[DVV]

Many of us here have probably talked much about this subject.
It still fascinates me, after all these years, after trying
so many amps, after making quite a few, even after desigining
quite a few myself.

What makes an amp sound good? What is required to make it
sound better than others? How much load tolerance do we need
and/or want? What response do we really need?

Do tubes sound inherently better than solid state? How can
a relatively low powered tube amp ever hope to track real
world dynamics with power outputs such as 8-10W per channel?

And consequently, what can we do to make the amps we already
have sound botter? What to look for, what to change, where to
stop?

The floor is yours.

Cheers,
DVV

[MaxCast]

the easy answer...parts and implementation.

Good topic, DVV.  We have a few amp designers here so I hope this becomes an interesting, civil discussion.

[DVV]

the easy answer...parts and implementation.

No Max, that's the obvious answer, but not an easy one. Think - if it were that easy, we would have many more good amps than we do now, and many people could design and make them.

I'm not nit-picking here, Max, or playing semantic games, just trying to get the topic off to a good start.

I have had the chance to audition several amps made of very standard, commericially available, even mundane parts, which sounded better to much better than quite a few amps using exotic and top quality parts.

In some ways, this is not unlike the specification game.

Good topic, DVV.  We have a few amp designers here so I hope this becomes an interesting, civil discussion.

Thank you, Max. As you say, I am counting on some of those people I am sure have much to say about it, especially Dan Banquer.

Cheers,
DVV

[Val]

How can a relatively low powered tube amp ever hope to track real world dynamics with power outputs such as 8-10W per channel?

This one can't be separated from speaker sensitivity, and it points to the more general answer that the amplifier and speaker should be complementary to one another, and seen as a system. Apart from room problems, a basic incompatibility at this level (power needs, tonality) is the main cause of disillusion with sound systems, in my opinion.

Regarding your other questions, the many answers that have been proposed through the years illuminate the fact that we still don't know enough. I think Hiraga was right when he argued for a smoothly decaying harmonic distortion series, as well as those who proposed less global feedback, but I am sure more is needed. There are new ways (Halcro?), new output devices (Clayton, Bryston SST, Unison Research Unico) and more intensive detail (Curl's JC 1 and Blowtorch) that help advance the genre.

Val

[ehider]

Here is a good set of basic design criteria to design a SS amplifier to sound superior to most of what is currently available:

Choose active components and topologies that exhibit the following excellent characteristics: Fast settling times, Excellent waveform symmetry and very high slew rates.

Choose output devices (or topologies) that allow the amplifier to better control the drivers of a speaker with excellent control of back emf by virtue of an extremely high dampening factor.

Choose output devices that swing high current values so that you can use fewer devices total per channel.

Physically orient the power supply transformer so that the magnetic field has the least amount of interaction with the front end and output circuitry.

Active current source (instead of resistor) all front end active devices.

Realize that the power supply (and all the wiring, rectifiers & regulators) are actually part of the amplifier's sonic signature (just like the other active transistors). Choose power supply components after testing many different brands of components listed above (instead of just "spec-ing" by current/voltage needs and low noise specifications).

Pay particular attention to power supply capacitors and their sonic signature (black gates or Jensen four terminal caps are typically considered the absolute best sounding).

Parallel all fuses in the signal chain with high quality film capacitors.

Do not use green solder mask covered circuit boards as the poor dielectric of green solder mask ink has a slight "blurring effect" on the overall sound.

Pay attention to both the input and output wiring and it's overall sonic signature (i.e. test different wiring to determine it's sonic character on an amplifier's overall sound).

  The above general ideas for obtaining a decent sounding SS amp are typically not prevalent in any one amplifier design (YET THESE IDEAS ARE BASIC NO BRAINERS!). IMHO if these basic ideas were implemented during the design faze of more amplifiers we would have a plethora of great sounding amplifier choices.

Good god it's 2003 yet many of above basic ideas still foreign to most amplifer designers! Do they all live in a bubble!!!  

[DVV]

Hi Eheider, all,

Here is a good set of basic design criteria to design a SS amplifier
to sound superior to most of what is currently available:

Choose active components and topologies that exhibit the following
excellent characteristics: Fast settling times, Excellent waveform
symmetry and very high slew rates.

Logical. However, full of pitfalls. Sometimes, devices which look
humble on the datasheet next to another, highly touted device,
perform FAR better in real life.

Example: Sanken RETs look fantastic on paper, and indeed measure
very well, yet when in actual circuit, their real life current
delivery is modest before onset of gross distortion. By comparison,
BD 249/250C (European versions of TI's TIP 35/36C, however somewhat
improved) look almost like cripples compared to Sankens, but in real
life will deliver far more current than the Sankens for the same
distortion. Sankens are rated for 18 amps, BDs are rated for 25 amps,
yet are far cheaper and also in plastic pack.

The usual "argument" that the Sankens are rated for 60 MHz and are
therefore better than the BDs, which are rated at ">3 MHz" is only a
half-truth at best, if one uses the output stage at unity voltage gain.
Not many amps will have a full power bandwidth of 1 MHz, let alone
more. But what is often overlooked is that those BDs have a Ton time
of just 350 nS and a Toff time of just 450 nS, inherently symmetrical
and far better than most other devices costing much more.

Almost the same argument could be used in favor of Motorola's great
pair, MJ150022/15023; they look modest on paper, but perform really
wonderfully in real life. And they are well neigh indestructible, you
need a sledge hammer to really hurt them assuming you have not subjected
them to excessive voltages.

I am not promoting any obove any other, but I am hinting that much, not
to say most, audio design today is about hip parts and hip devices, not
infrequently against all reason. Not that many don't know this already.

Choose output devices (or topologies) that allow the amplifier to
better control the drivers of a speaker with excellent control of
back emf by virtue of an extremely high dampening factor.

This should be simple enough - use emitter-follower configurations,
with a predriver-driver-output stage. This yields the fastest possible
response and low distortion, unless something is very wrong in the
actual implementation, in addition to a low output impedance.

Choose output devices that swing high current values so that you
can use fewer devices total per channel.

Well now, this is the king size dilemma, because it begs the question
just hom much power do you want? If you want the power amp to be the
ideal voltage source, then you have a doubling of power output with
every halving of the load impedance. So, how low do you want to go?
4 ohms? 2 ohms? 1 ohm? 0.5 ohms? Hey, I know you are not supposed to
use audio power amps to drive welding arcs, but some people out there,
like for example Electrocompaniet, actually specify their amps at 0.5
ohms.

Next, what's "power"? Is that on a continuous basis, or in impulses?
If in impulses, how long is an impulse, after which it becomes a
short circuit? IEC recoomends 20 milisecs, but I disagree with that,
my personal criterion is 100 milisecs. This is a VERY important
question, because it will change EVERYTHING - power devices, power
supplies, heat sinks and ultimately, the very size of the amp.

My personal response to this is continuous down to 2 ohms, below
that, impulse only. But, in my view, this applies to maximum output
voltage swing before clipping, or in other words, not the simple rated
continuous power, nut the actual impulse power the amp is able to
deliver. For a nominal 100W per side amp into 8 ohms, mine wiill deliver
about 150W in impulses before clipping, so into 4/2 ohms I would
expext ideally 300/600W in impulses (T=100ms). Which is why I use 600 VA
toroids per side in what is nominally a 100W/8 ohms amp, and why some
people give me funny looks.

But this rules out single ended output stages (I don't like the sound
of the latest UHC MOSFETs). It can be shown that using multiple output
devices keeps them far more in their SOAR, and hence provides better
linearity and lower distortion, especially under high power conditions.
Next, dividing your power stage into more trannies allows you to use
the heat sinks much more efficiently.

The objective problem you do have is matching those devices. But I would
refer you to the site of SGS-Thomson, there used to be a text on
solving this problem in several rather ellegant ways (ellegant because
they are simple and effective). Unfortunately, after my recent Windows
induced hard disk crash, I have lost the addresses.

I believe in multiple output device stages. My rule of thumb is to have
as many devices as required until my nominal 8 ohm power output is about
10% of the absolute device capability. Thus, for a nominally 100W per
side amp, I'll use four pairs of output devices. Notwithstanding the
above reference, I will match them anyway.

Physically orient the power supply transformer so that the magnetic
field has the least amount of interaction with the front end and
output circuitry.

Well, I stick my power supplies inside a full Faraday cage, so I don't
have any problems like that.

Active current source (instead of resistor) all front end active
devices.

Haven't done it any other way for the last 20+ years. But you are quite
right, much of the industry still sticks in a plain resistor and hopes
for the best. When they feel like flaunting it, they actually add a
zener diode with a parallelled capacitor too. Wow!

Realize that the power supply (and all the wiring, rectifiers
& regulators) are actually part of the amplifier's sonic signature
(just like the other active transistors). Choose power supply
components after testing many different brands of components listed
above (instead of just "spec-ing" by current/voltage needs and low
noise specifications).

Amen to that!

Pay particular attention to power supply capacitors and their sonic
signature (black gates or Jensen four terminal caps are typically
considered the absolute best sounding).

Agreed in principle, disagreed in component selection. When I want the
absolute best and money is of no consequence, I buy and use Siemens
Sikorel caps. Never saw, measured or heard any better - but at a price.
Not too well known in North America, just like Jensens are not well
known in Europe.

Parallel all fuses in the signal chain with high quality film
capacitors.

Logical and recommended.

Do not use green solder mask covered circuit boards as the poor
dielectric of green solder mask ink has a slight "blurring effect"
on the overall sound.

I must admit this is my very first time to ever see something like
this. Surely not all green solder masks are of the same quality, as
they are not from the same souce?

If not green, then what? Blue?

How about sandwich boards? Copper underneath, then a layer of gold
plating, then a layer of solder?

Pay attention to both the input and output wiring and its overall
sonic signature (i.e. test different wiring to determine it's sonic
character on an amplifier's overall sound).

Absolutely! This can't be repeated often enough (unfortunately).

The above general ideas for obtaining a decent sounding SS amp are
typically not prevalent in any one amplifier design (YET THESE IDEAS
ARE NO BASIC BRAINERS!). IMHO if these basic ideas were implemented
during the design faze of more amplifiers we would have a plethora
of great sounding amplifier choices.

Aha, but implementing these ideas costs money, even for megabuck
companies buying by the truckload, like Panasonic/Technics, Sony, etc.
Given the more or less standard multiplication factor of 2.5 minumum
(price at factory door to final retail price), typically 2.8-3, it's
easy to understand why the big boys will always prefer NJR's 4558, which
they buy at 10 cents a piece at 5,000+ to say AD826AN, which they
can't have for less than $1,80; that $1.70 will in the end turn out to
be a difference of (1,70x2,8) $4,76, or double that if more than one are
needed. Furthermore, a "saving" of $1.70 on a series of 100,000 represents
a nice sum of $170.000 - have you any idea just how much advertising
space $170K can buy? And they will always rather adevrtise than use
good parts.

So they invented "Special Editions", "Precision", etc versions, which
are basically the same thing, but with better components. As a random
example, look at Marantz CD6000 KIS; instead of a weedy transformer, they
throw in a 80 VA toroid. Wow! That's progress for you, really the art of
music making. Next they copper plate the chassis. Wow again, some
discovery, makes you wonder why they use aluminium in their top of the
rande model then. Etc, etc, etc.

Good god it's 2003 yet many of above basic ideas still foreign to
most amplifer designers! Do they all live in a bubble!!!

No, they live in the modern audio industry, and every engineer has two
hawks looking over his shoulder, one is an accountant, and the other is
a PR manager.

Let's face it, the modern audio industry really cares only about sales
figures, dollars and cents, the sound of it all has been forced at gun
point to move to the back seat. Real audio still lives on on the margins
of the audio industry, with the little guys, who are still trying to do
it the right way.

But I'm glad you asked the above question, it should explain to some
people here why I have such scorn for the mainstream audio industry.
Practically all of what you said above was listed in Prof Matti Otala's
legendary texts in IEEE in 1972-1975, when he first identified TIM and
produced methods to get rid of it, and when he noted that speakers were
very complex loads, and again produced methods to grapple with that too.

Cheers,
DVV

[ehider]

DVV,

Wow! What excellent feedback to my "basic" overview. Indeed, I couldn't agree with you more about using your ears as the final judge regarding topology choices. I will say though that too many designers I've talked to are very close minded to what my mentor calls "the important three sss" (slew rate, symmetry and settling times).

Regarding capacitors, my sources are telling me that nothing compares to the unique four terminal Jensen caps, nothing! If you've done direct comparisons of your favorite caps to these unique four terminal Jensens I'd be very interested in your assessments. It's a royal pain in the ass getting those Jensens, but at this time I trust my mentor and his obsessive gurus in terms of using these in a "best sounding" capacitor implementation.

In terms of your solder mask question, I was in the circuit board industry for over 4 years. There are a myriad of solder mask choices, but the best solder mask is no solder mask at all. Many of the smaller high end audio companies go with a basic plated no solder mask board due to low volume, in house basic board making equipment. Little do they realize that these sort of circuit boards always sound the best. In my experience there is no solder mask (other than Teflon dry mask, which is damn expensive and very rare) that is worth a damn sonically. All other masks smear the sound slightly. This is easily proven by making an evaluation board without mask and one with mask. The difference in sound is a pretty big eye opener  for most that have heard the difference

I also agree with you regarding individual component costs and why the big audio corporations just don't "get it". Economy of scale tends to have manufacturers always going for the compromised cheap and dirty choice due to the fact that they mostly focus on the purchase cost of thousands of parts at a time. To me this explains why the Japanese (for instance) have never come close to a high end manufacturer in terms of a great sounding amplifier/pre-amplifier offering. But what about the small boutique "high end" manufacturers. They don't get the benefit of economy of scale with them making so few of their products per year in comparison. They are surviving strictly off the fame of high end reviews and their reputations. IMHO these companies need to take a "no holds barred" design approach and NEVER substitute a .25 part for a $2.00 part if there is a definite audible difference  Yet I see these sorts of oversights again and again  Is it really a bottom line cost, or is it ignorance and stubbornness of the designer not paying attention to the little details

[JoshK]

Great thread guys!  I have always been very skeptical of a lot of marketing and claims, even from manufacturers on these sights (although a lot less so).  As a quasi-economist by practice (not by training) this is a text book example of using marketing and differentiation to create pricing power, thus allowing profit.  This is the downside to capitalism, it inevitably creates cheaper products, which are often cheaper in many senses.

It is good to (re)read these sorts of discussions to remind ourselves of what is important when all this pseudo-scientific marketing is thrown at us to ellicit our hobby dollars.

[DVV]

Great thread guys!  I have always been very skeptical of a lot of marketing and claims, even from manufacturers on these sights (although a lot less so).  As a quasi-economist by practice (not by training) this is a text book example of using marketing and differentiation to create pricing power, thus allowing profit.  This is the downside to capitalism, it inevitably creates cheaper products, which are often cheaper in many senses.

It is good to (re)read these sorts of discussions to remind ourselves of what is important when all this pseudo-scientific marketing is thrown at us to ellicit our hobby dollars.

Not great yet Josh, but it stands a good chance of becoming one. I was sort of wondering when you'd kick in, I knew you'd be along eventually.

This kind of thread is my thing, this is what I like doing, asking the basic questions. I have always agreed with James Bongiorno that we are not getting the right answers because we are not asking the right questions. But much as I think I know of few of those, it won't work without other views, and whether I agree with them or not is immaterial, of no consequence.

The idea is to lay it down on the line and let everybody make up their own minds.

Like you, I am sick and tired of hype and outright lies we are served on a daily basis diet. Let's get some locally present talent, and we do have quite a few VERY knowledgeable people here, and analyze the topic. Some of us, such as for example Dan (Baquer) and I, have different ideas on how a circuit should like like, but we agree completely about the basics, and it's the basics I hope to cover here.

No strings, nothing is forbidden, speak your mind and let's all discuss it. And it has already started with Eheider's posts, already he has my mind whirling and spinning - and God, how I love that! Learning is such a pleasure!

Cheers,
DVV

[DVV]

Hi Eheider, all,

DVV,

Wow! What excellent feedback to my "basic" overview. Indeed, I couldn't
agree with you more about using your ears as the final judge regarding
topology choices. I will say though that too many designers I've talked
to are very close minded to what my mentor calls "the important three
sss" (slew rate, symmetry and settling times).

Slew rate is relative in so many ways. In the early 80-ies, Sansui used
to advertise really way out slew rates, like 200+ V/uS as a routine
matter. In their AES papers, they proposed the idea that the only
important slew rate was that of the input stage, so they threw in some
FETs and got wild slew rates. The fact that the rest of the amp, and
especially the output stage, didn't have a hope in hell of following that
speed didn't bother them. Yet, common sense tells you that the ultimate
slew rate you will actually be hearing is that of the slowest section of
the amp, which will determine the usable slew rate.

Furthermore, be wary of how it's declared. A common trick in the industry
is to disable the input and output filters to get better numbers. Yet, when
you listen, they are in circuit, and consequently, your actual slew rate
is of course lower.

Now, to its significance. That it is important nobody disputes, but how
much of it do you need? It has been accepted from late 70-ies that if your
slew rate is 0.5V/uS per PEAK output volt, you won't have any problems
with it, no SID (Slew Induced Distortion). Cautious people double this
just in case. So, for say 100W/8 ohms, your RMS voltage is 28.3V and
your peak voltage is 40V, so a slew rate of 40V/uS will eliminate any
problems with that.

In practical terms, if one looks at the amp as a whole, i.e. input to
output, this implies a power bandwidth of about 80 kHz or so, and that's
a power bandwidth any idiot can get with a half sane design. Being just
three quarter idiot, I have no problems getting 100V/uS, implying a power
bandwidth of about 300 kHz, and I hasten to add, since I love to stick
my finger in hip designers' eyes, with those shoddy BD 249/250C
transistors, not the sexy and expensive Sanken devices. Don't ask me
what I do with fast types, like Toshiba's 2SC5200/2SA1943.

Symmetry is obvious, no sense in discussing it. It has to be there as
much as possible.

Settling time is a nice one. Oh, I agree with that, boy oh boy, how I
agree with that, and then some! For years, I've been saying that the
Analog Devices AD826 op amp was a great sounding op amp, and people
asked how come, when there are much faster op amps out there. Yes, I'd
say, but compare their settling times, and then come back and name any
op amp in the still-sane price class which can better it.

Problems with power amplifier settling times are of course far more
complex, especially after it went through a cycle of a large power
burst into an evil load - but it can be done, it can be made so that
it settles down very, very quickly. There are several methods, which
can be combined to good effect, and here again, it's the designer's
ears which come into play, to determine which combination gives the
most natural effect. Obviously, this also depends on your actual
circuit topology too.

But what I'm missing here is even mention about memory distortion.
This is a result of a semiconductor (tube, transistor, FET, MOSFET,
IGBT) having gone through several more or less violent dynamic peaks,
which went by very quickly (say, two or three crescendos), and the
device didn't have the time to fully discharge the last before the
next struck. As a result, we have blurring, loss of attack, definite
loss of response speed and ambience. This is a VERY simplified and
not too accurate interpretation, but that's the essence of it.

Companies like Lavardine (France) and Lundahl (Sweden) did a lot of
work on it - as did I, albeit unwittingly. I mean, I was in fact
struggling with this effect without knowing what it was, that it was
an effect unto itself - to me, it was just not the right kind of sound
and I was displeased and trying to set it right. I did, and years later,
had the opportunity to read that I was in the forefront of audio
technology without even knowing it.  Life sucks sometimes, my one
crack at world fame, and I didn't even know about it.

But seriously, it is stunning when you get it right. Everything seems
to become so crystal clear, so well defined, you wonder what have you
been wasting your time on before that.

Regarding capacitors, my sources are telling me that nothing compares
to the unique four terminal Jensen caps, nothing! If you've done direct
comparisons of your favorite caps to these unique four terminal Jensens
I'd be very interested in your assessments. It's a royal pain in the
ass getting those Jensens, but at this time I trust my mentor and his
obsessive gurus in terms of using these in a "best sounding" capacitor
implementation.

Four terminal? That would be the T concept? The idea that the current
is taken in on the cap, filtered and sent out on another pin altogether?
If so, that concept is also being used by DNM (Dennis Morecroft Products,
actually manufactured by Aerovox in England) and Philips, who also make
and sell T-caps.

This does yield better filtering, or so the theory goes, but I prefer to
use a decent line filter before the juice even gets into the amp, in which
case classic caps will do more then well.

Here's one for you - with good line filtering, the audible differences
between capacitors are greatly reduced. Not gone, but shrunk down to very little. The reason is obvious - in all cases, they work in a highly
off-loaded mode, which also covers up the existing quality differences,
as they need to filter much less and act as energy storage much more,
in which case their capacity and current ratings play a far more
important part than their absolute quality. Please don't misunderstand
this, this is NOT to say that it no longer makes any difference, there
are still other issues, such as for example speed, which no line filter
will ever be able to do anything about. Just reduced.

But that I'll want to try them, have no doubt. All the fun's in trying.

In terms of your solder mask question, I was in the circuit board
industry for over 4 years. There are a myriad of solder mask choices,
but the best solder mask is no solder mask at all. Many of the smaller
high end audio companies go with a basic plated no solder mask board
due to low volume, in house basic board making equipment. Little do
they realize that these sort of circuit boards always sound the best.
In my experience there is no solder mask (other than Teflon dry mask,
which is damn expensive and very rare) that is worth a damn sonically.
All other masks smear the sound slightly. This is easily proven by
making an evaluation board without mask and one with mask. The
difference in sound is a pretty big eye opener for most that have
heard the difference.

Well, you unleashed the worm, there's no other way for me but to try
it out. Thank you for the tip anyway, it would never have occurred to
me, quite frankly.

I also agree with you regarding individual component costs and why the
big audio corporations just don't "get it". Economy of scale tends to
have manufacturers always going for the compromised cheap and dirty
choice due to the fact that they mostly focus on the purchase cost of
thousands of parts at a time. To me this explains why the Japanese
(for instance) have never come close to a high end manufacturer in
terms of a great sounding amplifier/pre-amplifier offering. But what
about the small boutique "high end" manufacturers. They don't get the
benefit of economy of scale with them making so few of their products
per year in comparison. They are surviving strictly off the fame of
high end reviews and their reputations. IMHO these companies need to
take a "no holds barred" design approach and NEVER substitute a .25
part for a $2.00 part if there is a definite audible difference
Yet I see these sorts of oversights again and again  Is it really
a bottom line cost, or is it ignorance and stubbornness of the designer
not paying attention to the little details.

I beg to differ here. Kensonic Accuphase, for example, has given me much
listening pleasure in its day, and was, in my view, in line with the US
high end of the day. However, that was some years ago, say mid-80ies, I
sort of lost touch with them, and I have no idea what they are doing
these days.

Then there was the Nakamichi Stasis power amp. True, it used licenced
Stasis technology from Nelson Pass/Threshold, but hey, that was one great
amp in its day, say around 1993/1994.

As for your above text on the Japanese, I think it's all of that, plus
a little more. To me, there's no doubt such a thing as a Japanese school
of audio design exists. They tend to all go the same way. For example, for
years, high class Japanese amps had a FET input differential amp, followed
by a cascode amp, followed by a current mirror. They ALL did that and no
other way. Sansui broke that mould with some high class engineering
(remember AU-607?) most atypical of the Japanese, and Kenwood followed suit with their LO-M5 series, and then Luxman hopped on the bandwagon.

Right about now, I feel damn stupid - here I am, explaining circuits,
when I should put them on my site and let everybody download them. I
think I'll do that by the coming weekend, need some time to find and
organize all those schematics by Sansui, Rotel, Luxman, Kenwood,
Amcron (Crown), Phase Linear, Harman/Kardon, Yamaha, Marantz, etc. Plus some nice projects from Elektor, Europe's widest read electronics
magazine (in English, Dutch, German).

Cheers,
DVV

[Dan Banquer]

Well folks, from where I sit, I have broken just about every "high end" rule mentioned and not mentioned in previous statements on this thread. My experience is about the exact opposite of ehider's, and it appears many other people.
  As I outlined in my FAQ's on my web site, the "characterization" of harmonic distortion is  very important and 99.9% of the time neglected. What order and how much distortion an amp produces is crirical to how an amp wil sound. Ehider, and others take a few minutes and read the begining of the FAQ's on my web site. www.redesignsaudio.com
  For all the discussion on bandwidth and output transistors, I will tell you this. I use Motorola MJ15003 and MJ15004 output transistors. If I remember correctly they have a gain bandwidth of "only" 4 Mhz. Well folks the LNPA 150 is less than 0.2 db down at 20 Khz, and at full power into 8,4,and 2 ohms the 3 db point is 80 Khz. Go figure.
I have found that uniformity is a crucial issue, and is the basis for the definition of linearity. Most power amps out there are not uniform, because they were never designed to be. Note there is very little talk in this business of uniformity.
That uniformity also extends to output impedance, again I deliberatly set up the output impedance to be uniform up to about 8 Khz, which is the top end of the highest prime musical tone recorded. That uniformity also extends to distortion. Distortion should be uniform independent of power and frequency in the amplifiers linear range.
UNIFORMITY, UNIFORMITY, UNIFORMITY!!!! It's the basis of the true meaning of linearity.
For those of you who might think I am just a measurement freak; forget it. I listen a lot, I am an ex musician, and son of a classical musician.
I decided along time ago that if I wanted the electronics I was looking for I was going to have go about this myself. The business for all practical purposes is just an extension of that.
I have few hints for those of you working on your own designs. If you are using a fully complementary topology, do everything that you can to make it as complentary as you can on a DC and AC basis.
This is all for now folks; if I get some time later I will share some of my experiences with parts.

[Ferdi]

What a great topic! I am remembering a lot of my half-forgotten highschool physics and am also reminded of "the art of electronics" by Horowitz and Hill that a friend recommended.

Are there any good books on these topics or is amplifier design really a dark art?

edit: poor choice of words...

[ehider]

Damn this thread is getting good! There's quite a bit of expertise here that I can't begin to compare with    

In regard to the question about the rare four terminal Jensens, they actually make an Electrolytic capacitor with four terminals (to be connected like a two terminal capacitor, each of the two terminals are tied together when you wire it up). I can't speak of all the details about this cap other than it's unique internal construction is supposedly unlike any other capacitor which requires it to have four terminals. My guru buddies recently found it and are all going crazy upgrading everthing they own with these things. Many of these guys spent hundreds (some thousands) on Black gates prior to this Jensen discovery, so these must be quite a bit better for everyone to go to the trouble to change out to the Jensens. I too changed out on my pre-amp and DAC and was very surprised at the improvement with the implementation of these four terminal Jensen caps.

Dan,

I obviously can't speak to your explanations becuase I only know what I know and am NOT a commercial designer. Unlike you, I haven't gone down the path of creating my own amplifier company. I just don't have the expertise or scope with all the little details and discovery that take years to learn (and unlearn too  ). What I do have is a long time relationship with a genius electronic designer with impecable credentials and unbelivable accomplishments. As far as I'm concerned, he is one of the greatest pre-amplifier and amplifier designers that has ever lived. He  was the first designer to make a commercial budget pre-amp (circa 1980) that was as good as the best of the best (reviewers were shocked at his accomplishement   ). Additionally, he also "helped" other groundbreaking cable and speaker companies establish themselves as leaders in their fields, yet he was the creator of their intial breakthrough products This guy is veeeery well rounded and continues to blow me away with his simplistic elegant design solutions.

I've spent the last 15 years letting my genius buddy guide me through the myriads of mistakes that most amplifier manufacturers make. His abilities are sooooo beyond me (or anyone else I've ever met in high end audio) that it's a shame he's no longer in the business. What he managed to do though is build me (and a couple other select friends) a custom solid state amplifier that has trounce every amplifier that I've ever compared them too. That goes for the top Krell, Levinson, Classe, Plinius, Pass, Rowland and Audio Research. (Of course I'm sure there may be a better SS amp such as the Halcro, but I'm not willing to sell cocaine or become a hit man in order to afford those damn things). What was a reall eye opener is that he managed to make those custom amps for a mere paltry $1600! Many of the "tricks" he used were what I outlined in my prior threads. This is my basis for a way (but NOT the only way ) to make a great sounding SS amplifier.

Please be aware that I'm not trying to say that there isn't another way to make a great mousetrap. I just know that my mousetrap beats the living poo out of any commercially available amplifier at a fraction of the cost of what I could buy commercially. If that's not a great SS amp I don't know what is

[DVV]

...
For all the discussion on bandwidth and output transistors, I will tell you this. I use Motorola MJ15003 and MJ15004 output transistors. If I remember correctly they have a gain bandwidth of "only" 4 Mhz. Well folks the LNPA 150 is less than 0.2 db down at 20 Khz, and at full power into 8,4,and 2 ohms the 3 db point is 80 Khz. Go figure.

...

According to my Motorla data sheets, MJ 15003/15004 are rated at ">2 MHz".

The fact that your amp is -3dB down at 80 KHz whatever the load impedance strongly suggests two things:

1. That your amps have a very wide and uniform closed loop response, and knowing what little I do about your work, I would deduce that you go for a nice, uniform open loop response as well, and

2. That your effective response is limited by a filter rather than the amp electronics, probably an input filter, and I would deduce a first order one too (you wouldn't want to mess with transient response, would you Dan?).

By consequence, if your actual response is to say 200 kHz, and you limit it to 80 kHz, you greatly reduce possibilities of TIM, TID, etc. If 80 kHz also happens to be your open loop response, such a filter would in effect eliminate transient intermodulation for all realistic input signal levels, since the amp would not need to rely on the feedback loop to correct for mistakes (it would be faster in open loop than the fastest realistically possible input signal).

This is getting boring, Dan, but we agree yet again. My goal is always to go for a 100 kHz open loop response, then insert an input filter at around 80 kHz. Bye-bye to transient intermodulation.

Cheers,
DVV

[Dan Banquer]

Hi Dejan and Ehider;
  My experience with caps and power amps I think pretty much follow Dejan's. about 4 years ago I was requested by a customer to install Black Gate caps where I could in the LNPA 150. I limited it to the smaller caps on the power supply regulation board and a few on the amp board. To replace the large electrolytics (33,000 uf times 4) would have cost 2400.00 at the time. I did not replace those. After test and listening I was not impressed.
  Black Gate claims better E.S.R. at higher frequencys, which may well be true. However, I strongly suspect that if you have a reputable AC line filter already installed in the amp, then the power supply caps don't have to suppress the H.F. Trash coming from the AC line. A DeZoral Brick looks like the best possible I have seen to date. Corcom, Delta, and Schurter also make AC line filters in a can which can be gotten from places like Digikey etc. These are not as "stiff" as a DeZoral but do an adequate job of filtering H.F. Trash. Face it; it's cheaper to install one of these than pay an exorbitant price for Black Gate Caps.  The other nice thing about these filters is that they slow up your initial turn on surge current, with out causing problems in the transient response of the power supply.
  Just as a reminder, whenever you use an electrolytic cap, make sure to parrallel it with either a film or ceramic ( depending upon application). I will typically use a .1 uf .
  Now to attempt to respond to Dejan's quetions. I don't remember what my open loop response was for the LNPA 150. I do remember dropping the feedback from 28 db gain to 26 db gain and the amplifier liked that very much. It snapped right in so to speak.
  My present input filter has a 3 db point of about 340 Khz. There is also a filter after the voltage gain section of the amp.  I am thinking of reducing the input filter  to 235 Khz. Your right about the output transistors and bandwidth. They are a unity voltage gain and have plenty of bandwidth for this application. I also suspect that considering the gain bandwidth of the output transistors vary with supply voltage, having a fully regulated supply keeps that pretty uniform also. I have this thing about uniformity.
  About a couple of years ago I had somebody measure TIM  for the LNPA 150 on an Audio Precision. It measured 0.02%. Given the nature of the measurement, and your typical can't get out of their own way loudspeaker. That appears to be plenty of room. I have measured the slew rate at 60 volts peak to peak into 8 ohms at 5 Khz and got 25 volts per microsecond. Remember my rule of thumb on slew rate?
  Extending bandwidth on power amps has few drawbacks. the first is you will decrease performance of reactive load handling, due to more ringing and more out of band components, the second is the amp is more susceptible to out of band interference. Over the years the old concept of out of band rejection has become very important to me.
  One last thing before I have to go. I used to subscribe to Elektor Electronics magazine. I have tried to resusbscribe but they have not responded. Can you help here Dejan? They have a great magazine and many excellent audio projects, I should know. The basis for my D/A Converter comes from one of theirs and the basis for the headphone amp I am sort of working on comes from a pre amp design of theirs.  T.S. Geisberts is one smart cookie.

[DVV]

Hi Dejan and Ehider;
  My experience with caps and power amps I think pretty much follow Dejan's. about 4 years ago I was requested by a customer to install Black Gate caps where I could in the LNPA 150. I limited it to the smaller caps on the power supply regulation board and a few on the amp board. To replace the large electrolytics (33,000 uf times 4) would have cost 2400.00 at the time. I did not replace those. After test and listening I was not impressed.
  Black Gate claims better E.S.R. at higher frequencys, which may well be true. However, I strongly suspect that if you have a reputable AC line filter already installed in the amp, then the power supply caps don't have to suppress the H.F. Trash coming from the AC line. A DeZoral Brick looks like the best possible I have seen to date. Corcom, Delta, and Schurter also make AC line filters in a can which can be gotten from places like Digikey etc. These are not as "stiff" as a DeZoral but do an adequate job of filtering H.F. Trash. Face it; it's cheaper to install one of these than pay an exorbitant price for Black Gate Caps.  The other nice thing about these filters is that they slow up your initial turn on surge current, with out causing problems in the transient response of the power supply.

Quite so. Add a good line filter, and although it cannot be cheap, it will practically eliminate line noise and allow you to use cheaper, though still good quality capacitors to even or better effect than most expensive ones alone. Furthermore, no capacitor, or capacitor bank, can ever hope to match a decent line filter at anywhere near anything one could call sane money.

I for one take Eheider's point of a $1,600 amp being an outstanding value for money - THAT'S what it's all about, I think. Value for money, a category all but lost in the modern audio world.

  Just as a reminder, whenever you use an electrolytic cap, make sure to parrallel it with either a film or ceramic ( depending upon application). I will typically use a .1 uf .
  Now to attempt to respond to Dejan's quetions. I don't remember what my open loop response was for the LNPA 150. I do remember dropping the feedback from 28 db gain to 26 db gain and the amplifier liked that very much. It snapped right in so to speak. [/quote

 Been there, Dan. Exactly the same effect, I started out with 30 dB and soon enough, 25,8 dB was shown to be an optimal value. At that value, it sounds right, the way it should sound.

  My present input filter has a 3 db point of about 340 Khz. There is also a filter after the voltage gain section of the amp.  I am thinking of reducing the input filter  to 235 Khz. Your right about the output transistors and bandwidth. They are a unity voltage gain and have plenty of bandwidth for this application. I also suspect that considering the gain bandwidth of the output transistors vary with supply voltage, having a fully regulated supply keeps that pretty uniform also. I have this thing about uniformity.
  About a couple of years ago I had somebody measure TIM  for the LNPA 150 on an Audio Precision. It measured 0.02%. Given the nature of the measurement, and your typical can't get out of their own way loudspeaker. That appears to be plenty of room. I have measured the slew rate at 60 volts peak to peak into 8 ohms at 5 Khz and got 25 volts per microsecond. Remember my rule of thumb on slew rate?

No, but I will say this, you have a sane approach.

  Extending bandwidth on power amps has few drawbacks. the first is you will decrease performance of reactive load handling, due to more ringing and more out of band components, the second is the amp is more susceptible to out of band interference. Over the years the old concept of out of band rejection has become very important to me.
  One last thing before I have to go. I used to subscribe to Elektor Electronics magazine. I have tried to resusbscribe but they have not responded. Can you help here Dejan? They have a great magazine and many excellent audio projects, I should know. The basis for my D/A Converter comes from one of theirs and the basis for the headphone amp I am sort of working on comes from a pre amp design of theirs.  T.S. Geisberts is one smart cookie.

I am thinking of subscribing, but these days, they do almost everyhting elase but audio. I'll see about their collection of audio circuits, though, it's available on a CD locally.  If I get it, you get a copy.

Cheers,
DVV

[TheChairGuy]

<<What makes amps sound different?>>

Beer or other aged spirits is the correct answer!

[audiojerry]

Gentlemen, thank you for allowing the less knowledgeable to be privy to such great, high level discourse. Much of what you discuss is beyond my comprehension, but still very useful and educational.

Dan, I find it refreshing to see a designer and builder engaged in this topic, discussing your theories openly and not hiding behind a bunch of proprietary claims. Your amps sound like a good design and good value.
Would you be willing to provide a sample to a few Audiocircle members for review? I believe the feedback and interest generated from it would increase awareness of your company and might  provide a nice boost to your sales.  

For the not so well informed, could any of you scholars briefly explain what the impact slew rate, and rise and setlling times have an the sonics?

[DVV]

Hi Jerry, all,

Gentlemen, thank you for allowing the less knowledgeable to be privy
to such great, high level discourse. Much of what you discuss is beyond
my comprehension, but still very useful and educational.

Come on, Jerry, this whole thread was started for the benefit of you
"less scholarly" people. Dan and I have been discussing it on our own
for some time now,  BUT - and there's always a "but" - the problem is
that Dan and I see things in a very similar way. Too similar for a healthy
discussion. So more participants are needed to make things really cook.

But the real point is to try to give everybody interested in the topic
at least a small insight into the inner working of power amps, to try to
illuminate some of the often used, but generally little understood items,
specifications and aspects of power amplifiers, and later on, preamps.

Dan, I find it refreshing to see a designer and builder engaged in this
topic, discussing your theories openly and not hiding behind a bunch of
proprietary claims. Your amps sound like a good design and good value.
Would you be willing to provide a sample to a few Audiocircle members
for review? I believe the feedback and interest generated from it would
increase awareness of your company and might provide a nice boost to your
sales.

Now, that's what makes threads like this worth my while, when they get
people to think and act. If that happens, I don't mind the hours spent
on the subject. Good thinking, Jerry, and you especially, because if
memory serves, you are a tube man. Who better than a tube man to tests a solid state device such as those Dan makes? I never saw them, I never saw Dan, I never heard them, but as one (cursed?) well into the art, I pick
up things differently than laymen.

For the not so well informed, could any of you scholars briefly explain
what the impact slew rate, and rise and setlling times have an the sonics?

Here's my attemept.

Slew rate reflects the rate of exchange, or how well does the whole
amplifier track the input signal. Ideally, it should track the input
signal to the dot, but with its assigned gain. We describe this tracking
capability as slew rate, and express it in volts per microsecond. So,
when you say "10V/uS", what this means is that the amp can track a change of the input signat up to a speed of 10 volts per one microsecond.

In principle, higher numbers are better, but there are pitfalls. Slew
rates are achieved by increasing effective bandwidth, but there's a price
to pay. Increasing the bandwidth also tends to make the amplifier overshoot (i.e. go above the desired level), to ring (i.e. take its time to settle down to what it should be doing) and to act unruly when facing complex loads such as all loudspeakers are. This can, in extreme cases, lead to instability and even oscillation.

It is generally accepted that your slew rate is good enough when it is 0.5V/uS per every PEAK volt of output. For say 100W/8 ohms, peak output voltage is 40V, so a slew rate of 20V/uS is deemed sufficient. Personally, I like to have it at 1V/uS per every peak volt, in the above case 40V/uS.

Rise time is the time it takes the amplifier to go from 10% of its nominal
signal to 90% of its nominal signal. It is measured in microseconds, and
again, the wider the bandwidth, the shorter the rise time, but again,
with similar problems as above.

Settling time is the opposite of the rise time, or how long it takes an
amp to go down from 90% to 10% of its nominal signal. Again, the shorter
the better, but ideally, rise time should equal settling time for complete
symmetry.

The problem is that in the real world, rise times are easier to achieve
than settling times, because there is no such thing as a perfect conductor
or semiconductor. They always store a small charge, which takes time and
care to discharge fully. Therefore, most often, rise times are shorter than
settling times.

Beside all this, it is not at all the same thing just how you got all those
specifications. To simplify it, you could have a mediocre amplifier which
you could then hit with some heavy feedback. This would deliver the goods
nominally, but in real life, this would in fact cause as many, if not more,
problems than it would solve. But it would look great on paper.

Conversly, what Dan and I do is go for amplifiers which have decent open
loop bandwidths, i.e. responses even without any feedback. Then we add just
enough feedback to correct for circuit faults (and every circuit has them,
believe me); in other words, we use feedback to iron out a nicely washed,
white shirt, rather than using heavy duty presses to cover up smudges.

A small example. A typical commercial amp will have an open loop bandwidth (without feedback) of say 5 kHz. The designer will add 40 dB, or 100:1, negative feedback and will achieve say 500 kHz. So, on paper, he has 500 kHz.

I would make its open loop response 80-100kHz, and so would need to add just 6:1, or 15.6 dB, of feedback for the same effect. But since my amp has a wider inherent bandwidth, it relies less on feedback for performance, and would always sound better than the one above. If I now added an input filter with a cutoff frequency of say 70 kHz, no signal ever entering my amp could be faster than the amp itself under open loop conditions, and therefore I would be rid of transient intermodulation (well, it would measure something like say 0.005% or less, totally academic, guaranteed to be inaudible).

Next question, please?

Cheers,
DVV

[JoshK]

Thanks,

The last post helps explain a lot of things in this thread I didn't quite understand at first.