Which Capacitors for Audio?

[thor.zmt]

Tell that to any decent SE.

I do tend to nowadays use fully parametric EQ's on the inputs, in digital desks for live use. Because I notch out the acoustic feedback frequencies. That needs parametric EQ's with Q levels I have not seen in analogue parametric EQ's.

Even if you get a high enough Q, 4 bands are very limiting for that.

And I still need TONE CONTROLS on top, to shape the TONE. Which is what we need for broadcast and recording, as opposed to

A fully parametric EQ is overly complicated for that purpose and harder to use than the classic designs. On top of that to fit that on a desk, knobs get so small, I need tweezers to turn them!

So I normally like 6-8 parametric bands for notching out feedback (that are completely unnecessary fogrrecording music) and a classic 4Way semi parametric.

I never touch the frequency or Q setting during use, only for setup and soundcheck. In use I have a simple 4 Way tone that is immediately intuitive and simple to use, for live situations with bands that were not during sound check and need a lot of work to sound good.

If in the studio we really need a full parametric EQ to be "creative" (aehhm, yeah, right), well that's what George made the 8200 for, get one and patch it in.

I guess different strokes for different blokes. I believe things should be as simple and intuitive as possible while being serviceable, but no simpler.

Thor

 

[thor.zmt]

Voice coil based sensing is not good,
I found that out in the 80's and instead went to microphone on the cone

You conclude a general principle, on the basis of a failed experience.

No, I conclude so after analysing WHY the experiment failed. This:


BL varies with excursion. This unavoidable. The shape of the curve is adjustable, but you never get a flat line.

For the back EMF or a secondary coil would with the voice coil this nonlinearty becomes a distorted sense signal for the acoustic output.

We compound distortion and introduce avalanche conditions when the excursion becomes large, requiring extra protection.

There have been a few commercial attempts using voice coil sensing, the ones from Tannoy and Yamaha were failures in my view, being unable to outperform simple traditional open loop systems such as a REL Stadium.

Going from voice-coil sensing to microphone sensing and current drive is at least changing two parameters. Basic scientific approach says change only one parameter at a time.

I was engineering a submission to a design contest (which never took place, the wall came down first and I had escaped already almost nine month before the wall fell), not doing research.

But I first had actually switched Mid/Hi to current drive, it worked out great.

I switched to microphone feedback after reading about Backes & Müllers system in use in West Germany and found this was great. Adding current drive increased loop gain at LF without adding stability issues.

Still, MFB is a way to modify damping by electrical ways.

It is not. Because it extends far beyond damping. MFB is a broad spectrum "catch all" error correction using Black Feedback (should probably be actually Voight feedback, he patented this way before Black) around the transducer.

If I were in samantics, I would say that damping factor is a physico-mathematical notion that is neither strictly electrical or related to whatever impedance.

Is this meant to convey anything? I read it as meaningless non-sequitur.

Thor

 

[thor.zmt]

Current drive works when cone displacement is low. That's why it's good for midrange, or bass speakers where displacement is low by virtue of acoustic loading.

You are missing my point.

This driver, with it's single piece alu voice coil former would not move the cone much even if hit with a lot of bass. It was massively overdamped, due to the voice coil fromer acting as shorted turn.

So yes, doing this does work for bass speakers. And very well.

Thor

 

[abbey road d enfer]

No, I conclude so after analysing WHY the experiment failed. This

You jumped from voice-coil sensing, which is indeed flawed, to pressure sensing. In between, as I suggested, there is displacement/speed/acceleration sensing, which is, within the domain of application, almost instantaneous, so better suited for a feedback system.

But I first had actually switched Mid/Hi to current drive, it worked out great.

I contest the use of current drive for applications where there is large excursions, and response including resonance zone.

Is this meant to convey anything? I read it as meaningless non-sequitur.

OK, I won't make any reference to semantics, just say that we don't put the same value on words.

 

[abbey road d enfer]

You are missing my point.

This driver, with it's single piece alu voice coil former would not move the cone much even if hit with a lot of bass. It was massively overdamped, due to the voice coil fromer acting as shorted turn.

Tell me how a cone that doesn't move can produce loud bass, except if loaded with a particularly efficient horn.

 

[thor.zmt]

Tell me how a cone that doesn't move can produce loud bass, except if loaded with a particularly efficient horn.

You are missing my point again.


The cone does not move (with voltage drive) because the damping applied from the single piece alu voice coil former. The damping is so great, that the motor cannot generate enough force to overcome this. Yet at midrange frequencies the same speaker has very high SPL.

It means sufficient damping to get any desired Qm can be supplied simply by adjusting the thickness of the alu former, perhaps even just as a metalised overlay on a kapton voice coil former for example. Get the resistance right, voila you have the precise Qm you have trageted.

In order to get bass output we obviously need to shift the damping from "massively over damped" to "critically damped" or whatever particular mechanical damping we have decided is required to give the results we want.

As the voice coil former is always completely within the magnet field of the magnet, at least for functional excursions, this damping of the fundamental resonance(s) is much more linear than that provided by the amplifier "dumbing fuctor".

In fact, combining this small modification with an otherwise absolutely standard speaker and an otherwise absolutely standard amplifier converted to current drive can substantially improve the speakers linearity in for example a small bluetooth speaker and reduce intermodulation distortion etc..

Of course, simply creating a speaker with a Qm = target Qt cannot work with a classic voltage output amplifier. But that is not the aim here.

Thor

 

[abbey road d enfer]

You are missing my point again.

View attachment 111382

Nice drawing, but still, how a cone that doesn't move can produce sound.
Optimizing Q at the detriment of spl is not desirable.

 

[thor.zmt]

Nice drawing, but still, how a cone that doesn't move can produce sound.
Optimizing Q at the detriment of spl is not desirable.

I officially give up. You really, really, really do not get it.

The cone moves based on the force created by current in the voice coil and the counter force of the various elements forming suspension and in our case the voice coil former forming a shorted turn.

There is no optimisation to the detriment of SPL.

Given the same current in the voice coil and the same Qt for the driver (and all else being equal), the seamless metal(ised) voice coil former and current drive will have the same SPL and frequency response as the voltage drive version of the system with a non-conductive voice coil former. It is really, really basic.

The key differences will be that that the current drive version with metalised voice coil former will have no thermal compression, significantly reduced odd order and high order harmonic distortion and no reduction of the damping with large signal low frequencies, which in turn will help keeping the voice coil in the actual magnet gap and thus avoid excessive modulation of the midrange in common "bass/midrange" type systems, compared to the voltage drive version.

Even better all that at a cost difference that is essentially miniscule to zero in mass production and requires minimal changes to established production methods. All gain and no pain.

Anyway....



Thor

 

[living sounds]

A null test can tell you how different they are.

If the difference is small enough it becomes a never mind...

JR

I've done lot's of null tests with converters using DeltaWave.

https://deltaw.org
Often enough, it does reveal interesting stuff.

 

[living sounds]

Up and down sampling in plug-ins is limited to factors of 2, so use the traditional method of inserting zeros in the data flux or averaging adjacent samples, which results in MSB rounding error. Since it is most often done internally in 32-bit floating, I can confidently claim I can't hear it.
Stacking 8 plug-ins all with up/down sampling would result in the 3 MSB's in error, for a clean 22 bit resolution (32-bit has a resolution of 25). Even tracking at -20dBfs would leave 18 bit resolution, or 0.0004% accuracy. I don't think I can hear it either.
I have found that most problems with digital processing are wrongly attributed to inherent faults in the principles. Most of the times, these problems end up being due to improper implementation, either in the digital algorithms, or in the analogue support circuitry.

The problem is in the filters. Since you cannot do up- and downsampling without them, and designers want to preserve the frequency response, need it to be real time and use modest CPU, the steep filters with few taps used wreck the signal integrity.

A simple test: Use a plugin with switchable AA filters like Fabfilters, select a neutral mode (e.g. limiter with threshold all the way up) and put one each on identical DAW tracks, flip the phase on one and turn on AA on one track only. What is left on the sum bus goes through the whole frequency range (not just the ultrasonic filtered part) and is of rather high amplitude. And the effect is cumulative, so stacking compounds the problem. The degredation is audible with decent source files and converters with one instance only.

 

[living sounds]

I agree. One time in a massive queue at Beijing immigration (before the Olympics , a lot of rules suddenly changed and every foreigner had to get his paperwork resorted), I met an ABC (nominally American Born Chinese, usually also applied to US based and educated CBC's) who's job was head of the ADC/DAC chip development division for a famous US based maker, who's chip Ard one of the staples in pro audio.

We started chatting and found we worked in a similar field. In the end we went out for dinner at Yin Yong Restaurant for Duck, Cat Ears and Bear Paw.

And he came over to have a listen to the audio system we were making, which he thought sounded great.

We also talked about if he ever listened to the ADC/DAC Chips his people designed.

He looked at me quite like "is he mad" so I went and pulled out the EVM for the latest greatest chip out and plugged it in and set up level matching in the Amp ( < 0.5dB difference).

We played the same cut synchronised and switched between the two and his face fell. The EVM sounded, politely put, English way, "pretty good" (that's English for terrible).

He was even more shocked when I explained to him what DAC Chip my design used. We went to the famous mushroom hotpot after and talked a lot of shop, on what I thought were reasons for the difference is in sound.

There is a corollary here, quite a few years later this chip manufacturer introduced a fundamentally new range of DAC Chip's. And tarnation and blimey, they sound excellent. I actually designed them into several products. I also checked on the guy, he is still there.

I suspect, he started to listen, in more way than one.

Interesting! So which ADC/DACs would you recommend?

 

[thor.zmt]

Interesting! So which ADC/DACs would you recommend?

I have been using various Burr-Brown / TI "Advanced Segment DAC" parts, under software control and with some use of non-documented features and slightly non-standard analogue stages.

Other than that the CS43131/198 is really good as DAC. Especially if used dual mono and possibly using multiple DAC's per channel.

ADC I have been using TI parts, the PCM4222 is nice. The TLV320ADCX140 range is good for lower end gear.

I have been unimpressed generally with ESS.

Something I'd like to try is to build an ADC/DAC using the industrial 20-Bit / 1 MspS ADC/DAC Chips from Linear, these tend to be 50 Bux plus a piece, but we could get a 768kHz sample rate 20 Bit ADC/DAC without digital filter.

Add an FPGA and trade sample rate and bit depth with simple filters, so averaging two samples give us 384k/21 Bit, while averaging four gives 192k/22Bit.

With likely 300 USD+ in chip's alone this would be not a cheap device.

Something like this I might call:

Atlantique Macrosonics

Model T.R.E.S

And try to make it look appx like this:


Hehehehe...

Thor

 

[abbey road d enfer]

I officially give up. You really, really, really do not get it.

I'm too stupid.

The cone moves based on the force created by current in the voice coil.

Once you say the cone does not move, then you say it moves...

There is no optimisation to the detriment of SPL.

Whatever optimization conducted by passive means results in loss of efficiency.

The key differences will be that that the current drive version with metalised voice coil former will have no thermal compression,

By what miracle? Thermal compression is due to Re increasing. You think a flimsy piec of Al will significantly reduce thermal resistance?

significantly reduced odd order and high order harmonic distortion

Why?

and no reduction of the damping with large signal low frequencies, which in turn will help keeping the voice coil in the actual magnet gap

If you keep the voice coil in the field, you limit the excursion, which limits LF spl. Typically Xmax is a fraction of Xdam, about 10-15dB below. I used to design systems that operated at the limits of Xdam and max VC temp.

 

[scott2000]

Are we talking about these kinds of things?

 

[Newmarket]

Well I'm still not understanding if the cone is moving or not Maybe both simultaneously in different dimensions

 

[JohnRoberts]

The problem is in the filters. Since you cannot do up- and downsampling without them, and designers want to preserve the frequency response, need it to be real time and use modest CPU, the steep filters with few taps used wreck the signal integrity.

A simple test: Use a plugin with switchable AA filters like Fabfilters, select a neutral mode (e.g. limiter with threshold all the way up) and put one each on identical DAW tracks, flip the phase on one and turn on AA on one track only. What is left on the sum bus goes through the whole frequency range (not just the ultrasonic filtered part) and is of rather high amplitude. And the effect is cumulative, so stacking compounds the problem. The degredation is audible with decent source files and converters with one instance only.

That sounds like a case of designing the filters to look good on paper (specs) while cutting some corners.

For linear paths I have found high frequency multi-tone testing to be revealing. There is resistance to use test stimulus above the human hearing range. The secret sauce in multi-tone testing is that in band signals combine to create out of band rate of change.

Just thinking out loud....

JR

 

[thor.zmt]

Once you say the cone does not move, then you say it moves...

I stated no visible movement with a specific 15" Midrange (for front horn loading) driver when driving the speaker with a signal containing bass, which would cause very large displacement with classic 15" Bass Driver like EVM15B or JBL 2226.

If the cone did not move at all, there would no Midrange output either.

And you made clear you understood this by referring that enclosure systems that minmise cone excursion (horns, to wit).

So you are just being deliberately obstuse.

Whatever optimization conducted by passive means results in loss of efficiency.

Clearly you are mistaken, IF for the added mechanical damping electric damping is reduced.

By what miracle? Thermal compression is due to Re increasing. You think a flimsy piec of Al will significantly reduce thermal resistance?

No, but current drive means Re is no longer material. Current drive cancels Are variations contribution in the driver's acoustic output.

A conductive voice coil former with the correct resistance provides electromechanical self damping of the driver, so driving it with an AC current source doesn't cause LF peaks and poor transient response, due to the absence of electrical damping from the driving amplifier.

Why?

Basic laws of physics? Lentz among others?

If you keep the voice coil in the field, you limit the excursion, which limits LF spl.

Once the voice coil leaves the air gap output drops off so rapidly, we do not add appreciable LF output anyway. But cooling of the voice coil is comprised, potentially creating a thermal avalance condition that delaminates the voice coil glue.

Typically Xmax is a fraction of Xdam, about 10-15dB below. I used to design systems that operated at the limits of Xdam and max VC temp.

Interesting, what was the HD under those conditions? And compression?

At Xdam the BL is likely 10-20dB down on standard Xmax BL.

This may be tolerable in a pure subwoofer to be used below 100Hz. Even there it would be mostly a distortion generator at Xdam, with preciously little actual added acoustic output, as overdrive levels would be massive.

A servo loop could be used to make this usable in practice, again, Sub use only. The protection circuitry against thermal or mechanical avalanche conditions would be interesting. I'd not volunteer to do that without a serious DSP maybe not even with.

At the same time, why not specify a longer voice coil, drop midband efficiency (ultimate LF Output is down to Xmax and cone area anyway) and you have a Sub that has better inherent linearity with exactly the same 32Hz (or whatever frequency you pick) output?

Or, why not use an industrial rotary motor or a a set of linear motors a do a modern version of Danleys "ServoDrive". You can easily get a 1sqm carbon fiber honeycomb panel and get way past +/-1" linear excursion.

I have grave doubts about doing this with a full range speaker or a mid-woofer operating up above 1kHz. I am generally interested in the midrange, this where all the music is.

Subwoofers are basically air pumps that are more suited to industrial solutions. It's so trivial a problem in this day and age, I leave it to others to solve.

Richard E. Lord did a good job many years ago, two Chinese made copies of his "Storm" Sub bring up the low end in my living room. Mids are 10", 98dB/2.83V.

Thor

 

[5v333]

Once you say the cone does not move, then you say it moves

it must be due to the particle/wave duality of cones!

 

[thor.zmt]

Well I'm still not understanding if the cone is moving or not Maybe both simultaneously in different dimensions

It is moving, or there would be no acoustic output.

But next to "common mid-bass" 15" driver, the movement from even heavy bass signals is almost invisible, while a common driver would pump past Xmax (open air conditions).

This refers to the 15" drivers we used in the 80's as lower mid drivers (160...320Hz - 1.2kHz) in large horns. The cost a bit more than JBL or EV, but were much, much louder and cleaner.

I cite these simply as empirical proof that using a single piece alu voice coil former can provide prodigious levels of (electro) mechanical damping.

My proposal for an inherently damped driver to be used with current drive is separate.

For this I suggest to make Qm = desired Qt using a suitably coated, clad or otherwise processed coil former that provides the damping.

The resulting system allows conventional operation of the driver in acoustic terms, while driven with current, comparable to a classic voltage driven system, but with material improvements to objective and subjective signal quality.

At the same time the required changes to a "cheap generic" mass produced driver and the driving electronics are simple and extremely low cost, approaching zero in proper mass production.

As this system is mainly expected to operate as mid-bass or full range system, the cone will need to move as much as needed to produce the required LF output, I take that as read.

Application? Can be very wide ranging. Small Bluetooth speakers? Huge fully active concert systems? What is the technical difference?

Thor

 

[thor.zmt]

Are we talking about these kinds of things?

Not specifically, but could be. Arguably, these things are most in need of improvement these days. And affordable active (not in my) "Studio" monitors, that often make a cheap NAD Amp from 1990 and a pair of Warfdale Diamonds from the same time bought in the local charity shop positively "pro grade".

Large scale high end concert and venue sound systems tend be really not bad at all these days. Same for serious Studio monitors.

They need less help. Not that they cannot be improved further.

Thor