Is compression always a bad thing? Is there such a phenomenon as "euphonic compression" which is preferrred by some for its "fat" sound and non-fatiguing presentation?

I have heard tube amplifiers with deliberately engineered compresssive charactertistics, kind of a "retro" sound similar to the old days of recording on magnetic tape, pushing the gain to +3dB in order to achieve tape saturation on peaks and transients, in order to give a "fatter" sound.

Has anyone deliberately engineered soft compression into any front end electronics, interconnect or speaker cables, amplifiers, or loudspeakers? If so, could you explain the benefits of compression, how much to use, how to achieve it, some examples of current commercially product products which have compressive characteristics, either deliberate or otherwise? Do some driver magnet materials have compressive characteristics, such as iron, nickel, alnico, etc. whose compression is noticeable at normal listening levels?

I'm interested in utilizing euphonic compression in a loudspeaker I'm designing, just to see what it sounds like, if a certain amount is OK, what materials to use in xo or driver magnet to produce acceptable amounts of euphonic compression.

For example, how would a crossover be designed and what components would be used to deliberately compress peaks and transients? Would iron-core coils provide compression due to saturation effects? Are there any formulae or best practices in this regard, or is it all up to trial and error, trying various compoents until you hear the desired result?

Thanks for your help, rdb :-)

 Watch out for tweeter distortion - Lynn Olson 17:18:20 11/16/02 (1)
 Re: Watch out for tweeter distortion - RandyB 07:23:48 11/19/02 (0)
 Re: Euphonic Compression - tcain 09:46:40 11/13/02 (9)

...

Show full thread outline!

 Watch out for tweeter distortion - Lynn Olson 17:18:20 11/16/02 (1)
 Re: Watch out for tweeter distortion - RandyB 07:23:48 11/19/02 (0)
 Re: Euphonic Compression - tcain 09:46:40 11/13/02 (9)
 Re: Euphonic Compression - RandyB 16:32:56 11/13/02 (8)
 Re: Euphonic Compression - tcain 16:50:56 11/13/02 (7)
 Re: Euphonic Compression - RandyB 17:30:51 11/13/02 (6)
 Re: Euphonic Compression - tcain 19:47:35 11/13/02 (5)
 Re: Euphonic Compression - RandyB 21:10:06 11/13/02 (4)
 Re: Hello........... - Art J. 06:16:46 11/14/02 (3)
 Re: Hello........... - RandyB 08:05:55 11/14/02 (2)
 Re: Hello........... - Art J. 08:43:46 11/14/02 (1)
 I'll evaluate Zobel, thanks - RandyB 09:20:03 11/14/02 (0)
 Driver compression - John Escallier 20:20:46 11/12/02 (3)
 Re: Driver compression - RandyB 21:12:56 11/12/02 (2)
 Re: Driver compression - John Escallier 21:25:51 11/12/02 (1)
 Re: Driver compression - RandyB 22:17:25 11/12/02 (0)
 Re: Euphonic Compression - JOEY. 02:15:05 11/12/02 (2)
 Re: Euphonic Compression - RandyB 07:10:58 11/12/02 (1)
 Matching up compression characteristics - Duke 11:33:51 11/12/02 (0)
 Re: Euphonic Compression - Art J. 20:02:03 11/11/02 (5)
 Re: Euphonic Compression - RandyB 21:36:30 11/11/02 (4)
 Re: Euphonic Compression - Bruce from DC 08:35:42 11/12/02 (1)
 Re: Euphonic Compression - RandyB 14:28:48 11/12/02 (0)
 Re: Euphonic Compression - Art J. 06:58:40 11/12/02 (1)
 Re: Euphonic Compression - RandyB 08:05:33 11/12/02 (0)

I always associate compression with very fatiguing presentation.
Here is my favorite link on the subject that goes into some depth,
even though its mainly about CDs.

• George Graham on Compression

Thanks for the link, interesting article. The euphonic compression I'm thinking of would prevent only the top 1-2% of the loudest peaks and transients from being too loud when the volume is turned up to comfortably loud levels for all but these loudest peaks.

An example: 95dB sensitivity speakers being driven by a single-ended tube amplifier with lots of headroom, no clipping, turned up to realistic SPL's, but the loudest transients peaks are too loud for comfortable listening. Is it ever permissible to limit these peaks to reduce discomfort so that the rest of the program material can be played at a higher volume setting than would be possible if the peaks were not limited / compressed?

Reverting back to the magnetic tape example, where the recording engineer in pre-digital days would set the gain to +3dB so that the highest peaks would be compressed due to tape saturation, I was wondering if there are any amplifiers or loudspeakers that have this characteristic, whereby only the very highest 1-2% SPL levels of transient peaks are compressed to permit a higher volume setting for the 98-99% of the rest of the program material.

Does anyone know if there known coefficients of compression for coil, capacitor, and resistor materials and configurations for various frequencies and SPL levels? If there are any treatises on this topic, links would be greatly appreciated.

This is a phenomenon I have observed on some high-efficiency loudspeaker prototypes, 95-97dB sensitivity, some of the peaks are too loud when the volume is turned up to realistic levels, the rest of the SPL range below these highest peaks is fine. I'm trying to "tame" the sensitivity of the speakers on the 1-2% of the loudest peak levels, I guess, so that these peaks are not too loud.

Thanks, rdb :-)

Thanks, rdb :-)

First point, in my view is that compression is always a bad thing; however, if you are trying to engineer compression into a driver, you could

1. try using a very non-linear suspension by employing a suspension with a reasonably linear force/displacement characteristic at 'normal' excursions, but with a pronounced (much steeper) force/displacement characteristic at slightly greater than 'normal' excursions (i.e: your 1 to 2 % of the time). You would have to experiment with this.

2. if your driver uses a vented pole piece, block the hole, this will increase compression effects, but may also reduce power handling.

3. you could wind the voice coil in a non-unifrom way: tightly spaced coils in the normal excursion range (middle of the voice coil) and more widely spaced coils at large excursions (either end of the voice coil).

But what a waste.

Just out of curiosity, is the driver completely custom, or is it a Lowther (just a feeling) rebuild?

HAVE FUN

JOEY.

Thanks for the tips! I'm designing a range of 5-6 standmounts and 1-2 floorstanders, all with minimum 95dB senstivity, using a variety of woofers and tweeters:

Eminence Alpha 6 and 8, Beta 8cx and 10cx - custom doped
Eminence custom spec'd 8 and 10" - rubber surround, resin impregnated pulp cone with kevlar fibers
PHL 1240, 2460, 3451, and 3020 - stock drivers

Peerless 95dB dome tweeter with 1" deep waveguide
Eminence APT50 phenolic compression driver
Morel HD series high-efficiency tweeters
Raven 1, 2, and 3 ribbon tweeters

In my prototypes, the uncompressed transient peaks from the woofers are fine, it is in the treble where I want to tame things down a bit on the highest 1-2% of the transient peaks, so I'm looking for something I can do in the crossover . . . I don't want to dope the HF diaphragms, due to lowered resonance frequency and rolled-off HF response. So I'm looking for caps or resistors, maybe even a coil somehow in a non-standard topology to limit the highest HF peaks . . .

The voice coil winding configuration sounds promising, but I don't have small enough fingers or skill sets to rewire tweeter coils, and the Ravens don't use voice coils.

Thanks for your help, rdb :-)

Thanks very much, this sounds like an elegant solution! I'll research these devices and try some out.

Best regards, rdb :-)

Consider a brass marching band. It is loud and it has transient peaks like no tomorrow. Is it painful, unpleasant? No.

I suggest that "unpleasantness" and "too loud" as perceived by listeners to stereo systems are the result of overloads -- and the resulting distortion -- somewhere in the reproduction chain. And that compression, no matter how accomplished, is at best a band-aid solution to these problems.

Regarding tape recording. IIRC, the principal consequence of recording at "+" VU meter settings is an increase in harmonic distortion. That produces the "fat sound," just as has been demonstrated with amps that produce a lot of harmonic distortion when driven hard but not to the point of clipping the waveform.

In a multi-way system, matching up the compression characteristics of the different drivers would be quite desirable, in my opinion. Otherwise if the woofer compresses a bit more on loud peaks (which woofers will do when they exceed linear Xmax), then those peaks will have a brighter tonal balance. If this is what you're doing, then you are exploring some very interesting (and to my knowledge undeveloped) territory. As an amateur designer I've tried to match up compression characteristics, but frankly I never thought of adding a bit of compression to the tweeter. That is brilliant.

Have you run this by Alan Hulsebus? I wonder if he'd have any thoughts.

I can't offer any great ideas on how to pull it off, but I sure do wish you the very best of luck with your project!

Duke

The woofers in my prototypes are working fine without compression, it is the tweeters that are exhibiting some harshness on the loudest peaks. I don't want to change the crossover re slope or crossover point, as the current setup is fairly flat with a nice blend throughout the overlap frequency band. I would just like to limit the tweeter on the loudest 1-2% of transient peaks, I guess they are too efficient at the loudest levels relative to the woofer, and when the volume is turned up to realistic SPL's, there are occasional peaks which sound a little too loud in coming from the tweeter . . . this is what I am trying to tame. I'm looking into zener diodes per Art J's post, I'll experiment with them to see how they change the sound, if this is an acceptable solution or not . . .

I was just using the analog example to see how "deliberate" compression is being implemented these days . . . maybe there are some recording engineers out there who made the transition and figured out how to achieve the "fat" sound in a better, less distorted way . . . ??

Thanks, rdb :-)

The copper of the voice coil heats up as a result of power dissipation (don't forget, we're only talking 1-2% efficiency here for normal non horn drivers.) When it heats up, the resistance causes less current for the same voltage drive from the amp. The time constant is a thermal one, rather slow, not cycle to cycle.

If the different drivers are not matched in this respect, the frequency response will change with time at higher power levels.

The magnet structure will have some hysteresis, but that is to be avoided, it will cause all sorts of harmonics.

TTFN, John

Thanks for the background info . ..

The line of speakers I'm developing is designed to be used with low-powered SET amps from 1.5 - 10 watts, although they will handle upwards of 100 watts rms. With the low wattage of SET's, will there still be heat-related compression?

If so, how do you determine heat-compression compatibility between woofer and tweeter, so that they stay "in-synch", compression-wise?

Also, what if there is a whole lot of bass going on, and little up on top, wouldn't this make it very difficult to maintain both drivers in heat-compression synch?

Thanks, rdb :-)

That would be driver dependent. Some manuf's give the info.

##If so, how do you determine heat-compression compatibility between woofer and tweeter, so that they stay "in-synch", compression-wise?##

I would assume matching compression parameters. Unfortunately, different musical programs will give different power spectra.

##Also, what if there is a whole lot of bass going on, and little up on top, wouldn't this make it very difficult to maintain both drivers in heat-compression synch? ##

Same answer as last...But, I run a mobile rig for hobby/party stuff... Never too much bass. Gotta shake the martini's without bartender intervention.

I remember some application using light bulbs for compression. Don't remember what app, though. Light bulbs have about a ten to one resistance ratio from cold to hot. In series with a speaker, they'll be low resistance for low levels, and at high levels, will warm up and increase resistance. By playing with different wattage/voltage bulbs, you could conceivably tailor the "compression" in this fashion.

Keep this in mind: I run 300w/per channel for subs, my light rail attaches to the sub cabinets. When I run "hot", the 120 volt, 300 watt bulbs tend to blow out; they don't like vibration. I had to design an isolation mount for the lighting effects fixture. The 12 volt bulbs, with thicker filaments, didn't show that sensitivity.

TTFN, John

PS. IF you patent something using this, I want credit too.

What driver parameter provides heat-compression data? If two drivers have different coefficients of heat-compression, what can be done to put them in synch?

I guess this is where ferro-fluid comes in handy, to help reduce voice coil heating and subequent increase in resistance. If the woofer doesn't have adequate cooling, and a ferro-fluid tweeter is used, I guess this means that over 1/2 of all consumer loudspeakers have heat-compressive woofers and non-compressive tweeters, so after 1/2 hour of play, the woofer is compressive and the tweeter is non-compressive.

This phenomenon might mean that designers of loudspeakers should take this into account, and take measurements after the voice coil has had a chance to warm up. If so, this would seem to indicate that like electronics, speakers need to "warm up" in order to achieve their steady-state performance equilibrium, at least where the woofer and tweeter have different coefficients of heat-compression.

I imagine the average audio consumer wouldn't want to buy a loudspeaker if he thought that its performance would change depending on how loud it is played, how many watts are being used, what the temperature of the voice coil is, etc.

If a design could be implemented that would take heat-compression into account as between woofer and tweeter, which would maintain synchronized compression modes, this might be considered to be a value-added benefit, i.e., it might have marketing value to differentiate this type of design from everthing else. Or is this just too esoteric to matter to a typical audio consumer? It would have to be explained in accessible terms such as "iso-comp" or "equisterisis" or similar. The benefit would be less coloration at loud SPL's, so while dynamic compression can't be eliminated, it can be managed to prevent tonal anomalies.

Anybody out there working on anything like this, or has it already been implemented in a consumer loudspeaker? Or is this new turf, ready for some enterprising individual to plum the depths of the technology and come up with a new design paradigm?

Best regards, rdb :-)

Thanks for the info, I'll check it out . . . where do I get magnet wire, what is it made of, what guage, etc.? If I go into an electronics supply house, will they know what magnet wire is?

It sounds like the speakers are too dynamic to me, on the loudest transient peaks, and there is a "glare" or stridency, as if the dynamic range of the speaker is too great, too much difference between the soft and loud parts of the program material . . .. Is there such a thing as having too much dynamic range?

Best regardas, rdb :-)

This is why Directly heated triodes work best with high efficency. Tube circuits, especially push pull and hybrids etc. are prone to emphasising this upper midrange/treble resonance. SET's with Direct Heating have no resistors and capacitors in output stages and instead use chokes and transformers to couple the signal, far better tonality is the result.

Try some of the smallest magnet wire in the ratshack 3 pack for speaker wire. It does help "damp" some of the ringing. But damping and filtering are working in the wrong direction ultimately. The actual source of the problem lies not in the speakers driver but in the system IMHO. If damping electrically is needed, plain wire, either as a transfer function from amp to speaker or as a component of the speaker in the form of coil or tranny works better than splitting the signal into smithreens with ceramic or semiconductor. Follow link for introduction to Direct Heating. TC

• Exploring the wonders of Direct Heating

Thanks for the info on magnet wire, I'll get some and check it out this week . ...

I'm using DHT's now, PX25 and 300B SET amps, along with a Technics 100 wpc AV receiver (to check compatibility with SS mid-fi amps), Hafler 9505 200 watt mosfet, Shearne Audio 100 wpc dual-mono amp, and I'm hearing similar glare on all of these on maybe 1-2% of loudest transient peaks on my prototype high-efficiency speakers. I don't hear glare on less efficient commercial speakers being used for comparisons, Aerial Acoustic 10T's, various Reynaud models, various Westlake models, Heybrook 2000 series. I also don't hear the glare on various Loth-X high-efficiency models, Ion 1, Ion 2, Ion 3, and Ion 4.

I don't hear any glare or harsh peaks on any of the woofers I'm using, i.e., Eminence Alpha 6, Alpha 8, Beta 8, Beta 8cx, Beta 10cx, PHL 1240, 2460, 3451, and 3020.

The HF units I am using are Peerless dome tweeter with 1" waveguid (forget which model), Eminence APT 50 phenolic compression driver, Morel HD 37, and Raven 1 and 2 ribbon tweeters. I'm using first order xo's on on all but the Ravens, which have 3rd order (to provide steep enough slope per operating specs).

If I don't crank up the volume quite as loud as my preferred level, I don't hear the glare, only when I crank it up to realistic SPL's so that the woofer output is at what I feel is realistic levels . . .

Could it be that the tweeters are just too efficient relative to the woofers, that I need to add an l-pad or resistive attenuation on the tweeters? I'm going to try out some zener diodes to limit the loudest transient peaks to see if this resolves the tweeter glare, I don't know what the effect will be on overall sound quality with the zeners, I've never used them before . . .

Best regards, rdb :-)

Thanks for the Zobel / notch filter ideas . ..

When I started this project, I was just going to use a cap on the tweeter, let the woofers run full range . . .. design creep has added attenuating resistors on the tweeter, a coil on the woofers, 3rd order for the Ravens, possibly zener diodes, now I'm considering Zobel and notch filter . . . . Maybe I'll call the final design "Topsy", it just grew . . . . . .

Best regards, rdb :-)

I have been posting on Hi-Efficiency forum in the beginning stages, now I'm down to the 2nd and 3d order design issues, the basic designs are completed and I'm refining things, so I thought I'd see what the Propeller Heads think about a few issues, to get a fresh take on things . ...

Euphonic compression is used by several tube amplifier manufacturers to produce a "retro" sound, I was interested in finding out if I could put in just a little compression to get a "fat" sound, with the added benefit of maintaining a flat frequency response between the woofer and tweeter on the 1-2% of the loudest transient peaks.

I didn't want to use Zobels in this project, since these loudspeakers are designed to be used with flea-powered tube amps (and ss up to 100 watt rms), and I want to minimize power loss in the crossover. I originally wanted to use just a cap on the tweeter, and let the woofers run full range, but high frequency cone breakup and too much overlap forced me to add a coil on the woofers . .. I'm concerned that if I add a Zobel, it will suck too much life out of the few watts available on the lowest-powered SET amps, like 2A3, 300B, PX25, for example.

Best regards, rdb :-)

Thanks for the info re Zobel benefits, I'll definitely include Zobel topologies in the iterative design process, see if it leaves enough "life" with flea-powered amps . .. and how it improves driver integration and overall presentation . ... could be a desirable trade-off, fer sher!

If anyone out there has used a Zobel with flea-powered amps, please let me know results, if this can be done without making things too "polite", and if so, best practices for implementation . ..

Thanks, rdb :-)

For example, a tweeter crossover at 3kHz at 12dB/octave yields the *same* diaphragm excursion at 1.5kHz as it does at 3kHz. The diaphragm excursion will stay the same, in fact, until the tweeter encounters its own self-resonance (for a 1" dome, around 600 to 800 Hz), and the net acoustic rolloff increases from 12dB/octave to 24dB/octave. Unfortunately, there is usually a rise in response around the self-resonance, so excursion at resonance may actually exceed the level at 3kHz.

Will you hear this increase in unwanted excursion? Not directly, no. The acoustic response at 750Hz still some 24dB below the nominal, but the contribution to IM distortion is substantial. The spectral center of energy of most music is around 500 Hz, so the 500 Hz to 1kHz region intermodulates with the music that the tweeter is intended to play, above 3kHz.

To a first approximation, driver distortion is proportional to excursion, although this changes as the acceleration limit of the diaphragm is approached, then exceeded. Exceeding the acceleration limit of the diaphragm is classical "breakup", and can happen slowly or suddenly, depending on the cone material. More rigid materials like metal, carbon-fiber, or Kevlar tend to enter the breakup region more abruptly than softer materials like paper or polypropylene.

As you can see, first-order (6dB/Octave) crossovers actually increase cone excursion until first resonance, below which the net acoustic response drops at a 18dB/Oct rate. Second-order crossovers have constant excursion, and third-order crossovers decrease cone excursion at a 6dB/Octave rate (until resonance).

The tradeoff between tweeter excursion (and increased IM distortion) and degraded phase linearity is one of the more difficult choices facing the speaker designer. The only way to make an end-run around this is a digital crossover - but that opens another set of problems all on its own. (How do the early adopters of digital speakers feel about DVD-A and SACD?)

At any rate, getting back to the original posting, I would be wary of the tweeters entering an abrupt break-up region, with the onset masked by the louder output of the woofers. This can be quickly debugged by disconnecting the woofers and listening to the offending musical passages with the tweeters and associated tweeter crossovers alone. You maybe surprised at how much distortion you hear without the woofers masking the distortion.

Thanks for your info, I'm using 3d order on the Ravens, and don't experience as much (or any) of the annoying "glare" which I get with the other HF units, which are using 1st order xo's . . . I don't know if this is because the Ravens are just smoother, or if the 3rd order is helping to prevent the glare, and the 1st order xo's are permitting the HF units to be pushed too hard at frequencies less than 3-4 multiples of the HF unit's resonant freqency.

Question: What loudspeaker design software would let me get a handle on breakup phenomena, distortion, etc., to help figure out better what's up with this? LMS, CLIO, others, ???

Thanks for your input!

Best regards, rdb :-)