Can I obtain the best phase response without measurement?

Testing an amplifier by looking at its square wave response does not tell the whole story. You will need to look at the gain and phase at the unity gain crossover frequency to be sure you have an amplifier that is unconditionally stable. I have seen amplifiers that have a good-looking square wave response that misbehave at unity gain by not crossing through 0 dB or, passing through the unity gain point and swooping back up through 0dB at a higher frequency. This can especially happen with vacuum tube amplifiers that use an output transformer. Mr. Zobel has a fix for that sort of behavior.
 
  • Like
Reactions: presscot
If I were to obtain the best phase response without performing measurements, then consequently I would analyze along with both fouplace and larier. They both perform perfectly well without any mearement's data.
 
Last edited:
  • Like
Reactions: presscot
I have access to a gain-phase analyzer at my work place. It's an essential tool for optimizing loop stability for SMPS. The analyzer we have only goes up to 5 MHz, but that's good enough for SMPS and vacuum tube amplifiers. I bought an HP4194A gain-phase analyzer that goes up to 40 MHz, and will be setting that up in my basement lab with a GPIB-USB adapter.
 
  • Like
Reactions: wchang and presscot
Can I obtain the best phase response without performing measurements? If so, how to achieve it?

Yes, you can. Use a fullrange driver. 😀

...please don't beat me!


You can, but it's not worth the effort. You use a frequency generator, reverse the polarity of one driver and then change things on the xo and move the driver back and forth. The position it's the quietest (because of the interference/maximum cancellation), there is the perfect phase once you change the polarity back. You can change the frequency and 'scan' per ear and with the change of frequencies if there are still quet(er) frequency ranges. Rinse and repeat, it takes ages to do that.

Also, how did the speaker manufacturers in the 70s and 80s (pre-1985) manipulate the phase response at that moment?

In the 70s there were already very good measurements available for frequency response. Aside from angle measurements, you can see on the polar plots (vertical) the lobes. It took a lot of time and that made it expensive for high accuracy.

In the time before that, crossovers often had minimal phase characteristics and/or shallow filters.
 
  • Like
Reactions: presscot and wchang
Frequency response is one measurement one can take, but it's no substitute for a gain-phase measurement to determine the 0 dB crossover frequency and the phase margin at crossover. Having a clean square wave response is part of the picture, but again no substitute for a gain-phase plot, as the square wave response might look wonderful, but conceal wonkiness at the 0 dB crossover point. The square wave response won't reveal conditional stability at or near the 0 dB crossover frequency. I've seen amplifiers that have wonderful square wave response, but a refusal to go below 0dB, or having passed through 0 dB , then swoop up and re-cross 0dB at some higher frequency. You need both gain margin and phase margin to guarantee stability.
 
  • Like
Reactions: presscot
If we assume loudspeaker drivers are minimum phase devices, then you can use a FFT to derive phase from the measured frequency / impulse response.

The problem with this approach is it needs to be done for each axis of measurement. Software applications like VituixCad can do this for you (just check the minimum phase checkbox on the driver tab).

The problem here, is software like VCad (and VCad shouldn't solve this IMHO) do not support different X,Y,Z offsets PER OFF AXIS MEASUREMENT, which you would need to have such simulations being correct / close to reality. For most 2/3 way systems with drivers on a similar plane, this small but subtle (for example Z) change per axis doesn't really matter (to me). For the perfectionist or if dealing with large discrepancies, it could make a material difference and force different design choices (driver placement, driver choice, XO frequency and topology etc...).

As such a 2 channel system which measures the phase response - which effectively means, as you move off axis and the relative Z between acoustic centres changes between drivers, your phase measurement reflects this.

So - as others said. If you want accuracy in your measurements, it's much cheaper and easier to invest in a 2 channel measurement system than futz about with minimum phase approaches which will never be accurate, if accuracy is your goal.
 
I didn’t oppose the measurements but I was just curious how did the speaker builders do it in those days without computer/software aided?
Professional speaker designers measured multi-way loudspeakers individually, often using chart recorders, which would give a frequency response curve with as much detail as any modern measurement.
Here's the response of a 2x15" bass horn we took in the 1970s, note the perforations on the top and bottom of the paper which was dragged along as the pen moved up and down (like a VU meter).
On axis used a green pen, 45 degree off axis speakers vertically aligned blue, 45 degree off axis speakers horizontally aligned red.

Fiberglass Cruncher.png

You could smooth the response by putting some light finger pressure on the chart recorder pen 😉 .

Speaker builders could manipulate the phase response of the individual drivers by using different crossover components resulting in different slopes, and viewing the acoustic response to determine if the acoustic filter responses matched what they were trying to achieve.
Fine tuning adjustments of frequency and phase response by moving the driver's acoustic center location was also employed.

I destroyed a lot of high frequency driver's diaphragms with crossovers made without measurement that sounded "good enough" by ear in the shop, but were way off the design mark.
After that $500 mistake (about $2000 in today's money), measurement seemed to be prudent.

Art
 
Can I obtain the best phase response without performing measurements? If so, how to achieve it?

Here's what I would do - assuming it is for a 2-way speaker :

1 - use a 1st order crossover, preferably serie, all speakers balanced in level (by design or with a pad) and connected in the same polarity.
2 - choose speakers with large bandwidth, and no major incident in its frequency response, at least 1 octave distant from the planned FC.
3 - avoid to choose a transition frequency FC in an area where the speakers shows notable variations in their responses, or are too close to their resonant frequency (I favour FC=4500Hz, personally, as a convenient compromise when possible).
4 - place yourself as best to equal distances between tweeter and mid-woofer to your head (listening position, slanting the enclosure or align the speakers in the cabinet construction).
5 - place yourself in axis of the vertical plane to avoid vertical directivity problems that can occur with 1st order crossovers (or move the enclosures accordingly).

With all these conditions reunited, chances are that you will benefit an excellent phase, group delay and transient response...

T