Discrete Op Amp: Why is This Noisy?

[Joe Kramer]

Hey Friends,

I was expecting this circuit to be as quiet or quieter than a IC, but not so.  It's not very noisy by any stretch, but it hisses just enough to be noticeable, even when plugged into a quiet, low gain amp.  With apologies for the artwork, here's the schemo:

Code Select Expand


               D I S C R E T E   O P   A M P


                     9V
                      |
               +------+-----+
               |            |
               Z            Z
          5.6M Z            Z 100ohm
               Z            Z
               |            +------------------+
               |            |                  |
               |      +---- | -----------+     |
               |     c|     |c           |     |e
          1uF  |   b|/       \|b         |  b|/
   IN >---||---+----|\       /|---+      +---|\
               |     e|     |e    |            |c   1uF   36K
               |      +--+--+     |            +----||---N N N-+---> OUT
               |         |        |            |               |
               Z         Z        |            Z               Z
          560K Z    100K Z        |            Z 3.3K          Z 12K
               Z         Z        |            Z               Z
               |         |        |            |               |
               +---------+--------+-----+------+---------------+
                                        |
                     (2SC1583)          G    (2N5087)



The resistor divider at the input gives about half the supply voltage on the collector of the 2N5087.  The resistor divider at the output knocks the gain down to about unity.  The dual 2SC1583 and the 2N5087 are reputed to be low noise components, and all resistors are 1/2w metal film.  On the actual build, I decoupled the supply with a 33uF tant and a .1uF ceramic.  Also, there's a 470pf from the base (input) of  Q1 to ground.  Anyone have any ideas why this is not as quiet as it might be?  All help much appreciated!

Regards,
Joe   

[brett]

Hi.
I'm no expert, but I wouldn't usually like to put a 5.6M resistor to the base of a gain-stage transistor because of the amount of thermal noise the transistor puts in.  Check out "noiseless biasing" for info on simple setups that don't inject all that noise.

(A simple approach would be to connect the 5.6M and  560k away from the base.  From their junction, connect (1) to the base with a moderate-sized resistor (100 to 470k?), and (2) to ground with a 22uF cap.

I'm sure that the gurus will have other ideas, too.

have fun

[Coriolis]

Have you tried asking on this forum?

http://prodigy-pro.com/forum/

Not to shoo you away from the great minds on this forum, but these prodigy-people babble about discrete opamps all day long....
There are a handful of people over there that seem to have a lot of hands on experience with this, so maybe woth a try.

C

[WGTP]

http://aronnelson.com/gallery/KHE/Boss_ROD10_Schematic?full=1

Are you running it open loop?  If the gain is real high, that might cause noise.  Might also try different transistors to select for the most quiet.

Here are some Boss discrete op amps.

[Jay Doyle]

Well I can see a couple of things.

First off, a previous poster was right about that 5.6Meg resistor, it will inject a ton of thermal noise, not to mention limit the amount of current into the input transistor. For noiseless biasing and control over the bias, hook up a 50k trim pot with the outside legs to V+ and ground and connect the wiper to the input through a 100k-220k resistor, also decouple this bias supply with a 10-100uf cap from the wiper to ground. This way you can dial in the bias, and supply enough current to the input.

Secondly, the 100k on the emitters of the diff. pair is too big and limits the amount of current through the diff. pair. I'd drop it down to 22k.

Third, you need to connect the emitter of the output transistor to V+ independently, not in conjunction with the 100ohm resistor on the diff. pair.

And last, and probably most importantly, you ARE running this open loop which means max gain, which also means max noise.

The collector of the output transistor is the output and the base of the transistor in the diff. pair that DOESN'T receive the input signal is the negative input for the opamp. Your bias is so screwy with the 5.6Meg because you have the neg. input attached directly to ground when the opamp wants it to be at the same voltage as the pos. input.

You need a feedback path from the output to the negative input, just like in a regular opamp, for this to work properly.

Discrete opamps are quiter than regular opamps, you are just missing a couple of things that make your circuit an opamp.

Good luck,

Jay Doyle

[WGTP]

http://www.diystompboxes.com/pedals/schems/DiscreteBoost.pdf

And then you have this.

[Jay Doyle]

Good God. Has it really been almost five years since I designed that? 

Wow.

C5 is redundant and not needed and the layout is horribly big but in essence, that is it.

I hope it helps.

[R.G.]

Jay Doyle and Brett are right - you're injecting a lot of noise in your biasing. Change those biasing resistor to 1/10 of what they are now and put a 1M from their junction the transistor base. Decouple the biasing junction at the bias string with a 1uF cap.

Noise should go down some.

It's called "noiseless biasing".

[Joe Kramer]

Okay, guys, thanks!  I had feeling it was in the biasing (really I did!).  Ordinarily, I'd do as everybody's said and use a 50K trimmer as the divider from supply to ground, and then a 1M or so from wiper to the base, not forgetting the decoupling, of course.  I didn't do that here because I was trying to save space.  This op amp is part of a larger circuit (an analog delay) and it's really crammed into a small piece of perf board.  I appreciate the more involved corrections from Jay, but I'm probably not going to do a lot of reworking on this one, mostly because of the elbow-room factor.  Besides, as unlinear as the circuit probably is, it does sound very good (aside from the noise).

Thanks so much for the help Brett, Coriolus, WGTB, Jay, and RG!

Regards,
Joe

PS: I'll try to get back here with a lab report after I make the bias changes. . . .

[Coriolis]

PS: I'll try to get back here with a lab report after I make the bias changes. . . .

Please do! I'd very much like to do some experimenting of my own with discrete opamps, since they're low parts count, and - so I hear - leaps and bounds ahead of monolithics if done right.

C

[Sir H C]

Other bits, I prefer using a transistor instead of a resistor for the current source, helps make for a longer tail pair. 

There are some cool discrete op-amp designs out there, Dean Jensen has one, there are some in many old mixing boards.  The biggest problem is getting those input devices to match.  Best technique is to use ICs that contain matched transistors, relatively cheap and you can do an entire discrete op-amp with one.  Don't listen when people say that there is an IC there, it is discrete transistors.

[Jay Doyle]

Please do! I'd very much like to do some experimenting of my own with discrete opamps, since they're low parts count, and - so I hear - leaps and bounds ahead of monolithics if done right.

It really depends on what you are going for. For our purposes, most of the time, just use an IC. Unless you are looking for something specific and interesting, like using a Ge transistor as the output device, there isn't much advantage to a discrete unit. In fact, they take up a lot of space.

The do offer an interesting overdrive, it seems to be smoother to my ears. And you can specifically control certain elements, like slew rate, that in an IC are most of the time immovable.

BUT, they don't have anywhere near the gain of an IC and to start to get closer to that ideal, you start adding so many parts that the work gets a bit overwhelming. This makes building things like a discrete Tube Screamer kind of frustrating as the gain isn't sufficient to get the same amount of distortion.

One easy way to improve a discrete opamp is to add a Class A or push-pull follower output stage so that it can drive things better, including it's own feedback loop.

I would recommend Joe D.'s Diode Compression Opamp if you are looking for distortion.

[Jay Doyle]

Please do! I'd very much like to do some experimenting of my own with discrete opamps, since they're low parts count, and - so I hear - leaps and bounds ahead of monolithics if done right.

It really depends on what you are going for. For our purposes, most of the time, just use an IC. Unless you are looking for something specific and interesting, like using a Ge transistor as the output device, there isn't much advantage to a discrete unit. In fact, they take up a lot of space.

The do offer an interesting overdrive, it seems to be smoother to my ears. And you can specifically control certain elements, like slew rate, that in an IC are most of the time immovable.

BUT, they don't have anywhere near the gain of an IC and to start to get closer to that ideal, you start adding so many parts that the work gets a bit overwhelming. This makes building things like a discrete Tube Screamer kind of frustrating as the gain isn't sufficient to get the same amount of distortion.

One easy way to improve a discrete opamp is to add a Class A or push-pull follower output stage so that it can drive things better, including it's own feedback loop.

I would recommend Joe D.'s Diode Compression Opamp if you are looking for distortion.

[Joe Kramer]

http://aronnelson.com/gallery/KHE/Boss_ROD10_Schematic?full=1
Here are some Boss discrete op amps.

Hey WGTP: Thanks for posting this Boss thing--cool and worth a long look at.  Just the fact they went with discretes says something.  ICs don't squash the same way, and even in a clean context, they aren't (to my ear) as immediately musical-sounding.  The one I posted is very forgiving when overdriven with a line-level signal.  I like to think of it as a pseudo analog-tape sound.  Sounds even more "tape-ish" when built with Ge xstrs.

Hey Jay: I've seen some push-pull followers designs, but I'd like to try Class A.  Is there something you could point to as an example? 

Joe

[WGTP]

I thought it very interesting.  Jay, any speculation as to why Boss is using discrete op amp in a number of their distortions?  Notice that the OD III and Distortion use 2 op amps in series for more gain.

It is cool to look at the circuit variations and then the variations in waveform.  Dragonfly has them posted in the Layouts Gallery.

http://aronnelson.com/gallery/KHE/Boss_ROD10_Notes_Corrections?full=1

[Jay Doyle]

Joe -

Unfortunately, I don't have a specific example but I can describe it, hopefully adequately:

Essentially, you are tacking on a follower stage to the output of your discrete opamp. I'll use a JFET as an example, but it would work for any type of transistor. Connect the gate of the JFET to the junction of the collector and resistor on the output transistor. Drain goes to V+ and the source goes to ground through a resistor 1k-10k (the value matters less when you are using feedback but in your case, because you are running open loop, it directly affects the output impedence). The junction of that resistor and the source now becomes the output of your opamp.

Does that make sense?

WGTP -

You know, I can only make guesses. First, I definitely think that it is a different tone when overdriven, not as harsh as an IC, because the gain is limited and it tends to squash, or round of the corners instead of just slamming into the power rails like an IC. NOTE THAT THIS DOES NOT MAKE IT SOUND LIKE A TUBE (that wasn't for you WGTP  ) Second, it is more subtle because of the gain. Note that they are only used in the 'overdrives' not in the megablaster-type distortions. The Blues Drives has two discrete opamps in series. Third, and this is speculation, but it could be cost. Transistors, to a company like Boss, are probably a penny a dozen, which is four discrete opamps (without resistors), while ICs would probably still be a penny or two a piece. And lastly, those waveforms are interesting but I think that what you are seeing in the differences has more to do with the filtering involved than the distortion.

Jay Doyle

[WGTP]

Yes, I agree the waveform differences are mostly from the filtering.  I guess the low noise might also be an issues.  The first OD1 appears to be set up (diode wise) as the SD-1.  The ODII is more like a Tube Screamer.  The ODIII us 2 op amps in series with no clipping diodes, just Jfets distoring.  The Distortion is similar with diodes to ground at the end, like a DS-1.  The Fuzz does not use Jfets.  I need to find one of those cheep.   

[Gilles C]

I bought the Boss OD-3 when it came out because I prefer overdrives to distortion types of pedals, and because I liked the sound of the OD-3. And they use that kind of discrete opamp in it.

When Boss refers to that kind of discrete circuits, they seem to call it Natural Overdrive, which is more realistic than calling it Tube Sound.

I may be wrong, but I think they use the discrete part for the sound, and the ICs part for sustain, and the more extreme overdrive setting. They call it a dual-stage design.

At that time, I wanted to work on a circuit of mine with these specs, and wanted to know how a real pedal would sound, and use it as a guide for mine. I finally never built mine...

It's not my fault, I like discrete circuits... So, it's been 5 years since I built Jay Doyle's Discrete Opamp? Hummm, time really flies.

Gilles

[Coriolis]

It really depends on what you are going for

Actually I was thinking more of micpre's and such. This is also where my post about Prodigy-Pro forum comes in. 

Ch

[Jay Doyle]

Actually I was thinking more of micpre's and such. This is also where my post about Prodigy-Pro forum comes in. 

Oh man. I'm not even going to attempt to try to get into that.  I don't have the dicipline or ability to get into studio-type stuff. Maybe Gus will jump in here, that

This is what you are looking for:

http://anklab2.pirit.info/Schemes/Other/je-990.pdf

Not exactly a simple build. It is basically and IC opamp without the IC.