Adjusticator noise - stumped

[waltk]

I made this version of RG's adjusticator, and it works great except for a Shhhing noise when the gain is turned up all the way.  This is the high-Z input version, with a change using half of the opamp to buffer the bias voltage.  I used a 1K resistor for R12 that sets the top end boost at +20db.  For the last quarter turn of the gain pot, I get a soft Shhhing noise (that increases as I turn up the gain).  Does anybody know what causes this noise - or how to get rid of it?  Things I've tried: using all metal film resistors, using a higher-quality opamp, using a 22pf cap around R6 (this creates a low-pass filter with the 1M input resistor - right?).  Did I mess something up using half the opamp buffer to the bias voltage?  I've built a few of these, and they all do the same thing, so I don't think its just a sloppy construction mistake.  Any help would be greatly appreciated.

[R.G.]

I made this version of RG's adjusticator, and it works great except for a Shhhing noise when the gain is turned up all the way.  This is the high-Z input version, with a change using half of the opamp to buffer the bias voltage.  I used a 1K resistor for R12 that sets the top end boost at +20db.  For the last quarter turn of the gain pot, I get a soft Shhhing noise (that increases as I turn up the gain).  Does anybody know what causes this noise - or how to get rid of it?  Things I've tried: using all metal film resistors, using a higher-quality opamp, using a 22pf cap around R6 (this creates a low-pass filter with the 1M input resistor - right?).  Did I mess something up using half the opamp buffer to the bias voltage?  I've built a few of these, and they all do the same thing, so I don't think its just a sloppy construction mistake.  Any help would be greatly appreciated.

I think you may be listening to the input noise of that 1M input resistor.

Try changing to the 100K lower impedance one and see if that lowers the hiss. If it does, you may have to use that second opamp as a buffer so you can run the lower impedance version of the variable gain amp. Or we may be able to figure something else out. Try lowering the impedance and let me know how that works.

[waltk]

Thanks for the quick reply RG.  I tried a 100K in place of the 1M, and didn't notice a change in the hiss.  While trying to locate the source of the hiss, I've ended up socketing just about everything on the board - anything else I might try?

Here's the layout - if that helps.

[Processaurus]

Does replacing R3 with a wire help?  You normally run the vref right to the non inverting input rather than through a big resistor.

[waltk]

Does replacing R3 with a wire help?

Yep.  That was it.  Thank you so much for the help.  I'd be embarassed to tell you how many other things I tried to get rid of the hiss.  Obviously I'm still on the steep part of the learning curve.

So now I have a high-impedance booster/attenuator.  My intent with this is to use it as a pre-amp in front of chip amps, and so far it has worked great (except for the hiss).  The gain pot on the Adjusticator is the volume control for the whole amp.  None of the usual buffers/boosters work for this purpose because they can't attenuate - and unity gain makes for too much volume with the chip amp.  I guess I could have just stuck in a separate volume pot, but I like the idea of adjusting the volume by controlling the gain.

For anyone else who is interested, here's the updated schematic, layout, and pnp.  This is a modified version of R.G.'s Adjusticator.  For more information about the design, and how to change the gain limits, you can find it at www.geofex.com under "Circuit Sweepings|Adjusticator".  (If you object to any of this R.G., let me know and I'll remove it)





(Full size 600 DPI image here http://www.aronnelson.com/gallery/main.php/v/waltk/Photos/Adjusticator+HiZ4+PNP.png.html?g2_imageViewsIndex=1)

[PRR]

I dunno if I am missing a feature, but there are other ways to do this.

The inverter is a first-choice for +/- gain, but awkward for high impedance at low noise.

The bias-buffer is wanted for this gain-set plan, but it can be worked-out.

Given two BiFET opamps and guitar impedances and levels over +/-20dB gain range, I'd be thinking more like this:



There's headroom between guitar and 9V rail, take a little non-inverting gain up front, swamp noise in the variable gain stage.

Keep all inline resistors small compared to guitar: 5K naked pickup in bass, 60K with guitar volume pot half-way. OTOH large compared to TL072 happy load; U1 drives 1.46K but under 9V rail it won't be asked for over 2mA, and should be sweet/clean.

Wires to gain pot (actually the one to U2 "-" input) must be short.

AFAICT, TL082 is claimed lower audio hiss than TL072. I suspect they are selected from the same process, which has cleaned-up since the specs were carved in stone, and there may be no difference on average.

If you really want gain-drop to stop at -20dB, put 100 ohms in series with pot.

The biasing is a bit sloppy, opamp offset voltages add and stack. I figure 2mV chip error could be 30mV at U2 output, which is entirely OK here.

Input impedance is 5Meg plus stray capacitance, which should be less than another foot of cable, potentially much less. You DO want to use a shorting jack: un-plugged input hiss will be large.

Frequency response is 10Hz to near a MHz. A 1,000pFd cap across gain pot will roll-off supersonics at high gain.

Assuming 16 nV/rootHz hiss across 20KHz, 2.2uV hiss, output hiss is 25uV at gain=10, 2.5uV at gain=1. The thermal hiss of 5K-50K resistance (guitarish source) is 1uV-3.2uV. Taking 3.2uV, the noise figure is 2dB. Actual in-use NF is not a simple thing due to complex pickup impedance. However this compares favorably with good 12AX7, the benchmark for guitar pickups.

(The thermal hiss of 1Meg in front of an inverter is 14uV, which makes opamp hiss moot, and would be higher than many 12AX7s.)

Power and other detailing may be taken from R.G.'s details.

Before he was fired, Bob wrote-up some suggested op-amp experiments for beginners:
http://www.national.com/rap/meritbadge.html

[PRR]

D'oh! There is a 1-amp linear-pot way to do this.....

[waltk]

D'oh! There is a 1-amp linear-pot way to do this.....

I'm listening.

All I'm really looking for is a way to keep the high input impedance and +/- gain (-infinity to +20db) with full (guitar) frequency response (that doesn't roll off when you adjust the volume).  Your "non-inverting input followed by variable inverting" idea looks interesting.  Will it work?  What's the other 1-amp linear-pot way?

[PRR]

Nevermind.

The simple way gives a very S-shape dB versus rotation, +/-10dB covering 1-9 on the dial. There are "S-taper" pots made for graphic EQs to bend this nearly right, but most affordable strippable EQs use linear and overlook the kinks.

There is another way which can give a lovely curve, but overall unity gain gives internal gain of 6, which is too high for strong guitar under 9V rail, and it won't ever go to very high attenuation.

Going for one opamp was a folly.... the cost difference between '071 and '072 is nil. Linear pots are nice but 10K Audio is easy enough.

I'll stand on my  "non-inverting input followed by variable inverting".

I'd be curious, if you build both, if my musings about hiss levels are actually significant in real use. Design and practice can be very different things.

[R.G.]

I dunno if I am missing a feature, but there are other ways to do this.

The inverter is a first-choice for +/- gain, but awkward for high impedance at low noise.

Right on both counts. It necessarily implies a big input resistor, and that ups noise.

One general way to generate problems in electronics is to try to do too much with a single device, as you know. In many cases one can get away with stretching the limits of a single device's gain, impedance, frequency response, etc. and live with the minor issues. But sometimes the issues rise up and bit you anyway. DIY effects electronics is almost always a competition between getting the result you want while designing nothing that can't be done by someone who can barely solder and can't read a schematic. The Adjusticator is one of those compromises.

The correction for trying to do too much with a single device is to use more devices. In this case, where the high series resistance gives a noticeable problem, the answer from both of us was to use another device. Buffering the input to keep the high impedance one needs for a guitar pickup, then doing the variable gain in a second stage works.

I view this as Mother Nature telling me to quit playing around and do it right. 

[waltk]

Thanks for sharing your experience and insight!

you may have to use that second opamp as a buffer so you can run the lower impedance version of the variable gain amp

Buffering the input to keep the high impedance one needs for a guitar pickup, then doing the variable gain in a second stage works.

I'll stand on my  "non-inverting input followed by variable inverting".

I'd be curious, if you build both, if my musings about hiss levels are actually significant in real use.

Sounds like you're both recommending the "non-inverting input followed by variable inverting" approach.  I'm going to have a shot at this, and post my attempt (advice is welcome before I actually build the thing).
So my goals are:

• variable gain from -infinity to +20db
• high input impedance
• low noise
• flat frequency response
• a linear gain curve (or maybe an audio taper)

Don't care about:

• Leaving the output inverted
• making it a little more complicated

I guess this is a tall order, and I expect mother nature to intrude on plans.  I'll try this with a dual opamp, and go back to a simple voltage divider for the Vref.

[PRR]

> generate problems in electronics is to try to do too much with a single device

Sometimes serendipity happens. IMHO, some of "that Germanium sound" is not the Ge, but the too-too-simple circuits we had to use. Some sucked, a few were "bad" in "good" ways. There are several classic circuits which work sweet with extreme simplicity.

Yeah, I did 'know' that I might be juggling too many variables to push all corners of the envelope. It looked OK on a matchbook. Fortunately the idiot assistant speedily plotted all 270 degrees of rotation.



Both are drawn as "bridge" just to be elegant. The 1-ohm resistors are placeholders, should be short. I gave the problem to my IA as a "DC" problem 'cuz he's too old and dumb to plot gain and pot-param at the same time, and I'm too lazy to work-around. Obviously the idea works DC to infinity. For single-supply audio you need some caps to set DC and >20Hz gains different.

Out1 goes +20dB to -"infinity", but will be impossibly twitchy outside +/-12dB.

Out2 is semi-linear in dB, but won't go silent. And the real issue: if you work out the unity-gain condition, it multiplies by 6 then divides by 6. The result is right but hot guitar times 6 will exceed 9Vpp and fuzz. Might be fine with 18V or 27V supply. But still won't shut up.

> variable gain from -infinity to +20db

You ought to specify better. Infinity is a big number. Do you need to reduce a 2,345V signal to dead-zero? (Can't happen.)

Do you even need 20dB gain?

How much level does your load really need?

What range of levels comes off your sources?

Do you need this "infinite" range on one knob? Do you have a single continuous source which wanders in level ALL over the place? Or do you have a balsa banjo at 10mV and an active dreadnought at 2V, at different times/songs?

You are asking for infinity+20dB. Even if you are willing to cramp -20dB to infinity into the last crack, 40dB on one knob is always fussy.

What is often done, for inputs which must accept "any" source, is put 30dB on the knob (with maybe a fade-to-black in the last 10%) and a 10dB-20dB-30dB gain/loss switch. When you put down the banjo and grab the dreadnought, you click the switch two notches then trim the pot a couple points. That's faster and more repeatable than turning a pot from 9.34 to 3.56.

Determine your actual load sensitivity, and gain it down to ~~~50mV-100mV (for typical noise voltage, gives 85dB-95dB Signal To Noise, which is realistic for real systems in real rooms). For a PA power amp with 1.23V sensitivity, gain of 10 or 20 is usual. Take your hottest and softest sources. Gain/loss each to 50mV~~2V (about the max under 9V power). Figure out what gain/loss settings work for all your sources. Then put a simple volume pot between the switched-gain/loss input system and the load booster.

[PRR]

Twice as many parts, maximum 2V input, but can take sources over a ~~30dB range, apply 0-30dB volume control, and bring-up to power amp input level.

Also uses an affordable 3T switch, and puts the hi-med-lo positions in proper sequence.

[waltk]

Paul,

Thanks for the thoughtful and thorough response.  My brains would be leaking out my ears long before I could design a solution like this on my own.

The specs I listed were more of a wish list than requirements.  "-infinity" was a bit of an exageration, but I did want a big range on the volume.  The 3T switch is perfect to alleviate crowding on the volume pot.

For the input level of this, I expect it to be anywhere from straight gutar (single-coil or humbucker) to a boosted level (e.g. from a distortion stompbox).  The 2V max input level in your design should be fine.

For the output level, I wanted to be able to accomodate different chip amps - and get output close to zero on the low end, and drive the amp to its maximum (low-distortion) level on the high end.  I think the switch/pot will work fine for that.

So  - with your permission - I'll capture this schematic and make a PCB layout for it.  It'll take a few days though; I have to concentrate on work that I'm actually good at (and get paid for) for the rest of this week.

Thanks