[David Ellis]

Hugh,

You mentioned on my forum that phase shift of a loudspeaker will impact the loudspeaker.  I am curious about the subjective nature of this, and how the impact happens.  You really don't need to expound on the intimate guts of your amplifiers in detail.  Some general comments about the internal "why" regarding phase angle and amplifiers would be dandy.

I have only briefly experimented with it's subjective nature.  I have NOT performed a solid a/b comparison.  I have managed to try a few tube amps with very good success.  This first screen shot is a comparison between an old solid state Kenwood amplifier and a "new" Jolida 302B.  The Jolida is a push pull El34 unit with global feedback.  I... don't know how much feedback is present.  I couldn't get the spl levels identical, and this is the graph:



This was accomplished with a gated response, and there is obviously no problem with the tube amp suffering a degraded response using a gated response measurement.

I tried a Cary 572 amplifier in my system.  This amplifier is unmodified and sounds extremely mediocre   relative to the better stuff (i.e. Steve Harrisons 55wpc Nirvana + AKSA ).  While the Cary has all the instruments present, the grit is significant.  Tentatively, I attribute this to cheap electrolyitcs, but I haven't swapped them yet.  //  Nonetheless, the sounds is balanced and the amplifier presumably will measure just fine.  The instruments are completely present and in proper proportion.  

All of the subjective testing was accomplished on my 1801 that have this impedance curve:


My questions are:

1.  Will a gated(i.e. pulse) response during measurement "present" the phase angle to the amplifier? I am guessing "no", and believe my initial experiment was flawed.  Perhaps an MLS would work?

2.  Given the 1801's impedance curve, do you think flattening the upper or lower phase bumps will have the most significant impact for the speaker/amplifier interaction?  I do recall reading in a Valve amp book that a dampening factor of @15 will be fine, BUT there was NO discussion of phase angle/amplifier interaction.  What do you think?

3.  Do you believe the considerable phase angle of the 1801 would precipitate better results with your 55 or 100 wpc  amps?  Also, if the impedance were flat, which amplifier should perform better?

I'll be going on vacation this afternoon and won't return for @ 1 week.

Thanks in advance for your response Hugh.  I look forward to your comments.

Dave

[AKSA]

Hi David,

Weighty questions indeed.....!

1.  Will a gated(i.e. pulse) response during measurement "present" the phase angle to the amplifier? I am guessing "no", and believe my initial experiment was flawed.  Perhaps an MLS would work?

Possibly.  The issue is how long it takes for an impulse to create a phase shift at the output, and then whether or not this phase shift will destabilise the amp via the feedback system.  Destabilisation will be transient only, and normally results in well damped ringing.  It's quite likely not all pulses will precipitate this ringing;  it will depend on pulse length, periodicity and slew rate.

On a zero feedback amp phase shift at the load should not destabilise the amp.  It should be unconditionally stable.  However, the amp may not have the slew rate to handle very fast pulses, and this will show as intermodulation on music, but you really only need about 5V/uS for a power amp to avoid this problem.  Tube amps are good in this respect, but even with fb fairly resistant to oscillation under phase shift at the load.  Because of the hysteresis issues in transformers, conventional tube amps are usually limited to 15dB or so.  OTLs will use up to 50dB.  SS amplifiers can be anywhere from 50dB to 80dB.  The more the feedback, the more difficult it is to ensure stability.

SS amps with global fb are a separate category entirely.  Signal propagation through the amp is finite, and at some very high frequency, usually >500KHz, this delay approaches half the periodicity.  This situation means that negative feedback is transmuted into positive feedback, causing uncontrolled oscillation, almost always destructive of tweeters and output stages.  To solve this, we apply bandwidth limiting, so that the voltage gain of the amp is below unity by this very high pole frequency.  Any capacitive phase shift at the load will compound this internal phase change lowering pole frequency and taking the amp outside the Bode/Nyquist criteria and leading to oscillation.

2.  Given the 1801's impedance curve, do you think flattening the upper or lower phase bumps will have the most significant impact for the speaker/amplifier interaction?  I do recall reading in a Valve amp book that a dampening factor of @15 will be fine, BUT there was NO discussion of phase angle/amplifier interaction.  What do you think?

The problem with a peaky impedance curve is the underlying vectorial addition of a DC resistance and an AC impedance.  Only the first order crossover gives relatively flat impedance.  The arctangent of the ratio of resistance and impedance, IIRC, is the phase shift.  (I may be wrong here, but the point is that the existence of an impedance guarantees that current and voltage are NOT in step.)  This has repercussions for the safe operating area of the output stage, of course, but the phase shift will radically alter the Bode/Nyquist plot, inevitably leading to instability.

Damping factor need not be much more than about 25 in my experience.  A low DF can add a little reverb, decay by another name, and this is one striking feature of a tube amp over SS.  A very high DF seems to make SS amps sound a bit too tight, almost cold.  DF is of limited use, however, as it is by nature variable as it's a ratio of impedance.  Sadly (and inappropriately) it's only ever quoted at 1KHz into a purely resistive load!

A final point is that as impedance rises, current falls, and with it, sensitivity, since the speaker is driven by BiL - magnetic forces created solely by current, not voltage.  To the extent that impedance peaks alter sensitivity of the speaker across the spectrum, this is a bad thing.  The only ways I know to fix this problem are to use a shunt crossover network, or a 1st order crossover.

3.  Do you believe the considerable phase angle of the 1801 would precipitate better results with your 55 or 100 wpc  amps?  Also, if the impedance were flat, which amplifier should perform better?

The AKSAs (and the new Lifeforce) use phase lead networks which compensate capacitive loads very effectively.  This technique is much touted by John Linsley Hood.  The AKSA/Lifeforce amps will tolerate the Quad ESLs - they are extremely stable, and pretty well drive any speaker.  I don't believe reducing the phase angle of the 1801 would make much difference to the sonics - it will sound very good regardless, though the 55W would sound marginally better because the output stage is not as robust.  Tube amps, particularly zero fb amps, are more like current than voltage sources, and thus the peaks and dips in the impedance curve of speakers - and notably horns - do not much faze them.  Of course the geometry of a horn speaker is a band pass by another name, and this does contribute a peakiness all its own, but that's another story......

David, I'm not good at software, so can't match my comments with nice graphs, but since most good speakers use third order crossovers with two or three drivers I have set up the AKSA/Lifeforce to cope well with these types.  This is a tricky load at the best of times, so if the amp is stable AND very linear, it should sound very good regardless of load.

I hope this is helpful, and apologise for glossing over some of your deeper questions.  In closing, I have never yet improved the linearity of an amplifier without noticing an improvement in sonics.  Linearity is still a worthy goal.

Cheers,

Hugh

[andyr]

... All of the subjective testing was accomplished on my 1801 that have this impedance curve:


... Dave

Hi Dave,

I am merely an amateur but all I can say is ... Wow!!, what a seriously peaky impedance curve!!      Say, 6 1/2 to 29 ohms!

I cannot believe this is a good thing for either an amplifier to see or a passive crossover to try to cope with but it would be verry inneresting to know what is causing these massive impedance swings - is it the drivers, per se, or the crossover regime you've implemented?   

Are you able to do (or interested in doing  ) a quick experiment ... buy a cheap active crossover (analogue or digital) and replace the passive crossover, then do that impedance measurement again - and see whether it's any different.  If it's "better"     (less peaky) then it would seem the crossover is producing these nasties; if it's not, then they are intrinsic to the drivers you're using?  Which mebbe you need to change?

Regards,

Andy

[DSK]

Andy,
My knowledge in this area wouldn't even approach 'amateur' status but, from what I've seen, this impedance plot is not unusual. The lower peak is usually the box/port resonance, and the higher one is usually crossover induced ...if I understand correctly (very doubtful) 

The 1801's have been around for some time and almost everything I've read about how they sound has been quite positive. Usually along the lines of very smooth and easy to listen to, though perhaps not the last word in detail. Unfortunately, I've never heard them myself.

[AKSA]

Andy,

I confirm Darren's points here;  this curve is not at all unusual, and reflects a higher order crossover system (very likely third order) with not much driver impedance correction.

A properly impedance corrected crossover will normally show from 6R to not much more than 15R.

Wild divergences in speaker impedance do indeed make it trickier to drive and can awaken dormant instabilities in the amp.

OTOH, I too have heard this is an exceptional speaker, and I would never criticise this highly regarded design without listening extensively, examining the crossover with scrupulous care, and maybe trying a few changes - and doubtless I'd be crossing the same territory that David has covered anyway.

That said, I'm an amp man, not a speaker man.  I really wouldn't have a clue.........

Cheers,

Hugh

[stvnharr]

Darren, Hugh,
Perhaps I can clarify a little about the 1801 since neither of you have heard one, and I've built a pair.
The 1801 uses the Seas Excel W18E driver, one of the most detailed drivers around.  The Hiq OW1 tweeter is one of the nicest 3/4 inch tweeters you will ever find to listen to.  The 1801 is a highly detailed speaker!  It is a little weak in the bass however.
I know Hugh has heard the Orion, which uses the Seas W22E.  The 22 is similar to the 18, just goes a little lower, but you get the same detailed sound.

Clarified

[DSK]

Darren, Hugh,
Perhaps I can clarify a little about the 1801 since neither of you have heard one, and I've built a pair.
The 1801 uses the Seas Excel W18E driver, one of the most detailed drivers around.  The Hiq OW1 tweeter is one of the nicest 3/4 inch tweeters you will ever find to listen to.  The 1801 is a highly detailed speaker!  It is a little weak in the bass however.
I know Hugh has heard the Orion, which uses the Seas W22E.  The 22 is similar to the 18, just goes a little lower, but you get the same detailed sound.

Clarified

I'm very glad you posted this as, while reading it, I suddenly realised  that for some reason I had the Odyssey Lorelei in my mind during my previous post. Doh! Brain Fart!  Possibly because the Lorelei uses the SS 8545 and the 1801 designer discusses the SS 8545 vs the W18. http://www.ellisaudio.com/1801.htm
Sorry 'bout dat!