Zaph Audio SB12.3 or Troels' DTQWT?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Jump to Latest
I

InfiniteGain

Member
Joined 2005
Syncroniq said:
Ill stop here. I have a hard time clarifying myself in english.

At least I said is was wrong on some of the things. I still think many could have made the same conclusions as i. And still think the other things are relevant. Zaph should pay more attention to the timedomain when designing boxes, and do better graphs. Troels should write about his relasions with the bussiness (like is do on my page). I linked to a design like the one Troels made, that didnt have the rippels, so it should be possible.

But yes. I was wrong about the course of the ripple. It was the box, not the speakerunit. I still dont like it. But yes, i was wrong.
Click to expand...

Syncroniq,

I am not trying to get you to be quiet. However, as ScottG highlighted, it is dangerous to say that something is wrong with a speaker or its designer without associated proof. Something may be a tell-tale sign/indication of a few types of problems, but we cannot assume Troels, or any designer puts every measurement made on their website (if they have one) to enable us to be definitive based on our personal experience.

I just think there's a limit to how successful we can be at keeping sensitivity above 95db/W.... you might disagree and be able to build a speaker that measures better than the DTQWT and maintains that high efficiency. Your own competence is not the issue. The issue is: we should not denigrate a designer without proof.

For what it is worth, I think we should have some measurements on group-dealy for the DTQWT to address the concerns you raised over time-domain issues, but I would stop short of saying something in one graph is going to affect another area without knowing is there is an unmentioned part of the design that accounts for it in some way.

You did say you know Troels and other speaker builders in Denmark, so as there are probably a few DTQWT built in Denmark by some of them, maybe you could find somebody who could let you take some measurements and share them with us. This would bring more balance to the thread. You may well be right on the time-domain issues, but without proof I would abstain from saying 'there are' or 'there are not' issues in this area.

I know your CRAN works on impedance. The ripples on the DTQWT graph you posted was for impedance too. I was enquiring if you thought it would help without affecting overall efficiency.

So please don't be offended!
 
I

InfiniteGain

Member
Joined 2005
Nelson Bass,

That's nice work. I did not know about the Apple remote. It looks great as does the internal layout of the amp - close to the original in some areas. You seem to have more space in the units overall. I preferred the all metal chassis in the show pictures.

How long did it all take and do you have some kind of cost for the build? (PM me, or don't worry if you don't want to say).

I would certainly like to do a build like this. First I need to make my mind up and build a high efficiency speaker design that can be driven by this and the DPA300B clones, or any other SET with about 8 watts or more.
 
I

InfiniteGain

Member
Joined 2005
crazyhub said:
:bigeyes: Sorry but I don't agree and Troels either not; efficiency isn't lose in high order slopes vs lower ones; May I ask why you said that?
Click to expand...

Hi Crazyhub,

Glad to see some interjection :)

I am compiling some links to papers/research and information website to answer this properly so I can quote from those pages in a fuller response and so you know it isn't just me.

This area seems to be somewhat overlooked on the whole. I have seen another weblink in recent weeks where somebody asked of the affect of crossover-order versus overall efficiency..... nobody answered.

However, I will say that the quote was stated in the context of what Syncroniq proposed:
1) That you should flatten out the impedance response
2) Flattening out the impedance response requires an additional filter. (He says CRAN, I say Zobel).

Putting that to one side....

On a preliminary front, when we filter signal energy from reaching a driver, where does it go? Often, it is wasted as heat. Thus, the more effective the crossover is at attenuating signal energy and blocking it from reaching a particular driver, the more energy we have wasted. This means that an amp works harder to make 'real' watts appear at the driver terminals. (What you measure at the amp output before the crossover and what you measure after the crossover, at the speaker driver, can be different figures). It does not mean the speaker or its drivers are lowered in efficiency. However, I think for the most part, that reducing the efficiency of drivers, or making the amp work harder are tantamount to the same thing..... wasting energy! And wasting energy is reducing overall efficiency of the system (source + amp + speakers).

IOW, you can have a speaker where 1W at the speaker terminals will deliver a high db SPL. But you may require significantly more than 1W before the crossover (at the amp output) to get that 1W at the crossover output!

Does this phenomena have a name? Yes, "insertion loss" and it affects crossovers.

Higher order crossovers typically have more inductors, which are the worst offenders. This doesn't go in to the phase shifts and impedance variations by higher order crossovers, which require a more energy sapping zobel network, or notch filter to flatten out the impedance which will cause more losses.

Syncroniq argued that the impedance aberrations on Troels speaker needed sorting out, while I contend it's not going to be audible. I will also try and post some information to that effect later (probably over the weekend). My contention is that to keep efficiency Troel did the right thing and left the impedance alone!! His frequency response graph suggests little problem in the acoustic domain, so there is no need to over-engineer for a flatter impedance graph, or use higher order crossovers if it's going to mean more insertion losses and increasing the workload of a low-powered SET amp.
 
C

crazyhub

Member
Joined 2005
InfiniteGain,

Your statements make partly sense.
Partly because I believe that correctly designed high-order filters do not make lose efficiency or sensitivity to the whole system; the way an amp works is very mainly impedance (magn and phase) and back EMF related; high-order filters do not obligatory mean bad or difficult impedance load. Using more coils doesn't mean increased Joule effect if their DCR is maintained the same than in the 1st order series coil. Nor forgetting caps effects that do increase part of the bandwidth...:)
 
I

InfiniteGain

Member
Joined 2005
crazyhub said:
InfiniteGain,

Your statements make partly sense.
Partly because I believe that correctly designed high-order filters do not make lose efficiency or sensitivity to the whole system; the way an amp works is very mainly impedance (magn and phase) and back EMF related; high-order filters do not obligatory mean bad or difficult impedance load. Using more coils doesn't mean increased Joule effect if their DCR is maintained the same than in the 1st order series coil. Nor forgetting caps effects that do increase part of the bandwidth...:)
Click to expand...

First, let me apologize for the simplistic nature of the last post to you, I had not read your profile!

I agree with all that, just that higher order crossover give us other problems to deal with as Rod Elliot, who IMO relays the most informed opinion I have seen on such issues seems to imply (e.g. further phase shifts, increased distortions). Achieving no insertion loss partly depends on selecting the right drivers to be able to do that. Then measuring and doing the maths to re-work the crossover. Many DIYers would just get two drivers that have overlap in FR and approximately equal sensitivity and try to cross them without calculation as to whether they could use higher order and still achieve no insertion loss, or examine what problems they may create in impedance, phase shifts and inductor distortions. However, while no insertion loss is ideal once achieved, if we still have an impedance peak to remove with zobel, or notch filter, then we may introduce additional losses. On most speakers targeted at being driven by SS power amps, this is not an issue, but it is if we want to keep high db SPL for 1W from low-powered valve amps like SETs it is.

With respect to Syncroniq's tube 'friendly' design, he has a nominal impedance on 3.5 ohms and around 87db sensitivity. Now when I speak of sensitivity, or even efficiency, I use it like the majority of audio folk interchangeably despite this being technically inaccurate if we are to be pedantic. So this following link (Symphony Sound - Tube Friendly Speakers) will clear that up, but also shows a graph very similar to Syncroniq's Elysium Passage, just under where it says
"Here's another guy that doesn't play well with tubes:"
Click to expand...

To quote it's comments about the graph, it says:
"In this case the problem is really quite simple - the measured impedance for this speaker is unrealistically low throughout the vast majority of the audible frequency spectrum. From 80 Hz. to 50 kHz. this speaker measures under 4 ohms, and at several points it drops to 3 ohms. Even though the phase angles are relatively harmless, the impedance is so low that only a high inductive load would make a difference. This speaker requires an amplifier stable into very low impedance, and which is capable of outputting large amounts of current on demand."
Click to expand...

If we look at Syncroniq's Elysium Passage 2 speaker impedance graph, its almost identical where the black line is the impedance after an impedance levelling filter:
http://www.speakerbuilder.dk/images/pas20_imp.gif

Syncroniq's design (in my opinion) suffers from the same very low impedance and high current demand over the frequency range that the posted link says is bad for tube amps!

So what do we know about the DTQWT? Well, if the above information is correct, the ideal is high driver sensitivity, good phase angle alignment and not having huge impedance peaks over a short range of frequency. We can get this with a first-order crossovers if we can find the right drivers and build the right enclosure!

Amazingly, in the DTQWT, we have low order crossovers but insignificant impedance responses on the final impedance graph. Above 30Hz there is no peaking above 20ohm and above 1kHz, about 16-17ohm, but the variance is slowly changing here such this it will not be problematic for an amp, plus nominal impedance is as Troels says, around 7-8 ohms, minimum is 6, and we still have a high 94db sensitivity. The design is neither too current hungry, nor too resistive at any point. If Troels decided to put a zobel/notch filter to remove the midrange impedance peak, what would we realistically achieve (keep this in mind when reading the further links at the bottom of my post)?

Any further filter is going to affect the first-order response characteristics of near-perfect transient response and likely introduce an insertion loss, which in a very high sensitivity design we want to avoid (See Passive Crossover Network Design 4.1) I don't think we should overlook the virtues of a first order crossover for the sake of a flat impedance graph.

Rod Elliot's blurb on first-order crossovers:
Series vs. Parallel Crossover Networks

If we also look at:
Passive Crossover Network Design 4.3
mentions (and I paraphrase as his copyright doesn't allow me to repost his content) the zobel flattening the impedance with energy loss, but concludes that we should only use the zobel if impedance variation would cause a frequency response we cannot live with. Troels frequency response is excellent and the overall impedance not unwieldy for tube amps, especially given the maintained sensitivity, so there is no reason to add a filter to flatten the impedance and possible affect the good work at this point.

I agree with Rod's science on the matter as it's backed up in many places. So in Troels design, nobody should be worrying about the impedance response because we already have an acceptable frequency response and the impedance variations are too small to affect any amp (unless that amp is particularly poorly designed).

Additional, Rod Elliot's website on educating us about crossovers also lists inductors as poor for a myriad of reasons. They are a necessary evil, but notice the drawback in the sentences that end with a smiley face in section 4.3:
Passive Crossover Network Design 4.3
...adding inductors increase losses and increases distortion. That was the gist of what I was saying. As we increase crossover order, we introduce more inductors, more insertion losses and move away from focusing on sound quality because we get frequency flat at the expense of further phase shifts and impedance aberrations that make it harder to engineer a good solution. More inductors (higher order crossovers) are only necessary, if our frequency response is too poor. However, in the DTQWT it is not poor when using a low order crossover, and the low slope is nothing to worry about except with higher amplifier power over 50W (Rod seems to suggest this in Passive Crossover Network Design 4.1)

What we should realise is the DTQWT design goal is for low-powered amps including SETs. These will typically not exceed this power. Even most solid state listening will be in the 10-15Wrms range, although some transients will push power beyond 50W so momentarily as to be negligible. This may put some people who like it LOUD off, but as the DTQWT is so sensitive, it will be LOUD on very few watts!

I had intended a fuller treatment of these topics, but I am going to have a busy weekend. A list of web pages I compiled that deal with the issues raised are posted below for anybody who likes reading!

My personal conclusions for a high sensitivity design (and everybody makes different design trade-offs) are that if using a passive crossover:
1) use the first-order where possible. This is highly dependant on driver selection and cabinet design.
2) use a higher order where frequency response is too poor.
3) attempt to keep nominal impedance above 6 ohms with no impedance peaks that are excessively resistive.
4) if (3) cannot be achieve, use an impedance levelling filter, check for phase problems that may be audible

Why?
The first order crossover has some very ideal characteristics as Rod Elliot writes about. If higher-order means more energy losses, more re-work, more distortions, more difficulty in the time-domain, then do we really want to go there immediately because we selected two drivers that a high slope would make compatible? I don't... I'd select better drivers. Of course, 'perfect' drivers do not exist and it's very hard to get drivers to match well like in Troels design. In a more conventional cabinet, with less concern over sensitivity and high-powered amps for use, I would probably end up using an LR4 or going active. My personal design choices for high sensitivity speakers are of little importance when we have high-powered amps available, or ability to develop an active crossover. Where we have high-power on tap, we should try to go with higher-order crossovers and sort out impedance and phase problems if we can still keep distortions minimal, because any insertion losses (or total series resistance of the crossover) are going to be less significant from the amplifiers view point - so we can engineer for maximally flat frequency and impedance in such a situation, and we can pay more attention to rectifying phase, time and lobe problems. This usually gives us a lower sensitivity speaker with good accuracy, but that will not be happy with the lowest SETs.

Reading I had planned on referencing:
http://www.frazierspeakers.com/download/cross.pdf
Tech Topics: An Overview of Crossovers- Part 2
http://faculty.cua.edu/regalia/regalia-perso_files/sp-nov-87.pdf
Phase Distortion article
http://lib.tkk.fi/Dipl/2008/urn011933.pdf **
Power Handling Vs. Efficiency
Passive Crossover Network Design
Series vs. Parallel Crossover Networks
Lenard Audio - Education - X-overs. Time Alignment.
Lenard Audio - Education - X-overs. Time Alignment (part 2).
Understanding Crossovers
audio gear reviews - theory - crossover design
Audio crossover at AllExperts
Filter & Crossover Types for Loudspeakers — Reviews and News from Audioholics
WineBase - Audio and Home Theatre

** = the best research paper I have seen in a while.
 
Last edited:
C

crazyhub

Member
Joined 2005
InfiniteGain,

Impressive statements, quotes and links...

My main thoughts in this matter are:
1) Good cored coils do exist since 20 years or more here in Europe; I don't know since when they do exist in USA...but they exist...We only have to choose them carefully for the peculiar use needed.
2) I use them in series on woofers or in Z correcting networks and am unable to hear distorsions vs air cored ones; I do however use air cored ones on mid or mid-bass drivers.
3) I don't know your experience but from my own a nice sounding filter depends at 95% on preliminary study of the caracteristics of the drivers we want to mate and on the correct implementation and values of components (Z, phases, fr balance and on). This, largely before the quality or supposed quality of components or even theorical superiority of filter types.

But I admit not being overly interested by the theorical side of speaker-building...

Cheers
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Top Bottom