The Norton Amplifier
(VA3DIW)


The Norton amplifier has very nice performance.
It is a sort of CATV amplifier. The gain is flat over a decade or more. For HF the input impedance is little bit higher than 50 ohms. The gain is selected by turns ratio in the collector circuit and tap. For VHF, the input impedance is lower and needs some match. The isolation s12 is good up to few hundreds MHz. It can work in broadband mode from 2 to 200MHz.
On the lower frequency side, the frequency limit is set by the transformer inductance, leakage inductance, and capacitors.
The upper frequency side is set by the ferrite output transformer, collector capacitance including s21 drop, and transmission line length. It can cover few octaves and reach decade or two. The noise figure is good. Common base configuration has extended dynamic range in terms of P1dB compression, IP3, and IP2, compared to common emitter configuration.
For the least understanding:   Look at the schematic diagrams where is common emitter at the critical spot. It is the "Dutch Hottentot design culture" that will never work properly. Many people copy and publish the same crap. Once you build and use IP3 tester, you will start to understand.

The transformer turns ratio is:
N = M2 - M -1.

N ... collector to output tap turns
M ... output tap to Vcc, turns
E emitter input ... one turn, for lower gain oriented
gain = m2;   [in linear number]
voltage Gain = 20 * Log10(m2);   [in dB]

There is only one gain in dB. Doesn't matter if it comes from voltage or power. Some "NCDXF" folks don't know that. They claim power and voltage give different [dB] numbers. Of course, it is religion. What is really bad, is the mob attitude and ignorance of math.

Voltage gain in [dB] = Power gain in [dB].
If you use dB units, nothing changes.

G = 20.Log10(V1/V2) [dB]; voltage
g = 10.Log10(p1/p2) [dB]; power

magnet wire diameter 0.25mm


Vcc = 12V, Vces=0.5V, Ic= 25mA, 2N5109 from Central Semiconductor NY.
Yes. You can buy from Central Semiconductor. Check, if they ship to your location.

  E [turns]   M [turns]   N [turns]   nominal Gain [dB]   theoretical Zc [ohm]   Power [W]
  1   2   1   6.02   150   0.880
  1   3   5   9.54   400   0.330
  1   4   11   12   750   0.176
  1   5   19   13.9   1200   0.110
  1   6   29   15.56   1700   0.085



The "E emitter" input has one turn. If you swap the input wires, the amplifier will give higher gain - that's wrong. Oscillation.
The collector turns "N" ratio commands the gain. "M" goes to the Vcc. Select gain 6 dB, 9 dB, 12 dB and count with 0.5dB loss in the ferrite.
Consider the input transients from antenna with amplitude of plenty volts, then the ESD, then the lightning. I believe RF medium power bipolars (5 Watts) in metal case are more rigid for this application. Including the case thermal resistance. The 5-7GHz transitors are mostly unstable in the HF region. What is it good for, when it looks nice, but it oscillates?

You might find similar designs with the feedback in the emmiter. There are another 3-4 similar amp configurations with transformer used in cable tv CATV networks. I compared 2N5109 with 2N2222. The 2N2222 0.5 W transistor is little bit lazy, requiring higher collector current to get the gain-bandwith product. It has higher rbb resistance.
The metal case transistors are better than the plastic ones. There was a difference of 6-8 dB in IMD3 between the plastic and metal case. Same product, same parameters, different case. The plastic does not dissipate more power than the metal can, and has higher thermal resistance. The input was fed by IMD3 test signal generator delivering two input signals. Measured with HP spectrum analyzer. Transistor selection can be BFR34, BFR96, bfw16, bfw30, 2sc3355, 2SD636, 2N3866, and you will surely find another choice. Check for stability from s-parameters. Usually, the documented s-paramenters are from 500MHz up. HF bands are noisy, the noise figure won't make a difference.
The IMD3 mixing products drop with lower set gain, larger current (linearity), higher Vcc voltage, and better linearity transistor. Use of higher voltage will improve the IP3 in single digit [dB], not much. The amp. architecture makes the difference. If you drop the current, naturally there is less output power handling capability and the transistor delivers less power and the IMD3 products rise up. The two signal method is good enough. Ferrite limits are - maximum magnetic flux and the linearity. I have noticed some older sensitive HP EMI spectrum analyzers have limits for IP3 reading, with finite IMD3 floor. They get easily internally overloaded, even when using huge external input attenuator. You might get readings that are not fully correct. Check, if 3dB input level drop makes 9dB IMD drop. That's life.

I prefer ordinary transistors over the specialized HBT chips in specialized case, with specialized footprint, and temporary production/ replacement.
Once the semicondonductor foundry is bought out, belly-up, hit by the "Chineese virus", or the product becomes "obsolete" for unknown reasons, there are no spares.
There is a method how to measure the signal distortion using the EVM meter - error vector magnitude for linear signals such as OFDM, QAM, BPSK. Most of the companies rarely rent this machine. It is a small HP desktop mainframe computer.
Somebody thinks, there will be no IMD3 distortion in the SDR-DSP receiver unit. It is naivety and lack of information.
"I can get rid of these stupid transformers replacing them with the 3.3V /500mA drivers and buffers with noise figure of 15dB, 1GHz wide." They talk few months, big mouth. Then they are silent. You will see the intermodulations and the DSP image, harmonics, all, as well. Try >=14 bits. It will take few Amps from the power supply. The FFT is power hungry.
I saw designs with the 1GHz OPAmps. Calculate the noise. Specific calculation gave 100uV of noise, measurement gave 80uV of noise. :-) Nowadays, I am tired of fixing similar professional systems.
The HF test signal power levels used for measurement require high power splitter. Every ferrite material is different. Simulation is useless. You can try running scientific LTSpice till next Christmas, without any reasonable results. Consider Fair-Rite, MuRata, TDK ferrite materials, Neosid, Wurth. I got a nice two hole ferrites from Houston, Texas store. Strongly recommend that shop. Nice people, flexible attitude, good pricing, full customer service, in few days you will get it by surface mail.

The following article from ARRL is a description of consequences, when VHF/UHF transmitter transistors are used for HF frequency bands.
(Same thing Jack, what you have with Philips 7GHz transistors).
All RF bipolar transistors have a low frequency region with very high gain. It looks like rabbit ears. Then the gain drops, and follows datasheet values starting at 500MHz. It is useful to check the s-parameter matrix for stability area at the band of interest, and close by. Mostly the s-parameter table starts at 500MHz. Recent semiconductor manufacturers do not deliver any low frequency s-parameters any more. Enjoy the discussion what transistor is the best choice. The push-pull configuration with 5GHz transistors oscillates... You will find many XEROX websites, copying the mistakes.

Epcos toroidal ferrite cores
Warning: The ferrite edges are sharp. Epcos higher permeability ferrite conducts DC current. Sharp edge will cut the magnet wire insulation, resulting in shorted turns, reduced gain, and funny behavior. Everything counts. Electronics does not forgive anything. That's the lucky difference from politics.

Norton amplifier ARRL.pdf

VA3DIW

Resources:
[1] VA3DIW Dave's notes
[2] U.L. Rohde: Digital PLL Sythesizers
[3] Lankford: Common Base Transformer Feedback Norton Amplifiers
[4] Futurelek.com - smd and leaded RF transistors

MADE IN USA
The IP3 test signal source board is available for swap. Three boards left.
Build your IP3 tester during weekend on 2 x 2 inches board.