Zaph / Madisound ZRT 2.5 Problem

[5th element]

Ok so your measurements basically show a well designed loudspeaker. At least within the limitations of those measurements.

The left and right speakers are very well matched, always a good sign. The in room frequency response has a downward trend towards high frequencies, this is always helpful in producing a pleasant sound.

You've also got a dip at ~3kHz, rather than any kind of peak, another good thing.

The distortion measurements aren't all that helpful though. First of all your signal to noise ratio is poor, so we can't actually see how well things are performing.

To really test the tweeter you need to get the mic to about 10-20cm and run it at 2.83VRMS or so.

Do NOT collapse HD3-HD5 into one total HD box as that's completely meaningless. What we need to see are the individual levels of HD4 and HD5. These should be vanishingly low, if they aren't then you've got a faulty tweeter, or a fault somewhere else in the chain.

The impedance is telling. These are not the easiest of loudspeakers to drive between 100 and 1kHz. Coincidentally this is where most of the energy within music is located. An amplifier might be 4 ohm stable but that doesn't mean it is going to be happy driving that load. Still at the levels you are actually listening this shouldn't be a problem.

Around the 500-1khz mark (woofer territory in theory), on the distortion plots, we can see that the 2nd order trace and the 3-5th order traces converge. This should NOT be happening. 2nd order is at around -50-40dB down, nothing wrong with this, but the 3-5th should be below this, way below this, especially for 4th and 5th.

If the signal to noise ratio is what's limiting the measurement resolution (very typical) then this is not a problem. If however the 3-5th order harmonics are unreasonably high then something could be amiss. This may not be driver related, it could be anything, loose connectors, the amplifier, the room rattling. But you do need to make measurements of both the tweeter and the woofers. 2.83VRMS drive level and with the microphone 10-20cm from the drivers. We don't need frequency response accuracy here, or driver summation accuracy either, just a decent impression of the systems distortion. And separate out all of the harmonics. You could ignore second if you like, and for clarity, we've seen it already and it's relatively useless for solving this issue. We need 3-5th ideally.

Can you separate the woofers and the tweeter and take some measurements of the tweeter with the crossover attached. Likewise for the woofers.

As the man said.

I'm going to ask, since the measurements showed nothing- how is your room set up?
This could be the key,
Wolf

This harks back to what I was saying about my experience with hyperacusis and waveguides. Waveguides are a way of controlling the off axis response of the loudspeaker so that the room has less influence. (They also do other things but I'm digressing) The influence of the room hurts my ears and the typical 2-3kHz notch helps with this.

The room could quite easily come into play here in affecting how the loudspeakers are perceived.

Your post illustrates the divide between the "musical" vs. "analytical" school of speaker tuning. Should I tune my speakers to sound good on 20% of my music collection but sounding bad on 80%? Or should I tune it to sound good on 80% while making my other 20% sounding less than it should be? Is there a perfect solution?

I've never found this to be a problem. Bad systems sound bad with good or bad music. Good systems sound amazing with excellent recordings and simply good with bad recordings. They allow you to hear the bad in the bad. But for all bad recordings are worth, low bitrate mp3s, lost wax cylinder recordings from some century best forgotten, nothing has ever made my ears 'hurt', or made the music 'unbearable' or 'painful' to listen to. You're just aware of the limitations of the format.

I'm sticking by my original analysis here:

Low crossing a tweeter has nothing to do with sibilants. These are produced at around 8kHz. The only thing that low crossing an incapable tweeter will do is cause excess distortion. The aircirc tweeter is certainly not a tweeter where this will be a problem. Especially with 2.83VRMS measured at the loudspeakers terminals, where the tweeter might actually be seeing less than 1VRMS after attenuation.

I don't think frequency response has much to do with a good sounding speaker.

What? You can throw everything else out of the window if you like, but the number one criteria responsible for a decent sounding loudspeaker is its frequency response. Everything else can be amazing, but if that frequency response is a disaster it will not sound good.

----------------------------

How to fix this problem?

The first thing to do is actually do the distortion measurements as suggested above and see that the higher order harmonics are completely off the audible radar. If they aren't then something is faulty/amiss.

Listening fatigue, in my experience, is going to be caused by one of four things.

1) A wonky frequency response. This can mean a lack of bafflestep compensation or the tweeter level not being quite right, or just a terrible crossover design. None of these are really associated with what I've seen so far.

2) The power response of the loudspeaker interacting with the room in a way that hurts your ears.

3) An excess of distortion somewhere in the signal chain. Usually the result of things working too hard or being improperly used.

4) Your ears. When my hyperacusis started it was mild and got a little worse over time. I spent hour after hour after loudspeaker design, after crossover methodology until I actually realised, hey this is my ears. Realistically there was absolutely nothing wrong with any of the approaches I tried it's just my ears didn't like it.

Things then turned into a case of what can I do to actually make loudspeakers I can listen to. I figured this should be possible since my HD650s were still easy to listen to, but it took me ages until I discovered that waveguides were the answer, or at least a massive help towards my problems.

From a theoretical stand point the very low crossover on the aircirc tweeter should help prevent listening fatigue from reducing/removing the wide-narrow-wide dispersion profile that can make my ears bleed. The only downside to this is that the tweeter could potentially be stressed into excess distortion, but really? This is only at antisocial listening levels.

I've never had a problem with low crossing a tweeter and I've done it plenty of times, where I've traded a more even off axis response for a limited max SPL.

 

[Michael Chua]

.... It could be a certain distortion either by the tweeter playing low or coloration in the treble on the woofers... I dunno.

A flat response doesn't tell you anything. You need to measure the drivers separately to see what's going on underneath.

 

[merlinx76]

Do NOT collapse HD3-HD5 into one total HD box as that's completely meaningless. What we need to see are the individual levels of HD4 and HD5. These should be vanishingly low, if they aren't then you've got a faulty tweeter, or a fault somewhere else in the chain.

That is HD2 to HD5 collapsed. Which I believe shows in that region that nearly all the distortion is HD2.

 

[cnau]

I'm not sure if this helps but since the start of this thread I have put aside my ZRT2.0's and replaced them with 3-ways constructed of parts I had on hand.

With the ZRT's I could only play them equal to a loud level of conversion before a bit of distortion started creeping in. The 3-ways play much louder. I always thought it was my amp. Probably I am also overly picky and I usually just play at low volume.

The 3-way has a HiVi W10 woofer; Seas CA15RLY; and Seas 27TDFC crossed over at about 600Hz and at 5000kHz (a bit high but limited by my stock of crossover parts). All slopes except one are 2nd order electrical. Need to order some parts!

If your ZRT2.5's do not pan out I highly recommend attempting a 3 way. It was not as hard as I imagined it to be.

 

[Fernando R]

Sorry I have not gone all the way back to see if the amplifier was discussed. I have had amplifiers that would distort on 4 ohm loads. One set of speakers with parallel 6" woofers drove me nuts drove me nuts until one day I borrowed another amp. The distortion disappeared.

 

[Tweet]

Merlinx76,
Did you ever try putting a series resistor of say 2 ohms or more in series with each speaker cable when connected to your amplifier ? This is just to test if your amplifier is particularly impedance sensitive at high frequencies. There may just be some parasitic oscillations going on when the loading has become too reactive, resulting audio distortion at those frequencies.

C.M

 

[Lojzek]

I'd first clear whether this is amplifier's fault by trying the luck with a high quality SS design, an old school amp only, discretely built, not some HT receiver or chip amp like. I have had similar experience with Technics integrated amp that employed a chip for an amplifier.

edit: Haven't read the whole thread

 

[merlinx76]

I have tried a different power amp. I mentioned that 3 or 4 times at least now. I had a DIY Honey Badger for a little while that I was using. I know this thread is is getting long. We are just going in circles. I don't think we are really getting anywhere and this is something I'm going to have to figure out for myself.

 

[system7]

Well, we're just getting nowhere on this.

Zaph|Audio - ZRT - Revelator Tower

Here for the interested student are the FRD and ZMA files for the 18W/8531G00 which look quite reasonable to me.

I can't swear to this, but I believe they come from no less than John Krutke himself.

From what I can see, Visaton Boxsim, which I am using, finds a good correlation between reality and the Zaph measurements.

The impedance and shallow rolloff are the things that need improving IMO. And this is what I have addressed. Whether it's enough, remains to be seen. A difficult driver forces compromises elsewhere.

 

[andy2]

If you think distortion is the cause of OP problem than you probably live on another planet. I used to have a pair of Seas Mag. 6.5in. which had horrible cone break up but with proper xover , no brightness, no harshness, no sibillance. I can almost tell from looking at his measurements and my experience with tuning my own speakers, they are bright with higher than normal vocal sibillance both of which making listening unbearable. Sometimes you just have to face the music and either rework the entire xover or build another pair (more musical I hope).

 

[5th element]

I have tried a different power amp. I mentioned that 3 or 4 times at least now. I had a DIY Honey Badger for a little while that I was using. I know this thread is is getting long. We are just going in circles. I don't think we are really getting anywhere and this is something I'm going to have to figure out for myself.

We're getting nowhere because we don't know specifically where the problem is coming from. To some the frequency response would be absolutely fine, to others, such as andy2 below, it would melt his ears off. High frequency absorption/reflection, in room, has a lot to do with how different tweeters sound. In my room and with naked tweeters I cannot stand high dispersion types, such as 19mm domes or 19mm ring radiators.

We know this isn't the amplifier, the honey badger is plenty capable.

We can hazard a guess that this isn't the drivers because of scanspeaks excellent quality control. The limited distortion plots that you have shown indicate nothing gross is wrong, but we'd need to see HD4 and HD5 separately to be sure of this.

That is HD2 to HD5 collapsed. Which I believe shows in that region that nearly all the distortion is HD2.

That makes a lot more sense. Still it's completely useless because we need to see the details of the other harmonics to come to any conclusions.

If you think distortion is the cause of OP problem than you probably live on another planet.

I agree 100% with this. It's possible, if there were copious amounts of high orders showing up, but that isn't likely.

I used to have a pair of Seas Mag. 6.5in. which had horrible cone break up but with proper xover , no brightness, no harshness, no sibillance.

Ironically these actually sound pretty good when run full range and with an appropriate amount of bafflestep. Sure they sound unnatural but not as terrible as you might think! (At least to me).

I can almost tell from looking at his measurements and my experience with tuning my own speakers, they are bright with higher than normal vocal sibillance both of which making listening unbearable. Sometimes you just have to face the music and either rework the entire xover or build another pair (more musical I hope).

The elevated area between 3.5-8kHz could be responsible for that, but if you fill in the dip at 2.8kHz the speakers are essentially flat all the way up to 8kHz.

Merlin, have you tried listening to these near-field? Say sit 80cm-1m away with the speakers toed in with the speakers at ear height? Do not listen too loud and do sit in a comfy chair and relax, crouching doesn't count. Listening near-field is a good way to remove the room from the equation and therefore reduce the impact of the speakers off-axis response. This will allow you to simply listen to the frequency balance of the speakers alone and see if this presents itself with issues.

 

[merlinx76]

Merlin, have you tried listening to these near-field? Say sit 80cm-1m away with the speakers toed in with the speakers at ear height? Do not listen too loud and do sit in a comfy chair and relax, crouching doesn't count. Listening near-field is a good way to remove the room from the equation and therefore reduce the impact of the speakers off-axis response. This will allow you to simply listen to the frequency balance of the speakers alone and see if this presents itself with issues.

I tried that and it wasn't all that nice sounding. Not as good as when they are in position, especially now that I have a few acoustic panels.

I will measure individual drivers from closer as you said when I get a chance to try and get better distortion readings and will show them all. It looks like I had much more high order distortion than Zaph's measurements showed unless that was a result of my loud noise floor or something else I did wrong in my measurement. I have a feeling it could be my fault. I have to learn more about taking reasonable measurements and also interpreting them.

 

[akarma 76]

Scan 8531 is not worth the money and definitely is not the best choice for 2way design.

 

[Michael Chua]

What? You can throw everything else out of the window if you like, but the number one criteria responsible for a decent sounding loudspeaker is its frequency response. Everything else can be amazing, but if that frequency response is a disaster it will not sound good.

A flat response doesn't tell you anything. You need to measure the drivers separately to see what's going on underneath.

Hi 5th

I did not see your post because as I was writing, it wasn't there. Only after I reviewed the post later, did I realized that it may appear I was referring to your statement. I hope it did not cause any misunderstanding.

 

[system7]

I don't actually think this is a hopeless cause.

Zaph|Audio - ZRT - Revelator Tower

But it does fall into what the wise Lynn Olson calls a "LGWAG" speaker category. I mean, does anybody apart form me think that a second order 16uF and 0.25mH crossover is pushing a tweeter quite hard, albeit with a fairly hopeless LCR added.

And look at the level of the cone-breakup on the 6" bass which I modelled without great attention to the exact tweeter:

Where's the 5kHz cone-breakup notch? Gone AWOL.

A common but unsuspected cause of sibilance is crossing the tweeter too low, or using a shallow-slope crossover. Many designers - unfortunately, a lot of them in the high-end biz - forget that direct-radiator drivers increase excursion at a rate of 12 dB/octave. Thus, it takes a 12 dB/octave highpass filter to merely keep excursion constant in the frequency range between nominal crossover and the Fs of the tweeter.

For example, if the tweeter has a typical Fs of 700 Hz, and the intended crossover is 2.8 kHz (again, typical), it takes a 12 dB/oct electroacoustical filter to merely keep excursion constant in the very critical 700 Hz ~ 2.8 kHz range. Part of the reason that this range is so critical is that audibility of distortion is at a maximum in the 1~5 kHz region. (Perception of distortion similar to, but not quite the same as, the Fletcher-Munson curve.)

Staying with the same example, if the electro-acoustical filter is 1st-order (6 dB/octave), then excursion actually increases from 2.8 kHz on down, until 700 Hz is reached. Below 700 Hz, the excursion finally starts to decrease, but not very fast, only 6 dB/octave. This is troublesome because the maximum spectral energy of many recordings is around 300~500 Hz, so energy from this range can crossmodulate with the tweeter output.

This is why auditioning with little-girl-with-a-guitar program material and a full choral piece sound different. The LGWAG is spectrally sparse, and there isn't as much chance the tweeter will be struggling with IM distortion. Throw a dense, high-powered spectrum at the loudspeaker, though, and the tweeter will start to scream - and it is very audible on massed chorus as complete breakup.

At any rate, regardless of distortion of a particular tweeter (none of them are free of IM distortion), crossovers matter. Many designers want to take the tweeter as low as possible because the polar pattern is prettier and certainly measures nicer, but the inevitable price to be paid is more IM distortion resulting from increased excursion (the linear region is most tweeters is less than 1mm). Choosing a crossover is a difficult tradeoff between narrowing of the vertical polar pattern, IM distortion from out-of-band excursion, and how close the designer wants to approach the region of midbass driver breakup. The tradeoff is made more difficult when a rigid-cone (Kevlar, metal, ceramic, etc.) midbass driver is chosen, because the onset of breakup commonly falls in the 3~5 kHz region, right where the ear is most sensitive to distortion.

As you can see, the worst possible solution is a 1st-order crossover combined with a midbass driver that has a severe breakup region (Kevlar drivers, I'm looking at you). The 1st-order crossover fails to control out-of-band excursion, so program material in the 700 Hz-2.8 kHz region results in IM distortion in the tweeter's working range, while plenty of midbass breakup in the 3~5 kHz range gets through as well. And midbass breakup sounds the same as a bad tweeter, since the distortion and resonances fall in the same frequency range.

As a side note, most transistor amplifiers (including very expensive high-end products) go from Class A operation to Class AB around 1 watt. Feedback helps, but cannot fully overcome the two-to-one shift in transconductace as the AB region is traversed. In addition, thermal tracking is typically several seconds to a minute late (depending on the thermal mass of the heatsink and location of bias sensor), so the correct AB bias point is actually several seconds behind the program material. There are various sliding bias-tricks available (which avoid complete turnoff and associated switching transition), but they are all several seconds late. The more output transistors, the more AB transitions there are, since it is impossible to have transistors exactly match the switching transition - in production, they are matched for beta (current gain), but not usually for other parameters. Change the die temperature a bit, and the careful hand-matching goes away.

To recap, if you want lots of sibilance, use a midbass driver with severe breakup in the 3~5 kHz region (this is usually obvious from unsmoothed FR curves), pick a tweeter with limited excursion capability (not always spec'ed), select a 1st-order crossover at a low crossover frequency, and use an amplifier with a very large heatsink, many transistors, and somewhat unstable Class AB biasing (thermal overshoot). That should do the trick. Plenty of distortion from many different sources, even though the overall FR curves may look harmless.

This is one of the most enlightening things you will ever read about loudspeakers. What it's saying is that low crossover on shallow slopes makes for distortion. Simple as.

The suggestion of a three-way conversion is a good one. Two-ways have enough difficulties without adding badly-behaved 6" woofers to the mix. I have improved the circuit a bit with third order and impedance correction and series wired basses into the MTM realm, but only your ears will say whether it is enough. It's still a 2.2kHz crossover.

An 8" bass plus 1" tweeter can work spectacularly well with a really smooth polycone bass unit, doubtless at higher crossover point. The Harbeth Compact 7 ES-3 is an example, with metal tweeter as it goes. AFAIK, it's an electrical 3rd/4th order filter of some complexity. Damped ply cabinet. No Mickey Mouse stuff there.

 

[andy2]

Some of the diagnose seems a bit of a contradiction. First of all, an xover shouldn't introduce distortion unless there is something wrong or there are defects in the components. The xover may allow the drivers to operate beyond its range therefore let through more distortion. But the vibes I got from some of the veterans here is that there is nothing wrong with the design and low xover of the tweeter is perfectly fine so I don't see how you can say the problem is distortion at the same time saying the design itself is fundamentally sound.
If you still insist that distortion is the problem then the only source of distortion can only come from the drivers, but these are Scan Speak top of the line so it's unlikely that the drivers themselves are the problem (although personally I never worked with these drivers).
The other criticism is somehow a 6.5 + 1 is a compromise design, but there are a lot of very good 6.5+1 out there so everything just points to the xover design that is the problem.

 

[merlinx76]

I just want to put this here as food for thought. Quoted from Troels Gravessen's website regarding a 2-way with the 8531:

"I'm afraid this is where it's going to end for the time being. "Afraid"? Well, I'm not really sure this 8531 driver was meant to perform in a two-way system up to 2.5 kHz. Rather in a three-way system. A classical set-up would be the 8531 + the new 12M sliced paper driver from SS with a point of crossover at some 800 Hz -or a rather more modern three-way system with the 15M for midrange taking over at 300-400 Hz. The sound from this series-V4 is very transparent and with good tonal balance but I still think the 8531 has a certain "paper-sound". Too much lushness, if you get my meaning. Almost a classical JBL sound despite having a rather flat response contrary to the old JBL speakers (I'm thinking of the L100).

I feel very tempted to coat the 8531 driver, but applying coating material to a 300 US$ driver is a one-way ticket and you may end up ruining the driver. I'll have to consider this carefully. I'm sure some people will say that this is just what they want, and others will notice the "sound" (= colouration) from this 8531 driver giving too much sparkle to a piano note or a vocal performance. Running the SP95s-V4 from a Copland CTA505 valve power amplifier in triode mode and a home-made valve pre-amplifier makes this coloration less pronounced, so be sure not to have some aggressive front-end driving this speaker."


You can see his design here: SP95

Yes I realize, different XO at 2.5khz, but it is some food for thought. Just because it is a high end driver with low harmonic distortion & the final FR is quite good, does not paint the whole picture. Not saying I won't look into things further, but people act like those 2 graph are good so there is no reason for me to not love the speaker.

 

[wolf_teeth]

Are there any hard surfaces between you and the speakers? Coffee table or hardwood floors count.
Do you have any autocorrect EQ engaged like Audessey or such?

Just because the FR is okay doesn't mean nothing else is awry.
Wolf

 

[merlinx76]

One speaker does get floor bounce on the hardwood floor but even that bounce should be at least mostly blocked by a tiny side table right next to me. Nothing else between us. This problem has been present in multiple positions in 2 different living rooms. Again, it was even present when I had different boxes in an MTM configuration which is why I really lean towards it being a characteristic of a driver, or at least a characteristic of a driver used at this XO point anyway.

 

[wolf_teeth]

MTM speakers have a different lobing structure due to the MTM arrangement and the reflections are reduced for the same reason. Have you tried laying a rug on the floor, or moving the end table to next room? I know mine get harsh when the coffee table is empty and in between me and the speakers.

Wolf