Discrete opamps and gain...

[Moses]

Hi guys...

Yeah, sorry for all the numbers!! I mention these, mainly because I already have them kicking around the studio...

The focusrite currently sits in a 6u flight case with 1u of free space either side of it, and another 2u 2 channel preamp (which doesn't generate much heat)... Im pretty sure it's an overheating problem, because the only time it splats and pops is when it gets hot! It kinda makes me sad, because it was one of those 'lust' pieces of kit (for me!) before I got it... It would be sold as working/faulty!

abbey road d enfer, point taken. And nicely put! I have been scrounging schematics from around here and elsewhere for a while now.. I just need to fire up the iron and make good on my intentions!! Currently also digesting Mr Dove's console design guide at bedtime!!

Thanks!!

Mo

 

[tv]

Is it worth my while designing my own preamp, or more hastle than it's worth, especially with the JLM kits so affordable?

There is IMO no point in designing an utilitarian preamp vs. building from a kit - unless you're a die-hard DIY-er.

Otoh, spending your time on designing and tweaking an unique flavour preamp that you know you miss/want/need, that's something totally different.

 

[Moses]

Tv, that's very honest and much appreciated!!

Maybe I will just go down the kit route.. Learn as I go too!! And then, at some point in the future, I can start 'flavouring' as I desire!

But, to my original question... (digression my fault entirely!) In an ideal world (time and money not a consideration!) for a given piece of kit is it preferable to use an opamp or to design custom gain stages on a case by case basis? Im just trying to pick brains, and get different sides of the (technical!) arguements.

Thanks,

Mo

 

[tv]

Opamps vs. discrete ... for "utilitarian" preamps, there is imo no point going with discrete designs - there are very good opamps available today (OPA, LME, AD and LT series).

You can imo safely build a battery of opamp based pre's (from kits etc.), put IC sockets in them and "flavour" them by mixing and matching different opamps... (read datasheets for their max. rail supply voltage, so they won't blow by acident). IOW, this way you can (at least to some extent) "flavour" opamp-based pre's as well.

For the "vibe machine" pre's, going discrete is the only route - imo.


p.s. there is a "something inbetween" variant, usually called a "hybrid opamp" design - meaning that a particular gain stage consists of an IC opamp followed with an AB or A-class discrete follower stage. These combine the easy-to-replicate aspect of an IC design with a slight "discrete" sonic flavour (no matter what THD numbers say, there is a _slight_ sonic signature of a discrete bolt-on stage).

Perhaps, the latter "hybrid" designs would be good for you? Maybe search here for more - and I think that JLM offers a "hybrid" design somewhere - not sure, though.

Perhaps some other forum members will jump in and give more details.

 

[burdij]

As I recall from the mists of time, the original reason why discrete opamps were used in audio was because they were the only thing that could drive a line matching transformer with enough current to be able to put 30 V peak to peak into a 600 Ohm resistor. And the 30 V peak to peak was the thing that disqualified all of the available integrated opamps of the time (let's say the late 60's/early 70's) because their rails were pegged at +-15 V max. Anything higher would "let the smoke out". Of course, today, we have some integrated opamps that can deliver enough output current and voltage swing to do the job, but I think the general case is still true. Common opamps just don't have the guts to do the job where the job is delivering a normal +4db line level signal with average crest factor to a 600 Ohm load (that is still the standard, right?).

The "color" of those old recordings was the sound of those little transistors trying to supply immense amounts of current to a highly reactive load.

 

[PRR]

In my small recollection, chip opamps got into utility audio before custom opamps.

RCA made a quite good broadcast console full of 709 opamps, wired with opamp concepts. In the same period, many other makers were doing single-ended supply few-transistor discrete consoles.

The 709 had no short-protection, would die instantly. It also needed a complex custom compensation scheme. However once compensated, it had bandwidth to shame many later amps. It's not the lowest-noise amp ever built, but a good designer works around that; it was as quiet as my top of the line Langevin tube board.

Aside from needing boosters, and unstunning noise specs, the thing which made chips popular AND gave them a bad name is the 741 and its 558 dual. Low slew rate, and we did not really understand the ramifications of that. If you instead use the 301, uncompensated 741, and carefully pick signal levels and compensation caps, good performance is possible (down to 2K load); but many designs missed excellence.

The really good discrete audio modules make design so much simpler. And when you go so far as to design an excellent module, you may as well design it as an "Op amp"; it simplifies other system details, and can reduce the number of coupling caps.

Of course "anything" you can do discrete "can" be baked as a chip. There are some hurtles: the original low-cost process does not make great PNP, though exotic processes have come down to commodity price. Great audio performance may want large-area devices, but chip-makers tend to think in terms of small devices. The set-up costs are millions of dollars, and it is hard to pay that back from sales to low-price consumer gear or low-production hi-end audio markets. But in another world (ADSL) very high performance chips were made by the millions, and we can mis-use them by using only the bottom octaves of their MHz response. Load-cells need VERY low voltage noise, and the application is not cost-critical, so large-area chips are made that we can re-cycle into audio.

> is it preferable to use an opamp or to design custom gain stages

The answer is "yes".

Yes, which? Well, that's asking us to tell you how to be bigoted. Which is more fun at parties, blondes or redheads? Who is only good for ditch-digging, Irishmen or Belgians?(*) Asparagus or broccoli?

You can pick your bigotry and stick by it. Or you can try blonde Belgians digging broccoli versus redhead Irish ditching asparagus. Even if, at one time, my g-g-grandfather was an illiterate drunk laborer, things change. 709 is out of production. There are alternatives to '741. We now know enough to do good things with 741, when we must.

(*)I'm a little of both, so I'm picking on myself, not any Irish or Belgians reading this.

 

[tv]

Which is more fun at parties, blondes or redheads?

Speaking from memory and from my limited experience, my research has shown that blondes are more fun at parties, while redheads proved to be more fun after parties. Of course, in my research I didn't go so far to f.e. replicate the red snapper experiment - or anything along such lines. But, some of the subjects under test proved to be quite snappy..

 

[clintrubber]

Sounds like this thread is about to derail to Brewery territory 

 

[ppa]

By the way, ppa, that's why I feel it's a pity that your otherwise very nice DOAs are potted and proprietary: I would have loved to get under the hood. Oh well.

JDB.
[/quote]

thanks for the compliments. Regarding the potting and proprietary I'm sorry because 2050 has a "strange" frequency compensation circuit, different a lot from other DOA's and ICs, that gives high stability (2050 is stable at unity gain with more than 1 nF cap) but high slew rate. Moreover, 2050 has other things to say but is certainly too early to show its topology. However 2050 is designed like a DOA not like a IC op amp , so it sound very different from the IC op amps.
2050 has no roughness and also a vintage sound that's much far from IC op amps' sound.

Pier Paolo
 

 

[clintrubber]

Moreover, 2050 has other things to say but is certainly too early to show its topology. However 2050 is designed like a DOA not like a IC op amp , so it sound very different from the IC op amps.
2050 has no roughness and also a vintage sound that's much far from IC op amps' sound.

Pier Paolo
 

Take care there, wordings like this start to qualify for the White Market...

(just kidding  , but please add some more solid tech-info as soon as you see fit)

 

[abbey road d enfer]

Common opamps just don't have the guts to do the job where the job is delivering a normal +4db line level signal with average crest factor to a 600 Ohm load (that is still the standard, right?).

To my surprise, I have found that it may not be the standard anymore, particularly in digital recording. It seems many young sound engineers coming from the flourishing schools of audio tend to operate as close as possible to OdBfs on the PPM their DAW offers, which means, with most converters an average level of +8/10dBu with peaks of +18/20dBu. And 600 ohm is not a concern any more. Under these conditions, slew rate may be a more important factor than output current capability. The irony is that this signal is scaled down to +/-2.5v at the converter input.

 

[JohnRoberts]

600 ohm terminations were already fading when I started designing back in the '70s. It's a carry over from long telephone lines that need to be properly terminated, and tube gear with relatively high source impedance. No real justification for modern gear other than providing ability to interface with dino-gear.

+1 on irony,,, trend on codecs is moving even lower than 5v rails.

JR

 

[abbey road d enfer]

Aside from needing boosters, and unstunning noise specs, the thing which made chips popular AND gave them a bad name is the 741 and its 558 dual. Low slew rate, and we did not really understand the ramifications of that. If you instead use the 301, uncompensated 741, and carefully pick signal levels and compensation caps, good performance is possible (down to 2K load); but many designs missed excellence.

That reminds me of one of my very first mixers, in 1976, that used a lot of 748's. These need a 30pF compensation cap for unity-gain stability. The mixing amps had a manufacturing mistake: the assembler used ceramic compensation caps, marked 30P, assumed they were 30pF, when in fact they were 30nF (1000x). The mixer was assembled and before testing, the expert who was commissioned by the customer came, supposed to be a "golden ear", listened to it and said it sounded wonderful, so the mixer was delivered and installed as it was. When the expert had the time to do the objective commissioning, he was a bit embarassed by his findings. In fact, on normal music program, the large signal response was adequate.

 

[JohnRoberts]

Ouch...

Back in those days you could get decent slew rate with LM301s using feed forward compensation, but that only worked in inverting mode. Oddball compensation always made me a little uncomfortable. The TL07x and NE553x that came out around then are still quite respectable when properly used, while even the 553x compensation scheme is disconcerting to purists. 

Lots of -10dBV recording gear using slow (by today's standards) opamps, gave respectable results when operated at modest levels.

JR