Difference between IC opamp and Discrete opamp?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Osse

Well-known member
Joined
Dec 2, 2007
Messages
81
Here comes another newbie question for some nice pro to answer :grin:

I've been looking the web to see a detailed description on different types of op-amps, but I've just found on the IC opamp, could anyone explain how the discrete one works, and what the sound difference compared to IC opamp is, if there is any 'general' sound difference.

Thanks a lot!
 
This is like comparing store bought furniture to home made furniture.

Some home made furniture can be quite good and better than cheap store bought furniture. Very expensive store bought furniture will often surpass home made in most respects but will lack the pride of being able to say "I built that".

Unless you are skilled in the art you will be hard pressed to come close to the performance of better off the shelf opamps, I wouldn't try. Even if I could I'm not sure I could match the performance for the same cost. Sometimes as needed I may hang drivers on the output of an opamp for more drive capability, or add some low noise devices in front of an opamp for better noise performance, but I see no advantage to rolling an entire opamp.

I doubt this is the popular opinion.

JR
 
[quote author="JohnRoberts"]

I doubt this is the popular opinion.

JR[/quote]
:grin:


Here some more, from
http://www.dself.dsl.pipex.com/ampins/discrete/discrop.htm, 12 Oct 2001 (see the webpage itself for proper formatting):

Circuitry made with discrete transistors is not obsolete.
It is appropriate when:


A load must be driven to higher voltages than the opamp can sustain between the supply rails. Opamps are mostly restricted to supply voltages of +/-18 or +/-20 Volts. Hybrid-construction amplifiers, typically packaged in TO3 cans, will operate from rails as high as +/-100 V, but they are very expensive, and not optimised for audio use in parameters like crossover distortion. Discrete opamps provide a viable alternative.
A load requires more drive current, because of its low impedance, than an opamp can provide without overheating or current-limiting; eg any audio power amplifier
The best possible noise performance is required. Discrete bipolar transistors can outperform opamps, particularly with low source resistances, say 500 Ohms or less. The commonest examples are moving-coil head amps and microphone preamplifiers.
The best possible distortion performance is demanded. Most opamps have Class-B or AB output stages, and many of them (though certainly not all) show clear crossover artefacts on the distortion residual. A discrete opamp can dissipate more power than an IC, amd so can have a Class-A output stage, sidestepping the crossover problem completely.
 
[quote author="Osse"] could anyone explain how the discrete one works, and what the sound difference compared to IC opamp is, if there is any 'general' sound difference.[/quote]

The first opamps were very 'discrete" indeed (although their failures were not necessarily discreet :grin:). They used tubes, rarely had a noninverting input, and had a lot of drawbacks.

With the advent of sand state we started making reasonably small modules with discrete transistors inside, along with supporting resistors, capacitors, and sometimes diodes. At the same time monolithic fabrication of arrays of devices was in intense development, leading to some workable but generally inferior-performance integrated-circuit operational amplifiers. But the appeal of miniaturization and cost reduction was compelling for a lot of applications.

The circuit topologies (the way that parts are connected) derived from discrete opamps were applied pretty much directly at first. Then various brilliant people began to think outside the box and realize what could be done with monolithic structures. It was said that Bob Widlar bedeviled the drafting staffs he worked with, by forcing them to come up with symbols for devices that didn't exist as discretes.

At this time the circuit topologies began to significantly diverge. Of course the principles of operation of transistors and so forth didn't change, nor have they to this day, so the basic answer to your first question is, both IC and discrete amps work the same way.

Now, to the second point, we enter into the fuzzy realm of subjective evaluation and to what extent measurements correlate with sonic preferences. A very simple discrete design is apt to underperform a well-designed integrated part as far as our measurement tools reveal. Sometimes the higher distortion and other coloration may be perceived as desirable. Sometimes not.
 
quote: "...but I see no advantage to rolling an entire opamp."

Besides the Doug Self remarks, I would add one advantage that may be surprising:

Sometimes lower cost, given specific constraints on power consumption, noise, frequency response, and distortion.

Under the sleek roll-back cover and case of the portable powered speaker system, JBL On Tour, is a board packed mostly with discrete devices. Most of the signal processing is done with them. Given the detailed knowledge of source and output impedances and other performance criteria, it actually made sense to use tailor-made discrete topologies. The SOT23 devices are about a penny apiece in large volumes; resistors are practically free. Labor to run the pick-and-place machines and reflow ovens is cheap in the Far East, and if the vendor has a well-tuned process, yield and reliability are high.

These circuits are hardly general-purpose opamps of course. And the development time was substantial, justifiable for the anticipated volumes and revenues, but what would have made little sense for smaller productions.
 
From what I've figured out over the years, it's a misconception (as Brad suggests) to assume that the monolithic approach will always be cheaper. It will be if you're talking strictly about NJM4560s or similar, but when you get into the uber-low distortion op-amps made from LT or BB, discrete op-amps can work out far cheaper.

You can make (well, providing you have the right design and PCB layout) a pretty mean op-amp from say, a handful of BC550C + BC560C. These transistors cost around a penny each in quantity. By the time you've added everything up, you have a serious op-amp for less than the cost of a high-end IC. It won't measure as well as the IC, but it'll run on a wider range of voltages and *may* appeal in subjective tests.

A professional designer that I work with is always worried about making things 'too good' - it concerns him that much of the equipment with stellar performance figures is often labelled 'sterile' by end-users. He classifies many modern op-amps as being in that bracket. A DOA gives the experienced designed the opportunity to make something that would be difficult to commercially justify to the chip maker - it can be tailored to have a particular characteristic that may not look good on paper, but appeals to users.

I agree totally with John's furniture analogy. The super-expensive designer furniture from Scandinavia may not be to everyone's taste - some prefer a more 'rough and ready' aesthetic.

Justin
 
[quote author="thermionic"]
A professional designer that I work with is always worried about making things 'too good' - it concerns him that much of the equipment with stellar performance figures is often labelled 'sterile' by end-users.

...

I agree totally with John's furniture analogy. The super-expensive designer furniture from Scandinavia may not be to everyone's taste - some prefer a more 'rough and ready' aesthetic.

Justin[/quote]

I have those concerns too, as I'm close to completing a quasi-"high-end" phono preamp. My quasi-partner is polling other audiophiles extensively, and when he discloses some of my directions he often gets feedback to the effect of "topology X sounds dry and sterile". At the same time I don't want to be associated with something that measures horribly. Call me quasi errr I mean crazy :grin:

On the furniture analogy: sometimes a rougher finish etc. is still prized to some extent, if other qualities are there. I think of the harpsichord maker Martin Skowronek, whom people criticized for some relatively crude craftsmanship at times, but whose waiting list just got longer and prices higher and higher.
 
I wasted a great deal of time over designing phono preamps and I have one observation for you to consider. This point was really driven home when the exact same unit was reviewed by two different reviewers on two different systems and found dramatically different sounding (one liked it, one didn't).

Giving the reviewers some undeserved benefit of the doubt, capacitive termination can cause top octave frequency response variations in MM cartridges significantly larger than the typical design considerations of RIAA accuracy and subtle topology caused response errors.

I used to sell a little dipswitch board with a few polystyrene capacitors so users could dial in the cartridge termination for their system. Customers are not very receptive to doing this and that adjustment was not in this particular review unit. There may be some different ways to deal with different capacitive termination needs that would be compatible with the audio-phoolish mind set. The first thing I would do is review what the popular cartridges are in use, especially important what cartridges are used by reviewers. To some lesser extent what are typical tone arm, turntable capacitance figures.

Perhaps roll your own funny interconnect cables, whith longer/shorter higher/lower capacitance cable to more conveniently dial in systems. For X cartridge use "blue" cables, etc. If there is also an active interconnect cable market you may have to work around that. If there's an obvious cartridge/cable/turntable target, you can ship a dialed in nominal solution but life is rarely that accommodating.

While not as big a factor variable input R could make some audible difference. Uber slick would be variable input capacitance, but to do this mechanically would be physically large (like those old radio tuners). I'm uncomfortable with throwing some varactor diodes at the front end since in theory you could see a volt plus at that input while in practice more like tens of mV.

A variable input capacitance knob would give them something audible to adjust, that isn't make believe. :grin: If your designing for MC... never mind...

JR
 
> explain how the discrete one works

Exactly the same.

General Op-Amp Theory suggests that all op-amps are interchangeable; the whole idea is that function is determined by external parts and not by the amplifier.

Real life is not so simple. Long ago tube opamps ran on 250V and gave 1mA; when transistors snuck into the circus opamp conventions moved to 15V 10mA. If you swap modules between these eras you won't be happy. Some jobs are slow: integrating the average earth-movement in an earthquake, and some jobs are fast, and with other requirements this leads to different models each adapted to be less-imperfect for one job over another.

Chip IC opamps do the same thing as discretes, except there are trade-offs. You don't carve a chip-mask for an application which can only sell 1,000 units; but you can hire workers to assemble discrete amps in very small quantity. But once you commit to carving a chip-mask, you can justify a lot of brain-power to make that chip do well in many-many markets. Indeed you must, because chip costs are $1,000,000 for the first unit and $0.10 for every one after that. If you can double the market and the sales, you can cut the price in half; or just cut it to 3/4 and pocket a good profit.

Historically, the first designers in a technology have harsh limits. The first ICs did not have good PNPs. That eliminates several good tricks. But note that when discrete transistor op-amps were new, a good PNP was much more expensive than a good NPN, so PNP-free tricks had already evolved.

There was a Golden Age for discrete audio modules. In 1972, you could buy darn good discrete transistors, make good amplifiers. You could instead buy LM741 chips, which are excellent for many things, but "ehhh" or "arg" for high level or low noise high quality audio. We did use a lot of 741. They gave chips a bad rap. There were alternatives: 709 can be done well, 301 allows some extension, the 318 was mighty fast and reasonably clean. For some reason the LM741 got used a lot, too much.

The 5534 chip was designed for audio. Not that the market really justified it, but it was also good for industrial strain-gauges and more obscure things. It really does 9/10th of the things any discrete "audio op-amp" can do. Also 351/TL072 opamps do 8/10th of audio jobs at very low price. These been mainstays of large audio systems. Newer chips have come along. Some from far left field: pushig fast digital data down long phone lines, "DSL", requires a driver which can incidentally do great audio without the least strain. Now that every house has DSL that initial mask investment is paid down and the per-chip cost is under a dollar.

So you do discrete the same reason you make your own clothes or furniture. When maintaining older homes you make replacement beams by hand cuz commodity lumber is not the size the farmer used in 1837, and some of that 1970s gear is worth maintaining. Maybe to do it better than any ready-made product. Sometimes just cuz you like the feel of the chisel in your hands, seeing the electrons in little snips of hand-wiring.
 
thanks for those great responses, I've learned something today afterall... :twisted:
 
Does it really cost around 1-mil USD to tool up to make an analogue audio IC? I've always suspected it did, but I've never had it confirmed.

I have to say that, if that's the order of investment needed, it gives me respect for firms such as THAT for releasing niche ICs that are strictly for audio buffs. That's a big risk for such a potentially small market.

John - you do realise that someone from a hi-fi firm will read your idea through Google and release a phono pre with a big old tuning capacitor (providing they can find enough)? Having said that, how big would the fins be on the thing to give say, 330pF? I bet the hi-fi press would have a field day over it... I recently demo-ed a phono pre with a bias control on the front. It was FET-based and the control seemed to have a pretty odd effect.

The latest thing in phono pres is balanced input, or balanced throughout. Never mind the fact that, with modern transistors / high-end op-amps, the self-noise of an unbalanced preamp might be as much as 20+ dB below that of the inherent noise in the cartridge...never mind the surface noise of vinyl!! (Much as I love vinyl - I don't delude myself)


Justin
 
[quote author="thermionic"]John - you do realise that someone from a hi-fi firm will read your idea through Google and release a phono pre with a big old tuning capacitor (providing they can find enough)?[/quote]

If they can't find enough, they'll just put a big old wooden knob on it and charge $6000 apiece for them.
 
I don't care... ideas are cheap. I've got plenty more where that came from.

I'm not holding my breath. Doing something real and useful is foreign to the audio-phool world.

Isn't vinyl obsolete? I'm not asking if it's spinning back into fashion, only if it has somehow escaped the laws of physics that rendered it obsolete or no longer king of the sound quality hill some time ago.

I hear wax cylinders are coming back.. :wink:

JR
 
[quote author="thermionic"]Does it really cost around 1-mil USD to tool up to make an analogue audio IC? I've always suspected it did, but I've never had it confirmed.

I have to say that, if that's the order of investment needed, it gives me respect for firms such as THAT for releasing niche ICs that are strictly for audio buffs. That's a big risk for such a potentially small market.

John - you do realise that someone from a hi-fi firm will read your idea through Google and release a phono pre with a big old tuning capacitor (providing they can find enough)? Having said that, how big would the fins be on the thing to give say, 330pF? I bet the hi-fi press would have a field day over it... I recently demo-ed a phono pre with a bias control on the front. It was FET-based and the control seemed to have a pretty odd effect.

The latest thing in phono pres is balanced input, or balanced throughout. Never mind the fact that, with modern transistors / high-end op-amps, the self-noise of an unbalanced preamp might be as much as 20+ dB below that of the inherent noise in the cartridge...never mind the surface noise of vinyl!! (Much as I love vinyl - I don't delude myself)


Justin[/quote]

Some years ago (circa 1982) the cost of a chip spin at National Semi in the Interface Products division, according to R. V. "Balu" Balakrishnan, who had just gotten a job there, was a little over 100k. You got your samples and had already programmed a test jig. It was the day of reckoning and great anxiety. One very bright chip designer who had insisted on working alone, without peer oversight, had one too many failures at this stage and had to clean out his desk within an hour of the last one :shock: .

I saw a phono pre recently (a home-brew though) with big tuning capacitors as part of the RIAA compensation. Oh, and the standard AM radio tuning caps were 365 micro-micro-farads (as they said in those days).

Keith O. Johnson believed in balanced signals wherever possible. He was bringing out both sides of the cartridge (or tape head) back in the '70's in his home system. He had the preamp in the base of the 'table, and it fed a line-level stage and then twinlead at rather high voltages across the room to powered speakers of his own design. In fact everything was of his design except the 'table.

I was privleged to be his guest on one occasion, in the pre-CD era. The sound was quite awesome from vinyl, but truly spectacular when playing back master tapes on his modified three-track machine they were recorded on---the machine that was the mainstay of Reference Recordings for some years, until Keith had tinkered with digital audio enough to be happy with it.

As I oohed and ahhed, KOJ pointed out that the actual THD performance etc. of his system was relatively modest. But he was quite keen on his belief that the ear was remarkably sensitive to asymmetrical responses, and worked hard to avoid such.

Quite a character.

It's true that at best cartridges are not themselves ever going to deliver a gigantic dynamic range, even with a noiseless amplifier and noiseless loading resistor, until/unless we get ~room-temp superconductors. And of course this is all before the needle drops into the groove.

However, the balanced preamp actually makes it harder to get low preamp noise, except for the advantage of induced noise/hum of the non-stochastic variety being rejected through the reduction in common-mode signals. Again, the right way to go in many respects, since you can add extra silicon if you want to preserve your already-low preamp noise numbers, and provided you are still in the series-noise-limited regime.
 
[quote author="bcarso"]

Keith O. Johnson believed in balanced signals wherever possible. He was bringing out both sides of the cartridge (or tape head) back in the '70's in his home system. He had the preamp in the base of the 'table, and it fed a line-level stage and then twinlead at rather high voltages across the room to powered speakers of his own design. In fact everything was of his design except the 'table.
-----

However, the balanced preamp actually makes it harder to get low preamp noise, except for the advantage of induced noise/hum of the non-stochastic variety being rejected through the reduction in common-mode signals. Again, the right way to go in many respects, since you can add extra silicon if you want to preserve your already-low preamp noise numbers, and provided you are still in the series-noise-limited regime.[/quote]

That's interesting... I published a phono preamp kit back in 1980 and got pinged by some guy in TX (sorry don't recall his name) who managed to convince the patent office that he invented balanced input phono preamps. :roll:

I ordered a copy of his patent wrapper (the entire file) and the discussion between him and the "wet behind the ears" patent examiner was amusing. He apparently was completely ignorant of the concept. Less amusing that this guy was granted a patent. I sent him a copy of an old tube manual phono preamp using a transformer input, pointing out that it was "balanced". By law if an inventor becomes aware that his patent is invalid he is obligated to notify the PTO.

On the subject of noise and balanced vs. single ended inputs, I guess in theory there is nominally 3dB more but not for all input noise terms.

For the case of a bridging input termination (>10x termination vs. source) there is some cross cancellation of input current noise. In the case of approx 1.5k source impedance and 47k total termination (2x23.5k). More than 90% of current noise noise present at + input shows up at - input and vice versa. The noise voltage terms will still combine with the expected +3db so for JFET designs the savings is reduced.

Perhaps for a MC amp that is more likely to use bipolar this may make a modest difference. While it looks like there could be some cross cancellation of termination Johnson noise, this will be dominated by the source impedance shunt which is differential and won't cancel.

This does allow the designer the flexibility to play with normal vs CM input impedance. Instead of 2x 23k to ground, those input resistors could be 1M with a 50k in parallel with the cartridge. This would further improve the cancellation of current noise. Of course just because you can doesn't mean you should. More useful IMO for bipolar input device approaches..

YMMV

JR
 
It is amazing what gets through the patent office. At least they have patent purgatory now, where interested parties can view submissions for a while and complain if prior art is being missed. But who has the time?

I'll have to ponder that noise current cancellation argument a bit longer.

I remember this astronomer that was an instrumentation specialist, Pete Stockton if memory serves, who advocated differential circuits everywhere, despite their not being strictly speaking optimal for stochastic noise. But his experience with real-world interference, especially in the telescope environment, drove the practice. I doubt that the situation has gotten all that different today, and to make matters worse the observing is mostly done remotely, so you are not at liberty to go out and do a little tire-kicking of the equipment if you're getting peculiar data.

One patent, speaking of noise, that really bothers me is the one of Nelson Pass's that was mentioned in here as having been licensed to TI (?). First, it seems incredible that the topology has not been used before. Second, the argument for distortion reduction is plausible, but I think not very much for noise.

There was one Pass Labs advert a while back that made a claim for signal-to-noise, or perhaps it was described as dynamic range, that on the face of it was absurd. Also, when one of the big Pass amps was tested by Stereophile, the performance was quite shy of the specification, and rather than bloody the letters page (I had learned they don't much like critical technical discussions, after I sent a blast that pointed out errors in a tutorial written by S'phile staff) I corresponded with Atkinson via email.
 

Latest posts

Back
Top