Latency of solid state vs. valve amps?

[Keith_W]

I have a question for ASR members.

Today, I tried time aligning my speakers using Acourate. To my great surprise, the woofers were AHEAD of the midrange horn and tweeters. Meaning, the mids/highs had to be DELAYED with respect to the lows. I would have expected it to be the other way round.

I suspect that it may have something to do with the amplification I am using. The mids/highs are driven by a Cary CAD-211AE valve amp, whilst the woofers are driven with a SGR EL30S solid state amp.

My question: do valve amps typically have higher latency compared to solid state amps?

 

[amirm]

Is the tube amp transformer coupled to speakers? And or have input capacitors? If so, both of these would introduce phase delays although admittedly, I had not thought of them in this context.

 

[Keith_W]

Amir, the crossover circuits in the speaker have been bypassed. They are not in the signal chain. So yes - each amplifier channel is directly coupled to each individual driver.

I suppose I could perform an experiment where I reverse the amplification (i.e. solid state to mid/high and valve amp to the woofers) to see if it reverses the delay. But this would take hours of my time, and I thought it would be easier to just ask the forum

 

[Purité Audio]

Measuring error.
Keith

 

[Keith_W]

Measuring error.
Keith

I suspect that, too. I will repeat the measurement and see what happens.

 

[RayDunzl]

My question: do valve amps typically have higher latency compared to solid state amps?

I'm devoid of any direct measurement experience here, but it doesn't mean I can't come up with some dumb (or obvious?) questions...

Delayed how much?

Is the delay constant across the frequency band?

What is being used to split the frequencies before the signal hits the amps?

Is the measurement mic equidistant from the speaker drivers?

 

[Keith_W]

I'm devoid of any direct measurement experience here, but it doesn't mean I can't come up with some dumb (or obvious?) questions...

Delayed how much?

Is the delay constant across the frequency band?

What is being used to split the frequencies before the signal hits the amps?

Is the measurement mic equidistant from the speaker drivers?

1. The mid/tweeter is delayed by 88 samples at 65536 taps. I am not sure how to convert this to ms.

2. I am not sure how to check this, or even which frequency was being used to generate the impulses. All I hear is a 10Hz - 24kHz sine wave sweep. Perhaps other users of Acourate might chime in (Dallasjustice?)

3. Digital crossover in the computer, generated by Acourate, into an 8 channel DAC (RME Fireface UC)

4. Yes, the microphone has been aligned with Acourate's microphone alignment function in the horizontal plane. In the vertical plane, the mic is placed in the listening position - i.e. at tweeter height.

 

[amirm]

Amir, the crossover circuits in the speaker have been bypassed.

I was referring to components in your tube amp Keith, not crossover.

 

[RayDunzl]

The mid/tweeter is delayed by 88 samples at 65536 taps. I am not sure how to convert this to ms.

Formula: 1/f * samples = delay in seconds

88 samples at 44.1k = 1/44100*88 = 0.00199 seconds - 2ms - about 2 feet delay in the air for CD rate.
88 samples at 96k = 1/96000*88 = 0.00091 seconds - about 10 inches delay in the air for 96k rate.

2. I am not sure how to check this, or even which frequency was being used to generate the impulses. All I hear is a 10Hz - 24kHz sine wave sweep.

The phase display would show delay by frequency.

Here's mine (SPL and phase in REW), after AcourateDRC, flat above 250Hz, rising delay below:

3. Digital crossover in the computer, generated by Acourate, into an 8 channel DAC (RME Fireface UC)

Ok.

4. Yes, the microphone has been aligned with Acourate's microphone alignment function in the horizontal plane. In the vertical plane, the mic is placed in the listening position - i.e. at tweeter height.

Hmm... The tweeter/mid would be closer to the mic (I assume), so, not the source of the delay you mention.

Maybe you are reading the delay information backwards... The woofer is 88 samples later than the tweeter could make sense, because it would be a little farther from the mic.

 

[RayDunzl]

If you really want to pursue this, I would:

Reverse the function of the SS and Tube amp - tube on woofer and SS on tweeter - take a measurement.

Or, Tube on left side, SS on right, and measure.

 

[Blumlein 88]

I don't think you are seeing propagation delay (latency) differences in your amps. If it were enough to cause 2 milliseconds difference you could not use any feedback and have stability. As both tube and SS amps successfully use feedback the circuits aren't slow enough in propagation to cause an issue I don't think. So the delay, if real and not measurement artefact, lies elsewhere.

 

[DonH56]

Yes, there can be delay/phase differences among amplifiers, and I would expect greater phase shift in a tube amp than a SS most of the time but there are a myriad of variables at play. As Blumlein 88 said, however, 2 ms is a lot of delay and unlikely to be just the amplifier. Note crossovers (digital or analog) add phase shift so it depends upon where you (or Acourate) measure (at what frequency) to determine the delay.

I can no longer hear the last (upper) octave of a sweep going to 24 kHz, alas....

 

[Cosmik]

A phase shift is not a delay or a way of seeing into the future. A high pass filter will give you a phase lead. This does not mean that anything has been delayed or is travelling back in time from the future to the present. It is merely that the gradient of the input waveform is being partially differentiated.

 

[Keith_W]

Formula: 1/f * samples = delay in seconds

88 samples at 44.1k = 1/44100*88 = 0.00199 seconds - 2ms - about 2 feet delay in the air for CD rate.
88 samples at 96k = 1/96000*88 = 0.00091 seconds - about 10 inches delay in the air for 96k rate.

Thank you. The sample rate I used was 48kHz. That translates to 1.8ms.

I suspect that I have it backwards, but the measurement in Acourate is quite definite. The woofer peak arrives 88 samples before the tweeter peak. The horn peak arrives a tiny bit later - 2 samples after the tweeter.

Given that it is "only" 1.8ms (and well below the 20ms Haas threshold), it is no wonder that correcting this made hardly any audible difference when I tested it last night.

This is my first effort at time aligning the speaker, which was why I was surprised to discover this. It is also highly probable that I made a mistake somewhere in the procedure, even though I was sober and trying to follow the instructions as closely as I could.

As for everyone else - thank you for your contributions. I will repeat all the measurements tonight and take another look.

 

[amirm]

Definitely keep us posted Keith.

 

[RayDunzl]

Thinking more about it:

The woofer peak arrives 88 samples before the tweeter peak.

No, the woofer peak arrives 88 samples before it it is supposed to arrive, which is much later (relatively) than the tweeter peak.

Example: 88 samples early at 20Hz, 20Hz = 50 milliseconds for a cycle...

There are 48,000 / 20 = 2,400 samples in a 20Hz cycle.
There are 48,000 /20,000 = 2.4 samples in a 20kHz cycle.

The woofer peak at 20Hz should occur at the 600th sample time, relative to the starting time of its cycle. If it is 88 samples early it occurs at the 512th sample.

2400 samples of 20Hz is 360 degrees, so 512 samples is 76.8 degrees and the peak (90 degrees) is 13.2 degrees early (in sample times). 13.2 degrees phase shift. that has a "reasonable" ring to it.

The woofer peak, though, at 20Hz, occurs 512 samples after the tweeter peak, using the same relative starting time, which, at 20kHz, would occur approximately at the first sample time. The woofer peak doesn't arrive before the tweeter peak.

I would think Acourate goes through this sort of calculation, during the sweep, knowing when a cycle at some frequency starts, to know when a cycle peak should arrive, comparing to the measured arrival time, adjusting for the overall static delay, to calculate relative delays and timing.

Does that make any sense, hard to type an explanation. I'm much better at waving my hands and drawing pictures.

----

Maybe I have this all wrong.

How do you define "time alignment" between the turtle (woofer) and the hare (tweeter)?

 

[Sal1950]

 

[DonH56]

Where phase really matters is at the crossover point. You want the woofer and tweeter in phase, working together, where they are both playing the same frequencies. Well away from there it doesn't really matter (well, it will hurt the impulse response, but in general you can tolerate a lot an octave or more away from the crossover frequency). At the crossover frequency, if the woofer and tweeter were completely out of phase, you'd get no sound (for perfectly matched everything). In the real world you'll get a dip with the depth and width dependent upon the crossover and driver characteristics. Audibility of that is up to the listener, natch, but it can be pretty obvious.