Single transistor phase splitter questions

[earthtonesaudio]

I'm having trouble finding info about the single-transistor phase splitter.  The kind where you have equal valued emitter and collector resistors, and take signal from both outputs.  ...So I have some questions that my web searching has so far not answered.

1. Bias?  Do you just bias the base to 1/2 supply voltage and call it a day, or is there some optimum point depending on the transistor (if using BJT, JFET, or MOSFET?)
2. Emitter degeneration/bypassing.  In the phase splitter configuration, will the "emitter bypass cap" still have the same function of raising gain for signals coming from the collector?
3. Other things I should think about?

Thanks in advance!
-Alex

[frequencycentral]

Hi Alex,

There's some info on both these pages, you've probably seen them already though.

http://www.tpub.com/content/neets/14180/css/14180_40.htm

http://www.uoguelph.ca/~antoon/tutorial/xtor/xtor2/xtor2.html

[R.G.]

1. Bias?  Do you just bias the base to 1/2 supply voltage and call it a day, or is there some optimum point depending on the transistor (if using BJT, JFET, or MOSFET?)

Let's reason this out. For a clean phase splitter, you would presumably want the maximum undistorted signal swing on the output. If we were not splitting, only amplifying, then the best bias point for this is roughly half the power supply because that lets the active node swing up half the supply and down half the supply without limiting at the ends of the power supply. The biggest possible signal is the full power supply if we have a perfect transistor and no losses. If we put an emitter/source resistor in, that DC voltage subtracts from the available power supply, so we then want the collector/drain at half the remaining power supply after we subtract out the loss in voltage to the DC on the emitter/source resistor.

In a splitter, we want two outputs, each as big as possible, and each equal. That means that the biggest possible signal on each phase splitter output is half the power supply peak to peak; this is because in one direction the active device is off and the two outputs are at full power supply and ground, and the other way they meet in exactly the middle of the power supply.

Since each phase is half the power supply peak to peak, the static, no signal, bias point must be one fourth of the power supply for the emitter/source side and 3/4 of the power supply (i.e. 1/4 of the power supply down from the top) for the collector/drain signal (presuming NPN or N-channel devices).

So what is required is a bias point which puts the collector/drain at 3/4 of the power supply and the emitter/source at 1/4 of the power supply. Now you have to decide what bias is needed on the control node to get them there.

With a bipolar, this is easy. The base must sit at one Vbe abvove (for NPNs) the emitter. For JFETs, the gate must be the right amount of Vgs below the source. For enhancement MOSFET, the gate must be one threshold voltage above the source. All the JFETs not only have larger offset voltages than bipolars, they are highly variable. Bummer.

2. Emitter degeneration/bypassing.  In the phase splitter configuration, will the "emitter bypass cap" still have the same function of raising gain for signals coming from the collector?

Another thought experiment. With equal collector and emitter resistors unbypassed, you get equal AC signals on both emitter and collector because the same current goes through equal resistors. If you leave the DC conditions but in effect short circuit the emitter resistor, the AC current now goes through the capacitor, not the emitter. Where you had a gain of one to both emitter (has to be, from base to emitter is an emitter follower) and collector (has to be, same current through an equal resistor), now you have larger AC currents in the base because the emitter resistor no longer limits base current. So the same base current swings the collector more.

When you put a capacitor across the emitter resistor, the emitter stops swinging, and there is little or no signal at the emitter. The signal all appears at the collector. And the gain is high. But there is no phase splitting action possible. If you split the emitter resistor into two resistors with a cap to ground from the middle of them (or use the wiper of a pot) then the gain at the emitter goes from unity (cap/wiper at ground) down to zero (cap/wiper at emitter). The gain at the collector goes from unity (cap/wiper at ground) to some large number (cap/wiper at emitter). In between the signal at the emitter is the inverse of the cap/wiper position ratio.

3. Other things I should think about?

The Art of Electronics - Horowitz and Hill.
Loading on the collector affecting AC gain.
Cap values and frequency response changes.
Transistor self heating.

[earthtonesaudio]

Thanks guys!

I tried out some numbers in this nice little sim, also helpful:
http://www.daycounter.com/Calculators/Transistor-Bias/NPN-Transistor-Bias-Calculator.phtml

[puretube]

I love the 1/3rd - 2/3rd biasing...

[earthtonesaudio]

I love the 1/3rd - 2/3rd biasing...

Does this actually work?  RG's post had me thinking the output signals can't overlap. 

[R.G.]

RG's post had me thinking the output signals can't overlap. 

They can't. Think about it - the way you make output signals is to have the voltage across the active device go between the full power supply (that is, it's completely off) and 0V, when it's acting as much like a short circuit as it can.

With bipolars, there is some apparent overlap which comes when you overdrive the base. The C-E voltage goes down to a few millivolts and the collector is actually at a lower voltage than the base. When the base gets driven a diode drop higher than the collector, the base-collector junction starts conducting and base current pours into the collector (and emitter, they're at about the same voltage). Then the collector voltage goes up again because you're pulling it up with the base current.

The signals can't overlap, though.

[puretube]

I love the 1/3rd - 2/3rd biasing...

Does this actually work?  RG's post had me thinking the output signals can't overlap. 

Little less "headroom", but they stay off the rails...

[earthtonesaudio]

Okay I think I get it now. 


Thanks everyone who chimed in!  I have a fully functional phase splitter on the breadboard at this very moment, behaving exactly as I hoped it would.

[Gus]

look at the schematic

http://www.diystompboxes.com/smfforum/index.php?topic=56690.0

[frequencycentral]

Okay I think I get it now. 


Thanks everyone who chimed in!  I have a fully functional phase splitter on the breadboard at this very moment, behaving exactly as I hoped it would.

Just out of interest Alex, could you post a schematic of what you settled on? Thanks.

[earthtonesaudio]

No problem... This is possibly the most beautiful schematic I've ever drawn, so you'd better sit down...




Transistor is BC338.  I measured the collector and emitter voltages, and they were in the ballpark of 3/4 and 1/4 supply voltage but I forget exactly what they were now.

RC lowpass on input for shunting RF.  As shown it should filter out frequencies 1MHz and higher, but I'll probably eventually increase that cap to 1n for a 15kHz rolloff.

The variable resistor is for blending between unity gain signal from the emitter, and boosted signal from the collector (there's a mixing stage later on).

[JDoyle]

The 56k/27k bias string in the Ross Comp sets it up so that the emitter is biased near optimal for a phase splitter as it is used in the envelope detector - plus you can tap off of it with what ever value you want and get 'low noise' biasing by bypassing the junction of the 56k/27k with a cap.

Your biasing scheme of 1Meg and 470k is too high of a resistance, in my opinion, to have on the input of a BJT - lots of noise plus it is feeding the BJT very little current making it over-sensitive to lots of issues. Again, just my opinion.

Regards,

Jay Doyle

[earthtonesaudio]

Good points Jay.  I haven't noticed that it's overly noisy but I'm curious about the "issues" you mentioned.  What sort of issues?

[Pedal love]

The base voltage here works out to roughly 3/4 of the supply.

[R.G.]

Your biasing scheme of 1Meg and 470k is too high of a resistance, in my opinion, to have on the input of a BJT - lots of noise plus it is feeding the BJT very little current making it over-sensitive to lots of issues. Again, just my opinion.

Good observation Jay. A better thing to do would be to divide the resistors by 100 (i.e. 10K and 4.7K) with a BFC to ground, then run a 330K to the base of the transistor. This makes the so-called "noiseless biasing" setup.

[earthtonesaudio]

Well, less noise is a good thing.