Wednesday, November 30, 2011

The Rotary Woofer

I first read about it more than 20 years ago, the Eminent Technology TRW-17 Rotary Woofer.  Yes, I can't afford it (not planning to open theme park).  Peter Montcrief in the International Audio Review called it the only true subwoofer.

They have a downloadable audio test CD.

They argue that there is no clearly defined low frequency limit to human hearing.  Instead, human hearing simply becomes less sensitive at lower frequencies.  At 5 Hz, it takes an astounding 115dB simply to be audible.  For cone speakers, this is extremely difficult, it takes more and more surface area to reproduce such low frequencies.  Of course, it was exactly what the TRW-17 was designed to do.

(I'm not planning go get one, they're very expensive and require custom installation.)

Dialed back 26 Hz boost to 0.1db

On Wednesday morning, I was finding it somewhat difficult to crank the volume level on Bass Ecstacy by Bass Erotica (a very low-bass-heavy album).  So then it occurred to me that after I extended the deep bass to 16 Hz this weekend by adding one subwoofer port plug on the left side and two on the right and lowering the low frequency cutoff accordingly, I may well have been adding back in much of the circa 26Hz bass the boost had been intended to restore.

So I dialed the 26 Hz EQ (in the DCX 2496 crossover) back from 3.5dB, where I had set it most recently, to 0.1dB, effectively disabling it without removing it.

Then the extreme bass became more tolerable, and I was able to raise the level by a comfortable 10dB, making the lyrics and effects much more open sounding.  I was even dancing.  But I think the bass still needs more refinement.  I believe I can still listen to this very difficult album louder in the master bedroom, where I have custom equalized the bass with parametric and graphic EQ using a Behringer DEQ 2496.

Might be time to bring back the keyboard, which was put away on Saturday Night to prepare for a party on Sunday.  I was even thinking about using a real oscillator, but might be nicer to program a frequency offset slider into my keyboard sine program.


Back on Friday night, when I found another bass port plug for my left subwoofer, I measured the effects of decreasing the low frequency cutoff frequency on 16Hz playback, and the effect of adding the one plug.  Adding the plug did increase output by a few dB, as did changing the cutoff to the one-plug recommended setting (which is 18 Hz).  Adding the plug had a bigger effect than turning the cutoff to the two-plug recommended setting (which is 16Hz).  I had the numbers written down, but they did not get copied to this blog.  Now I've found a second plug for the left subwoofer, and thinking about doing a whole matrix of measurements.  But I need to replace the C cells in my Genrad 1933 SPL meter, which conveninently has a "flat" setting and 1 inch microphone.  I was measuring levels between 89 and 100dB SPL at 16-20Hz.  I believe I boosted response at 16Hz from 89 to 92dB.  The bass plug not only increased 16Hz response, it flattened the hills and valleys between 16 and 25 Hz.

SVS shows the flattest response (ruler flat) to 20Hz with no port plugs.  They show 2.5dB loss at 20Hz with one plug, but improved extension to 15 Hz before cutoff.  They show 6dB loss at 20 Hz with two plugs, with little visible knee in the curve, but lots of drooping, so response isn't actually extended until you get down to 13Hz or so.  Three plugs produces 8dB loss at 20Hz, with no indication from the curve that this offers additional frequency response extension at any frequency.  According to their graph, I think I would prefer the 1 plug modification, very little difference at 20Hz but nearly flat extension to 15Hz.  The two plug seems to sacrifice too much response at 20Hz, but seemed better when I measured on the right speaker.

Here is the 

Tuesday, November 22, 2011

House Curve discussion

Here's a discussion (or really, an introduction to the concept) of a House Curve, what I generally call "room curve" (coming from the Tact tradition of Room Correction System preamps).

He proposes an interesting way of setting curve by making 100 Hz and 30 Hz sound equally loud.  (Just due to human hearing response, the 100Hz would naturally sound louder if played at equal physical loudness, and the effect is greater at lower playback levels.  So this is obviously increasing deep bass response over the flat or natural response, a point Wayne doesn't make or belabor.)

I find the "discussion" (actually, there isn't any discussion, except Wayne raises and addresses some issues in his own writing) rather lacking.  The rationale for house curve is very flimsy.  I suspect that HTS editors feel take house curve as a given need, didn't need much convincing.  Wayne is at his best shooting down other bogus ideas, such as that a house curve is needed because of the industry's "X curve".

Despite having this gnawing feeling that rationale for a room curve (other than flat) is at best circular reasoning, I count myself as a believer in having a house curve and Wayne's ideas (such as the 30Hz, 100Hz test) are sensible even if his arguments for them are weak.

I'm thinking the way to think about this is to consider that every room has a reflective signature, and a room curve is chosen to make music more intelligible given that reflective signature, they type of music it is, the type of speakers, etc.  Start with this as "hypothesis 1" and I think it's fairly obvious.

"Hypothesis 2" goes farther, making some specific claims.  Flat average response falls flat because it lumps together direct and reflected sound, which the brain is somewhat capable of perceiving separately.  You would think the direct sound should be the flattest, if you equalize the total sound, which includes proportionately more bass, you will make the direct sound component of it lighter in bass.

Aha, but we do have ways of mesuring or computing direct vs reflected sound.  And a system can be designed around the goal of flat direct sound.  Has been I'm sure.  And what is the result?  I don't know, hypothesis 2 could be wrong.

"Hypothesis 3" takes a different but similar view.  Instead of direct vs reflected analysis, our brains are assumed to have real-room-response correctors.  When we go into any room, we start correcting the sound to fit our perceived sense of how the room itself boosts that (in modal patterns in the bass).  Therefore, the recorded sound played back should have those same boosts.

The problem, however, with taking H3 seriously is that flat-played-back sound will indeed get the room boosts added to it.  That is what the natural room boost does to all sounds.  So from this perspective, a "house curve" would be adding to this.  But why should boost be added more to played back sound

The keyboard oscillator

For testing speakers, particularly subwoofers,  I find that nothing beats the conveninence and flexibility of a keyboard synthesizer.  Most Kurzweil "programs" (what others might call soft instruments) are very complex.  But it's not hard to cook up a sine wave oscillator useful for testing subs, building on the simple "Default Program" number 200.  Here is what I generally set up:

Tone: sine wave
Pressure sensitivity: 0
Control #8: volume
ADSR: sustain 100%, else 0%

The area between C0 and C1 (16 and 32 Hz) is interesting.  I think my EQ boost in this area helps restore strong response in this difficult region.  But each different note causes a different feature of the room to start rattling.  Only below 18Hz or so does it seem like keys do nothing.  And there I wonder if I haven't programmed my SVS PB13 to cut out too high.


I can't reiterate enough how useful a keyboard oscillator is for system adjustment.  It just sits there calling me to plunk a few more notes, checking out some other aspect of the sound.  Changes I've made so far:

+5dB at 27Hz changed down to +3.5dB

Room mode cut: 44Hz -9dB Q: 2.2 (sounds all to the better)  This is the most important room mode to tame within the subwoofers operating region where it is most important to eliminate high Q resonances (the sub tends to stimulate room modes far more than the panel resonances can almost be ignored on the Acoustats).  Back before I started using Tact correction in the Living Room (January 2010, I remember it well if not memorialized here) I had been using two tuned resonance cancellors at something like 38 and 45Hz.  The keyboard makes it easy to check these things out and be sure you haven't gone too far (I know 9dB is a lot but it is vaporized in the wind of the resonance).

PB13 subwoofer tuning: I've put in two plugs into left sub ports (leaving 1 of 3 ports open) which permits me to dial back the sub low frequency cutoff (lfc) two notches (I had previously dialed back the lfc from 20Hz to 18Hz anyway.)  So now I'm choosing the 16Hz cutoff and I'm doing so within recommended usage.  Putting the lfc in in a lower-than-recommended position now buys very little difference (1dB) at 16Hz: measuring (with GR 1933 meter) I get 69 or 70dB at 16 Hz, 69 with recommended filter setting, while 70dB is about the average response level 20-80Hz, though response clibs to 72.5dB at 18Hz.  As it now stands, 70dB at 16Hz is barely barely audible (the lower setting changes this to barely audible); the strong 72.5dB at 18Hz is nicely audible.

Monday, November 21, 2011

New house curve

I started running the Tact Room Correction analyzer for measurements on Friday evening and into Sunday.  I decided that my living room system is measuring so good it does not need full system correction (and I will need to get better about pasting from measurement into target curve).  However, I made several important adjustments.
Final Adjusted Response, Both Channels, Nov 20

First I moved speakers slightly back and out by about 4 inches, the most available with current positioning of Belkin PureAV power conditioner.  This seemed to bring a slight additional improvement in image coherency,   The highs are also adjusted to be slightly hotter, with listening position just barely off the Acoustat beam, and flat response in the highest frequencies.  The previous speaker position is marked by tape.

To reduce bass blooming around 100 Hz, I backed down the subwoofer crossover from 84Hz to 71Hz, and changed the slope from 24LR to 48LR.

I lowered panel crossover to 80Hz, not wanting to make it lower for speaker durability.  I tried several crossovers, but decided I liked the thinking behind 24LR the best.  Years ago measurements showed the LR24 as having the nicer looking impulse than LR48, but I wonder about that now and think the the Tact impulse itself has multiple cycles, so I would have to do actual impulse measurement with other program..

To increase bass in the range 22-30Hz, which looked notably sucked out in both channels, I added a bandpass filter to the Behringer subwoofer outputs, 5dB of boost, center frequency 27Hz, Q 2.2.  It could use more boost in left channel than right, but crossover currently has this set to stereo mode.

I increased delay for panels relative to bass slightly.  I get the best measurments of bass impulse when turning off crossover.  Then put sub in one channel, panel for same channel in opposite channel, can get picture of bass vs panel impulse.  I also did this method of using two channels to measure one channel for checking crossover frequency response.

Resulting delay ia now 0.85mS in right channel (was 0.70) and 0.75 in the other (there is an extra 0.10 mS delay for right subwoofer as compared with left, same as before).  This is consistent with having listening position closer to the front, so relative distance to subs from listening chair has increased slightly.

I also tried dropping the bass level from -7.0 to -8.0, but quickly decided I wanted more bass.  Funny in the crossover picture the sub bass dominates, but that's the way it sounds best.

Monday, November 14, 2011

Coherent Imaging and making the speakers disappear

More than two months ago I moved my supertweeters out of the living room.  They were contributing to a gridlock which made it impossible to move the Acoustats.  Since I am now (since early this year) listening from a position much closer to the Acoustats, it was seeming like they might be too far apart for the close-up listening position.  The angle between the speakers from my head was more than 60 degrees, and while I was still getting a center image, beyond the center things seems a bit vague,

Finally on Sunday evening I started moving the Acoustats laterally in toward the center.  First about 5 inches in on either side, then a few more inches which reached the maximum point I could move them inward because of my electronic equipment.

The new position also allowed me to move the listening chair even closer in and still get a stable center image, and moving closer in gave me nicer bass.  The nicer bass is because I am closer to both the Acoustats and the subwoofers, and because I'm moving away from the center of the room where all room modes have their main cancellation and there is a big bass suckout.

But there were two problems now.  The image started to get a shrunken quality, with the soundstage no longer seeming life sized.  And now, some instruments seemed to be playing right from the speakers themselves.

To fix those problems, I moved the Acoustats slightly back and slightly to the side.  Because of electronic equipment, notably the tuner and the MSB PAD-1 which converts the tuner output to digital, I couldn't move the speakers back more than about 4 inches.

But that 4 inches made a magic difference.  Now the speakers themselves were not longer clearly the source of as many instruments as before.  Instead, the position of those instruments moved forward, to the the same depth as the center of the image.  Thus the center of the image was no longer by itself, there is now a right-center and left-center, and much of the music appears to be coming from a plane about 3-10 feet back from the speakers.

So I'm glad I started these speaker moving experiments because I now think I may have some of the best imaging I've ever heard.  I plan to move some of the equipment so I can move the speakers even farther back from the listening chair, and possibly more the the side as well, for an even better, more lifesized image.

One thing very peculiar was that I need to dial in about 0.23ms of right channel delay to make the center image work.  Either that, or move physically closer to the left channel, so that it seems I am way off center.  I dial in this overall delay very conveniently using the Tact, though it can't be correct that way as it affects both subwoofers and speakers alike.  I tried muting the subwoofers, and I still needed that 0.23ms of delay.  Notably when I muted the subs, I also noticed that the notes in the bass line for Spanish Harlem began to sound equal, though at a much lower level.  That's very strange also because when I muted the subwoofers I have very little bass response below 85Hz where the subwoofer crossover is, and that is basically where all the first 3 bass fundamentals are.

The need for delay is very puzzling.  It might represent some early reflection, or some difference between the speakers, such as the fact that I replaced the 40 uF cap on the right side with a nice 630V solen film capacitor, but haven't made that same change to the left side.

***** Update

The next day, the need for a 0.23msec right channel delay disappeared.  I was listening to a Cactus Pear recording in which violin sounded slightly to the left, and piano more toward center.  I was thinking at first this was a demonstration of the wider center image (including left center and right center) I bragged about yesterday, but to be sure I tried headphones and realized violin (played by Stephanie) was supposed to be in the center.  Dialing back the delay to zero fixed the problem.  I recalled some of the songs I had listened to on Sunday night, and there too the need for right channel delay disappeared.

I have to believe this was either a temporary threshold shift or some similar problem with my own hearing.  It's true I was listening fairly loudly, I noticed I had heated up the Aragon amp to 150 degrees F.  But I noticed the need for delay before cranking up the volume.  Maybe it had to do with my cold, or a temporary earwax configuration.

Sunday, November 13, 2011

The Bass Line on Spanish Harlem

One of the great mastering engineers had a suggestion for tuning the bass response of a monitoring system.

Spanish Harlem on the Rebecca Pidgeon album by Chesky.

One thing for sure it is a very appealing track I don't mind listening to over and over again.

But hearing the first 3 bass notes, I wonder "are they actually supposed to sound the same loudness"?  I've now tried all my systems and the Koss Phones.  All but my Bedroom system (which was actually tuned using Spanish Harlem...) play the notes in slightly increasing loudness, with the third being notably louder. The third note also has an especially "open string" quality.

Now I see what the notes are, they are G, B, D.  Indeed, it seems like the D is the open bass string D, and it should sound naturally louder.  Even if the notes were somehow played exactly the same level, we should hear the D louder due to increased hearing sensitivity.  Now, finally, I'm looking at the opening in Wave Editor, a nice Mac program, and I see that the three notes are clearly in increasing loudness, with the third being way louder the the previous two.  Here are the peak levels:

G1: -23.2dB (49 Hz)
B1: -17.6dB  (61.7 Hz)
D2: -14.1dB  (73.4 Hz)

Even those numbers, suggesting a 9dB increase, don't do the view in the Wave Editor window justice.  The D is way louder, also with the sustained ring that comes from being an open string.  The opening G looks pitifully small.

Could it be I've mistuned my bedroom system for this?  The bass line does sound very nice, probably the nicest, on my bedroom system, which has highly hand-tuned parametric filters intended to provide smoothly increasing room gain down to 16Hz.  (Has good response to 13Hz, by the way, with SVS 16-46 sub.)

Then again, maybe that was the idea, to get nice bass boosting room curve.  It turns out that actual flat response falls flat.  A bit of increasing low end boost, seems to sound the most natural (not to mention powerful) for some reason.

In kitchen I can play the synth bass in Garage Band and the 3 notes sound like identical loudness.

Saturday, November 12, 2011

Parasound HCA-1500A quiescent

After about 2 hours of idling, the Parasound HCA-1500A is drawing 65 watts.  Heatsinks seem to be measuring max temperature around 108 degrees F, mostly mid 100's, measuring through the highly ventilated top cover, might actually be a bit higher with cover removed since cover may be confusing my IR probe.

After overnight idling, the draw is 63 watts.  Heatsinks are like before, with max around 107.

At low levels, Aragon amp is Class A

I was surprised to see power consumption hovering around the quiescent 160W (+/0 4W) playing KPAC radio at medium level today.  I had Tact volume set to 82.1 and Behringer level set to -6.1 (bass is at -7).

If power consumption doesn't increase, it's operating like Class A.

This might not be a surprise with many speakers, but the Acoustat 1+1's pull voltage and current unlike just about anything.

My Aragon 8008 BB Warmup

Time, total amplifier power consumption, emitter resistor voltage
1min, 140w, 21mV
2min, 160w, 25mV
3min, 175w, 28mV
4min, 186w, 29.5mV
6min, 192w, 30.3mV
8min, 190w, 30.1mV
10min, 187w, 29.8mV
?, 175w, 27.9mV
?, 171w, 27mV
20:30, 168w, 26.5mV
30, 158, 26.1mV

All measurements taken from right channel (what Klipsch calls "outside" channel).  While Klipsch information suggests inside channel is to be biased higher, I measure higher temperatures (by about 3 degrees F) on the outside channel, therefore I suspect my early production unit has the same bias (much higher than Klipsch specified) in both channels.

Beyond 30 minutes, the voltage continues to fall to about 24.5mV, the starts rising again toward 26.5mV peak.  After 10 hours of idling (with covers on) there is still a little oscillation, but it's close to 160w and therefore presumably 26mV.  The temperatures in the middle of the heat sink measure between 130 and 136 degrees F, it gets cooler toward the edges, down to 125 or so (but possibly also measurement error, given that my IR probe has some width function.

Thus, there appears to be a damped oscillation, the "mass" effect probably coming from thermal mass, and the "loss" coming from convection.

When playing music at moderate level, the heat sinks do not appear to get much beyond 136 degrees, and in fact cool down with music at moderately low level.  I haven't yet seen temperature above 137 degrees.

Now Klipsch called for 8mV for inside channel, and 12mV for outside channel, but that doesn't fit with my numbers at all.  And in fact it doesn't seem to fit with the Klipsch specification of 120w power consumption for 8008 mkII (same as BB) amplifier.  Looking at my numbers, an easy extrapolation shows that 120w would correspond to emitter voltage of 17mV.

From that number, or any of these numbers, given the assumption of equal biasing in both channels, we can figure out how much power the output transistors are dissipating at idle and therefore how much the rest of the amplifier is dissipating.

I'm going to use one of my best numbers, the 30 minute reading of 158W and 26.1mV.

26.1mv across 0.33 ohms is 79.1mA
79.1mA across 140V (+/- 70V rails) is 11.1w  (Note: I didn't measure rails. AC was 123V)
11.1w for 6 transistor pairs is 66.6w
66.6w for 2 channels is 133W
That means the rest of amplifier must have been consuming 25W, quite plausible.

Now that I can estimate the actual factor emitter voltage (17mV), I can calculate the maximum class A power for an amplifier with that level of bias current.

17mV across 0.33 ohms is 51.5 mA
51.5mA for 6 transistor pairs is 309mA
Maximum Class A average power into 8 ohms is 2RIb^2 1.53w (or 3.06w peak)
Maximum Class A peak power at any impedance is Vr(2Ib) 43.3w  (Vr is 70v)
Maximum Class A average power at any impedance is 21.6w
The impedance for maximum Class A power is V/I  113.3 ohms

That would seem to be the Class A specifications for a factory biased 8008 BB or 8008 Mk 2 operating at 120W idle.  It would have less than half of the 8 ohm Class A power as mine, but still well exceeds 1w.

On the other hand, an amp biased according the the Klipsch memo at 12mV inner channel and 8mV outer channel would do this:

12mV across 0.33 ohms is 

Thursday, November 10, 2011

Now I need ESP950 to Stax Amp adapter

They are available commercially through A Pure Sound at $140, seems like ripoff.  They will also modify your Koss with new cable and Stax plug for $150, and that seems like a lot better value.

Here's a thread about where you can buy the 5 pin Stax Pro Plug.

It's said to be the same as 6 pin microphone connector, minus the center pin.  One manufacturer is WBI.  Allied used to sell them.  It may be called "6 pin XLRM"

Koss E-90 Energizer Inside

The inside of the small Koss E-90 energizer is fully stuffed with two circuit boards.  It's easy to get the above view simply by removing the back cover.  From there, however, disassembly gets tougher, you have to remove the feet (which would presumably need to be glued back on), the volume control, and the front headphone jack to get the two circuit boards out.  I didn't feel sufficiently motivated to do that.

The bottom board appears to be mainly power supply, and ends in a row of electrolytics that couldn't be much bigger and still fit.  The upper board appears to be the amplifier, and seems stuffed with small transistors, resistors, and other parts.  Notably toward the front there are a few mylar caps which might be easily replaced with polypropylene.

Everything I've seen continues to convince me that the power supply is almost certainly a switching design (down to the choke in the back similar those seen in all switching supplies), and most likely the amplifier is a kind of switching type (actually Pulse Width Modulation) too.  One interesting feature is the ribbon cable that connects bottom board to top board.  This is independent of all the signal connections, which are quite obvious.  Is this because the top board needs many different independent power supplies?  That may be part of it, but it could also be that the top board ultimately synchronizes with the bottom board through multiple phases of a high speed clock.

Now PWM amplifiers are VERY efficient, and usually quite capable in delivering full power at any frequency, attributes that appear in the E-90.  But where they fall down is that there is not as much resolution in the upper frequencies as in the lower frequencies, because as the audio frequency approaches the switching frequency, it can only be constituted from a smaller number of up and down choices.  This is a problem with all 1-bit-like systems, including the DSD system used with SACD, despite the 2.88 Mhz sampling rate of DSD, a huge amount of digital processing called "noise shaping" is used to shift the noise (i.e. low resolution) from upper audio frequencies into the supersonic.  Earlier PWM systems attempted to use frequencies as low as 500kHz, which is obviously way inadequate.

Although Infinity and Sony made PWM amplifiers in the golden 1970's, the idea became unpopular either for real or rumored reasons.  Bob Carver has always wanted to make the highest power highest value and highest efficiency amplifiers, and said he tried to make a good sounding PWM design, but gave up, and instead went with the rail voltage switching designs he dubbed Magnetic Field amplifiers with characteristic flair (i.e BS).  The rail voltage is switched to whatever the current signal level requires, allowing the rails to stay as low as possible, therefore requiring the amplifier to dissipate as little power as possible (most of the energy going into the load instead of into amplifier heatsinks), even less than an Class AB amplifier, though otherwise it is made just like a Class AB amplifier.  Actually the idea seems a bit rube goldberg, but Carver got it to work by using chokes to store energy sufficient to kick the rails up when needed, and that's where the Magnetic Field name truthfully comes from, but reading Carver's ads you would think of something entirely different.  Since the late 1990's, however, electronic technology has advanced to the point where very high speed PCM digital systems could drive a PWM amplifier much more accurately than previous analog systems could.  So PWM amplifiers are back on the scene in a big way, even made by high end manufacturers such as Tact and Rowland Research (the latter formerly known mainly for building massive high bias amplifiers).

But now that I've also mentioned the rail switching design, it is entirely possible that the E-90 is a rail switching design also; it would provide the needed efficiency and eliminate the need for significant heat sinking just as in Carver's Magnetic Field amps.  Koss is rather mum about how their E-90 actually operates, giving only the barest of specifications to a direct inquiry.

Stax, in contrast, has gone for Class A electrostatic headphone amplifiers.  Class A amplification is the most linear and needs the least corrective feedback to work decently well.  It is also the most inefficient, consumes the most power, and needs the most heat sinking.  But many think it's the way to go for audio because it is the highest quality.  Stax is the commercial leader in making electrostatic headphone amplifiers, it's a veritable giant compared with the tiny perfectionist operations that make amps like the Blue Heaven.  That being said, even Stax is fairly small as electronic companies go, and even smaller since the 2000 bankruptcy in which all other business lines were dropped but headphones, their most well known product, were continued (thank goodness).

Stax tube amplifiers may be the most famous and popular among it's top line products, and many feel give the right harmonic balance to the otherwise slightly bright sounding phones, Stax has made transistor Class A amplifiers for headphones since the 1970's, and many feel the transistor amps offer the greater ultimate resolution and clarity.  My view is that the Stax tube amps aren't worth the money.  The tubes used do not seem to me like the right choices.  From reading many reviews (never tried one myself) it seems the tube amps seem deliberately down rez'd to soften the music.  If you want tubes, get a true perfectionist tube design like the $4995 Blue Heaven I described in an earlier post. But there is little such complaint to be made about the Stax transistor amps.  They seem honest designs, if not as far out as some made by elves.
Stax srm 1/mk 2
I have now bought a Stax SRM-1 MkII Pro for $325 (the low price coming from the fact that the seller can't actually test, but he is allowing me a 7 day trial and examination).  That is a Class A solid state amplifier similar in concept and execution to Stax's latest top-of-the-line amplifiers (which cost up to $2400), but made from 1982-1995.  It's said to be better than the current lower end Stax amplifiers, if not quite as good as the current top-of-the-line.  Picture below is from article by Ken Rockwell which you can read at this link.

Wednesday, November 9, 2011

More thoughts about Koss ESP950

I've been reading this great thread about headphone waterfall plots.

The author has measured Koss ESP950 and likes them a lot.  They have very flat and smooth frequency response compared to most, and very good lack of stored energy in the upper mids and highs (just some tiny wriggles, similar to but possibly even better than most Stax electrostatic headphones, which are typically much more expensive).

There is considerable stored energy in the lower mids and upper bass.  This is possibly attributable to the less-than-open enclosure.  The author strongly disagrees with the proposition that the Koss is bass weak, in fact he thinks it has a slighly dark sound, partly from the orientation of the stored energy, and partly from rolloff in the extreme highs.  However, there is rolloff in the very deepest bass.

He believes the ESP950 sound much better through a midrange Stax transistor amplifier, the 323 (which sells for about as much as the entire ESP950 package, with other stax amplifiers selling for far more).  He feels it opens up the sound compared with the Koss E90 amplifier, however waterfall plots show essentially no difference in the mids and highs.  There is one measured difference, a significant extension of the deepest bass.

I've studied some response curves, and decided Koss would sound a tad better with a mild boost starting around 2.5k, and possibly more up higher.  Tonight I tried 2dB and 3dB boosts at 3k, which is the lowest my Tact digital pre will do an easy treble boost from.  These boosts are surprisingly subtle, but both seem to make the sound lighter and have a more open quality, more like my old Infinity ES-1 headphones.  I'm sticking with the 2dB boost for now on the least harm principle, though 3dB might actually be better.  And guess what, bass instruments sound less muddy also.

One thing nice to know is that apparently the Koss E90 doesn't do any special equalization.  Other than the deeper bass response shown with the Stax 323, the frequency response is the same.

Elsewhere, I've been reading many many blogs devoted to making electrostatic headphone amplifiers (usually for Stax, but often for the Koss as well).  Often these amplifiers are very elaborate, and even the parts cost might exceed the price of all but the most expensive Stax amplifiers.  There is an enormous and unexpected number of websites and articles devoted to such projects.

Above is picture the first tube amplifier from about 10 years ago by Kevin Gilmore, who went on to design others.  He complained that the Stax SRM-T1S tube amplfiers use 6FQ7, which isn't really up to the job.  He uses 654A's.

Above is pictured the prototype of the famous Blue Hawaii amp also designed by Kevin Gilmore.  It's a hybrid transistor/tube design with EL34 outputs with current source loading, and power supply in separate chassis, and the whole package looking mind boggling.

Above is the commercial version of the Blue Hawaii and another electrostatic amp.  Kevin Gilmore also designed the KGSS (the KG is Kevin Gilmore again) can be ordered from HeadAmp!  Well, actually the KGSS does not appear on the order page, but the Blue Hawaii does, as does a fancier but ultimately not as serious looking all tube amp, the Aristaeus.  You don't actually order these because they are cheaper than currently available Stax units, the Blue Hawaii, for example, appears to run $4995.00, and they want 1/4 down payment up front, that's about double the price of the top Stax tube unit, but it looks worth the extra cost in sophisticated parts and design (read the Blue Hawaii link above to hear Kevin Gilmore describe it).  HeadAmp says they make their own products in the USA and can customize just about anything.  I suspect they'd be able to build a KGSS if you really wanted it.  It appears that unlike Gilmore's prototype, the commercial Blue Hawaii has two pairs of outputs for two heads.  Four EL34 tubes in all.

One much simpler change involves merely changing the AC wall wart.  The Koss E90 amplifier is powered by a small 9VDC 1amp wall wart.  This provides rather high impedance power.  Unloaded, it actually puts out 12VDC, though the constant draw of the Koss brings it down.  Some people report significant improvement using lab grade power supplies instead, with DC voltages as high as 12V.  I'm wondering if lower impedance power might help with the bass response.  Most E950 tweakers, however, think a whole new amplifier is needed.  Removing the back felt is said not to be a good idea, that's the only defense against dust getting on the diaphram, a potential cause of the infamous E950 squealing that may be the number one reason units are returned to Koss.

I got my Koss a nice Stax HPS-2 stand (I was worried about spilling a drink on the coffee table and mucking them up) and Stax CPC-1 dust cover, very nice.

Class A power from Class AB Amplifiers

I started a thread about calculating Class A power from Class AB amplifiers at DIYAudio, one of my favorite sites.  It was inspired by the fact that I calculate the Class A power from my Aragon 8008 BB as 3.58 watts per channel into 8 ohms (as described in an earlier post).

(Note: the full Class AB power is 250 watts, and about 500 watts into 4 ohms, and even more into a still lower impedance, so the amp has plenty of power available, but the question is specifically how much "Class A" power the amp can produce with all transistors operating in a linear range, therefore having the most quality)

Meanwhile, Mondial and Klipsch (the first and second manufacturers of Aragon amplifiers) claimed it had a maximum of 26 watts Class A power.

Much lively discussion ensued.  It turns out my calculation of the Class A power into 8 ohms was correct.  Furthermore, if my amp used the lower bias suggested in a memo from technical support, it would have 0.8 watts of Class A power into 8 ohms.

And it turns out that Mondial and Klipsch were correct also, at least for one channel of an Aragon 8008 BB...but misleading.  It's quite possible to get 26 watts of Class A power from a factory specified Aragon 8008 BB, but you must use a higher impedance than 8 ohms, probably more like 50 ohms.  So this is not really a relevant specification, but rather a way of crafting a technically correct but deliberately misleading specification.  Also, given the lower bias, it appears likely their "Class A" power was actually peak power, not average power.  So two tricks were used to craft a misleading specification.  And possibly two more tricks, read on.

The most reduced formula for Class A power calculation is 2RII where I is the fixed quiescent idle bias current, and R is the load resistance.   Since this I could be confused with the current-available-in-class-A, which is twice as much as the idle bias current, I've now decided a better description is 2RIb^2 ("two R Ib squared").  This is an algebraic reduction of the more transparent calculation (based on P=I^2R, Ia=2Ib, and Pa=Pp/2):

2*Ib  *  2*Ib * R / 2

This formula yields "average" power (sometimes incorrectly called RMS power, see the link above for some discussion of that), that's the reason for the final 2 divisor (from Pa=Pp/2).  Peak power would be 4RIb^2.  So obviously increasing R gives higher power and reducing R gives lower power.

But the increasing the R value runs into a limit when the full Class A current available to the load (2I where I is idle current) causes the voltage across the R to reach the maximum voltage available from the amplifier.  In the case of my amplifier, that occurs around 65 volts.  Given maximum current available in Class A, 0.946 amps, and maximum voltage, 65 volts, it is straightforward to calculate peak then average power:

Pp = 0.946 * 65 = 61.49W

(R = e/i = 65 / 0.946 = 68.71 ohms)

Pa = Pp / 2 = 30.75

OK, so my 8008 BB seems to exceed the factory "spec" of 26Wpc maximum Class A power (into specially chosen resistance) by just a tad.  But the amplifier idle bias current recommended by tech support is more than 2 times lower.  So an amplifier with that level of bias could not meet factory specification even with any specially chosen resistance.  There are two possible interpretations:

1) The factory idle bias current was actually much higher than specified in the memo (shown in earlier post 12mv inner channel and 8mv outer channel), and just short of what my amplifier has.

2) The factory "maximum class A power" specification was actually a peak power specification as well as specially chosen resistance specification.

I think I believe interpretation (2).  Assuming that is correct, what would would the bias current need to be?

Pp = 26W, therefore 2Ib = 26 / 65 or Ib = 26/130

Ib = 0.20a

Therefore emitter resistor voltage for 6 transistor pair 8008bb would be

0.33 * 0.2 / 6 = 11mV

Just a tad below the "inner channel" specification of 12mV emitter resistor voltage recommended by Aragon technical support.  But it seems the outer channel specification of 8mV would never make it, nor would the same values get there with the 4 transistor pairs in an Aragon 8008 ST:

0.33 * 0.2 / 4 = 16.5mV

So even with two tricks used to get the maximum specification, it only applied to the inner (left) channel on an Aragon 8008 BB, unless the bias was set higher than specified in the memo (which is possible, it's quite possible that tech support recommended an especially conservative bias to prevent subsequent failures for people who already had one problem).


My Aragon 8008 BB has 3.58 watts Class A power into 8 ohms, and 1.79 watts into 4 ohms, and therefore meets the "first watt" requirement popularized by Nelson Pass, whereas a factory specified unit would not.

My amplifier is biased more like an Aragon Palladium amplifier (just slightly more, actually).  But since those amplifiers are bridged monoblocks, and mine is not, that actually gives me another advantage in Class A power at 8 ohms and less, since each half of a bridged amplifier sees only half of the total load.

So mine is actually better than an Aragon Palladium.  It's an Aragon Platinum!

Saturday, November 5, 2011

(unconfirmed) Aragon Bias Adj Instructions

Here are the Aragon bias adjustment instructions, from some unconfirmed tech support document.  See my very important (of course) comments below:

2000 & 3000               4-6 mV EACH CHANNEL

8008ST & BB             INNER CHANNEL-12mV             
and MKII                   OUTER CHANNEL-  8mV                  

8008X3 & X3B    8-10mV EACH CHANNEL                   

8002                            8-10mV EACH CHANNEL                   

PALLADIUM        II            INNER CHANNEL-25mV             
& 1K                          OUTER CHANNEL-20mV               
                                    W/ TOP COVERS ON-HEATSINK TEMP-118 DEG. F.

4004                            6-8mV EACH CHANNEL                
2004                            4-6mV EACH CHANNEL    

DIA150,A125X5       6mV EACH CHANNEL                


I realize now this is a very ambiguous document.  What exactly does "run amps under load" mean, how much power output, continuous sine or music test, and into what load or efficiency speakers?  And what is the "operating temperature" ?

This is far more important that a reader might assume.  The heatsinks in the 8008 and Palladium series amplifier have very large thermal mass.  If they are heated up sufficiently, 2 minutes is no where near long enough to cool down enough to approach the stable bias level.

An hour or more  might be required.  While I determined my outer channel emitter resistor voltage approaches 26mV in a long term idle warmup, when I run my amp for 30 minutes at moderate level into killer load (Acoustat !+1's) it gets hot enough to roll the voltage to 14mV for a many minutes; I watched it for 15 minutes or so and it hadn't significantly moved upwards.  This is because high heatsink temperature causes the bias compensation transistor to cut back the idle current to some minimum level.  Only when the heatsinks have cooled down sufficiently is the normal idling bias restored.

So if you start out with heasinks at a relatively high temperature, and allow only two minutes, what you may be setting is actually the minimum bias level, not the long term level.

Friday, November 4, 2011

Up and down and up with Aragon 8008 bias level

I've now watched the Aragon bias level over more than an hour.  It may be on the high side of normal for this amplifier, possibly due to long term drift of some kind, but the bias circuit seems to be working and the high bias may be part of why the amp sounds so good.

Within a minute, the outer channel bias is at 21 mv (NOTE: I mean the voltage measured across the 0.33 ohm emitter resistors, I use the term "bias" for that measurement in this post.)  20mv is specified for a Palladium amp, which has the same size heatsinks; 8mv is specified for the 8008 by Klipsch tech support, but I'm wondering if these specifications aren't deliberately conservative.  At two minutes, it's roughly up to 25mV.   At that time, there is about 160 watts total quiescent power draw, same as I measured after leaving on overnight (though, based on later data, I believe that corresponds to 26mV long term average bias).  24mV is the specified value for the inner channel on a Palladium, so it can't be too bad, if actually being used the amp will dissipate much more power than that.

From there, the bias keeps rising until it reaches a peak of 30.3mV at 6 minutes, with total power draw of 192 watts.  It slowly starts falling, reaching 29.8 at 10 minutes with 187 watt draw.  At 20 minutes, it'd down to 26.45 with 168 watt draw.  At 30 minutes, it's down to 26.1 with 158 watt draw.  Here my notes stopped, but I recall it got down to about 24mV (140W or so) and started rising again, peaking just over 26mV and falling again.

Apparently there is some oscillation caused by the thermal mass of the heatsink between the power transistors and the bias regulator transistor.  Over many hours, as every part of the amp rises to its long term most uniform temperature, the oscillations should largely damp out, theough I recall still seeing a few watts fluctuation around roughly 160 watts quiescent after 10 hours of idling.  There may also be some effect caused by airflow from my HVAC system cycling on and off.

When the amp is actually operated, as I discovered last night, the bias cuts back.  I observed 14 mV bias after playing music at moderately high level for 30 minutes, at which point quiescent power draw was below 140 watts.  That must help the heatsinks from getting too hot.

All in al, I think this is about right for me, about the highest bias possible for this amp, but probably enough downward bias regulation and heatsink capacity to handle any loads I will actually throw at it.

If it becomes a problem, say if I can't party all weekend without tripping thermal cutout, I can adjust bias downward to the specs suggested by tech support, which must have been intended to handle exactly that sort of situation.  I do have an above average line voltage and far below average efficiency speaker.  But I don't tend to listen with average levels above 90dB, and even continuous use on my part is not likely to cause rise to cutout temperatures, especially because I run the bass below 84 Hz on my subwoofers and not the panels.  (When, years ago, I caused my Parasound HCA-1000A to thermally shut down, I was not only running the bass on the panels I was attempting to increase the highs with some equalization.)

Clearly the amp should not be left on all the time, it wastes a lot of power, shortens lifespan of parts, and serves little purpose because the amp warms up to full bias anyway within 2 minutes.  But my fear that amp still hadn't reached full bias after 30 minutes of operation yesterday, would rise to 160W overnight, and then just keep on going upward relentlessly were unfounded.  Actually, the amp reaches full bias after about 2 minutes of operation, overshoots by 25%, and then has some damped oscillation staying within 10% of full (26mV) bias, and cuts back bias by 50% under heavy use until the heat sinks cool down.  These are indications of well thought out design, not failing circuitry.

Now I can also calculate the amount of linear Class A operation.  26mV across 0.33 ohms is 0.0788 amps.  The total current across 6 pairs of transistors is 0.473 amps.  Twice that can be delivered to speaker in Class A: 0.946.  Peak power into 8 ohms is 7.16.  RMS is 3.58.  While the amp is playing, it heats up and the bias, and therefore the Class A power, drops a bit.  If I cut the bias back to tech support specs, I'd get less than a watt in Class A.

Wednesday, November 2, 2011

Class A Watts

Everybody seems to get this wrong.  Nelson Pass explains a hypothetical case very clearly, but when it comes to his own amp, he is wrong too (OK, just very slightly wrong, you could call it rounding error compared with the liberties most manufacturers take).

I think what's happening is there are two definitions of Class A.  The textbook definition (linear his credit Nelson Pass uses that one), and the stretched to cutoff definition, and most people use that.

Start with my Aragon 8008BB idling at 160W, with heatsink temperature at 132 degrees F (I'm sure it shouldn't be hotter than that).  First of all, since this is a stereo amplifier, that means the idle is 80W per channel.

The RMS output maxes around 45V, which means the rails must be about 67V, about right for a 250W into 8 ohms amplifier (which it is, the 200W spec into 8 ohms is conservative in order to allow the doubling into 4 ohms).  Update: the PS schematic says 70 volts, I will correct the following analysis for that.

Assuming that all the quiescent power is dissipated in the output transistors (which would be a limiting case maximizing bias current and therefore Class A operation), this means the current running through from the top rail to the bottom rail for each channel is 80W / 140V or 0.57 amps.  If you simply summed all the transistor currents, you'd get 1.14 amps, but I don't think that's a helpful way to think about this, since half are in series with the other half.

In Class A operation one side goes from 0.57 amps to 1.14 amps while the other goes to zero.  So the peak output current in class A is 1.14 amps.  That makes the Class A peak to peak power 1.14 * 1.14 * 8 or 10.4 watts peak to peak, or 5.2 watts RMS.  Very far from the 26wpc Aragon claims.

I suspect this is because cutoff to zero doesn't actually happen on one rail side until the other side is conducting considerably more than 2x the standing bias current.  So Aragon/Mondial/Klipsch hooked up scope, and observed the level where cutoff to zero was happening on one side.  But this is non-linear operation to cutoff, and wouldn't work without a push-pull feedback amplifier design.

But it gets worse.  Not all power is dissipated in the output transistors.  The actual bias specified for the amp (in an unconfirmed memo at DIYAudio) is 12 mv across 0.33 ohm resistors on the inner channel, 8mv across 0.33 ohms on the outer.  That would mean, for the inner channel, the bias current per transistor is 0.036 amp, or 0.22 for all six on one side.  That would mean 0.44 amp peak class A current, or 1.6 peak to peak class A watts, or 0.8 RMS Class A watts.  BTW, the power transistor dissipation for that channel would be 30.8 watts, with 20.5 watts for the other channel, for a total consumption (not including other circuitry) of 51 watts.  It is hard to imagine the remaining 109 watts dissipated elsewhere, so I'm sure my amplifier must be drawing more bias current than that!

My amp bias seems may be similar to the  Palladium II according to the memo, which is specified as 25mV for the inner channel, which would correspond to bias per rail side of .456 amp, and therefore (assuming +/- 70V rails) per-channel dissipation of 64 watts in the power transistors of that channel, and 51 watts in the outer channel having specified bias voltage of 20mV.

Update: it looks like my amp is indeed close to the Palladium specification.  In the outer channel I noticed bias voltage around 22 mV after 5-10 minutes of warm up when the total power dissipated was 166 watts.  Unfortunately, not thinking them important, I didn't write these numbers down, so this might be a somewhat inaccurate recollection.  Later, after playing the amp for a half hour, and 20 minutes of stabilization, the recorded bias voltage was 14mV with power of 132 watts.  I didn't measure the 6 hour warm up bias, where the quiescent dissipation reaches 160W, and possibly having the cover off for measurement affected the results, but 20mV +/- 2mV would be consistent with these findings.  The palladium specifications would account for 115w of quiescent power, the where does the remaining 45w go?  Power supply transformers and filter caps, and two large driver boards could easily account for that.  Solid state preamps can use as much as that.  Also I may have larger rail voltage than spec, since my AC power is around 123V.

I'm thinking that the memo from Tech Support does not necessarily reflect what the factory did.  Tech support was possibly giving out more conservative bias setting numbers than required for two reasons: (a) concern that repair shops wouldn't do the proper days long burn-in as done at the factory, and (b) those people who have already had trouble with their amps, which is likely why they would need servicing, should probably use lower bias to increase the safe operating area of the amp.  In short, Aragon didn't want to have to see those amplifiers again.  On the other hand, they want the amps at the local audio salon to blow away everything else with superior high bias sound quality.

What I'm seeing could also be a result of bias drift or failure of some kind.  Drift seems more likely than failure.  Looking at the bias circuit in the 8008 schematic, there do not appear to be any electrolytic capacitors which would otherwise be a potential source of partial failure.  The bias seems to be mainly controlled by semiconductors, which would either work or be broken and not work at all.

For Nelson's hypothetical amp rail to rail bias of 0.75a which is 100W dissipation across a rail to rail voltage of 130 (+/- 65V), I get the same 18W pp or 9Wrms he does.  He gets this exactly correct in the hypothetical example, though he doesn't mention that this hypothetical example must be a mono amplifier, or that his X250 is a stereo amplifier and therefore dissipates only a bit more per channel.

Now lets think about the X250 amplifier which idles at 270W, which means 135 per channel, assuming 65V rails (which is possibly too low).  That means 1.04 amp from rail to rail.  Peak to peak power is therefore 2.08 * 2.08 * 8 or 35 watts, and RMS power is 17.5 watts.  Not far from the 40/20 he claims.
But the rail voltage is probably higher and the rest of the circuitry dissipates enough to make it more inaccurate.

Perhaps he meant the X250.5 amplifier (but forgot to say so).  It idles at 350W, which means 175 per channel, which means 1.35 amps from rail to rail, which means 58W peak to peak and 29W RMS, assuming the rails weren't also bumped up a bit from +/-65V.  Accounting for higher rails and the rest of the circuitry would bring this right in line with 40/20.  But this amplifier wasn't sold until several years after the post I linked above.

One other thing to notice is that because a square law applies, Class A power goes up as the square of the bias current.  A 1 amp bias current would produce 4 times as much class A power as a 0.5 amp bias current.

To be scientific, I should re-test the amplifiers

Now I realize that most likely I had the Parasound attenuated by 6dB by its volume controls.  Since at the time, the Behringer digital gain was set to +0.4dB, and since the Aragon has 1.5dB less gain, the new exactly equivalent digital gain setting would be -4.1dB.

While I was listening to We Want To Be Loved and Imaginary Day, I actually had the gain set to -6.9dB.  In other words, I had dialed in a relative 2.8dB of subwoofer boost.  No wonder the lows sounded so spectacular.  Later, while listening to Hotel California, I changed the setting to -6.5, which would correspond to 2.4dB of relative subwoofer boost.

I do not want to give this boost up, I think it needs at least that much.  I hadn't adjusted the sub for a long time, and with changes in EQ and listening position, clearly the new bass level is better.

So to be scientific and fair to the Parasound, I should hook it up again, adjusted with the same relative bass levels I am using now.  I'm not sure I'll bother with that.  To be really scientific, I'd use my ABX comparator.

Re-setting the panel level

I've been agonizing over finding the correct level setting for the Aragon which drives my acoustats.  I could have just used the level I had been using for the Parasound (+0.4 dB) adjusted upwards for reduced amp gain of the Aragon (+1.5dB) except I had been attenuating the Parasound by some unrecorded level, and I made the big mistake of twiddling the Parasound gain adjust knobs to maximum for frequency response testing before testing their previously set level.

Yesterday I had figured I set the Parasound gain to somewhere between -6 and -12dB.  I figured I wouldn't have done 6dB, because I know that is a maximum series resistance condition.  So 12dB?  That sounded too high.  I was beginning to think I had used 10dB.  But that's an odd one, the infinite repeating decimal of 3.33.  I wouldn't have done that, would I?

Meanwhile, I was listening to various music, and readjusting the level to suit each one.  Last night I decided Hotel California didn't sound clear enough in the mids, so I increased the level from 6.9 to 6.5 (neither of which made much sense thinking about how I might have attenuated the Parasound, so I worried about that.)

For 6dB previous attenuation level on the Parasound, the new level would be 4.1dB.  (I tried that for awhile but thought it too thin.)

For 9dB attenuation, the new level would be 7.1dB (close to the 6.9dB level I had been using for Imaginary Day).

For 12dB attenuation, the new level would be 10.1dB.  Thinking I might have chosen 10dB attenuation (despite the numerical difficulty), I listened to Fleetwood Mac Rumours DVD-Audio (which sounded better than I'd ever heard it) at 8.1dB.

Mind you in making these adjustments what I'm really doing is setting the relative bass level, because that is saying constant.  At present, the supertweeter is not connected (and its level hardly matters anyway).  The subs cross over at 85 Hz.  There is no simple way to set sub level, the response below 300 is highly muddled due to room modes and there are peaks and dips as large as 20dB, so you can't just set a matching level at some particular frequency or pair of frequencies.

Of course, there's no guarantee that I set the level of the Parasound correctly, it's just that over time you readjust until it's about right and sooner or later forget to make any further adjustments until something like this happens.

The last thing I listened to last night was the Rebecca Pidgin Spanish Harlem track with acoustic bass.  A famous mastering engineer said he always used that to set bass levels; you want the acoustic bass notes equal in level.  At the 6.5dB level I had gravitated toward, the third (highest) bass note sounded too loud, way too loud.  So then I dialed in 8.1, which sounded considerably more natural.  Of course I may also be fighting room nodes at the high bass note, I have often noticed it sounds loudest in the living room, so this is hardly a perfect test.  Also, the third note sounds like an open string, which inherently sounds louder and different.  But I didn't think it should sound like what I heard at 6.5.

Now this morning I thought I'd go back through this website and find out if I had ever recorded the Parasound attenuation level.  I had set it when experimenting with attenuating all amplifiers to reduce noise from the Behringer, which was most noticeable on the supertweeter but measureable elsewhere.

Indeed, after experimenting with 3dB Harrison Labs attenuator, which I thought made the midrange sound funny due to excess loading on the Behringer, I ultimately set the level via the Parasound level controls to 6dB, and that now seems to be the last and only documented case of setting the Behringer level control.  At the time, I opined that 9dB would be the max level for attenuation, to keep the Behringer from getting into its more distorted upper output level range.

So now it appears that if I wanted to duplicate the previous situation exactly, I'd now set the level to -4.1.  But that sounds thin, and I may like 8.1 better.  Effectively, I'm adding 4dB boost to the subs compared to previously.

Quiescent Power: 158W Max Heatsink Temp: 133 degrees F

Those are the measurements this morning.  Last night I hooked up Kill A Watt meter to Aragon.  When turned on without playing, the wattage started around 90, quickly rose above 100, and then slowly up to 170 or so.  Finally, as the amp warmed up, the wattage fell to just above 150.

The heatsinks seem a bit warmer than yesterday morning, with a maximum around 133, the average is around 132, dropping to 128 at the sides.

Tuesday, November 1, 2011

Aragon 8008 BB...A Groovy Class A Amplifier

On Halloween 2011, after all the kids had come and gone, I finally finished measuring and then hooking up my newly acquired Aragon 8008 BB amp.  At first, it seemed a trifle less detailed than the Parasound HCA-1500A I had been using.  And I had considerable difficulty setting the proper level, as during the set up process I turned the volume controls on the Parasound so had no proper measurement of the previous as-adjusted gain level.  I guessed I had the Parasound attenuated by 6 or 12dB, or somewhere in between.

It didn't take long, however, to realize how wonderful this amp sounded.  Like the Krell FPB 300 (which is sitting in the corner waiting to be repaired) it has a wonderful 3-dimensional deeply layered soundstage.  The soundstage is also wider to the left and right than the Parasound, and phase tricks in the recording (as done frequently on the first album I listened to, We Want To Be Loved by Grouse) became mind-blowing.  The mids and lows are palpably real, percussion impact is incredible.  I can feel whacked back into my chair.  Deep bass was surreal.  And this despite the fact I cross over the Acoustat 1+1's at 85 Hz to self-amplifying SVS PB13 subs, so you wouldn't think the amplifier for the panels would make that much difference on deep bass.  But the improved bass seemed to be there at any levels I set the Behringer digital gain adjustment to.  Finally, while listening to Hotel California on DVD-Audio I made the last gain adjustment, to get the highs in proper balance to the subwoofer bass, but I only raised (?) the mid level by 1dB.  In between Grouse and Eagles I listened to Imaginary Day by Pat Metheny on DVD-Audio.  This amp sounded so good I just couldn't quit listening until 4am.

I left the amp turned on overnight, and found the heatsinks to be 130 degrees F in the morning, measured with my IR probe set to "max".  That surprised me at first, but would be about right for an amp biased up to 26 watts class A, as Klipsch technical support said.  Nelson Pass designs his amplifiers to have 55 C temperature at the heatsink fins.  He says that is 60 C at the transistor itself, and you can run hotter, but he prefers not to.  55 C is 131 degrees F.  It looks like whoever last set the bias on this amp, either at the factory or elsewhere, knew what they were doing.

I didn't mean to suggest that this is an engineer's idea of a true Class A amplifier, but 26 watts of Class A power is a lot compared with most Class AB amplifiers which have just a few watts of Class A power.  Even Nelson's Class A amplifiers don't stay in Class A up to maximum output.  Only my Krell comes anywhere near this.   The Krell plateau biasing maintains class A operation up to full rated output, 600W per channel, into 4 ohm loads.

Stereophile also uses the Class A designation to describe the best audio products.  Actually, I think they only rated the 8008 ST as Class B, but IMO the 8008 BB deserves a Class A.


Before the amp was delivered last week, I had reason to worry I wasn't actually being sent an 8008 BB.  UPS indicated the shipping weight was only 68 pounds.  The 8008 ST is said to weigh 67 pounds and the 8008 BB is said to weigh 76 pounds.  The ebay seller told me he would take the amplifier back if it proved to be the wrong model.  The first thing I did when the amp arrived was try to weigh it by the "net" method, standing on scale either holding the packaged amplfier or not.  This was not easy, as you might imagine.  But I tried this in both front entry hall and the kitchen and both times I came to a net packaged weight of 81 pounds.  That would be perfectly consistent with a 77 pound amplifier, in double (but light) boxes with bubble wrap and styrofoam pellet cushioning.  (Those pellets are a pain to deal with, and I don't blame Krell for charging an "environmental disposal fee" if people incorrectly ship an amp to them that way.)  Despite somewhat flimsy boxing, neither the boxes nor the amp was damaged during shipping.  UPS must be getting better about that.  And don't tell them about the scale error they apparently made.  I think the box must have been resting slightly on the sides of the scale.

Then after unboxing I counted the transistors while observing through ventilation holes in the bottom.  I counted 12 large (very chunky looking) Toshiba transistors per channel, just as the BB version is supposed to have.  Then I took the cover off, and verified that there are two separate power transformers, and 4 x 35,000 microfarad capacitors.  Very nice big Cornell Dublier capacitors, mounted to PCB with 4 diode bridges.  I had already seen from the eBay photos that the amp has XLR and RCA inputs.  So there is no doubt it is either the BB version or equivalent in every way.  The circuit boards look to be a nicer material than typically seen in electronic equipment.  All in all, the amplifier looks first class on the inside.  It looks fine on the outside too, though fairly basic except for nice 1/3" thick faceplate.

The top cover (which covers the transformer and capacitors) has some damping material applied.  Fortunately that part doesn't get very hot, but the damping material should be examined once and awhile; it was probably added by previous owner.  Also, the factory feet had been replaced with single felt pads.  I felt that didn't allow enough ventilation underneath, so I added 3 more felt pads to each foot.

Before moving the amp into place, I measured power output into an 8 ohm load using Tektronix scope and B&K sine generator.  I connected scope probe to hot side but didn't connect ground, expecting amp to be grounded through common grounds.  However I was wrong about that, though I was able to do clipping measurements as intended the lack of ground connection meant that high frequencies were very rolled off and therefore at lower levels in the scope display.  I found the amplifier put out 45 volts from 20 - 20khz.  But I was still concerned about high frequency roll off in the amp itself.

A couple days later when I set the amp into place, it's only a little wider than the Parasound, I did a dedicated frequency response measurement measuring both the input and output of the amplifier at 20, 200, 2k, and 20khz.  I also did the same tests on the Parasound.  I still detected more HF loss than expected, at -0.13dB.  The curve posted at Stereophile shows only about that much loss at 50kHz.  However, the Parasound measured even more loss, at -0.15dB.  Then I realized that my B&K oscillator has about 3% distortion, probably mostly third harmonic, and what is happening is that it is the harmonics that are getting rolled off more than the fundamental.  To do this measurement properly, one needs to use a very low distortion oscillator or a good filter to prevent the harmonics from contributing differentially to the response measurements.  I do have such an oscillator, my ST1410A, but it's hard to move because of all the junk in my computer room at the moment.  I suspect the rolloff at 20kHz is far less than 0.13dB, and even if it were 0.13dB, that would be less than any 44.1kHz sampled digital sources.

At 2kHz, the Parasound with volume controls turned fully up had 1.5dB more gain (8 ohms,  about 5v output).  Usefully, the Aragon takes larger input voltages to achieve the same output level, though not by much.  Unfortunately I didn't measure the Parasound before turning the volume controls up, I could only guess they had been turned down 6-12dB.  So I had to set the crossover level controls by guess and ear.  I also played pink noise and did a 1/3 octave measurement, which was looking flatter than I remembered.

During testing, I made the amp mute relay shut off output when I inadvertently inserted DC levels of about 2V.  The B&K generator has a DC offset adjustment, and when the upper output level is chosen, the offset increases by 10x.  I previously had misadjusted the DC level to 200mV in the lower level.  With no input, the Aragon servo loop maintains very low DC offset, about 1mV.  But if significant DC is applied to the input, the amp quickly mutes.  When the DC is removed, the amp takes a few seconds to unmute.  This is actually a very useful protection feature.