Discrete vs opamps... - Page 10

A counterpoint from AMB:

"Why IC opamps?

Early opamps had limited bandwidth and speed, high noise floor, inadequate output drive capability, or other characteristics that made them sonically inferior to fully-discrete amplifiers.

Things have changed dramatically as technology advances. Today there are a number of very high performance opamps that are quite appropriate for high performance audio.

An IC opamp has all its parts formed on one monolithic substrate, with excellent matching between internal transistors, resulting in excellent temperature stability, precision DC and AC behavior, small size and reduced parts count in the overall design.

The recommended modern FET-input opamps for the M³ amplfier all possess very low noise and distortion, wide bandwith, speed and low offset. When configured as a class A voltage gain driver stage for the discrete output MOSFETs (which presents a light load to the opamp), the result is an excellent combination of both worlds. Along with with other carefully designed details, this topology meets the simplicity and high price-performance ratio that are fundamental goals of this project.

The old belief that IC opamps cannot rival the performance of an all-discrete topology dies hard, and the M³ is an accomplished counterpoint, both in terms of measured characteristics and actual listening comparisons."

Obviously they are trying to sell their opamp based product, but there is some truth. The possible superiority of discrete designs may remain, but it simply isn't that clear cut any more.

not entirely crazy but that statement is.

The "guessing" has zero to do with actual fact and the "fact" is no chip opamp can drive a headphone directly in pure class-A period no matter how big of a heat sink you try to weasle on the plastic parts

Again not possible at these current levels and no amount of wishing it were not so will change the facts and physics.Power is heat and the higher the power the higher the heat and is WHY methods other than class-a of amplifictaion were invented in the first place !
Even A/B has proven too costly and generates too much heat for modern electronic devices and is the sole reason for the other newer formats.Not better sound but cheaper to build and easier to cool.

It was a joke for those who know that a chip is an etched silicon substrate and to illustrate how ridiculous some of this conversation has become where "experts" chime in with "feelings" while being too lazy to actually read some background.arguing to do it just because it feels good.

Look at any class of integrated circuits and you find what ? Different packaging for different power levels.There is not much doubt when you have a "small signal transistor" along side of a "power transistor" which is which but has anyone given any thought to what the actual difference is ?

The Package the substrate is wrapped in and no more


A package meant for high power must disperse heat so you will find exposed metal on the plastic which allows the designer to add enough heat sinking to draw the heat away from the actual silicon which would rapidly burn up without it.even the lowly three-terminal-regulator I.C. comes in different packages depending on output current levels.

TO-18 up to 100mA (with a heat sink)
TO-220 up to 1.5 amps (with heat sink )
TO-3 up to 5 amps (with a heat sink)

same part,same function,different package for different levels of power which means different levels of heat

look at your central processor chip,your RAM,your sound card,all with big mamma heat sinks with itty bitty fans and for what reason ? So they do not detonate as soon as you turn on your computer.
Heat is the enemy of solid state and is why small electronic devices use switching power supplies (on or off rapidly but NEVER on all the time) and opamps run in class-B for any appreciable output and also why you will never see an opamp drive a mechanical device like a relay directly but only through a transistor buffer for the additional current.

Size and area of heat dispersal equates directly to power delivery and no way around it.

Actually the earliest "Op-Amps" used 12AX7 tubes in a differential input stage mated to a cathode follower output stage and low bandwidth or noisy they were not.Even today the 12AX7 is used in some very highly regarded extremely low noise phono and microphone stages stages.The "crappy" opamps most often referrred to were the 741 series because they were considered the first "universal" be all end all gain stage in a box but never were they considered an audio device even if used for such by lazy designers trying to save a buck.
It wasn't until PMI showed the way that the opamp arrived as a device useable in high quality audio but those parts were extremely expensive then so again did noyt see much use.I actually have a mic preamp I built years ago using PMI OP27's which is identical to the modern Analog Devices OP27 because in fact ADI IS PMI but even they are have moved away from serving a tiny niche market and into serving the bread and butter market of the portable device-cell phones,MP3 payer,etc and that means low heat which means high efficiency which in the end means as far from class -A as you can get and still be considered an analog chip.
Opamps are moving into the digital domain as is most all things audio and again it is ALL about the heat and class-a is a furnace so off the list of new devices

actually yes but this also depends on your standard of "high performance".Is it the multichip low noise recording console ? The portable headphone amp where you MUST have a low battery drain so efficiency of operation is a must ? A computer sound card ?

How many would pay $5,000 for a preamp that used an Op-Amp as its main circuit ? My money says not too many so the opamp has its place but the ultimate hardly and i don't care what you hang off the output in the way of a gain stage.Price points and value is what it is all about but the truth is a $1,200 amp rarely sounds as good as a $10,000 amp and this goes beyond a pretty face or fancy parts and directly to circuit topology.

All else being equal and assuming a competant design low level gain stages are on a scale from worst to best

1-Op-amp
2-Op-Amp class-A biased to around 2mA with a discrete or chip buffer stage
3-All discrete Class-a
4-All Class-a Triode
5-All Class-A DHT Triode

Stepping up to power amps it is just about the same but here the speaker becomes the main variable.

1-Chip amps
2-all the rest depending on your drive requirements so for instance a solid state behemoth of 200WPC may be what you need if you have a large multidriver low efficiency system but maybe a 45 SET amp if your speakers are a simple zero crossover single driver high efficiency type with pentodes/tetrodes/PP Triode all having a place depending on the speaker.

Headphone amps ?

Just a gain stage with a driver stage on the output and no more so it is "group-1" that has the most in common with the headphone amp

Do a search people instead of using conjecture on what "might be".The truth is out there and being the truth does not vary no matter who it is with the opinion.Simple physics says you can not have a true Op-Amp drive a headphone directly in pure Class-a and why I initially made the statement at the beginning of this thread that head-to-head a discrete Op-Amp will whip a monolithic chip Op-amp in sonics because class-a operation is superior for music and no opamp can drive a headphone in class-a operation

w00t, finally finished reading the 7 page thread. Lots and lots of food for thought =P


Edit: Decided to stay on topic in the end.

Just my experience. I owned a Pimeta(opamp) and a Gilmore Lite(discrete) at the same time. I sold the Gilmore Lite and kept the Pimeta, then I sold the Pimeta for an Mcubed. Though it is obviously a great little amp I found the GL not quite to my tastes in my system as it seemed to exacerbate the upper end of the Onix where the Pimeta seemed to tame it a bit.

While it is true that amb and morsel have emotional investment in the M3 and sometimes promote it as being great, they are not really "selling" anything, nor is it a "product" of theirs. It's not like Rockhopper or MisterX pays 5% in royalties every time he sells a M3

valid only as far as these two amps head-to-head go but not as a definitive statement on what is the better way.

for instance.take the same two amps but change the opamp in the Pimeta or the operating points and/or passive parts in the gilmore's and you end up with two entirely different sounding amps and maybe the results would be reversed.
Just about any topology can be made to sound good or made to sound a particular way to please a certain crowd but Class-A is inherently good by the nature of the topology.With the former you have to use every trick in the book to get sound approaching class-A but with the latter you have to actually screw up to get sound approaching the I.C..

Why ? Because the Class-A amp is a steady state device and everything else is not.The "A" topology is in the "on" state from the moment it is powered up to the moment it is powered down.The others are on "idle" until there is an input signal then the amp itself must respond to that signal by turning "on" and that means always playing catch up instead of just passing through anything present at the input.

Obviously there are good examples of chip amps and piss poor examples of discrete amps but that is about desicions made by the designer and not what is by nature superior and I also beleive most could not even do a single transistor single stage Class-a amp and have it sound as good as a cheesy simple opamp based amp but only because they are not capable of doing anything other than plugging in an opamp where the work has already been done,not because it is a better way.

Rikr, no offense but your posts are coming across (at least to me) as a little angry. I do not believe I have done or said anything offensive here. I am trying to get to the heart of the issue.

There is room for conjecture about what might be possible. All the arguments you have made are perfectly coherent, however that doesn't mean that there is no use in arguing what can be done in theory to bring an opamp up to a new level of performance. What I am trying to get answered here is, if TI decided to package their OPA627 in a package that was designed to dissipate large amounts of heat through metal to a heat sink of your choice, would that theoretically allow biasing to pure class A, and secondly would that pure class A biasing result in an improvement in sound over 2mA class A bias? If you did then have a monolithic opamp in class A (which you imply is somehow technologically impossible, which it is not, it is simply not currently packaged to do so (what's the difference? I think quite a lot - to me there is a big difference between a circuit being unable to do something because it is actually incapable of performing that task due to it's circuit design, and being unable to do something because it is not put in the right thermal conditions. Of course this only applies if one is talking in the theoretical domain, but I am here) would the sonic results be closer to your idea of what discrete class A can do?

The implication of this thread is that there is something fundamentally bad about a monolithic opamp with respect to audio reproduction. You quote a lot of fact, which is all of course true, however the only thing that you have so far said about what is provably and factually inferior about a monolithic opamp is that it cannot be biased into class A. That may be enough to prove superiority, but what I am trying to understand here is if the class bias was equal between the two, could an opamp be brought up to a level playing field.

If, hypothetically speaking, it were possible to put the OPA627 in a package such that it can be run in full class A without blowing, it would sound better than putting 2mA of bias current through it.

But that's in theory, I don't know if it would be possible to do that with the OPA627 without redesigning the silicon substrate itself, and if that were done it may no longer spec out as an OPA627 depending on what changes need to be made.

stating facts is now called angry ?


I beleive otherwise.If someone states factual evidence or facts as they know them to be it is up to the person with a counterpoint to back up their side with something.

already stated over and over many times by many persons :

The best opamp you can find operated Class-A as far as you can go within the limits of the chip then have an external stage take over.Simple and sweet.The rest is impssoble so I have as usual so stated it rather than see five pages of "lets try this" .
It maybe not a way to popularity but damn sure it is the truth and to be more blunt,should be the goal here not personalities or wishful thinking

because it will never happen is why !

I get a minimum of twenty electronic trade print magazines so I KNOW where the industry is headed and beleive me,what you would like to see happen has about as much chance as two moons rising tonight...NOT.

since this "wish list" is an impssibility why waste bandwidth discussing the impossible when there are those on the sidelines who through not knowing any better will also weigh in until there is five pages of vapor ware to no point ?
You want "theoretical" the DIY forum is available.But if you want factual information that is what the Amplifier forum is here for.Facts or listening tests or product announcements but not "maybe we can open up a opamp and make it do things it was never designed to do".Again realism says not.

WE are not semiconductor manufacturers so have no say in what is offered and WE are so low in the food chain of importance when the chip market is considered that it surprises me ANY opamps work well for audio and many of those that do will very soon be going "obsolete" or at the minimum to SMD which will make even the headfi past time of "opamp rolling" a fond memory for many.

[QUOTE]The implication of this thread is that there is something fundamentally bad about a monolithic opamp with respect to audio reproduction. You quote a lot of fact, which is all of course true, however the only thing that you have so far said about what is provably and factually inferior about a monolithic opamp is that it cannot be biased into class A. That may be enough to prove superiority,
the implication of this thread is that someone asked what is better : Discrete or I.C. and when answered truthfully has gone all over the map without a single counter that has as of yet any validity.The only thing offered so far is "I like the pimeta better the the gilmore"



I realise it can be hard for some to "visualize" from the written word just what the amount of heat generation we are tlking about here but beleive me when I tell you there is no way a DIP package plastic opamp will ever be able to tolerate pure class-a operation and the so called "power opamps" in TO-220 and TO-Can style packages that can produce higher levels of output power are CLASS-B devices and even then require huge heat sinks to get the power to flow.
Can these be made into Class-a devices ? Maybe yes maybe no but my experience with these chips is they are all crap for audio use except for (just like the low powered opamps) as a driver stage for a higher power discrete output stage where the device can be used in a larger area of class-A if you peg the output transistor to a supply rail (through bias) but why bother ?

Why would someone want to take a device that is severely comprimised and force it to try and be something it is not when the easier and cheaper route (and in the end better sounding ) would be a real front end mated to an stage designed for audio use ? These devices are designed to drive selenoids and motors,not audio transducers, and as such are built towards gobs of current which does not have to be linear across a wide range of frequencies but just THERE to muscle whatever it is the chip is driving.
no regard to output Z over audibly significant frequencies,no regard to slewing and damn sure no regard for "sound" quality and that is supposed to be the topic here,what actually sounds better.

If factual content is called "anger" then I plead guilty.

Guilty because no way do I sit by while some very wrong statements are made when there are so many who have no clue trying to understand the "why" of things and for them muddying the water is counter productive and my thought is there is enough BS out there in audio land that for me to add to by not staing my piece just will not do.


Call me impulsive

[QUOTE=rickcr42]not entirely crazy but that statement is.

For normal - not so specific informed - readers this is unreadable! Sorry, but who are you communicating at! I always expect that - how difficult the topic - the writer has always the responsbility to give clear messages so one could follow the message!

Class is now in session,be seated please

for this illustration on semiconductor packaging and heat/power dispersal i will use the Linear Technology LT101 monolithic buffer to illustrate a couple of points.

first up is the rough data :


LT1010 - Fast ± 150mA Power Buffer

20MHz Bandwidth
75V/μs Slew Rate
Drives ±10V into 75Ω
5mA Quiescent Current
Drives Capacitive Loads > 1μF
Current and Thermal Limit
Operates from Single Supply ≥ 4.5V
Very Low Distortion Operation
Available in 8-Pin miniDIP, Plastic TO-220 and Tiny 3mm × 3mm × 0.75mm 8-Pin DFN Packages

Potential uses :

Boost Op Amp Output
Isolate Capacitive Loads
Drive Long Cables
Audio Amplifiers
Video Amplifiers
Power Small Motors
Operational Power Supply
FET Driver

*the above are straight from the LT1010 buffer data sheet

damn fine specs for audio driving duties on the surface but digging a bit deeper we see some "cracks" starting at the edges :

.


So there it is folks.You can have it in a standard Op-amp sized 8-Pin miniDIP package but if you want to use the chip to its fullest capabilities you need to go up to the "Plastic TO-220" or "Plastic DFN" packages to "aid in reducing operating junction temperatures !"
Same part,on first glance same specs but when you dig the truth comes out.That truth is there are limits to how much heat can be dispersed for each package to keep the operating junction temperature within acceptable limits.To illustrate this I have linked to pics of the actual packages this buffer comes in.

First up is the all plastic standard Op-amp sized substrate housing.The LT1010CN8 package :




Next up is the LT1010CDD package which has if you will notice an area underneath where there is a metal pad.This pad is meant to attach to some form of heat sinking in order to have better power handling (heat rejection) but if not used is in reality no better than the above CN8 package.the exposed metal "paddle" is only as good as what it is attached to :



finally we have the ultimate heat dispersing package available for this particular device.The TO-220 "Power Tab" package.again notice there is a metal area but this time it is very thick and very large when compared to the other to packages.
Because of the size and area of this metal surface it will perform without a heat sink as well as the CN8 package with a heat sink but to take advantage of all it has to offer this "Tab" must be attached to a large metal surface or heat sink otherwise the heat can not be sucked away from the actuall on chip transistors and in this case will cause a total shut down of the device if out of spec (current limit shut down) :




If you compare the three you can easily see there is atually no compraison and even though the beginning part number is identical for each the end suffix is what really tells the tale.If you will notice the "TJMax" specification for each you will see the TO-220 version can handle a tad over 5X the heat of the DIP version and just under 2X of the CN8 version and this is very meaningful because it goes directly to physical size and ability to handle heat without shutting down or in extreme cases detonating and is why there are limits to what you can expect from an Op-amp in the area of Class-A current delivery.
It can not be done because there is no way to get the heat out of the package in a way that will prevent the chip from self destruction and is why they are only biased into Class-a for a portion of the output and then the power is handed off to an external device such as the above LT1010 to fill in the rest !

As an example,here is an Op-amp being used as the front end gain stage with an LM334 current source used to bias it into class-A up to 2mA where it is then "passed on" to the LT1010 buffer to perform the actual driving duties.I have used this buffer biased up to 40mA class-A and I can tell you flat out in order for it to get there it needs some serious heat sinking or it shuts down.

This is why I made the above staements about "impossible to do" because as you can clearly see it is !


Class is now dismissed.For your homework I suggest you read this pdf file and especailly the part near the end about which device you need to use to get what level of performance :

http://www.linear.com/pc/downloadDoc...28,P1219,D1683


If I was mean spirited I COULD say "here's mine,where's yours" but I will play nice and not

according to WHO ?

You ?

I am to be held to some high standard while those with no clue can post whatever with no actual facts to back up statements and it is me who has a responsibility ?


I think not.I owe you people nothing and can post on any topic I want at any time i want,just like everyone else

Rick, I still don't quite understand what the real life consequences of all this is supposed to be.

Does this mean that a discrete amp like the GS-1 will perform above a new Headroom Max? Or does this mean that a discrete amp like the GS-1 will definitely perform above an equally-priced op-amp based amp like the Emmeline HR-2 or Corda HA-2?

Or does the 'discrete > op-amp' statement only apply to high-end?

Or are these only technical hypotheses pertaining the potential of discrete amps to be superior?

Thanks.

the initial question asked in the thread topic is what it is all about that once I responded to ended up going all over the map and NOT on my end.
I only answered questions raised with valid answers but then I guess I really should have known better considering the crow here.i answer and it is me going off topic.Again typical and why I shun this joint like it was a contageous disease

Have fun

rick out

I read about five pages of this thread and then just gave up and jumped to the end because I thought I'd like to comment on an aspect of this discussion. So please excuse me if I cover something that was said late in the thread:

In a perfect universe, I agree with Rick that a well designed descrete, class-a biased circuit will kill whatever IC based design you could come up with. And it is basically for the reasons Rick states: you can't truly get a chip to run class A, and you can't tweak the insides of the chip for ultimate audio erformance. I'll add (and this has been mentioned, but bears repeating) that forcing a chip into "class-A" design with current pumps is NOT really a class A design. In my experience, however, it can improve the audio quality moderately.

However, we don't live in a perfect universe.

As also mentioned previously, as soon as you do a really fine descrete design you have to start matching parts. This is expensive. Also, really fine descrete designs usually don't have good output protection and temperature drift compensations in them because they decrease sonic performance. So, you ae more likely to blow up, or need to have service on, an audio hot-rod descrete amp. As Rick also mentioned earlier, if issues like cost, portability, and reliability are not in the picture, a well designed descrete is better.

But those issues ARE normally in the picture, one way or another. And as gpalmer said, what's guarrentying that the descrete design you get is a good one? I've seen lots of descrete designs that sounded like poop.

So let me sum up what I think the novice might want to know about descrete vs. monolitic designs:

1) In a perfect world, descrete designs can achieve higher levels of performance.

2) In the real world, IMHO and at a given price point, with low to medium cost headphone amps (up to, say, $800), good op-amp designs are just as likely to deliver the sonic goods as good descrete designs, and are more likely to be reliable.

3) In cost no object products, I still don't think it's as easy as saying a descrete design is always better. At this level, personal listening preference is very important, so your pick should be based more on your ears than what the design topology is. For example, we (at HeadRoom) simply can't do a descrete design and still have a crossfeed circuit, so if you want crossfeed you'll have to live with some percentage of the electronics being monolithic. Also the tube vs. transistor question rears its head.

Lastly, I want to say that Nelson Pass is one of my personal design heros. In understanding his designs you'll find that simply saying class A is the only way to go only presents a very small part of the picture. The devil is in the details with descrete designs, and a lot of homework (in terms of listening tests and confirming reliability) is needed for someone to make a long-lived, satisfactory product selection.