Things Coming Together; Bode Plot, DIY Isolation Transformer & Peltz Oscillator

[mawyatt]

Here's something that might be interesting for folks that like to experiment with electronics, or maybe just have some fun

Tying the Bode Plots with DIY Isolation Transformer and the simple 2 Transistor Peltz Oscillator together.

About Bode Plots:

https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg4275751/#msg4275751

https://www.eevblog.com/forum/testgear/siglent-sds2000x-hd-12bit-(published-for-chinese-domestic-market-only)/msg4286026/#msg4286026

About DIY Isolation Transformer:

https://www.eevblog.com/forum/testgear/diy-transformer-for-use-with-bode-plots/msg4182022/#msg4182022

About Peltz 2 Transistor Oscillator:

https://www.eevblog.com/forum/projects/simple-sinusoidal-oscillators/


The general idea is to measure the Closed Loop Response of the Peltz Oscillator utilizing the Isolation Transformer described above. The scope utilized was the Siglent SDS2104X+ with an external AWG (SDG2042X).

The Peltz Oscillator described above used a pair of 2N3904 NPN transistors, with an 470uH Ferrite Core inductor, a 0.22uF Film Capacitor and a 510 ohm resistor. Vee bias was set to -1VDC. The Isolation Transformer secondary was placed between the collector of Q1 and the Base of Q2, providing a DC path so as to not disturb the DC oscillator bias.

Note that the phase aberrations in the upper left corner are due to the oscillator actually running during the lower frequency portion of the Bode Plot routine, this was observed by monitoring the oscillator output with another scope during the plot routine. This speaks highly of the Synchronous Sampling Technique (or as some prefer Frequency Selective Sampling) utilized with Bode Mode in the SDS2104X+. Since the scope must "pull out" the desired waveform from the on-going oscillation, conventional sampling would have the wanted waveform obscured by the oscillation waveform and be difficult to correctly sample.

Simulations of the Oscillator Closed Loop Gain and Phase in LTspice schematic and plot are shown below.

This simple setup allows one to experiment with various bias points and component values and observe how they influence the oscillator behavior, and compare with simulations.

Have fun


Edit: Changed the oscillator inductor to 100uH (measured 97.3uH) and added plots.

Best,

[moffy]

Nice correlation between sim and real life. This is a fun project, thanks.

[mawyatt]

Thanks, fun project indeed!!

Here's a couple more with the 100uH inductor, these are time domain scope plots and simulation. Note the indicated frequency on the scope plot vs. the pervious last two Bode Plots (sim and measurement).

Best,

[moffy]

It's almost enough to make you think SPICE can be trusted. That is tongue in cheek as there are lots of gotchas between SPICE and the real world, but nevertheless the agreement is excellent, and as you mentioned previously, oscillators can be problematic in SPICE.

[mawyatt]

Spending the later part of our career designing chips in SOTA processes such as IBM SiGe BiCMOS one develops additional skill sets since these chips can't effectively be breadboard like conventional circuits. Some of the additional skills are becoming very familiar with how the simulators work under the hood and being able to quickly understand when the simulator is fibbing, and the same goes for models!! Spending a couple years and a few $Million USD designing a new chip in a new process, then another few $Million USD for a fab run creates an environment where errors of any type are career limiting!!

SPICE can be your friend or worst enemy depending on how well you manage and understand it's behavior.

Best,

[mawyatt]

Out of curiosity we did a couple types of analysis of the Peltz oscillator.

Here's a simple AC linear analysis and the net result shown below using a frequency independent transconductance model for the transistors.

Vout/Ve = -gm1*Re*Z*(gm1 + gm2)/(1-gm1*gm2*Z*Re), where Z is the parallel impedance of the LC network {Z = jwL/(1-w^2*LC)} as seen from the collector of Q1, where Q1 is common base and Q2 is common collector (emitter follower), and Vout is from the collector of Q1, and Ve is the emitter junction. This shows the gain goes to infinity (necessary for oscillator and remember linear AC analysis) when 1-gm1*gm2*Z*Re goes to zero, or gm1*gm2 = 1/(Z*Re) and at a frequency of where 1-w^2*LC goes to zero, or w = Sqrt(1/(LC)), or f = 1/(2*pi*Sqrt(LC)).

Since gm is Ic/Vt, where Vt is kT/q, then oscillations occurs when 1-(Ic1/Vt)(Ic2/Vt)*Z*Re goes to zero, or Ic1*Ic2/Vt^2 = 1/(Re*Z). Since Ic1 ~ Ic2 because Q1 and Q2 both conduct same current thru Re and since they have similar Vbe and Vce assuming high beta.

Ic = Vt(Sqrt(1/(Re*Z))

So the minimum Vee to sustain oscillations can be found by including the drop across Re and Vbe. This makes some sense because the transistor voltage to current conversion, or gm, must overcome the emitter resistor Re and LC tank losses.

For a 2N3904 transistor with Is of 1fa, Re of 510 ohms, Z at resonance of ~900 ohms (L = 449.6uH, C = 214.7nF), we get a collector current of Ic = Vt(Sqrt(1/(Re*Z)) ~ 39.2ua for a Vbe of Vt*Ln(Ic1/Is) or ~ 648mv, assuming junction temperature is ~35C.

So Vee is 2*Ic*Re + Vbe or ~ 688mv which is minimum Vee to sustain low frequency oscillations with Re, L and C. This agrees with a measurement of 692mv at a frequency of ~15.94KHz (calculates as 16.2KHz).

Best,