Print Page | Close Window

A Replacement For The Op-Amp

Printed From: Graham Slee at Hifi System Components
Category: DIY AUDIO
Forum Name: Graham's Wrinkles
Forum Description: My hints and tips on how to squeeze that little extra from an old
URL: https://www.hifisystemcomponents.com/forum/forum_posts.asp?TID=4344
Printed Date: 22 Sep 2018 at 7:18am
Software Version: Web Wiz Forums 12.01 - http://www.webwizforums.com


Topic: A Replacement For The Op-Amp
Posted By: Graham Slee
Subject: A Replacement For The Op-Amp
Date Posted: 23 Oct 2017 at 10:16am
Since changes in electronic component manufacturing techniques started happening circa 2004 because of RoHS I became aware of a change in sound quality when using operational amplifiers in existing products as well as new designs. I initially put this down to the change in chemistry of the interconnections within them, but over the years I have not been able to prove this. Therefore there had to be an alternative explanation. I am now pretty sure it was due to the shrinking of chip size, in particular the use of (sub)miniaturised chips developed for surface mount in leaded (DIP) packages. It tends to make financial sense. I am also pretty sure some of the older DIP op-amps which were inherently RoHS compliant continued unchanged in manufacture but many are now heading for discontinuation in favour of their shrunken counterparts. The shrink alternatives exhibit a rather coarse sound signature, so for the benefit of our hearing I think it time to explore the alternatives to operational amplifiers, which means taking a discrete approach. One of the big problems here is the number of transistors that have been lost to progress, but there is still a healthy supply of some types where transistors must be used, such as in power amplifiers, and also in developing world products.

This topic is about my search for a suitable alternative and how that alternative might change how audio designs are configured, hopefully for the better.


-------------
Not simple enough for Google-Bot to understand...



Replies:
Posted By: RichW
Date Posted: 23 Oct 2017 at 12:53pm
Interesting topic.
Surface mount component technology has all but taken over the electronics manufacturing
industry with some very small components now around. It must be a headache & a worry to
rely on the continued availability of traditional through-hole ICs.



-------------
Elevator, Accession, Majestic, Reflex M, Solo ULDE, CuSat & Lautus.


Posted By: JamesD
Date Posted: 23 Oct 2017 at 6:15pm
Given that the latest surface mount audio op-amps have very low distortion figures and contain huge numbers of transistors is this where the slightly higher distortion of a discrete solution with a simpler signal path results in better sound?

-------------
Aren't ears brilliant


Posted By: Graham Slee
Date Posted: 23 Oct 2017 at 8:39pm


The above is a chip or die found inside a DIP package. Actually this is the chip for a 555 timer IC but it serves to illustrate what I'm talking about. From my distant past I remember breaking a 741 open and the size of the die (or chip - the sliver of silicon substrate and the components etched upon it) was roughly 3mm square. A small signal transistor die was around 1.5mm square. Today's op-amp shrink sized dies/chips will be a lot smaller and probably the size which a transistor die/chip took up. I am not going to try breaking one open, it is obvious from the size of a SOIC package and the SSOP which is even smaller, that a 3mm square die cannot be incorporated.

As sizes go down speed increases. That goes for logic ICs which now react at nano second switching speeds. This makes them offenders regarding emissions and active components which emit are often less immune to interference into the bargain.

Why speed goes up with a reduction in size is because paths are shorter and therefore less inductive. You don't need a coil for inductance - a straight wire has inductance - a circuit board trace has inductance - an etched path on a die has lower inductance and so does its interconnecting wires with the outside world.

Shrunken op-amps share this reduced inductance and so become more open to high frequency interference (which is why so many are now being offered 'EMI hardened'). Their bandwidths are controlled for particular gains - that hasn't changed - but inputs and outputs will more readily accept RFI.

The main characteristic of an op-amp is its voltage gain. Gains of 80 to 110 dB are quite common. Their bandwidths can range from a few MHz to many tens of MHz. The possibility of instability has risen due to their susceptibility to EMC disturbances because of their shrunken chip sizes, and at the same time the sources of EMC disturbances have multiplied, and so have the radiated powers thereof.

Measurably and in simulation they still perform the same. However, no test gear will ever match the resolving power of the human ear (those in good order that is). It would be fanciful to think otherwise, but plenty do.

My recent work to do with the topic Understanding Record Repro (https://www.hifisystemcomponents.com/forum/understanding-record-repro_topic4323.html) has shown me what a difference the removal of a high gain op-amp - being replaced by a rather "lossy" simple transistor stage followed by a more benign and older op-amp - can achieve. The sound quality is much more acceptable, and this is because of a number of factors:

1. The transistor gain is limited by its nature to 40 x Vc, which on a 9 volt rail is approximately 160 or 44dB - this is open loop before the application of negative feedback. With the negative feedback required to have a reasonable distortion measurement the gain is only about 10 (or 20dB). There is only around 24dB of loop gain compared with 60 - 90 dB using an op-amp for the same purpose.

2. The transistor chip size is similar in size to an entire op-amp but is only used for the single function of a transistor. If carefully chosen such that there is no shrink transistor version, the interference problem should be no more (or as it used to be). I think it will be difficult to shrink medium power transistors so this might be a good avenue to explore.

3. The sophisticated op-amp this replaced was a three stage device to obtain the large overall gain required, and several techniques are used in these designs for stability which largely demonstrates that they are easily upset by disturbances.

4. The post transistor stage op-amp is an older and much simpler two stage design with fewer compensation components (namely one capacitor), so little to go wrong here. I will say however that the shrunk size package exhibits a bit of 'nervousness' in its sound quality compared to the now obsolete DIP package it replaced.

Op-amps are basically multiple transistor circuits designed for a number of applications. Their voltage gain is in the order of 100,000 making itvery accurate to set much lower gains using 'simple' resistor dividers.

A single transistor does not offer this. To get the same, one would have to make a circuit out of many transistors which is basically the same as in the op-amp package - which seems silly, but some manufacturers do it for no objective improvement. It looks good to the uneducated masses.

For audio however, the complexities of an op-amp or a discrete version isn't really necessary. What is required is the ability to set a particular gain and know it is going to do that gain for the frequency spectrum required.

Much of this investigatory work existed just prior to the op-amp revolution which tended to put the blinkers on it. Designers (including the quite talented) tended to go with the flow and adopt the op-amp reasoning that it was better, or if not, just as good. And I for one agree with them. That was before the proliferation of powerful RF (EMC) disturbances and the shrinkage of packages required for what we call progress.

I have therefore been taking a trip down memory lane and looking at a number of configurations. My favourite is the old DC coupled pair (used in the Proprius voltage amp) which I learned in the mid 70s, and was used in numerous 60s and 70s audio products. However, it is not easily implemented as is an op-amp, and many of these designs changed to op-amps in the 80s.

Much of the groundwork which I feel is necessary in audio today was done by researchers such as Butler, and developed-on by such as John Linsley Hood and Douglas Self.

I will illustrate the type of stages I find useful in the coming posts in this topic.

-------------
Not simple enough for Google-Bot to understand...


Posted By: Graham Slee
Date Posted: 24 Oct 2017 at 7:54am
An op-amp internal circuit looks like this

Op-amp internal circuit

This is an OP27 op-amp. Transistors Q1A, Q1B, Q2A and Q2B are the differential input which in turn drive another differential stage via Q21 and Q22, which is Q23 and Q24. Finally it is turned single sided by Q26 whose output drives a complimentary output stage which has similarities to a power amp output stage. The other transistors shown serve as current sources, sinks and mirrors, and in fact there are more transistors than shown but their circuits are simplified to the circled arrows which depict further current sources.

This is a three gain stage op-amp: differential; differential; single ended (prior to the output stage), and it has to be that way to obtain the open-loop voltage gain which is 1.8 million!

It needs such a high gain for the user to be able to set an accurate gain using a simple resistive divider (shunt or series feedback) provided that the final gain is some 40dB (100 times less) than its open-loop gain, and usually much lower.

The compensation network which helps prevent it oscillating is R5, R9, R12, an unnamed resistor, C1, C2, C3 and C4. Part of this also speeds up its slew rate - enhanced slew rate that is - and it achieves 2.7V/uS which isn't all that good.

Many a phono stage has been manufactured using the OP27, and the sound is OK but not very inspiring, and this is due, in my opinion, to its complication and its necessarily complicated compensation, plus its slew rate enhancement.

This next circuit is the TL071

TL071

Although an FET input it still has the differential input which then delivers its gain to a single sided stage before its output stage. It is a two stage op-amp. It's open-loop gain is lower. Its compensation is simply the capacitor shown in parallel with the diode. And used as a phono preamp it sounds far better than the OP27. However, it is noisier and so a little hiss will be heard with the stylus out of the groove, but I'd rather have that than the contorted sound of the OP27.

The op-amps have differential inputs which a single sided signal doesn't really need. The noise reduction it offers is really intended for differential cancellation. Some swear it is necessary for a single sided signal, but I beg to differ. Do we really need a differential input?

Going back in history we find this single sided discrete gain stage by Butler from a 1965 article on cascading transistors for greater gain and linearity

Cascode by Butler

Excuse the drawing conventions and the inverted (by today's standards) power supply. It is a cascode circuit where the upper transistor assists the lower transistor to provide more gain, and a third transistor lowers its output impedance (an emitter follower).

Its open-loop gain (R1 set to zero) is around 300 but with something like 30k it will do a gain of ten and be quite well behaved. And for low gain my opinion is this is all that is really needed. In fact it was used in a 1979 article by Doug Self for a high quality preamplifier as one of its stages. This is the kind of thing I am exploring to replace the op-amp wherever possible in any new design.


-------------
Not simple enough for Google-Bot to understand...


Posted By: JamesD
Date Posted: 25 Oct 2017 at 8:42pm
Has the damage already been done to the music I wonder given that recording studios are probably using op-amp ICs in their equipment?

-------------
Aren't ears brilliant


Posted By: Graham Slee
Date Posted: 26 Oct 2017 at 6:51pm
Originally posted by JamesD JamesD wrote:

Has the damage already been done to the music I wonder given that recording studios are probably using op-amp ICs in their equipment?


Not only that but the control room acoustics might not be properly balanced - Mike Oldfield recorded QE2 in his living room because his studio upstairs wasn't ready in time.

Some recording engineers might have had high frequency hearing loss.

Also, some records were recorded to sound good on mass-market (cheap) music centres, and often have their highs artificially bright - Mike and the Mechanics "Living Years" for example.

Then there are records where one side is beautifully balanced, and the other side atrociously balanced like Gerry Rafferty's "Right Down the Line".

And as I've been saying a long time, the most popular cartridge type is MM, so what type of cartridge will be set up on the lathe's playback arm? And do we realise that MMs have a 0.5dB - 1.0dB lift around 10kHz to 13kHz, which might have been compensated for? MC by comparison doesn't.

So you have to have a large music collection in many different formats, and different cartridges when it comes to vinyl, and also different speakers and headphones, to gradually educate your ears using circuitry which will do its best to get it right.

It ain't easy.








-------------
Not simple enough for Google-Bot to understand...


Posted By: Graham Slee
Date Posted: 27 Oct 2017 at 8:56am
Because of the work by Butler on the feedback cascode with emitter follower (earlier post) we have a chance of producing a high gain discrete transistor stage.

If we take an 18V power supply and bias a single transistor for symmetrical output it will have a collector voltage of 9V, and from 40 x Vc we find it will give us a gain of 360. What we need is at least a gain of 1000 which is 60dB so we can apply 40dB of that as negative feedback to obtain low distortion for a gain of 10 (20dB) amplifier (40dB suggested by Harold Black of Bell Labs in 1927 which still holds true).

Butler would not have had the luxury of simulation we have since his work, and as it is extremely important to have high frequency stability especially in this RF dominated world of today, we find his circuit needs much improvement.

The following simulator circuit was the result of many hours of empirical juggling:

Inverting Discrete Opamp

It features the current doubling emitter follower current source (Q4 - or should that be 'current sink'?) suggested by Self. This not only lowers its output impedance but makes its output more linear - it is not relying on an emitter resistor to pull the voltage down on negative excursions. This linearity improves the performance. You will also notice several small value capacitors and these are needed for stability as the dominant pole doesn't dominate sufficiently - this being a problem of the cascode as far as this particular gain goes.

And this is the simulation result for bandwidth and phase:

Inverting Discrete Opamp Open-Loop Bandwidth

As expected, and it's the same with op-amps, the frequency response is very limited, but we have around 75dB (x5600) gain. With negative feedback of 40dB or more we can expect a much wider and respectable frequency response. But what we are looking for right now is bandwidth with stability, and we can see that we have a 0dB bandwidth of 7.5MHz and a phase margin of 81 degrees. We also have a gain margin of better than 20dB (factor of 10). This is the gain at -180 degrees, which is 360 degrees in an inverting amplifier like this.

However, unlike the op-amp, if we apply negative feedback by selecting a ratio for Rf and Rin we will find that stability is upset, and there will be a spike in the LF response.

Unless we use the same techniques as used in an op-amp it will always be like this. But we want to try and get away from the problems we think the op-amp dishes up, so we will therefore find that each stage we make using this (and similar discrete circuits) for a particular purpose, will have to be adjusted or tuned to behave for the gain we want - as they always were.

However, it makes a good starting point from which a number of desirable gain stages can be made.



-------------
Not simple enough for Google-Bot to understand...


Posted By: Graham Slee
Date Posted: 06 Feb 2018 at 8:06am
This next stage is the classic 2 transistor multiple feedback design using an extra transistor for a low impedance output.

It cannot be considered as a discrete op-amp but it is a great fall back for where a larger output swing is required than an op-amp which is limited by its supply voltage.

Here it is working on a 36V supply but by adjusting values it can operate at much higher voltages. The transistors would have to be suitably rated.

Classic 2 stage discrete amplifier

Variations on this configuration have featured in numerous audio products since the 1960s and the sound is often far more natural than can be obtained from most op-amps.

The Cs on the first transistor are input interference filters.

Bias voltages might seem a trifle difficult to discover but the clue here is R2 whose voltage supplies bias via R10/12 to Q1.

There is only one AC negative feedback loop which is via R8. All other negative feedback is at DC due to bypass capacitors.

Input slew rate is determined by R11, 8 which form the factor "m" in S = 0.3mFt. This is an op-amp formula but works for this circuit too. The expected slew rate for its bandwidth would be quite low otherwise.

Unlike op-amps this stage is all class-A.


-------------
Not simple enough for Google-Bot to understand...



Print Page | Close Window

Forum Software by Web Wiz Forums® version 12.01 - http://www.webwizforums.com
Copyright ©2001-2018 Web Wiz Ltd. - https://www.webwiz.net