d e c e m b e r 1 9 9 7 3 V A L V E

E d u a r d o d e L im a ’s m in i m o n i t o r f o rS E 3 0 0 B a m p s

w h y d o e s ze r o -f e e d b a c k s in g lee n d e d s o u n d s o g o o d ?

v a r ia b le t im e d e la y r e la y

a le t t e r f r o m t h e N im i t z

V A L V E

v o lu m e 4 , n u m b e r 1 2 D e c e m b e r 1 9 9 7

Robert Root’s Eurydice style preamp

V A L V E 6 d e c e m b e r 1 9 9 7

w h a t is t h e t e c h n ic a la d v a n t a g e t o z e r o -

f e e d b a c k s in g le e n d e dd e s ig n s ?

By Lynn T. Olson

This oft asked question was posed on the SoundPractices “joelist” recently (thanks to SP Editorand Publisher Joe Roberts for creating the list asa place for all of us bottleheads to share ourinsights), and one of the most lucid replies tothis question was proferred by V&T News EditorLynn Olson, who gave us permission to print ithere. Keep a copy of this in your pocket for thenext time some NFB push pull bully tries towave distortion measurements in yo’ face....B.

Several reasons, all measurable with the rightequipment:

∗ Cleaner spectral response, due to an all-triode signal path with no loop feedback.Triodes have the cleanest spectra of anyamplifying device ever made, especially ifyou weight the harmonics as mentionedabove. Feedback multiplies low-orderharmonics into smaller but morenumerous quantities of high-orderharmonics. See articles by NormCrowhurst in Glass Audio for themathematical derivation of this.

∗ Crossmodulation. As a long-time speakerdesigner, I know that speaker drivers arewicked devices, cheerfully storingresonant signals for many millisecondsand then sending them right back to theamplifier. If there is zero loop feedbackthe trouble stops right at the plates oremitters, otherwise it goes to the inputstage and cross-modulates the desiredinput signal. The usual engineeringassumption that the feedback summingnode has zero distortion is lazy thinkingand is simply not true. If the summingnode has non-zero distortion the driverresonances will cross-modulate with theamplifier distortion. This cross-modulation is partly responsible for thenotorious matching problems betweenamps and speakers... worse, some of thevery best speakers are the most reactiveloads. If you want reactive loads, checkout ESL's or horns. If your amp can'thandle reactance gracefully, then you arestuck with Magneplanars, which arelargely resistive loads.

∗ Greatly reduced problems with TIM/slewing distortion, which is an especiallyunpleasant form of signal-dependent timedispersion. Feedback generates large errorovershoots when the amp nears clipping,slewing, or Class AB transitions. Theseovershoots can saturate low-current inputstages, which then lose the ability to drivethe Miller capacitance of the followingstage. See articles by Matti Otala and manyother authors in the AES Journal on thistopic.

∗ Feedback greatly worsens the audibility ofclipping, which is a serious problems withamps that don't have a 200 to 300 wattoutput capability. Music typically has a 10to 20dB peak-to-average ratio, so 90dBaverage levels require peaks of 100 to110dB. If the amp/speaker system is notcapable of that, it is highly desirable thepeak is gently compressed, not hard-clipped. The worst case is a feedbackamplifier with output devices that saturateor become misbiased during clipping; notonly does feedback hard-clip theamplifier, it may take a long time for theoutput devices to sweep out the charge,cool down the silicon die, regain normaloperation, and return to a low-distortionregime. This stretches out clipping andmakes it extremely offensive, in themanner of a transistor radio.

The effects discussed above interact in complexways. For example, a very common problemwith many transistor amps (nearly all, actually)is momentarly loss of phase margin stabilitywhen the amp clips. In the severe case, there areactually quick bursts of 2 to 20MHz oscillationwhen the amp clips into a speaker + cable load.(If this goes on long enough, the output devicesand the tweeters will fail. This is one of the mostcommon reasons that solid-state amps blowup.)

Assuming the above problem is addressed by thedesigner, when the feedback amp clips, the TIMdistortion will still greatly increase until theoutput transistors leave saturation, cool down,and re-enter their linear region. As long as theoutput device is outside the linear region,overshoots in the feedback network are verylarge and easily overload the input section,which then in turn has to recover fromsaturation. What starts as a simple overloadproblem in one part of the amp spreads to thewhole amp thanks to feedback ... and none of itshows up in a THD measurement due to thetransient nature of the problem. The amp can bevery unreliable, sound terrible, yet measure very

V A L V E 1 0 d e c e m b e r 1 9 9 7

STYLE

a mini monitor forsingle ended

300Bsby Eduardo B.E. de Lima

d e c e m b e r 1 9 9 7 1 1 V A L V E

This loudspeaker, which I ended upcalling STYLE, was originally designedbecause I needed a small speaker thatwould play with my 300B SE ampli-

fier. One of the areas where SE amplifiers excelis in the small details and decay of the sounds.Sometimes this can be appreciated even betterin small rooms at low listening levels. There-fore it seemed natural to design a smallspeaker to work with SE amplifiers in smallrooms.

This speaker was one of the speakers I havedemonstrated at the VSAC. It has around 90 dbof efficiency with a very, very easy inputimpedance.

I believe that the very best performance withSE amplifiers can only happen when the wholesystem of amplifier, output transformer,speaker driver and box is very well matched.This requires the loudspeaker be designed forthat particular amplifier, or less frequently,the other way around. This includes takinginto account the amp output impedance, dis-tortion levels, the transformer primary induc-tance and who knows how many other inter-related things! This may explain why most ofthe time it is more effective (and a lot morefun!) to let past experience, some intuition andmostly the ear do the whole job and find thiscorrect match. After all, the ear must be thefinal judge.

But although this speaker was designed to beused with my amplifier it should work verywell with most typical SE amplifiers. The bassalignment is targeted for amplifiers with anoutput impedance of 3.0 ohms, and it willwork well with amps with Zout in the range of2.0 to 4.0 ohms. At the frequency range thisspeaker is expected to reproduce, any primaryinductance bigger than 10H (for a 3K primaryimpedance) will work well. This means youmay use good quality small transformers andget good sound. This speaker was designedaccording to some ideas that are presented inGlass Audio (3/97 and 6/97) and that werethe subject of my seminar at VSAC 97. For itssize and efficiency you will have surprisinglydeep and controlled bass when they are usedwith a typical SE amp.

The Box

The box is very simple. It can be made withMDF. I have used 5/8” all around and 3/4”for the fascia. This fascia should have roundedsides. After some tests I found these materialsjust right for this box size, with no need forfurther bracing. If you want to try, thicker

materials and bracing may be used. This maybe good or bad, and experimenting is the onlyway to know. Remember to keep internal vol-ume constant. If you decide to use thickermaterials please change only the depth of thebox. The frontal dimensions should remain thesame. The box should be lined with 1” foam oflow density on all sides except the front baffle.Of course you can also experiment with thisuntil you get the best result. The duct shouldbe made of a 5cm (2”) curved PVC tube about18cm (7”) long. It must be a curved duct,otherwise it will not fit. Using a smaller diam-eter will allow you to use a straight tube. Thiswill work but it is not the ideal. Always keepthe duct as free as possible. Do not allow thefoam to block the area close to the duct.

The Drivers

I have seen no projects using the AudaxHM130X0, but this speaker with a TPX cone isa remarkable mid/woofer for our application.It has a very smooth midrange and just theright parameters to be used with high outputimpedance amplifiers. It probably has not beenused frequently because the calculated idealbox for common low Zout amplifiers will beridiculously small and the speaker could onlybe used as a midrange. But the driver has theXmax and power handling to be used as a smallwoofer. The Seas T25-001, with its silver coiland very good subjective sound, was mychoice for the tweeter. It has complementedthe Audax in a seamless way. Although notcheap, these are not terribly expensive units,each one being around $60.

The Crossover

This is a mix of the minimalist approach withsome impedance correction. The Audax unit isconnected straight to the amp. The impedancecorrection network keeps the impedancecurve in the mids and highs very smooth, with8 ohms being just about average. The mea-sured frequency response is very flat, beingwithin +- 2.5 dB through most of the range.

I have used no diffraction compensation in thecrossover. this makes this speaker sound betterplaced closer to the rear wall, as it is expectedto sit in a small room. the best position issomething that only experimenting in yourown surroundings can determine. I shouldalso recommend that you them on 24” to 28”stands placing the point halfway between thetweeter and the woofer at ear level. This is notvery critical, but certainly helps. The fre-quency curve shows the response from 400 Hzup, without any smoothing. This response

V A L V E 1 2 d e c e m b e r 1 9 9 7

changes very little with high outputimpedance amps. At frequencies below 400Hzthe response is affected by the measurementsetup limitations and by the value of the out-put impedance of the test amplifier.

Conclusion

I believe you should build this speaker with allthe normal care that high performance equip-ment begs for. The drivers and crossover com-ponents should be very well fixed. The cabinetshould be very well made and the driversshould be mounted flush with the front sur-face.

I should say that this speaker is intended to beused with high output impedance amplifiersand that it will sound bass shy with the aver-age transistor amplifier or tube amps usinglots of negative voltage feedback.

I hope you try this little speaker. It has beenthe source of a lot of great listening for me andI believe it fulfills its intent very well.

opportunity knocks once

Doc B. is getting too busy for his own good.Between several joint ventures, the kit biz, andputting on VSAC, he can barely get VALVEdone every month.And so Electronic Tonalities/ VALVE is look-ing for a person with skills in desktop publish-ing who would have an interest in assisting inthe monthly production of VALVE. We needsomeone who can come in to our office on amonthly basis and work with Doc to plan outeach issue, and then complete the transcrip-tion and layout unassisted, to a strict deadline.The pay won’t be too astounding, but on theother hand, you will only need to be working afew days each month, you get to be part of avery hot audio publication, you will get to bein on cool audio developments as they happen,and you’ll get an awful lot of opportunities totalk to some very well known audio personali-ties. This offer won’t last, call us quick if you’reinterested, 360-697-1936.

d e c e m b e r 1 9 9 7 1 3 V A L V E

and Eduardo B.E. de Lima.One pair of these loudspeakersmay be built by the reader fornon commercial purposes only.

d e c e m b e r 1 9 9 7 1 5 V A L V E

Above: Crossover schematic for the STYLE mini monitorBelow: 400Hz to 20kHz response of the mini monitor

V A L V E 2 2 d e c e m b e r 1 9 9 7

Doc B. and Sakuma san show off the 304TLspresented in appreciation to Sakuma san for hisefforts to show us how important capturing the

emotion and energy of the music is in how we designour equipment. I suggested these might be useful for

his next preamp project....Photo by Takuji Yamamoto, who put in many long

hours preparing for our guest’s generous efforts tobring U.S. bottleheads a taste of Japanese audio

philosophy.

Here’s a shot that helps explain all thecomments you hear about the highlight of the

show being the people who attended.From left to right:

Eduardo B.E. de Lima, next, a gentleman whowas in the Direct Heating party, and whose

name I am ashamed to say I missed, Lynn T.Olson, Takuji Yamamoto, Reid Welch, and

Susumu Sakuma.Photo by Mitsuru Uchida

Queen Eileen in VSAC HQ, somebody elsewho put in unbelievable long hours,

handling all the business aspects of VSAC,taking care of travel worn bottleheads

and feeding the troops at our first amp design class.Photo by Mitsuru Uchida

D o t h e s e V S A C p ic s e v e r s t o p ? This month’s photos are by Takuji Yamamoto of the Direct Heating group

and Mitsuru Uchida of Sun Audio