More audiophoolery: discrete op-amps better than ICs!

[coppice]

One of my side hobbies is collecting weird audio stuff, forget expensive cables, some guys claim that you can affect the sound by changing the knobs on the front panel. 

You've never tried changing a volume control? The effects are really profound.

[David Hess]

Degeneration will lower gain and increase noise significantly.

Lowering gain would be a good thing for audio amplification.

Emitter degeneration will decrease the open loop gain and not the closed loop gain.

As to increasing noise significantly: how significant is your "significantly"? is it 0.1db significantly or is it 10db significantly? is the noise in an amp going from -300db to -200db due to degeneration?

This depends somewhat on circuit details but 60 millivolts added to the difference in Vbe is typical from what I found online although I have typically used more.  Just winging it, if the collector current is raised to make up for decreased transconductance, then 22 ohm emitter degeneration resistors will more than double voltage noise in the input stage and the doubled collector current will double it again so that is a 6 dB increase right there without even including the doubling of bias current which will double the voltage noise through the input resistance if it is not lowered.  That is a 9 dB increase total.

Meanwhile if I have a matched input pair, I can throw all of that out and get a lower input bias current as well.

This link does not cover noise but gives some idea of the issues:

http://www.douglas-self.com/ampins/dipa/dipa.htm

One person's "significantly" can differ significantly from another person's "significantly".

For designs which are not limited by input noise this would not be an issue.

Degeneration doesn't need to be resistive, see "noiseless feedback" for example.

Within the context of an operational amplifier, I was referring to emitter degeneration of the input differential pair and I was unable to find any noiseless feedback examples which would apply.

Also, to be clear: dynamic range will drop (gain / noise ratio) for increased degeneration, but total noise (output v_n) will drop.

If resistors are used for emitter degeneration presumably to ameliorate the effects of mismatched transistors, then input referred noise has to go up.

[coppice]

Discrete-better-than-integrated seems to be a belief shared by the audionuttery community:

http://www.6moons.com/audioreviews/totaldac3/1.html

As I see it, a DAC is made up from a load of precision resistors. Are they actually being driven by the 74HC574 next to them? Even if that is not the case I wonder how they'd achieve 24 bit resistor driver matching...

I really doubt that this dac can do even 12-13 bit with monotonicity of output. Not to say that analog and digital is highly mixed all over the board.

Back when DACs had to be made in ways like that, the good ones were ovenised. You can trim something up to 16 bits or so, but you can't easily do it over a reasonable temperature range.

[SirNick]

I've been a fan of Radiolab for a while, listening to whatever I happened to catch on the radio.  Decided to begin at episode 1 and work my way up (it's good company while tied up at the bench).  This parallel-to-improve-noise thing reminds me a lot of the episode entitled "Emergence".  They consider ants, neighborhoods, and society as a whole, and find that individually we are dumb, but en masse, we average out to be rather smart.  Examples given include a game where a group of random folk are asked to guess the weight of a large animal.  Most of the guesses were way off, but the combined average of all of them was within a couple pounds.

Interesting stuff.  The best part for me is that, for every ridiculous audiophile concept or product that numbs the mind to consider, there's an episode of Radiolab that restores my faith in humanity's combined intelligence.

[hamster_nz]

I love this sort of nonsense and messing with the people who believe it! I want to make the suggestion to Wendell and Lenora at evil mad scientist to make a discrete 741 kit like they make the 555 kit. They already have the legs and such! Good learning type kit.

What like "The XL741 Discrete Op-Amp Kit" (http://shop.evilmadscientist.com/productsmenu/762)?

[n45048]

This thread reminds me of a high end Hi-Fi store where a friend of mine works (lovely guy). I went in there to grab a decent set of speakers (probably going to settle for a set of Focal's, but I digress).

One of the other employees was talking absolute nonsense, but I just let him go, didn't want to lead on that I knew more than he did. I also had to laugh at the $200+ price tag for a few metres of speaker cable.

Don't get me wrong, they had some very nice, reputable gear among all that 'sales-pitchy' marketing bullshit, but still.

[David Hess]

Discrete-better-than-integrated seems to be a belief shared by the audionuttery community:

http://www.6moons.com/audioreviews/totaldac3/1.html

As I see it, a DAC is made up from a load of precision resistors. Are they actually being driven by the 74HC574 next to them? Even if that is not the case I wonder how they'd achieve 24 bit resistor driver matching...

All of the HC logic outputs will match each other closer than their absolute value of about 25 ohms so it is possible that the R-2R ladder resistance is an insignificant load or at least taken into account.  I do not see how they could get 24 bit monotonic resolution though.  I might believe 16 bits.

I really doubt that this dac can do even 12-13 bit with monotonicity of output. Not to say that analog and digital is highly mixed all over the board.

Back when DACs had to be made in ways like that, the good ones were ovenised. You can trim something up to 16 bits or so, but you can't easily do it over a reasonable temperature range.

That thing is insane.  I imagine some of the old hybrid DACs from Analog Devices looked a little like it but without the extravagance and more realistic specifications.

With all of those 0.01% Vishay resistors per channel I could see getting into the 16 bit range if they were graded but I assume it does some sort self calibration.  You used to be able to buy precision R-2R networks but offhand I do not know of anybody who makes one good enough.

[SeanB]

Even a 12bit network is best as a hybrid with laser trim for the voltage ( not per digit, that you design in and test that you deposit the film evenly) and then drive with something that has good matching. Doing it with random resistors will not work for monotonicity, unless you match resistors to at least 6 decimal places. Even then it will drift due to power dissipation differences.

[dannyf]

8-bit is the practical limit for typical 1-5% resistors;

10-bit is the practical limit for 0.1% resistors;

beyond that it is practical not possible to do with even precision resistors.

[tszaboo]

8-bit is the practical limit for typical 1-5% resistors;

10-bit is the practical limit for 0.1% resistors;

beyond that it is practical not possible to do with even precision resistors.

Well there is always the bulk metal foil, which they make with 0,001% accuracy. For a price.
But in any case, R-2R networks have an absolute huge code change glitch. I dont see how it would be practical for audio. Probably much more practical to take an industrial DAC and connect it to I2S / DSD with some glue logic.

//The TotalDAC is absolute bullshite. The amount of effort put into a fundamentally wrong concept is just creepy.

[calexanian]

I love this sort of nonsense and messing with the people who believe it! I want to make the suggestion to Wendell and Lenora at evil mad scientist to make a discrete 741 kit like they make the 555 kit. They already have the legs and such! Good learning type kit.

What like "The XL741 Discrete Op-Amp Kit" (http://shop.evilmadscientist.com/productsmenu/762)?

Exactly like that one. hahahaha.. I actually visited them a few weeks back when I was in the area. Great people. 

[VK3DRB]

You'd have to have rocks in your head to believe this rubbish...

http://www.machinadynamica.com/machina31.htm

These clowns should send Dave some rocks for a review...

Genuine reviews?
http://www.machinadynamica.com/machina19.htm

[G0HZU]

Have they started on food products yet?

eg special £££ chewing gum that stimulates the ears and improves the soundstage and all the usual guff. You could have different flavours designed to improve the perception of classic music or pop or heavy metal.

The really expensive versions would come in a gold wrapper that seals all the cleverness inside.

[Richard Crowley]

But those magic crystals are so sensibly priced!  They could have charged 10x and the true-believers wouldn't have thought twice about it.

That 741 kit is an educational tool, of course. Nobody proposes that a discrete retro-discrete implementation of that antique design is practical. Even if you implemented it in a DIP package.

OTOH, the pro audio world has a major case of retro-philia.  Back in the mid-20th century they were making 1.25 inch square potted modules of discrete op-amps because the monolithic products at the time (the 709 and 741 typ.e parts) had major problems with high-quality audio.  So API (Automated Projects Inc. http://www.apiaudio.com/) started making the 2520 module in the same form-factor as the modules used by the analog computing people back in the day.



API is still selling audio gear, but I think they have moved on from the old discrete opamp modules.  However you can still buy clones and even kits from companies like CAPI (Classic Audio Products of Illinois http://classicapi.com/) and SLL (http://www.scottliebers.com/)



And a highly respected home-built mic preamp (John Hardy, http://www.johnhardyco.com/) still makes his own 3rd generation discrete audio op-amp, the 990, designed by Deane Jensen of Jensen Transformers. (US Patent #4,287,479)  Last year Mr. Hardy was forced by parts availability to finally switch to SMD construction to continue using the SSM2212 super-matched transistor pair that is the heart of the design.  He publishes a very interesting and comprehensive document about the 990 discrete opamp.  Recommended reading:  http://www.johnhardyco.com/pdf/990.pdf

[dannyf]

the 990

The 990 is, by the numbers, a very unremarkable opamp. Nothing about it stands out and many modern amps, at a tiny fraction of the 990's cost, can out perform the 990 on the numbers.

Yes, people who have listened to them love them - I am one of them.

I have multiple amps in my possession, from the el cheapos to some of the more expensive ones (Signature and Parasound). I spend the most time listening to a set of JLH1969 -> 4 transistors, single-ended input, all npn (n-ch) output and good for 5w on a good day.

Yet, I always liked how it sounds.

Human perception is very interesting, let alone rationale. It makes no sense to take a pure academic approach to it. Whatever floats other's boats...

[David Hess]

The 990 is, by the numbers, a very unremarkable opamp. Nothing about it stands out and many modern amps, at a tiny fraction of the 990's cost, can out perform the 990 on the numbers.

The 990 input noise has only relatively recently been matched by monolithic operational amplifiers.  Before amplifiers like the LT1028 became available, the LM394 and MAT02 were used to make low noise input stages in combination with monolithic amplifiers which were unmatched by monolithic amplifiers used alone and you can find the designs and specifications in old application notes for the LM394 and MAT02.  High open loop gain, CMRR, PSRR, bandwidth, and slew rate tend to go along with this.  You can still get lower input noise than is available in a monolithic amplifier by paralleling LM394s or MAT02s which might seem unreasonable but that was included in the old application notes as well.

Considering that it takes an LT1028 style operational amplifier to equal some of the specifications of the discrete 990 (chiefly input noise) and the LT1028 has not been around as long as the 990, I would not consider the 990 useless.  My preference though would be an LT1028 or similar at less than 1/5th the cost with a discrete output stage to increase drive capability and ameliorate thermal effects.  I usually get along with LT1007 style amplifiers though when input noise is not critical which is usually the case and I still have a whole tube full of LM833s that I am working through.

I am surprised though that there is not a JFET input version of the 990 using something like an LSK389 or LSK489 monolithic dual JFET in place of the LM394 or SSM2212.

[diyaudio]

couldn't believe the stupidity and ignorance!

It takes stupidity / ignorance to not see that a discrete opamp ****CAN**** out-perform an IC opamp.

Whether this particular discrete opamp out-performs an IC opamp is a separate discussion.

Building a discrete pre-amp stage and comparing it to an IC package (LM833N, NE5532 ect..) has no benefits to the ear for analogue playback material (sound), both these "audio circuits" will employee a dominant pole to force loop gain stability, high performance figures will be attenuated to meet the attenuated criteria to prevent self oscillation and reduce gain considerably at high frequencies. 

Measurement equipment and figures renders a different and valid story, unfortunately we don't listen with measurement equipment, but with our ears.

Discrete stages are more valued in class-d comparator stages, which yields faster slew rates dv/dt for quantization and noise reduction where ultra high speed events are taking place, this is of benefit and relevant.
   

   

[daqq]

Well there is always the bulk metal foil, which they make with 0,001% accuracy. For a price.
But in any case, R-2R networks have an absolute huge code change glitch. I dont see how it would be practical for audio. Probably much more practical to take an industrial DAC and connect it to I2S / DSD with some glue logic.

//The TotalDAC is absolute bullshite. The amount of effort put into a fundamentally wrong concept is just creepy.

And then you need just as matched push/pull drivers... 24 of them. Per channel. At those levels, even PCB traces might add something... but I'm sure it's OK, a trained ear is sure to hear that if it's OK or not.

[SeanB]

Opamp labs.........

Or a regular opamp and a discrete power output stage, which is enclosed inside potting resin in a steel can with an octal base. Took me a day to get one out of the can to replace the output transistors. Claiming 100ma drive capability and using 2N3907 transistors. Turns out they are not happy doing that with video though, though they supposedly are rated for it, or it was the 32 150R loads attached to the opamp. Replace with 2N2219 and 2N 2905 with clip on heatsinks and all is well again, though it would not fit back into the remains of the can though.

[dannyf]

a regular opamp and a discrete power output stage

An opamp + a diamond buffer is pretty much the ideal amp, only exceeded by a CFB xDSL driver -> incredibly fast, high current capabilities, huge slew rate, and no gain-bandwidth limitation.

The price you pay is poor DC performance.

[atferrari]

This stuff is quite normal in the world of audio woo-woo. You REALLY want to see nutso? Try this:

http://www.machinadynamica.com

Particle Accelerator Ion Gun

Amazingly cheap, I think!   

[coppice]

This stuff is quite normal in the world of audio woo-woo. You REALLY want to see nutso? Try this:

http://www.machinadynamica.com

Particle Accelerator Ion Gun

Amazingly cheap, I think!   

Must be German technology. It looks like its from Herr Dryer.

[Zero999]

The hair dryer really made me laugh.

I understand how some people prefer technically inferior technology, such as liking the sound of an old track, played on vinyl, through their ancient valve amplifier but no one should allow themselves to get brainwashed by that bullshit.

If you're old enough, vacuum tubes amplifiers will deliver the best sound to your ears.
The best electric guitars amps are made with tubes...

With a musical instrument amplifier the situation is different to a hifi amp. A solid state guitar amp for example sounds much the same as the volume goes up and up till finally (and suddenly) it hits the brick wall of hard clipping. Truckloads of negative feedback add to this sudden onset of grating distortion and clipping. With a vacuum tube amp, generally the distortion climbs gradually with the volume, and then right at the end it gets worse(?) quicker, but nowhere near as sudden as with a solid state amp.

The benefit of this is that the musician can use this gradual drift into distortion as an extra dimension of expression, much the same as a softly played piano has a gentle sound, but a firmly played piano has a much more urgent sound because of a different harmonic structure. The amp contributes to the timbre of the sound at the whims of the artist. Definitely desirable with a musical instrument. Really bad news for a hifi. It's horses for courses. Each device has it's own benefits and drawbacks.

Thank you for explaining that to me. My usual response to those who say valves amplifiers don't give nasty distortion when overdriven, like solid state amplifiers, is to never allow your amplifier to go into clipping but now I can see why some people do this.

How about a hybrid amplifier, designed so the valve stage goes into clipping long before the ultra low distortion solid state stage clips? Then the guitarist can get the best of both.

[dannyf]

to never allow your amplifier to go into clipping

You can never assure yourself of that, unless you know the music before you hear it.

You can do some basic math but assuming a reasonable crest factor, you need tremendous amount of headroom for a solid state amplifier to play the same music than you do with a valve amplifier.

Another way to say the same thing: a comparably rated valve amplifier sounds much better than a solid state amplifier playing the same music.

[wraper]

to never allow your amplifier to go into clipping

You can never assure yourself of that, unless you know the music before you hear it.

You can do some basic math but assuming a reasonable crest factor, you need tremendous amount of headroom for a solid state amplifier to play the same music than you do with a valve amplifier.

Another way to say the same thing: a comparably rated valve amplifier sounds much better than a solid state amplifier playing the same music.

To be fair. For a comparable size or price you can get a solid state amp with at least an order of magnitude more output power.