Simple transistor phase splitter (180 deg) for 1-30MHz?

[neazoi]

Hi I want to build a simple phase splitter for sine wave for 1-30MHz. If that possible with a single transistor or other discrete transformerless design?

My experiments with simple transistors show that depended on the frequency there are amplitude variations and large phase errors.

 

[BradtheRad]

A transistor is a phase splitter in the sense you can take outputs from collector and emitter.
Adjust values so as to obtain desired amplitudes. If desired change 0V ground to a negative supply voltage.

 

[FvM]

My experiments with simple transistors show that depended on the frequency there are amplitude variations and large phase errors.

Too large load capacitance, too low transistor fT, emitter/collector resistor too high?

 

[FvM]

A test shows the effect of transistor choice. BFR92A is SMD equivalent of popular BFR90 5GHz transistor.

The circuit needs to be modified if you want to drive a lower load impedance. It might be easier to use additional common collector buffer amplifiers in this case.

 

[neazoi]

A test shows the effect of transistor choice. BFR92A is SMD equivalent of popular BFR90 5GHz transistor.

The circuit needs to be modified if you want to drive a lower load impedance. It might be easier to use additional common collector buffer amplifiers in this case.

View attachment 168044

Notice at your simulation phases are not exactly 180 degrees out of phase. This is what I experience myself. Also the phase moves a bit as the frequency changes (1-30MHz). Why is this happening and can it be corrected?
Maybe the phase does not move but if may be from the increased distortion (in the oscillators I have tested) and it seems optically that the phase moves.

 

[FvM]

Please read the text comment in the schematic. Phase error with general purpose transistor 2N2222 is 17.5 degree, with BFR92A it's only 1.5 degree.
Main causes are transistor capacitances (mainly Ccb) and limited transit frequency fT.

 

[neazoi]

Please read the text comment in the schematic. Phase error with general purpose transistor 2N2222 is 17.5 degree, with BFR92A it's only 1.5 degree.
Main causes are transistor capacitances (mainly Ccb) and limited transit frequency fT.

I only have a BFR91A here, I guess it does not matter a lot. I see they are pretty similar

 

[FvM]

Yes, it should work too. A word of warning, these GHz transistors are likely to self-oscillate in your circuit at several 100 MHz, a base series resistor of 50 to 100 ohms often helps to supress oscillations.

 

[neazoi]

Yes, it should work too. A word of warning, these GHz transistors are likely to self-oscillate in your circuit at several 100 MHz, a base series resistor of 50 to 100 ohms often helps to supress oscillations.

These 100K output shunt resistors, can I completely remove them and use instead series resistors, 500R for the 0deg and 1k for the 180deg output?

 

[FvM]

The output capacitors and bias resistors are primarily for simulation purposes. You need to check DC bias and load conditions of your actual application. I can't suggest details without knowing your load.

 

[danadakk]

You can always use a dual DDS and get exact variable phase shift of very fine
resolution over a broad freq range.....Needs a simple processor interface to
manage it. This of course overkill if all you want is a 180 degree shifter.

https://www.element14.com/community/groups/open-source-hardware/blog/2018/08/11/building-a-direct-digital-synthesis-dual-channel-signal-generator

Regards, Dana.

 

[neazoi]

Yes, it should work too. A word of warning, these GHz transistors are likely to self-oscillate in your circuit at several 100 MHz, a base series resistor of 50 to 100 ohms often helps to supress oscillations.

Ok this is an idea for a 360deg phase splitter based on your circuit (relative phase between output signals).
It works on 4 variable capacitors, but it does not work on the 4 varactors in place.

I want to place there the 1sv149 varactor, can you help me import this to the ltspice?
Also I want to place a stereo potentiometer, one gang will provide voltage to the 2 top varactors and the other gang will provide voltage to the 2 bottom varactors.
When the 2 top varactors see a maximum voltage, the 2 bottom varactors should see minimum and vice versa (differential).

Can you help me a bit on this, I do not know how to make it in LTSPICE.

 

[vfone]

An integrated Gilbert cell, as the cheap NE612, would provide a stable and accurate phase shift over the entire frequency range.

 

[neazoi]

An integrated Gilbert cell, as the cheap NE612, would provide a stable and accurate phase shift over the entire frequency range.

I cannot simulate it.
I think that the mixer will produce many unwanted products in the RF.

 

[vfone]

If input level is low, the mixer will behave just fine. I am using 10mV input signal in the simulation.
The main reason using an integrated Gilbert cell is to get very low phase difference vs frequency range, as you mentioned you get in the first post.
Depending by the type of Windows you use, the SA612.asy file that I attached, should be placed in the Program Files\LTspiceXVII\lib\sym instalation folder, or in User\\Documents\LTspiceXVII\lib\sym folder. LTspice needs to restart after this operation.

 

[neazoi]

A test shows the effect of transistor choice. BFR92A is SMD equivalent of popular BFR90 5GHz transistor.

The circuit needs to be modified if you want to drive a lower load impedance. It might be easier to use additional common collector buffer amplifiers in this case.

View attachment 168044

FvM, see this modification I did in your circuit. I have decreased the input signal as well as the emitter resistor.
The output is still a clean sinewave, but the average collector current is 61mA, which is far bigger than the Ic of the device (25mA).
But the simulation still shows a clean sinewave at the output.

Is it safe to use it as it is, or do I need to take into account the Ic separately from the simulation results?

I guess I could use a BFR93A (Ic=90mA) instead with the same results?
But I need to be sure I am doing the right thing above, so as not to produce meaningless simulation results.

 

[FvM]

Rating for continuous Ic has to be observed. Either adjust the bias point or use a transistor with higher current capability.

 

[vfone]

In real life, using a single transistor, and mainly due to the Miller effect, the phase shift error vs frequency will not be as good as in the simulation.
The parasitic capacitance between input and output (so the phase shift) will change with the bias of the transistor and with frequency.
An integrated Gilbert cell, mainly due to its symetry and because all the active components are in the same package, will take care somehow about this issue.

Miller effect - Wikipedia

en.wikipedia.org

 

[BigBoss]

If two transistor solution is acceptable, a Differential Pair with one Port is Grounded will serve perfectly Single-Ended to Differential Converter so 180 degree Phase Shifter at Low Frequencies.

 

[biff44]

balun

1:1 CORE & WIRE Transformer, 0.15 - 400 MHz, 50Ω | ADT1-1+ | Mini-Circuits

Mini-Circuits is a global leader in the design and manufacturing of RF, IF, and microwave components from DC to 86GHz.

www.minicircuits.com

over your frequency range, less than 2 degrees of phase imbalance