups, but it helps a lot to know what I can expect.

[WaveshapeIllusions]

Wired it up, and it indeed functions. There is less of a volume drop than just having the output go into the transformer. Bass response is better than that test too. Cool.

[PRR]

> So I guess it's not a good idea?

Maybe not a great idea to give-up in 4 hours, a concept with so many angles. I putzed-around all day hunting for examples, but you gone and done it.

Let's see what I have that R.G. didn't touch-on.

> I've not seen this done with transistors

Bah. MILLIONS of radios used this scheme:



Note that Q2 is wired essentially just like your tube-amp. (Yes, B+ and B- seem to be other way round; this is arbitrary, and this is a transitional design.)

Another:

http://www.kadiak.org/tel/altec456b.jpg

Q2 Q3 are parallel because this is days of small transistors so they doubled-up to get Pro level output.

I spent some time (with timeouts) looking for non-bogus transformer and FET examples. They are rare. Rheinfelder published some ideas about 1971 (not on the Web). I did a 2-FET transformer distortion box in the late 1970s, and about 1982 designed a SE rig for a contest promoting HEXFETs. (Neither were noteworthy.)

> primary resistance is rather low, this may make for a rather low gain

The miracle of the transformer is that audio-band impedance is *higher* than DC resistance.

An ideal transformer is zero at DC and infinitely low frequency, infinity at infinite frequency, and in-between between.

> a small 1:1 audio transformer

Insufficient data. While transformers do not "have" an exact impedance, they are always designed for certain goals at some approximate impedance. What impedance was suggested for this transformer??

> a small 1:1 audio transformer

I'm gonna guess you found a modem transformer. These give very-good response from 300Hz to 3KHz, when worked as 600 ohms each side. They are reasonably flat from 100Hz to 10KHz..... when worked near 600 ohms.

The actual impedance, typically:

30 ohms DC to 2Hz.
500 ohms at 100Hz. About 5,000 ohms near 1KHz.
Ideally rising past that, but really falling to zero at infinite frequency (capacitance always sucks). Depending on design and impedances, there will be some peak or peaking in the 1KHz-100KHz range, then falling.

The naked transformer peaks and falls. For "flat" response a transformer must see a sufficiently low resistance on one side or the other! In preamps driving unknown but high impedances, we "always" use a low-resistance triode (12AU7) to flatten the response. In Power amps, the heavy load defines the response. In the cheap radio above, Q2's collector transformer is loaded on the secondary with sucky Bases, about 700 ohms each which fits the 3KCT secondary rating.)

> Should I add resistance in series with the drain?

Does nothing. The Drain is a nearly infinite impedance (unlike a low-R triode). Why would you increase it? If you drive true 600 ohm loads (telephone or old studio gear), the load will swamp the transformer's impedance curve and give flat response over some frequency range. When the load is high or unknown (amplifier input is always too-high-Z for good transformer loading), then you put a resistor across the primary (or NFB, or mixed-feedback as in the Altec 456b, a tricky design; and kills flavor).

> a tubey preamp without using tubes

I am reminded of making a Porterhouse Steak using only tofu and beans. (If you are vegan: making a tofu/bean dish using only chopped flesh.) Yes food is food and amplifiers are amplifiers, but the flavor may be different.

OTOH, this isn't rocket science experimentation. You are only risking a few dollars worth of parts. You certainly should play with it.

[WaveshapeIllusions]

Indeed I did get a modem transformer. 600-900 ohms. I was a bit concerned about the frequency response of it, but the loss of bass isn't terrible. It is definitely saturating, as there is a kind of fuzzy growl behind every note. It is also rather middy in tone. I like strong mids, so I am not bothered by that. I think I'll have to play it a bit to see if I want to keep it. I'd say it's more interesting than the Fetzer Valve to me. I could set it up to have a proper load on the secondary...

Thanks for finding other examples of it this in use.

[GFR]

look at the schematic of the Univox U65RN or thhe Giannini U65G

http://www.univox.org/pics/schematics/u65rn.jpg

The phase splitter is transformer based. It was very underrated at its time, but I really like the way it sounds.

[WaveshapeIllusions]

look at the schematic of the Univox U65RN or thhe Giannini U65G

http://www.univox.org/pics/schematics/u65rn.jpg

The phase splitter is transformer based. It was very underrated at its time, but I really like the way it sounds.

Oh, that's clever. I like that. It's almost like the setup in a passive DI. Very accurate balance between both signals since and you only need one driver rather than two.

[R.G.]

Historically, the "totem pole" arrangement for output devices driven by a transformer was the common style for power amplifiers for a few years, from perhaps 1959 through 1966 or 1968. The time period is just my guess based on seeing schemos of amplifiers and guessing about their design time.

It was a solution to the problem that good complementary NPN/PNP power devices did not yet exist.  During that time period one could get high power PNPs in germanium and high power NPNs in silicon, but germanium PNPs and silicon NPNs are not complementary, and don't combine all that well in a feedback amplifier. It vanished as soon as the Sziklai (or complementary feedback) pair was popularized to let you make an effective opposite-polarity power device with two power devices of the same polarity, in the quasi-complementary circuit. Designers really, really wanted to get rid of the expensive, heavy driver transformer with it's limitations on the circuit design.

That power amp circuit is a special interest of mine, as all of the Thomas Organ Vox amplifiers used it, as well as other amplifiers from other companies around that time.

[B Tremblay]

The Fender Musicmaster Bass amp also used a transformer phase splitter:
http://ampwares.com/schematics/musicmaster_bass.pdf

[R.G.]

The Fender Musicmaster Bass amp also used a transformer phase splitter:
http://ampwares.com/schematics/musicmaster_bass.pdf

Yep, they picked it up from tube amp practice. That's a conceptually diffferent circuit, though. If you look at the schemo, The tubes have cathodes to ground through a resistor, and each plate is connected to one end of an output transformer. They're actually running with voltage gain, in common-cathode connection. The totem pole/stacked output devices run in common collector/emitter follower mode without an output transformer.

You can do the same thing with tubes, and in fact there were some early tube amplifiers which used the driver circuit to run paralleled output tubes without an output transformer, direct coupled to the speaker. This at least moved the problems with a transformer back to a lower power level where it could be dealt with less expensively.

Notice that there is no feedback from output to the power amp input on that one. It takes three time constants to make something oscillate. A transformer is two of those all by itself. Two transformers is four, and it's hard to get by without at least one more, so you are very, very restricted on the open loop gain you can use without the thing becoming an oscillator. The classical tube guitar amp output stage has one transformer, and one time constant in the grid capacitor driving the output tube grids.  That all by itself has two ramifications. One is that the open loop gain is restricted to some value that's dependent on the tubes, capacitors, and output transformer to be stable, so the amount of negative feedback available to ensure consistency from unit to unit and suppress distortion is also limited. Also, the feedback has to be direct coupled, as it is to the PI in almost all cases.

Engineers live on the maxim that what can't be controlled must be made irrelevant. They were forced to exert a lot of control on the output transformer, because it could not be made irrelevant.

[tca]

Just to add some more to the collection (from 0.5W to 4W) with the transformer specs:





Cheers.

[R.G.]

Yeah, that's the transistor version of the tube push-pull circuit with an output transformer. The transistors are running in common-emitter, so they have both voltage and current gain, not just current gain like the common collector/emitter follower connection. That version was the first adaptation of the tube two-transformer circuit, and is nearly part-for-part, allowing for different values.

[PRR]

We ARE getting off-topic (transformer loaded JFETS), but....

> you only need one driver rather than two.

Astute comment.

However the driver is the "same" as the one in my "MILLIONS of radios" plan. Only the output differs: the push-pull has been split and half-flipped which potentially eliminates the output transformer.

> output devices driven by a transformer was the common style for power amplifiers for a few years

With few exceptions, transformer driven was the rule. Transistors are eXpensive. Costs a lot to get the gain you need. The CE connection gives best Power gain. But its input impedance is very much lower than the output of a driver stage working on the same/similar supply voltage. Today we just add another transistor(s). But when they might be $7/pop, we didn't--- a $5 transformer to optimize the impedance ratio was a buck cheaper.

> The totem pole/stacked output devices run in common collector/emitter follower mode without an output transformer.

Nearly all of these (including that VOX) run common-emitter. Drive is applied between emitter and base (via low-Z base-bias network and small emitter resistor). Collector-emitter voltage has nearly no effect on output (in the VOX, via 2200/22 ohm bias network, 100:1, but the stage voltage gain is only 4/0.5 or 8 so the 2200/22 divider does nothing for AC gain).

Grid transformers used to be THE way to push/pull tubes, but fashion changed around 1938 (largely with 6V6/6L6 tubes). The MusicMaster is the last small non-Boutique transformer-driven tube amp. It was driven by the economy of avoiding a 3rd small tube, and perhaps some old surplus iron found in an LA warehouse. (The Fender 300 is another example, but huge, and with an added reason to tranny-couple).

[R.G.]

Nearly all of these (including that VOX) run common-emitter.

Actually, the stacked totem pole output stage is common collector. It's a bit complicated to figure that out on the transistor with its collector on the output pin.

Or did you mean the driver transistor?

[PRR]

I beg to differ.   EDIT.... argument withdrawn.

[R.G.]

You have complete freedom to differ. I've only ever found four textbook evals of the circuit, and all four of them say the same thing - common collector. I suspect that it's not really one animal or the other, probably a hybrid or a chimera.

I'd welcome your enlightened analysis of it. I've thrashed it for a long time, and come to the conclusion that it's best described as a current amplifier, not a voltage amplifier. But that's just me.

[PRR]

> textbook evals of the circuit

Curiosity piqued. Do you have cites/hints handy?

FWIW: This form is shown in Transistors I, RCA, 1956, but without theory. It also cites US Patent 1944216 which, as you might guess by the low number, does it with tubes (circa 1936).
_________________________________

Getting BACK to WaveshapeIllusions' question.....

A transformer and a tube (or whatever) together, you have to find parts that work together.

Full-range audio transformers "prefer" to be low impedance. Full bass at high-Z raises costs (fine wire). Hi-Z always forces compromises in the treble response, controleld by heavy thinking and added complication. R.G. touches on this in his #11.

OTOH, tubes (and some other devices) prefer to be high impedance. Low-Z is like a wide highway. More road-tar, more cathode-stuff, more money.

This is less-true for BJT devices. And low-Z JFETs are sure possible. However, a scan of the types of JFETs commonly available (small RF, small switching, audio, stompbox cloner supply) show that most have voltage/current values appropriate for higher impedance.

Let's try 1K impedance. If you have a 10V supply, and want to use most of that in a 1K load, you need 10V/1K= 10mA of current. Actually for clean class-A work, you idle near 10mA then swing up almost to 20mA and down to almost zero mA.

So you want a JFET with "Idss" a minimum of 20mA.

That's the extreme end of Idss specs for the commonly available JFETs. And even a part number which could have Idss as high as 16mA or 24mA, could also have Idss as low as 2mA.

We are a factor-of-10 low in available current for 10V (or 9V) supply. Let's instead consider a 10K transformer winding.

These were once VERY common. Look back at my "MILLIONS of radios" schematic. It shows 5K load on the driver-- designs varied from 3K to 20K depending on other factors, but 10K was often used and still in production.

The part-numbers in the picture have been in production since 1960 and are stocked at Mouser for a few bucks each.



With 10K load on 10V (or 9V) supply, we need Idss a minimum of about 2mA. Many-many "small RF" and "small switch" JFETs will do 2mA. (The main exception is the electrometer FETs such as used inside electret mikes.)

The source bias for 1mA will typically be 1V-2V. FETs vary widely. A 1.5K bias resistor is a good first try. Breadboard it, compute V/R. then change R to get near 1mA.

The "10K:2K" transformer is NOT any specific impedance alone. It will work best if it looks into these impedances, at least on one side. In this case it is simplest to load the output. This also is how tube pentode power amp OTs are loaded (heavy speaker on the secondary).

The rated "10K:2K" impedances are "low fidelity". Bass cutoff as high as 300Hz, two octaves off the guitar's bottom. Experiment with lower values for "2.2K", as low as 470 or so which may reach the lowest guitar note. This will also affect breakup level.

Maximum output (clipping, where you want guitar to peak) is near 14V p-p primary, 6V p-p or 2V rms at secondary. This is still 5 or 10 times more than you really need into a guitar input. The added resistor and pot allow you to bring it down so you don't burn the input of the guitar amp. You will want to play with these values.

The tone at beginning of breakup can be shifted from more-sour to more-sweet by changing either load or idle current. This is where you can really get lost in too many changes.

Gain from gate to 2K load is low, 2 or 3. Gain to pot wiper is lower, perhaps 0.2. You need additional gain of 5 just to get unity gain. This could be a resistor-coupled JFET voltage-amplifier stage. Since this is an "effect", you may want more gain, such as yet another R-C JFET stage. And a volume/gain control after the first stage but not as late as the gate of the last stage.

Note the similarity to Fender 12AX7 Champ, Epi Valve Jr, little Kays, Herzog, and a host of other small tube amps: volt-gain, volume, volt-gain, power-output.

[WaveshapeIllusions]

PRR, I like the schematic you have there. Mine is pretty close to what you drew up. The only difference is the output network is not there and the transformer is 1:1. I'm going to play with loading the secondary. What is the 390k resistor for though? A set level of attenuation?

As far as frequency response, ideally I'd like to get near 60 Hz (for bass use). I know the saturation limit is frequency dependent, but for small signals I would figure it's less of an issue. With the current set up, low frequency response is not horrid. Compared to dry signal, the fundamental of C1 is about 6 dB down. The second harmonic isn't affected by much. This was after accounting for gain, of course.

I should put a load on the secondary and perhaps a source follower buffer stage to keep things consistent. I've noticed a decent variation in saturation characteristics depending on what I connect it to. I'd like to keep a certain amount of saturation going as I like the sound so far. It sounds mostly clean with some soft grit in the lower frequencies. Probably because it's saturating around the fundamental but not the higher harmonics? Kind of like a crossed-over distortion?

Since it's for bass use, treble response is something I'm not concerned about. Less treble is usually nice, since it makes for a warmer tone. So that means I'd want one with high inductance. I'll probably go with what you suggested, PRR on the transformer. 10k seems about right for the drain load, it's what I've usually ended up with. Thanks for all the help.