[ TNT | Who we are | Listening tests | HiFi Shows | Factory Tours | HiFi topics | Tweakings | Inter.Views ]

The TNT HDOA - High Dynamics Op Amp

How to overcome op amp limitations

[Italian version]

Two of op amps' greatest faults generally listed are very high inherent gains and relatively poor dynamics in terms of output voltage swings. As for their high internal gain, that's the way they are, and even though measures can be taken to tackle this problem, they usually yield rather complex surrounding circuitry. Using the "keep it simple" method, this will not do, and is thus best left alone and a careful op amp selection is recommended.
In addition, they are often incapable of delivering larger currents sometimes needed for driving low impedance lines (low in context of audio, i.e. impedances below 10 kOhms). Usual absolute limits are quoted as 50 mA, but in practice, only smaller values can be reliably achieved.

So why op amps?

However, regarding their output voltage swing, relatively simple measures can be undertaken to overcome this problem with excellent results. But, before you do any such thing, ask yourself - why use an op amp in the first place, only to have to subsequently solve their inherent problems?

Because integrated circuit op amps offer certain advantages which could be very useful in audio design and cannot be overlooked. For one, the very fact that they are integrated, i.e. everything is on the same substrate, guarantees good matching between stages, as well as a very uniform drift - not at all like discrete circuits, which can suffer from problems of mismatching components and sometimes strange drifts, stray capacitance, and such like, not to mention the wide variance in what are supposed to the same components.

The second very important reason is that ICs exist - and at very reasonable prices - which exhibit excellent noise figures, this in turn yielding good signal to noise ratios, so important in audio, and not at all easy to achieve with discrete circuits.
Some integrated components, such as for example PMI's MAT 02/03 dual supermatch transistor series, offer differential noise levels below the theoretical limit of resistor noise. To the best of my knowledge, no other component in discrete technology, let alone tubes, can even approach such limits.

The third reason is their compact size. Because they are small, it becomes possible to construct very compact circuit boards, greatly shortening signal path lenght, and thereby also decreasing any possibility of stray capacitance, which may lead to increased distortion at best, and oscillation at worst. The designer still has to be very careful, but his chances of making mistakes are significantly reduced.

Now, wouldn't it be nice if we could have an op amp with a large voltage swing of say 80V peak-to-peak, with say 200-300 mA of current available? And with very low noise? And with reasonable speed of not less than say +/- 10V/uS? And not too expensive, thank you?

Yes it would - and here it is. Introducing the HDOA - High Dynamics Op Amp.

High Dynamics Op Amp (HDOA)

Basically, the circuit uses a well known specimen, Precision Monolithincs Inc. OP37 ap amp. This is the decompensated version of the well known OP27 chip; it is stable with gains greater than 5, but in return, instead of a typical slew rate of just 2.7 v/uS, it offers typically 15 V/uS.
Since its slew rate is a function of its output voltage swing, increasing this swing will also increase the slew rate - good news for audio, even if 15 V/uS is already enough, technically speaking.
OP37 was chosen for several reasons:

  1. It is a low noise chip, offering typical noise figures of 3.5 nV sq.rt. of Hz, and low noise is a must in audio;
  2. It is readily available, very important for self-builders, and has a reasonable price to boot;
  3. It is relatively fast "as is", right out of the box, again good news for audio, and in this circuit, its speed is both welcome and somewhat increased;
  4. Its typical current requirements are modest, around 3 mA only, which makes it easy to construct low noise supply rails (also cheap);
  5. I happen to believe, on basis of my experience, that nobody makes audio chips as well as PMI and AD, and OP37 is a classic, and
  6. I just happened to have 30 or so pieces lying around the place, doing nothing.

[Circuit diagram]

As shown, the circuit will quite easily drive a very low impedance line, but it is not recommended for lines with impedances below 50 Ohms. On the other hand, it will drive any power amp I have ever seen or heard of to its maximum power output, and beyond, to self-destruction if need be.
It can also be used as a headphone amplifier; if your phones have an impedance below 50 Ohms, reduce the power supply rails to say +/-24V (still enough swing to make your eyes pop in and out with the tune). It can be used as a preamplifier, but for that, you'd need two such circuits in series per channel, with volume and balance pots in between. Since each is connected as an inverting amp, two in series will produce a non-inverted output.
Above all, this is a highly versatile circuit. Power supplies can be anywhere between +/-18...40V, possibly even higher, but I haven't gone over that limit. However, it will need to be adjusted for different voltages.
The output mode can also be varied from class AB to pure class A, but this also involves some planning and adjusting. But, let's look at the circuit.
OP37 accepts the signal at its inverting input, while its noniverting input is connected to the signal ground. R1 defines its input impedance, in this case 33 kOhms. That's high enough to enable relative immunity from preceding electronics' possible higher output impedance, yet still low enough not to invite too much trouble from other possible noise sources, such as RF. Overall gain is defined by dividing R2 with R1 (in this case, app. 10:1) and it is inverted, i.e. with a minus sign before it.
For more gain, decrease the value of R1 - don't touch R2 unless you are proficient with electronic ciruits. If your needs are exotic, put two resistors in parallel to obtain the required balue.
R3 and R4 stabilize the output as constant impedances. C3 and C4 serve to stabilize the output of the circuit and are options. By that, I mean that the shown values help make the amp more stable with 50 Ohm lines. If you intend to drive only higher impedance lines with the amp, say 10 kOhm and more, both capacitors become optional. You can change the values or do away with them altogether. However, you may also experience oscillation, so removing them is not recommended.
Q1 and Q2 act as voltage stabilizers for the op amp. BC 639/640 transistors are shown, but you could just as well use BC546/556B types, or BC140/160 if you like, depending on the bias current you want flowing through the output stage.
The diode following the zener is there to thermally stabilize the voltage, and the added benefit is that it compensates the voltage drop on the transistor, thus enabling the IC to have truly 15V, not 15-0,6V as would be the case without them. Another useful aspect of this arrangement is that the supply voltage to the IC is constant no matter what you use for rails, and you can use anything from +/-18 to +/-40V - whatever it is, the op amp will always have +/-15V, period.
R9 and R10 have two jobs. One is to enable the op amp to receive the current it needs to operate and the other is to bias Q3 and Q4. By changing its value, we also allow more or less current to flow under no signal conditions through Q3 and Q4.
It works like this: the more current you allow to flow through Q3 and Q4, the greater the bandwidth and the lower the overall distortion, and vice versa. On the other hand, the more current flows through these transistors, the hotter they become, so make sure you have some coolers at hand. The GREATER the value of these resistors, the more current flows through Q3 and Q4, and vice versa.
As a general principle, R9 and R10 should have a value between 100 and 180 Ohms. No sense in overdoing it, as you gain nothing but problems to cope with. This leads us to a key question - when is an operating class a true class A?
The answer is really quite simple - when your quiescent current (i.e. current used with no signal applied) is equal to what it needs to be for a full power swing into a load. Thus, if your load will have 10 kOhms or more, that's one thing, but it's a different matter if you drive a say 50 Ohm load. However, start experimenting with values of 100 Ohms and work your way up.
The greater the R9/R10 value, the more bias current will flow through the output transistors. Therefore, start low and slowly work your way up, not vice versa.
For a 10 kOhm load and a +/-35V peak-to-peak swing you need very little current, no more than 3.5 mA, while for a 50 Ohm load you need 700 mA for full power class A operation. Obviously, we have several variables here - load impedance and supply line voltages being the most important. The lower the load, the more current we need; the lower the supply voltage, the lower the output swing, hence the lower the power dissipated.
Obviously, while BD241/242C type transistors are shown (100V, 40W, 3 MHz), if you need less output current while still being in true class A operation, you could easily opt for less powerful transitors, such as BD139/140 (80V, 12.5W, 50 MHz) for less demanding applications, like line drivers and preamplifiers. For headphones and low impedance lines, higher power types are recommended for safety of operation, if for no other reason. If you want greater power output, you could use Motorola's MJE 15030/15031 transistors (120V, 8A, 50W, 50 MHz).

Basic characteristics

As shown, the circuit will have an output voltage swing of +/- 33V, enough to drive any power amplifier to self-destruction. You won't need all that, to be sure, but you will also be free of any compession under real life conditions.
The frequency response is flat to -1dB beyond 100 kHz and is within +/- 0.2 dB 20...20.000 Hz. That should be flat enough for everyone, especially since these figures were obtained at full power.
The signal-to-noise ratio, assuming 1% metal film resistors, should be better than -90dB below 10V output (unweighted), and can be improved upon (see later).
Slew rate will vary from 15 to 40 V/uS, depending on several factors, such as the circuit's power bandwidth and output voltage. Either way, it's good enough for audio purposes.
THD was below the 0.01% limit of our test equipment even at a 11Vrms output, by which time your power amplfier has long ago clipped itself to Happy Hunting Grounds.
No sense in bothering with anything else.


Well, if you try it you'll soon discover this is a different ball game. Compared to a standalone op amp driving a typical audio line input with a 20 kOhm input impedance, which is way above the problem line with an op amp by itself, you will first notice the full, deep and powerful bass lines. No wonder, that's where the current goes.
You may also notice that the ambience is better rendered, more soundstage depth than with the IC alone.
But it's the high range that will really surprise you - very unlike any IC op amp I have ever heard to date, though of course I haven't nearly heard them all. But there is an ease, a feeling of freedom than no IC has ever given me. And enough warmth to suggest possible tubes hidden away somewhere.
Forget compression and distortion - they don't apply to you any more, at least not in any meaningful terms. I mean, 0.01% - how far should we go to make already inaudible still more inaudible?


This basic circuit can be adjusted to many purposes; let's take a look at the most common ones.

A) Preamp
If you want to build a preamp, you will need two such circuits per channel, the first acting as a buffer, and the second providing the actual gain. In that case, you will need to install one dual pot for volume and two single pots for channel gain after the first curcuit, and then the second circuit. For that purpose, I would advise power supply rails of say +/-30V, no more.
The reason is as follows - with 60V peak-to-peak, you can use relatively low noise transistors at the output of the first stage, such as BC 639/640, simply because there, you still don't need much current, since it will be driving relatively pure resistance loads (pots). Q1 can then be BC550C, Q2 BC560C, Q3 BC640 and Q4 BC639. R9 and R10 should be 100 or 110 Ohms, and R1 should be 47, 51 or 56 kOhms - make sure it's no greater than 56 kOhms, since OP37 is stable only with gains greater than 5.
I would advise 47 kOhm, which is high enough as the overall input impedance. But, if your buffer has a gain of 6, and your amp also has a gain of 6, your overall gain is all of (6x6) 36. Assuming amplifier input sensitivity of 1.5V, that would make your input sensitivity something like (1.5:36) 42 mV! Thus, I would advise using an OP176 for the first stage - it has only a little more noise than OP37, but is faster and is unity gain stable, so it can have less gain (2 or 3 would be just fine).
I would also advise installing low pass filters on the input board, with a cutoff frequency of say 100...150 kHz, possibly less.
The second stage could then be as shown, but use BD 139/140 transistors instead of BD241/242C. Make R9 and R10 100 Ohms. Install a high pass filter after the amp, connecting in parallell 1uF polypropylene, 1uF polycarbonate and 2.2 uF bipolar, with a 100 kOhms resistor to ground.

B) Headphones amplifier
Use as shown, possibly with a 10 Ohm/2W resistor in series with the output. Try several values for R9 and R10, as this will change the class A to class AB crossover point, but also the dissipation of the output transistors Q3 and Q4. Make sure they have decent heatsings screwed on, use silicon paste liberally. I haven't tried it as a headphone amp.

C) Low impedance line driver
Use as shown, but substitute Q1 for BC140 with a star-shaped heatsink and Q2 for BC160 also with a star-shaped heatsink on. Change R9 and R10 values as appropriate (150...180 Ohms), possibly, in extreme cases, substitute Q3 for MJE 15031 and Q4 for MJE 15030.

Power supplies

Their capacity depends on what you plan to draw from them. The more current you need, the more complex it's likely to become. However, since the supply voltage for the IC is fully regulated, it will remain constant no matter what you decide regarding the power supply. This in turn allows you to avoid any voltage regulation and go for "passive" power supplies. Some people think they sound better than regulated power supplies.
There are so many available projects that I decided not to supply yet another power supply schematic. I do suggest you look over the virtual battery operation principle on Andrea Ciuffoli's web page http://www.geocities.com/ResearchTriangle/8231/. Use all of it or some of it, if you like.
Power supplies are a wide topic and cannot be covered here and now; anyway, they merit a special edition all by themselves. Do remember that the worst place to save money is on the power transformer.
As for shown bypass capacitors, please don't change the tantalum caps for some other fancy type. Trust me, if tantalum is good enough for professional measuring instruments, it will be good enough for your and my audio. Nobody has come up with better decoupling caps so far (though they are not much good for anything else).

Parts List


Go easy on changing values of R9 and R10. Remember, the larger the value, the higher the bias current of Q3 and Q4. Start with smaller values, like say 100 Ohms, which will decrease bias current. Put your finger on the isolated, plastic part of the transistor and feel its temperature change. If it starts to heat up quickly, turn the power off and install smaller resistors, say 91 Ohms. If it stays stable and heats up moderately after say 15-20 minutes, then stays more or less the same, you can leave it like that.
If you insist on pure class A+, meaning still more current than is needed for normal class A operation under given circumstances, it's a great idea to install some heatsinking on Q3 and Q4. How large and good, and hence expensive, depends on how far you intend to go.
If you want to be able to switch on saying: "Warp speed, Mr. Sulu!", then good U-profiled heatsinks will suffice; but if you switch on shouting: "Tora! Tora! Tora!", then high efficiency heatsinks will be required.
Whichever you use, don't forget the mica insulation and use silicon paste very liberally; bolt sinks on securely, make sure nothing wobbles.


You are free to tweak to your heart's content, of course. Here are some suggested possibilities.

You can use another IC op amp instead of the OP37. I suggest you try out OP176 and AD825. They will provide a very different sound, OP176 being warmer and spatially very good, while the AD825 will bring in tremendous speed, clear up your treble, but with a slightly colder, more detached sound than the OP176.
This is my opinion, you may think and hear differently. Also, there should be no problems using other chips, from reputable sources such as Burr-Brown, Motorola, National Semiconductor, Texas Instruments, SGS-Thomson, etc.
Remember that whenever you have frequency compensation, to get the most out of any one chip, you will need to adjust that compensation to that particular chip, under your particular circumstances. There is no way anyone can foresee all possible combinations, however, the basic diagram shown will work for OP37.
Using another, faster chip, such as AD825, which is almost 10 times (!!!) faster, will neccessitate adjusting the frequency compensation if you want to have the full benefit of the extra speed.
These two capacitors (C3 and C4) in fact improve circuit stability. Another combination of op amp and transistors, driving higher impedance lines (15 kOhm and more), may in fact allow you to throw them out altogether, or drastically reduce their values. But this has to be tried out for every specific configuration, as there are far too many variables which are impossible to know in advance.
That's it. Happy building.

© 2000 Copyright Dejan Veselinovic - http://www.tnt-audio.com

How to print this article

[ TNT | Who we are | Listening tests | HiFi Shows | Factory Tours | HiFi topics | Tweakings | Inter.Views ]