Another discrete (germanium) op-amp

[Passaloutre]

What's causing the extreme asymmetry at high gain?

[samhay]

Many op-amps have some asymmetry in how close they can swing towards the power rails. This is a somewhat extreme example:
The circuit can swing to within about a diode drop of the negative rail (ground here), but it can't get very close to the positive rail.
You can get closer to the positive rail if you increase the current through the circuit (R4+R5). You can offset this somewhat by playing with the trimmer to bias a little closer to ground - you could do this by ear, but would be more accurate if you used an oscilloscope.

[Passaloutre]

I built your layout, but it doesn't work. The problem I think is that the layout is so tight, all the metal transistor cases are touching each other. The transitor case in a 2N404 is internally connected to the base, so the whole thing is shorting through the cases. Short of covering all the transistors with heat shrink, this layout isn't going to work for me. All of my transistor bases are sitting at the supply voltage.

Oh well, at least it sounded good on the breadboard. Should be fine for anyone with plastic transistors.

For what it's worth, it looks like you switched the output 1k and 10uf on your layout relative to your schematic. I'm not sure it matters, since they're in series

[Passaloutre]

All is not lost! I was able to slip these little heat shrink condoms on every other transistor to insulate them from each other (you can see the size difference between the 2N404 and the 2N2222 at the bottom). Tomorrow I'll set the bias trim and box it up.



I'm glad this one didn't end up in the box of failed circuits, because it sounds really good. I didn't change the bandwidth from the original schematic, I kinda like the loose bass, and my attempts to tighten it up killed it's character.

I suggest anyone who wants to build this spread out the layout a bit. 8 transistors is a lot for that much space, unless you have TO92 packages.

[samhay]

Excellent - it's working now?

For those who wish to build this with more portly transistors, a few tweaks to the layout should make this a little more comfortable for all involved (note: layout not verified!).


(larger version: http://i.imgur.com/xO4XjYE.png)

[Passaloutre]

Yes, it's working wonderfully. And has surprisingly little noise, even outside the enclosure at high gain settings.

You can consider that first layout verified, though like I said I reversed the 1k and 10uf, to make it match the schematic.

[samhay]

Excellent.

>I reversed the 1k and 10uf
They are in series and nothing is connected between them, so the order doesn't matter - they are equivalent.

[Frank_NH]

Thanks Samhay and Passaloutre for this excellent and interesting topic.  It just so happens I've been fooling around with a simple JFET/PNP discrete "op amp" similar to the one from the Blues Driver (below).  I simulated this in LTSpice and it showed some promise, so I created a full tube screamer type circuit out of it (including an op amp stage for the active TS tone control).  I used 2N5457s for the JFETs and a 2N2906 for the PNP.  I have it on my breadboard now and it sounds great (and best of all - no JFET drain trimmers!).    The sound to me is a little smoother than a conventional tube screamer, but I haven't done any real A/B comparisons yet.   I'd like to try improving the "op amp" stage somewhat, and to investigate how different FETs and PNP transistors affect the sound.  It seems to be pretty forgiving of JFET/PNP variation as I just pulled four 2N5457s and two 2N3906s from my parts collection at random and it fired right up (may have gotten lucky there).  I can post my Spice schematic if people are interested.

[Passaloutre]

I'd like to see the schematic! Are you the same Frank_NH from the gretschpages?

[samhay]

Frank - that circuit snippet seems to pop up quite often, which may be testament to it being worth a look!

[Frank_NH]

Yep, I'm on the Gretsch Pages too.     I'll try to find an upload host for the schematic image since you can't post image files here.

BTW - I first saw this simplified "op amp" topology in the excellent Stompboxology issue called "Going Discrete", available here (see Fig. 14):

http://moosapotamus.net/files/stompboxology-going-discrete.pdf 

[Passaloutre]

try imgur.com. It's dead simple. I'm otter on gretschpages

I think I had seen that stompboxology article before, but it makes a lot more sense to me now than when I first read it a few years ago. Still I've learned a lot from this thread, as I've not dealt with opamps very much, and certainly not their inner workings.

Sam, do you have a name for this circuit? Just so I know what to call it when people ask about my awesome booster

[R.G.]

Some general observations:
- the point of feedback-stabilized amplifiers in general is to force the insides of the amplifier and any quirks it has to not be "visible" outside the amplifier, only the amplification set by the feedback components
- feedback amplifiers in general and opamps in particular do his by having much more open loop gain than they need, so they excess gain can be "wasted" in covering up the internal open loop faults
- the bane of feedback amplifiers is oscillation; a stable feedback amp is one which has a low enough gain so its compensation network can get its open loop gain under unity before the internal phase shift gets to 180 degrees
- another way to say that is that the bane of feedback amplifiers is compensation; the open loop gain of the amplifier must be knocked down under unity by the time the internal phase shifts add up to 180 degrees.
- compensation nearly always cuts the open loop gain dramatically as frequency rises
- a discrete opamp that is tolerant of subbing in many different transistor types means that it is either low phase shift or low frequency response, or both
- the point of using a discrete opamp instead of an IC one nearly has to be to let some of the "funny stuff" from the open loop gain shine through, not being covered up by the feedback gain of the opamp; accordingly, if that's what you're after, you want the open loop gain of the parts in the open loop gain path to be low; low open loop gain covers up the internal quirks less well
- germanium in general has poor high frequency response, which is another way of saying high phase shift at modest frequencies
- accordingly, if it's stable for germanium in the diffamp or transresistance/voltage gain stage, it's overcompensated for silicon - which may be OK if what you want are to hear the quirks
- clipping behavior is by definition where the forward gain of the opamp circuit runs out as it gets near a power supply; so the higher the open loop gain, the more likely clipping will be sharp cornered and the same; the lower the open loop gain, the more of the softness of any clipping from the internal stages will show through

I think that using low gain devices in a discrete opamp will give more possibly good sounding "funny stuff" without feedback covering it up.

[Passaloutre]

This was a super interesting post. I'm having to read through it a few times to figure it out. This part:

I think that using low gain devices in a discrete opamp will give more possibly good sounding "funny stuff" without feedback covering it up.

When you say low gain, do you mean low gain in the grand scheme of transistors, or do you mean particularly low-gain germanium transistors? The one I just built has ~60s hfe (measured using your transistor tester from the fuzz face article) transistors in the signal. I've got some with gains in the 20s-40s, would those be better?

[Frank_NH]

OK.  Here's what I'm playing with on my breadboard.  Basic tube screamer without the buffers.  Op amps implemented using the simplified Blues Driver "op amp" topology.  Asymmetric clipping.  Slightly modded tone filters.  Sounds pretty good through my Roland BC-60 set clean, like a slightly smoother TS-9.    No deliberate JFET matching was done for the differential pairs, though that would probably be desirable.  I'm going to try some different transistors to see how robust the design is to part variations.

One of my goals with this project is to experiment with JFET overdrive designs that don't require trimming drain resistors.  All this thing needs is a bias voltage at the first stage.  You could add a 50K trimmer to allow offset bias, but that may not be needed for a simple tube screamer as long as you are reasonably close to half supply.

[Frank_NH]

Hopefully, my schematic image in my previous post is showing in the above.  I notice that on another computer it was filtered out (not showing)...

I can try hosting it on Google Drive if need be...

[Passaloutre]

I can see it fine! Looks good. I can agree with the sentiment about trimming JFETs. I like the tone of a lot of JFET designs, but I can't be bothered to install and tune all those trimmers.

[Frank_NH]

Thanks.  So far, the design seems fairly forgiving of transistors (e.g. tried some vintage 2N5138 PNPs and it worked just fine).  The sound is very much mid range tube screamer.  May put the standard TS BJT buffer at the output just for fun.  Not sure if the primitive op amps can be improved for this application...

[samhay]

RG:
- the bane of feedback amplifiers is oscillation; a stable feedback amp is one which has a low enough gain so its compensation network can get its open loop gain under unity before the internal phase shift gets to 180 degrees
- a discrete opamp that is tolerant of subbing in many different transistor types means that it is either low phase shift or low frequency response, or both

Good point. Phase shift should be modest (and simulation agrees) when the transistors are not frequency-limited.
I don't know what will happen in the real world, but I'm guessing the cap across the feedback resistor aught to keep any high frequency oscillations under control.

- I think that using low gain devices in a discrete opamp will give more possibly good sounding "funny stuff" without feedback covering it up.

That's the hope.

[samhay]

Sam, do you have a name for this circuit? Just so I know what to call it when people ask about my awesome booster

Not really - I have boxed it up, but haven't added labels yet. If Another Germanium Discrete Op-amp Booster doesn't roll off the tongue then feel free to call it whatever you fancy.