(Informal) Chat with Dr. Earl Geddes of GedLee Audio

[Gorgonzola]

The tube amp thing actually came up later in the video. I sounded like Dr. Geddes view is that while tube amps do have higher 2nd harmonic distortion, it's usually not high enough to overcome the very high masking our ears have for such distortion. So, people probably don't like tube amps for that higher distortion, but rather for some other reason, likely (imo) psychoacoustics(ex: placebo). Actually one of my main takeaways from the talk is just how incredible our ears are at masking low order harmonic distortion. Dr. Geddes mentioned one case where it was up to 20%, but was still too small to overcome the masking.

Well it's not fashionable around ASR to admit that you can hear differences, but I have to admit that very often I can. I've never owned a tube power amp but I have owned three different tube preamps.

Subjectivist audiophiles will insist that that tube -- and some s/s components -- will produce a smoother sound and what they best describe as a "layered soundstage" or sometimes as a "holographic effect". This is very consistent observations for what that may be worth. They apparently don't perceive these effects as distortion per se but rather as pleasant, even more like "live" performance. Personally I think their perceptions have some basis in reality though they are artifacts produced whether by of low-order harmonics or some other effect not yet quantified.

Sure, might be my imagination -- can't prove it isn't -- but I have heard it myself from my tube preamps and one, (quite expensive), s/s power amp. In my non-scientific observation there is a distinct correlations between higher 2nd and/or 3rd harmonics a measured, e.g. here at ASR, and these smooth/warm and holographic effects.

Opps. I have to admit that I can sometime hear subtle differences when rolling tubes or, for that matter, op amps.

 

[yourmando]

Excellent interview, @hardisj. You are putting out the best reviews and YouTube videos.

Very interesting. I always thought the very large and efficient speakers sounded cleaner at high due to higher sensitivity and lower THD.

It looks like that’s not necessarily the case, as the perception or distortion or harshness is actually not linear as SPL increase.

You guys talked about group delay sounding worse as volume increases.

But I don’t think you guys ever touched on why the large, highly efficient speakers seem to sound cleaner at higher SPL, If it’s not really due to lower distortion like THD or IMD.

So if it’s not the lower distortion, why do the large, efficient speakers seem to sound cleaner and more dynamic at higher SPL? If not lower distortion, what else could it be? Thanks!

 

[KSTR]

So if it’s not the lower distortion, why do the large, efficient speakers seem to sound cleaner and more dynamic at higher SPL? If not lower distortion, what else could it be?

It actually is the lower distortion which makes them sound better... distortion comes in many many variants.

 

[richard12511]

Yep. I used to test compression as part of my max SPL testing and provided the data. But I don’t think many understood the results and it was never really discussed. It just didn’t seem to catch on like I expected it to. So I decided it wasn’t worth continuing given the time and complexity of that particular test method.

However, as he and I were talking I realized I resume providing that data but instead change up the presentation a bit to make it easier for people to understand. I can use essentially the same method I already use for my current compression tests by tweaking it a little bit to add a round of wideband stimulus between sweeps at a fixed output level (maybe 86dB @1m and possibly again at 96dB or so; I haven’t decided yet). I have already created the template for it and will try it out this weekend. I will likely call the two sets of data “instantaneous” and “long term” compression testing. The former being what I already provide. The latter being the new set of data.

I think my problem with the SPL data you used to provide is that I had no way to contextualize and thus compare what I was seeing. It was hard for me to tell if something was good or bad. Your new compression graphs are very easy to read and easy to see when something is bad(not a straight line). I think if you can find some way to get in graphical form with some sort of benchmark reference it could be very helpful.

Honestly, I know that's a lot of hard work, and previously I would say it's not worth it given what you already give us. However, hearing Earl Geddes talk about it made me really curious again, especially for small bookshelf speakers.

 

[yourmando]

It actually is the lower distortion which makes them sound better... distortion comes in many many variants.

Well, that’s what I thought until I saw this interview

More specifically, I don’t think they discussed why exactly the big, high efficiency speakers sound better at high SPL. What specific thing is responsible, if not the normal THD or IMD measurements?

 

[hardisj]

Your new compression graphs are very easy to read and easy to see when something is bad(not a straight line). I think if you can find some way to get in graphical form with some sort of benchmark reference it could be very helpful.

Done.

I have a bunch of speakers on hand for upcoming reviews and had already tested them all. Some reviews were already published. So, I went back and re-tested all the ones I have on hand still. They are: Neumi BS5P, Klipsch The Fives, Klipsch Heresy IV, JBL 305P, Mackie CR3X, and PreSonus Eris 3.5.

I'll give you guys an exclusive sneak peak.

...


Long Term Compression Tests
The below graphics indicate how much SPL is lost or gained in the long-term as a speaker plays at the same output level for 2 minutes, in intervals. Each graphic represents a different SPL: 86dB and 96dB both at 1 meter.

The purpose of this test is to illustrate how much (if at all) the output changes as a speaker’s components temperature increases (i.e., voice coils, crossover components).

The tests are conducted in this fashion:

1. “Cold” logarithmic sine sweep (no stimulus applied beforehand)
2. Multitone stimulus played at desired SPL/distance for 2 minutes; intended to represent music signal
3. Interim logarithmic sine sweep (no stimulus applied beforehand) (Red in graphic)
4. Multitone stimulus played at desired SPL/distance for 2 minutes; intended to represent music signal
5. Final logarithmic sine sweep (no stimulus applied beforehand) (Blue in graphic)

The red and blue lines represent changes in the output compared to the initial “cold” test.

and here is the data from the Klipsch The Fives (a similarly sized powered 2-way speaker):

 

[ta240]

The tube amp thing actually came up later in the video. I sounded like Dr. Geddes view is that while tube amps do have higher 2nd harmonic distortion, it's usually not high enough to overcome the very high masking our ears have for such distortion. So, people probably don't like tube amps for that higher distortion, but rather for some other reason, likely (imo) psychoacoustics(ex: placebo). Actually one of my main takeaways from the talk is just how incredible our ears are at masking low order harmonic distortion. Dr. Geddes mentioned one case where it was up to 20%, but was still too small to overcome the masking.

So, that would make the SINAD rating not quite as important as the THD numbers, correct?

 

[617]

Interesting. Maybe try with a passive speaker? Also are you adjusting for the mic heating up? Just kidding.

 

[hardisj]

Interesting. Maybe try with a passive speaker? Also are you adjusting for the mic heating up? Just kidding.

I'm in the middle of running the scripts for a couple of the powered speakers which I tested in passive mode (the cheapies that have the amp built in to one speaker and a speaker cable output to the other passive counterpart). I had to write new scripts to process this new data and produce the graphics.

 

[hardisj]

Here's the payoff for this data. We can see just how different results can be when we compare two extremes:
1) Klipsch Heresy IV
2) PreSonus Eris 3.5 (testing the passive speaker)

A reminder of what this test is:

Spoiler: Click for test parameters

The below graphics indicate how much SPL is lost or gained in the long-term as a speaker plays at the same output level for 2 minutes, in intervals. Each graphic represents a different SPL: 86dB and 96dB both at 1 meter.

The purpose of this test is to illustrate how much (if at all) the output changes as a speaker’s components temperature increases (i.e., voice coils, crossover components).

The tests are conducted in this fashion:

1. “Cold” logarithmic sine sweep (no stimulus applied beforehand)
2. Multitone stimulus played at desired SPL/distance for 2 minutes; intended to represent music signal
3. Interim logarithmic sine sweep (no stimulus applied beforehand) (Red in graphic)
4. Multitone stimulus played at desired SPL/distance for 2 minutes; intended to represent music signal
5. Final logarithmic sine sweep (no stimulus applied beforehand) (Blue in graphic)

The red and blue lines represent changes in the output compared to the initial “cold” test.

First, the Klipsch Heresy IV:

PreSonus Eris 3.5:

 

[hardisj]

The one trend I've already observed is that compression is very prominent at the frequencies where port resonance(s) exist. And the above data of the PreSonus (as well as the Neumi and the Mackie data) show this in folds.

 

[617]

There we go! Do you think a 10 minute test would change anything? I'm curious if it sort of hits a limit.

This data is fascinating by the way. Even in the tests that show less than 1db shift, sometimes the speaker gets louder and then quieter.

Also in the JBL and PreSonus we see deviation at the crossover region, I wonder why that's the case? They don't have active filters heating up. Tweeter VC?

 

[617]

Also I feel like I've been reading endless handwringing articles about speaker 'break in' and loosening suspensions and this is the first good data I've ever seen on tonal changes from compression and surprise surprise, it's way more complex than a woofer having a few glasses of wine and a joint.

 

[McFly]

This is exciting.

 

[puppet]

@617 ..it's the passive filter components going out of spec.

 

[Duke]

The solution Earl uses for the thermal effects he describes starting around 27:00 is a "brute force" approach: His speakers have a LOT of thermal headroom, such that the drivers are being operated at a small fraction of their rated AES power handling. For instance the 15" woofer he uses in the Summa has a real-world efficiency of about 95 dB/1 watt (after the crossover removes the unwanted peaking north of the crossover region), and is rated at 1000 watts AES (2000 watts "continuous").

On paper, this implies 125 dB at 1000 watts, and 115 dB at 100 watts (1/10th the AES thermal power handling). In the real world, thermal modulation and/or thermal compression would almost certainly be significant up around the AES rated power (1000 watts), but would probably be negligible down around a tenth of that (100 watts), so as a rough ballpark estimate I think we can assume negligible thermal effects up to at least 115 dB @ 1 meter. (Thermal modulation is a short-time-constant effect dominated by rapid voice coil heating on peaks, while thermal compression takes longer to set in as it includes the [temporary] loss of magnetic flux due to heating of the magnet itself over time).

The inductor on the Summa's woofer is an air-core unit rated at 500 watts, and the resistors he uses have plenty (but not crazy overkill) power handling for their applications. Some years ago I briefly assembled Summas for Earl so I have some familiarity with them.

In contrast, the well-respected Scan-Speak 7" Revelator midwoofer is rated at 85.5 dB/1 watt and 70 watts AES (140 watts "continuous"). This implies 104 dB at the full AES rated wattage, and 94 dB at 1/10th the rated AES wattage. (And in case anybody's wondering, the 7" Revelator is about the same price as the B&C woofer Earl uses in the Summa.)

So as a ballpark estimate, the Summa has 115 - 94 = about 21 dB more compression-free headroom than a speaker which uses the 7" Revelator. Does this matter? I suppose the answer depends on the application and on who you ask, but in general I think it does. (I haven't looked at tweeters in this post but if we compare high quality prosound compression drivers with high quality domes the trend is similar.)

One comment about "loudness cues" from distortion: I've listened to Earl's system in his room several times, at least once so loud that he and I literally had to shout into one another's ear to communicate (and we were sitting side-by-side), yet the speakers NEVER sounded harsh or hurt my ears. I have only heard that sort of lack of harshness at very high levels from ONE other speaker system (big Pipedreams) which was about twenty times the price, and then only when driven by amplification (400 pound GamuT S300) roughly eighty times the price of Earl's Pioneer receiver.

 

[617]

@617 ..it's the passive filter components going out of spec.

We need thermal imagery!

The solution Earl uses for the thermal effects he describes starting around 27:00 is a "brute force" approach: His speakers have a LOT of thermal headroom, such that the drivers are being operated at a small fraction of their rated AES power handling. For instance the 15" woofer he uses in the Summa has a real-world efficiency of about 95 dB/1 watt (after the crossover removes the unwanted peaking north of the crossover region), and is rated at 1000 watts AES (2000 watts "continuous").

On paper, this implies 125 dB at 1000 watts, and 115 dB at 100 watts (1/10th the AES thermal power handling). In the real world, thermal modulation and/or thermal compression would almost certainly be significant up around 1000 watts, but would probably be negligible down around 100 watts, so as a rough ballpark estimate I think we can assume negligible thermal effects up to at least 115 dB @ 1 meter. (Thermal modulation is a short-time-constant effect dominated by rapid voice coil heating on peaks, while thermal compression takes longer to set in as it includes the [temporary] loss of magnetic flux due to heating of the magnet itself over time).

The inductor on the Summa's woofer is an air-core unit rated at 500 watts, and the resistors he uses have plenty (but not crazy overkill) power handling for their applications. Some years ago I briefly assembled Summas for Earl so I have some familiarity with them.

In contrast, the well-respected Scan-Speak 7" Revelator midwoofer is rated at 85.5 dB/1 watt and 70 watts AES. This implies 104 dB at the full AES rated wattage, and 94 dB at 1/10th the rated AES wattage. The 7" Revelator is about the same price as the B&C woofer Earl uses in the Summa.

So as a ballpark estimate, the Summa has 115 - 94 = about 21 dB more compression-free headroom than a speaker which uses the 7" Revelator. Does this matter? I suppose the answer depends on the application and on who you ask, but in general I think it does. (I haven't looked at tweeters in this post but if we compare high quality prosound compression drivers with high quality domes the trend is similar.)

One comment about "loudness cues" from distortion: I've listened to Earl's system in his room several times, at least once so loud that he and I literally had to shout into one another's ear to communicate (and we were sitting side-by-side), yet the speakers NEVER sounded harsh or hurt my ears. I have only heard that sort of lack of harshness at high levels from ONE other speaker system (big Pipedreams) which was about twenty times the price, and then only when driven by amplification (400 pound GamuT S300) roughly one hundred times the price of Earl's Pioneer receiver.

Earl's comments in his 'Home Theater' book show his preference for high output systems but realistically I think there is some middle ground. I have Abbeys and the output is in excess of what I need personally. I think I could get sufficient headroom from a large conventional speaker, perhaps something with an 8" midrange and a robust dome tweeter in a big waveguide. Having said that, if hifi aspires to realistic dynamics, using sound reinforcement drivers is really the best way to go. Frankly it's the compression driver which seems to make the big difference, the HF articulation and momentum is totally unlike dome tweeters. Hard to describe, it's just very lifelike, but of course you can't seperate the compression driver from the huge WG.

 

[richard12511]

Done.

I have a bunch of speakers on hand for upcoming reviews and had already tested them all. Some reviews were already published. So, I went back and re-tested all the ones I have on hand still. They are: Neumi BS5P, Klipsch The Fives, Klipsch Heresy IV, JBL 305P, Mackie CR3X, and PreSonus Eris 3.5.

I'll give you guys an exclusive sneak peak.

...


Long Term Compression Tests
The below graphics indicate how much SPL is lost or gained in the long-term as a speaker plays at the same output level for 2 minutes, in intervals. Each graphic represents a different SPL: 86dB and 96dB both at 1 meter.

The purpose of this test is to illustrate how much (if at all) the output changes as a speaker’s components temperature increases (i.e., voice coils, crossover components).

The tests are conducted in this fashion:

1. “Cold” logarithmic sine sweep (no stimulus applied beforehand)
2. Multitone stimulus played at desired SPL/distance for 2 minutes; intended to represent music signal
3. Interim logarithmic sine sweep (no stimulus applied beforehand) (Red in graphic)
4. Multitone stimulus played at desired SPL/distance for 2 minutes; intended to represent music signal
5. Final logarithmic sine sweep (no stimulus applied beforehand) (Blue in graphic)

The red and blue lines represent changes in the output compared to the initial “cold” test.

and here is the data from the Klipsch The Fives (a similarly sized powered 2-way speaker):

Wow, the JBL 305 actually does great there.

 

[KSTR]

@617 ..it's the passive filter components going out of spec.

Well, non-air-core inductors will start to saturate (during a high power section, not after it) but all other parts in an XO are stable. The XO parts are sure not the reason for the changes seen.
It's VC temperature only which spoils things. A copper wire almost doubles its resistance when approaching 200°C and this easily throws off any XO alignment that hugely depend on VC DC resistance.

There are secondary effects as well and some of which are permanent. After a long high-power test (30mins++), reaching thermal limits (when the VC really starts to smell) most often with cheaper drivers with undersized magnets there will be a permanent efficiency drop as the magnet has demagnetized a bit, this will also shift acoustical alignment.

The loss can come from sheer VC current but also due the heat if you stress it really long enough to seriously heat up the complete driver assembly. Neodymium can't stand as much heat as ferrite but even ferrite degrades.

Therefore, once the speaker has fully cooled down again (may take several hours) it is a good idea to remeasure the cold state again and see if it's still the same.

---------

@hardisj, please consider using periodic pink noise which can be the stress stimulus and the FR test stimulus at the same time. Or at least, if you're using a log sweep, take a descending one and as short as possible (but I think you've addressed that already). Tweeter VC's thermal time constants are very short. Small AMT's for example fully recover within a second or two.

EDIT: Brain fart deleted ;-)

 

[Music1969]

Really great interview. One of the greatest audio interviews / chats I've ever heard.

I could tell from your facial expressions, you were learning a lot of new things (just like all of us).

You are on your way to being the next Earl Geddes, keep up the great work Young Padawan.