Power Signal Generation

Status
Not open for further replies.

sabu31

Advanced Member level 1
Joined
Jan 28, 2011
Messages
467
Helped
12
Reputation
24
Reaction score
13
Trophy points
1,298
Activity points
5,358
Dear All,

I want to generate a power signal of a particular waveform. I am attaching the figure of the required waveform. What should be the configuration of L-C circuit along with pulse circuit for this to be realized. Typical voltage level of 400-500V.

 

Hi,

Don't know for LC.

A suggestion could be to use a gated square wave (monostable or astable, you'll know which you need of the two) that contains the five clock-period signal. It could then be that signal is fed into a non-inverting amplifier biased at mid-supply, or an inverting amplifier on a dual supply, which is then fed into a variable gain amplifier, whose gain is set/driven/modulated by an RC rise and fall, RC triggered when the x5 pulse starts. Something along those lines.

There must be easier ways than that, though.
 

What is the period/freq of that response in your posted pic ? I assume its a SINC
waveform ? What is power level you are trying to achieve ? Frequency accuracy ?

Two basic methods, classic lookup table to DAC, the other PWM followed by filter.


PWM :



Then G this up, class D if audio kinds of signals. Or class AB if distortion sensitive.

If you are real highs freq there is always density modulation, see attachment.




Regards, Dana.
 

Attachments

  • Generation_of_Nyquist_sinc_pulses_using.pdf
    120.3 KB · Views: 145
Last edited:

There should be something simpler, in my opinion. The power level will be in 200-300W. At high voltages of 400-500V. The frequency frequency between the rise and fall is around 55kHz
As a basic trial simulation, I tried Square wave with LC filter in LTspice
I am able to get second half of the waveform . But how to make it symmetric is challenge.
Once the design aspect is clear, then can translate into higher power level.
 

Attachments

  • Doc2.pdf
    85.4 KB · Views: 146

Pulse parametrics :

Continuous -

1) You want period individual sine variable ?
2) Modulation period, overall frame width, variable ?

Triggered -

1) Single pulse or N pulses, you control N ?

An example, substitute mentally your sinc pulse for the sines in pic (the 4 sines is the frame)



2) You want period individual sine variable ?
3) Modulation period, overall frame width, variable ?

Timing accuracy for both continuous and tiggered, for the individual sine and whole frame ? +/- 50%, +/- 1%.........?


Regards, Dana.
 
Last edited:

The waveform attached shows sine wave. But my requirement is the waveform as shown in post 1.
Is there a general reference regarding this. Relation between the pulse width of square wave and the L and C components etc. Because once the design equation is known I can modifying based on available component for initial trial.
 

The waveform with gauss shaped envelope can be obtained as pulse response of a higher order band pass (multiple coupled LC resonators). A simple LC circuit won't reproduce the symmetrical shape.
 

    sabu31

    Points: 2
    Helpful Answer Positive Rating
Answers to the questions ?
 


Ampere level calculates to be 1/2 A. Do you plan to send 1/2 A through a series LC or tank arrangement?

Using standard formulas for reactive impedance, calculate L & C values which present 800 ohms at 55 kHz.
Then using the standard formula for resonant frequency, see whether these tentative L & C values are realistic for using in your circuit.

Does anything say you must run 1/2 A through an LC arrangement? What if you were to build a low-power sine wave oscillator around capacitors?
 

Hi,

I hate to tread on FvM's delightful and appropriate solution, it's really nice and simple (but I'll bet it's not 'simple' to calculate component values) and what you were asking for. Great!

After a rather - unfortunately - odious morning of three pathetic iterations of an idea, this seemed the closest to the image in post #1... but I think it isn't even what you want - I gather that you want a step response and not a continuous pulse train... Here is a load of old..., I mean, a conceptual design using a quad OPA376, VG1 and therefore Vo are ~55 kHz, and to modulate VG1, VG2 is (obviously) ~11 kHz.

Generating tidy triangle waves is actually the hardest part and for the sake of demonstrating the idealized concept I ditched astable 555s and integrators and gladly resorted to artificial voltage generators. The triangle could be e.g. dual comparator triangle wave generator (would be another quad package) or with a square wave oscillator followed by an integrator, which would mean a second quad op amp package for the 55 kHz and 11 kHz triangle waves, or just some VCO IC.

To turn Vo on and off at will, gating the sine wave signal to only start when VG2 is at it's lowest point was easy with a comparator set to trip around 600mV and a BJT to invert the inverting comparator output fed into an AND gate (a third IC package...), but finding a way to shut it off after precisely 5 sine waves using artificial signals from a square wave generator was not fun to spend time on.

WRT the 500V, I would assume that feeding Vo into a suitable BJT or MOS transistor used to turn on and off a power MOSFET would again be easy, but the devil's in the details of biasing the transistors correctly.

WRT this conceptual design, I haven't dwelt much on the topic, just thrown a thing together that maybe works outside simu-land, and feel there must be a more efficient way of doing this than using two and even up to three quad OA ICs - besides the beautiful iteration above of FvM's that looks like a tailor-made solution.

I'll stop wasting your time now.


 

I hate to tread on FvM's delightful and appropriate solution, it's really nice and simple (but I'll bet it's not 'simple' to calculate component values) and what you were asking for. Great!
I'm basically illustrating my previous comment that the requested waveform is a bandpass pulse response (or a synthesized waveform, multiplying sine with gauss pulse). I found that a filter with "soft" characteristic, specifically Bessel gives best results, that 30 % relative bandwidth is roughly giving the intended shape and that third order (three complex pole pairs) produces already a useful waveform, although not exactly symmetric.

A passive LC filter with a specific transfer function can be designed following the instructions in filter text books, e.g. Analog and Digital Filters, Design and Realization, Harry Y.-F. Lam, or by using a design tool.
An important point to consider is that any load connected to the filter output will modify the filter characteristic and respectively the waveform. Before starting with a actual design, one must know the load specification. The other problem is with input voltage, to get 450 V output, you need to apply a considerably higher input pulse or step.
--- Updated ---

You increase output voltage by applying a pulse train to the filter, although this slightly widens the response, see example.

 

Attachments

  • Bessel3rd.zip
    872 bytes · Views: 115
Last edited:
apart from a high power op-amp ( 150V, 10A type ) driving a transformer, with a clever sig gen ( or FvM's ckt ) driving the amplifier

I would posit that FvM's ckt is likely the only way to get the required 500V pk ...
--- Updated ---

high power op-amp ....

 

Thanks FvM, your posts answer my queries directly. I need to study further regarding scaling to higher power levels .
Also is it not possible to generate the waveform with a single pulse as compared to three as shown in your waveform?
Just for curiosity whats the software used for plotting/simulating waveform shown in post #13.

Is the load 50 Ohm (R) or it can be any value?
 
Last edited:

It's of course possible to generate the waveform with a single input pulse or step. Purpose of the pulse train is to increase the ratio of output waveform to input pulse magnitude, as discussed. You can vary the cycle count in the LTspice simulation to try yourself.

A LC filter design has to be calculated for a specific combination of source and load resistance. Standard is to have source and load termination,, resulting in a different set of L and C values. Filter impedances can be easily changed by scaling all L and C elements.

I used Nuhertz Filter Solutions to design the filter and visualize the time response,
 
Hi FvM,
I tried in LTpsice by increasing the number of pulses, however after some time it becomes purely sinusoidal. How do I increase the number of transitions (oscillations in output) and have only one peak. What is the relation between pulse width/ period/number and filter output
 

.


OP, do you want the waveform single event triggered or recurring ?

Do you care, if recurring, about inter symbol delay ? If so do you want it variable ?

Freq of modulation and freq of carrier ? You want those variable or are they fixed ?
If so what are they ?

Any comments about accuracy of parameters ?


Regards, Dana.
 

Status
Not open for further replies.

Similar threads

Cookies are required to use this site. You must accept them to continue using the site. Learn more…