Friday, March 11, 2011

Optimal Source Impedance....

For years now I have tried to explain (preach? scream?) that tube amplifiers do not have "slow" or "flabby" bass. This common misconception originated when solid state amplifiers came onto the market and the speakers boxes began to be tuned to the lower output impedance of the new amplifiers. If a box is optimally tuned for a solid state amplifier's one ohm source impedance, then it will be severely under-damped (woolly, slow, flabby, etc) with a zero feedback tube amp of a few ohms output impedance.

But what happens in the upper regions? Do compression drivers need to be driven from an optimum source impedance or is lower always better? After all, lower output impedance means a higher damping factor. We have all been told that damping factor must be as high as possible. Well, I am saying that just is not true... please prove me wrong! I was measuring anyway, so I took these plots:

Impedance plot of a JBL 2435 on LC600 horn:

I know that the mighty Bell Labs actually specified a driving impedance for their Fletcher system. It basically was the root mean square of the peak and valley of the impedance plot. They published that this driving impedance was required to meet their frequency response specification. We do not quite meet that formula here, but look at the difference in the region from 1kc to 2kc when the driver is driven directly, through a four ohm resistor, and through a twenty ohm resistor.... and this region from 1kc to 2kc is precisely where the speaker impedance peaks, dips, peaks... driving impedance will change frequency response.. that we already knew...

Frequency response plot of JBL 2435 on LC600 horn - ignore everything above 10kHz, new computer and mic calibration has not been corrected.. sorry...

This is a tight zoom on an impulse response of the JBL2435 driven directly from a solid state amp, through a four ohm resistor, and through a 20 ohm resistor... I did not move the speaker for any of these tests... just clipped in a resistor on the amplifier end...

oh... I know... I should run a square wave through there... I wonder if my measurement gear is up to it... stay tuned!



  1. Jeffrey:

    People generally believe that the Q of a speaker system is important only in terms of bass resonance, where you get to modify the box parameters in order to optimize the Q. However, it is just as important in the design of the rest of the speaker. When driven through the 20 ohm resistor you get typical underdamped response - the leading edge is reduced but the follow up harmonics are rich, especially when compared to the size of the initial transient. The straight in response is probably over-damped, very large leading edge, but the follow up is a little shorter time wise than the others (you need to compare where the transient starts to where it settles down. All of them settled down about the same time in your graph, but the direct started later than the others).

    This is why the Slage autoformer/transformer for attenuating mids/highs changes the characteristics of the speaker so much, it is because of the source impedance has drastically changed.

    We all know that putting a big ass resistor in line with a speaker kills the life. Well, now you know why, it is terribly underdamped at that point. And that is why proper crossover design is such a bear. You have to worry about dB, phase, and damping across the whole range to optimize a speaker. And that is not easy.

  2. If you eq the system so the response is the same, the impulse responses will all be the same as well.

    If you are using a loop back on one channel in your measurement system, you should be able to get a good square wave response regardless of your sound card, although once again if the response of the driver is not flat, that will change the shape of the square wave, so really all you're seeing is an effect of changing the frequency response.

    This really just emphasizes what you mention about Bell labs spec'ing output impedance of the amp - this needs to be a spec for the speaker for it to work as designed.