Tube amplifier analysis

Wow, I'm amazed how many posters here don't recognize the difference between constant phase shift and constant time delay.

Bonus points for anyone would can construct any arbitrary phase shift from a

Tim

-- Deep Friar: a very philosophical monk. Website:

Well...put the design up in SED or confess that it ain't so easy..

At a single fixed frequency? Trivial. Name your frequency and phase shift.

Amusing side bar: Create a "filter" that has constant 50us delay for all frequencies.

...Jim Thompson

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

I should be more specific: using ideal amplifiers, i.e. adders and resistors. You aren't allowed reactives -- those are, presumably, already in the phase shifter (if it "is" a phase shifter). General, so it has to be independent of frequency.

Part 2: now do it with a 90 degree phase shifted signal (and its original). Trivial.

I believe they call an approximation of those, "spool of coax". ;-)

Tim

Williams' filter book has a section on allpass 90 degree active phase shifters. It's not hard to do a wideband, unity-gain, 90 degree shifter with RC networks and opamps. Ham radio operators do it all the time for phasing-method SSB. Once you have a 90 degree shift, simple scaling and summing can give any angle you like.

Digitally, use a Hilbert filter and, again, a bit of summing.

John

Without something to actually "phase shift" you obviously can't do it.

With reactances you can do almost anything. Trivial over about an octave... growing massively if decades of range are needed.

Indeed... strict summations, though NOT a linear pot ;-)

Student exercise: sin(theta + phi) = A*sin(theta) + B*cos(theta)

Solve for A & B as a function of phi ;-)

Actually quite easy to do for f < 1/tdelay

...Jim Thompson

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

Pffbt, that's hardly "all frequencies". At least coax (good coax) works up to a few hundred MHz, given slope compensation (although I'm not sure about the comp's corresponding delay).

For literally *all* frequencies, nothing beats free space. But you'd better have an all-pass antenna.

If you mean "all frequencies of interest" (such as frequencies phase shifted only a little), an RC is probably a good enough approximation. 50us means f

Putting on professor cap: Make that "all frequencies of interest" ;-)

How many feet of coax for 20us ?:-)

I used the 20us example because I actually built (many ages ago) a "pop" filter for records that delayed everything (audio) by 20us so I could look for the spike and analog-switch it out ;-)

I can remember using TTL delay devices that had spiral wound coax inside to emulate _very_ long chunks of coax.

In fact the MIT labs had such coax so that delay and interference could be viewed with a crap scope (~1960).

...Jim Thompson

Now we're debating definitions, specifically the definition of "phase" for a non-sinusoidal waveform. It's like arguing whether a mixed voltage is "AC" or "DC"

If you build a black box that applies a 180 degree phase shift to any sine wave, it will invert any waveform.

Rich is defining 180 degree phase shift to be half-period time delay. In our shop, we'd refer to that as "delayed by 180 degrees" where "180 degrees" is understood to be a measure of time, half the period of the waveform; that's unambiguous. No linear box can do that for a general waveform; you'd need an adaptive delay line.

Consider running Rich's pulse waveform through an all-pass network that phase-shifts anything by 90 degrees; the result will look a lot differerent from a 1/4 period delay line.

John

Single-mode fiberoptics. DC-to-GHz of bandwidth and attenuation of a fraction of a dB per km.

John

Phase shift is a frequency-domain concept. When working in the frequency domain, you treat each sinewave component individually.

When one talks about a "180 degree phase shift" in relation to a complex waveform, it's shorthand for saying that the phase shift as a function of frequency is given by

phi(f) = 180deg, for all f

Only to the extent that the circuit you're using to do it is imperfect. If you did it perfectly, it *would* sound the same.

"greg" Phase shift is a frequency-domain concept.

It is absolutey a time domain phenomenon.

With a "phase shifted" wave - one easily notices that amplitude peaks and zero crossings are no longer co-incident in TIME with the input wave.

... Phil

If you separate a complex waveform into its individual frequency components, and shift each one 180 degrees, and recombine, you've just inverted the waveform.

An all-pass network will shift every frequency by any desired angle, without changing amplitudes. I don't know if it would sound very different.

John

"greg"

With the former, when altering the input wave frequency the output wave remains perfectly inverted.

With the latter, the amount of phase shift will vary and so too will the amplitude in most cases.

Only a COMPLETE FOOL tries to do THAT !!!!

The term " polarity inversion " is the one that best used when speaking about wide band signals rather than individual sine wave frequencies. However, the phrase: " 180 degrees out of phase" has gone into the tech lingo in many areas to mean EXACTY the same thing.

In any ** real context ** - there is rarely any confustion in meaning with

..... Phil

in=20

for

works up=20

about=20

better=20

Yep. Now if only we had affordable end electronics that was good for that.

However it does not change the phase of all frequencies by the same amount. At least that is what the math and the example circuits show.

The lithium niobate modulators can be had for a kilobuck or so (less on ebay) and a pin diode good for a couple of GHz is $20 or so. But the modulator is nonlinear (light-out goes as sin^2 of voltage, as I recall) and a tad tricky to drive. Direct-modulated lasers are usually nasty.

FM or equivalent into a laser diode could get you to 100 MHz maybe, cheap.

People sell these things as RF links, like for received signals from an antenna on a tower, down to a receiver in a shack on the ground.

John

It's easy to make a network that shifts all frequencies, over many decades of frequency, by 90 degrees, flat on amplitude. Just opamps and RCs, cascaded allpass networks with staggered center frequencies. Once you have 0 and 90, resistive mixing will make any other desired angle.

John

"John Larkin = a LIAR "

I doubt it is even thereticvally possible.

Cascasded APFs produce large phase shifts at the output - ie several cycles.

..... Phil