R.G's Low-Fidelity / Telephone / Noise Pedal - Noise section debug

[chi_boy]

I made a layout for the Low-Fidelity / Telephone / Noise Pedal from R.G.'s site and for the most part it works fine.  The filter section work perfectly per R.G. schematic, but I am having a problem with the noise source side of the circuit. 

This is my first attempt at a layout from schematic, so I suspect it is something I did, but I've been looking at this for 2 nights and I'm stumped.  It's only six parts, and not complex.

Would anyone be willing to look at the noise portion just to make sure I didn't miss something silly in the layout?  That portion of the circuit is circled in the image.  I used 2n5088's for Q1 and Q2 and a .1uF for the unmarked cap per R.G.'s comments on the forum. 

When I swing the panner to the "noise" side I hear a faint hiss that is relatively quiet, almost silent.  I was expecting to hear a little more racket with more volume, but maybe my expectations are the problem.  I will likely experiment with different transistors, but I would like to know that the layout is right and that my expectations are reasonable before doing that.

Thanks for any help.

George

R.G.'s Schematic



My version:



My layout:

[R.G.]

I made a layout for the Low-Fidelity / Telephone / Noise Pedal from R.G.'s site and for the most part it works fine.  The filter section work perfectly per R.G. schematic, but I am having a problem with the noise source side of the circuit. 
...
When I swing the panner to the "noise" side I hear a faint hiss that is relatively quiet, almost silent.  I was expecting to hear a little more racket with more volume, but maybe my expectations are the problem.  I will likely experiment with different transistors, but I would like to know that the layout is right and that my expectations are reasonable before doing that.

I don't see anything wrong at first blush, but I wonder if I messed up the resistor values on the second transistor, or if the Q1 you have is breaking over at too high a voltage. It's also possible that the 2N5088 you are using for the noise transistor is just too quiet. Modern semiconductor processing has gotten much *too* good in many ways.

Can you test the pin voltages on the noise transistor and the noise amplifier transistor and post back?

[caspercody]

What does this pedal do?

[chi_boy]

R.G.  I'll check those voltages tonight and post. The 2N5088's are new Cetral Semiconductors. I verified pins against the datasheet and the part has the pins printed on the face. If the resistors look good, I also have some 2N2222's I could try.

[R.G.]

What does this pedal do?

It creates a low-bandwidth, noisy audio output from a high bandwidth, quiet one. The idea is to make the output sound as though it went through a bad-connection telephone line.

There are two parts: one is a bandwidth limiting filter, the other is a noise source. Well, OK, three parts - there is a pan-pot to let you blend smoothly from pure low-bandwidth audio, no noise, through partial mixtures of the two, and all the way to pure hissy noise.

R.G.  I'll check those voltages tonight and post. The 2N5088's are new Cetral Semiconductors. I verified pins against the datasheet and the part has the pins printed on the face. If the resistors look good, I also have some 2N2222's I could try.

What I think may be going on is that the amplifier transistor may be running out of amplifying room and not getting enough voltage output. I'm right now going through some sims to figure out why I picked those resistors in the first place. It's been a while since I did that design. 

If the voltage at which the noise transistor's base-emitter breaks over is too big, it will saturate the gain transistor and not amplify the noise much. I'm trying to find the range of voltages that this works with. I can't imagine not having done this when I first did the design, but hey, sometimes I just do a forehead slapper.  Getting the pin voltages on the transistors will tell me directly if this is the issue with your circuit and transistors, too.

[chi_boy]

R.G.,

The voltages are as follows:

Battery: 9.38

Q1
E:  6.41
B:  6.98
C:  6.49

Q2
E:  6.98
B:  0.00
C:  N/C

Thanks again for revisiting this.

George

[R.G.]

Battery: 9.38

Q1
E:  6.41
B:  6.98
C:  6.49

Q2
E:  6.98
B:  0.00
C:  N/C

The simple way to fix it is to change the 22K to a 68K or 75K. Your circuit's probably fine. The voltage from Q2 is high enough to saturate Q1 so it can't put much of anything out. My bad!  Try upping to 68K to 75K - shoot, you could use 82K, 91K, maybe 100K - instead of 22K and see if you don't get more noise.

[chi_boy]

The first resistor I touched was a 100k, so it was the lucky one.

The voltages changed to this:

Battery: 9.38

Q1
E:  8.42
B:  8.38
C:  8.53

Q2
E:  8.38
B:  0.00
C:  N/C

Oddly, the "noise" wasn't much different.  Still remains pretty quite when panning.

[R.G.]

Hmmm. There's more happening there.

Do two things.

One, try a 220K instead of a 100K.

Two: let me do some simulation and figure out what's going on.

Obviously , the first one is faster. The second one is for me to figure out why the base-emitter of the noise transistor appears not to be breaking over like it should. No way it should be going to over 8V without breaking. Which means that Mother Nature is trying to beat me on the head with something I've forgotten. 

[chi_boy]

R.G.

The 220k made a huge difference.  Plenty of noise with lots of volume.  Sounds just like the white noise on a "snow" screen on T.V.

I'm embarrased to say I may have made a connection error when testing the 100k.  I'm going to put it back in to see if it was me or not. 

The voltages for the 220k are:

Battery: 9.37

Q1
E:  8.61
B:  8.45
C:  8.97

Q2
E:  8.45
B:  0.00
C:  N/C

[R.G.]

Is there any chance you have the 2N5088's in there backwards?

The reason I ask is that the breakover voltage of the 2N5088 base-emitter seems to be about 4.5-5V from the specs I found. The voltages you're reporting show it going to over 8V, which should not happen. I'm just guessing that with the transistor reversed, the collector-base junction would not be breaking over at all, and that would account for what we're seeing.

Otherwise it's back to the simulator for me.
===================================
Just as I hit send, your reply came up.

I'm glad it's working, but the comments above still apply - boy is it weird that a 2N5088 would have an 8.45V base-emitter breakover. "White noise on a TV screen" is what it's supposed to be, all right.

And no need to apologize for any errors. I'm still hunting down why it didn't work with the first set of resistors. 

[chi_boy]

I'm embarrased to say I may have made a connection error when testing the 100k.  I'm going to put it back in to see if it was me or not. 

OK, I feel better.  I didn't miss it after all.  I put the 100k back in and it is quiet again.

The "noise" that is generated is a nice smooth white noise type sound and is pretty consistant.  Do you think there would be any benefit to changing the transistors to something like a 2N2222 or 2N3904 to get a different character to the noise?   Maybe more snap crackle and pop?

Also, I used MC1458 opamps for the filter section.  They are a dual 741.  I'm not sure I like them yet, but they do add just a touch of distortion to the filtered signal.

Also regarding the pins, anything is possible, but I checked those several times.  The transistors are printed with EBC right under the part number.  I also included the orientation in the picture.  I may be reading the schematic incorrectly too.  EE gave me fits in school.  I can't imagine why I would be messing with this stuff now.  I must be drinking too much.  

[R.G.]

Different transistors will give different noises, but what everyone was trying to get was a smooth hiss...   

There are different noise mechanisms. The predominant one you're getting is avalanche noise, which is the smooth hiss. Noise that snaps and pops is shot noise - novel name, eh?    And there is another noise mechanism that generates noise which gets bigger as frequency lowers, called burst or 1/F noise.

My proto on this thing ( I found some of my notes) worked OK as reported in the schemo. But it may be that transistors have gotten better again. I'll have to do some testing and perhaps update that circuit.

I'm glad that I could easter-egg it in for you. If you filter the noise before panning it in, it will change the effect too. I was trying to get the smallest, simplest circuit I could.

[chi_boy]

R.G.

Thanks so much for your efforts tonight.  This really is a fun little pedal and I am tickled that it is working.  I will play a little on my own now and tweak a little now that it works.  As far as the noise goes, the smooth white noise with a little op-amp distortion may be the perfect combination.  Time to let it ferment and see how it sounds in a day or two.

Thanks again,
George

[Brymus]

Glad you got it working
I am very interested in this.
Would this be something you could do the "wish you were here" intro with ?

[chi_boy]

Would this be something you could do the "wish you were here" intro with ?

I think it would sound good for "wish you were here."  That is actually closer to the sound I was going for myself.  That intro has a little more static and whine and almost has a popcorn popping quality.  This noise generator is more of a white noise, so it woulnd't be exact, but definatly close enough for rock-n-roll.  Now that it's working, I am planning to try it with more bass cut.  The filter is set at 300hz, but I want to play with it and see how it sounds at 600hz, 800hz, and maybe 1khz.

[birt]

if you happen to have a metal can 2N2222 for Q1. solder a long wire to the rim of the can and connect that to a switch. switched to ground it's stock, switched to the enclosure (NOT grounded) it's an antenna that will pick up more noise and random radio.

[PRR]

> breakover voltage of the 2N5088 base-emitter seems to be about 4.5-5V from the specs I found

That may be a minimum?

It has been a looong time since I worked with these things. IIRC, in the 1970s, you could bet a beer the E-B breakover would be 6.3V-7.5V and never go thirsty. That's some kind of natural for the processes they used then. Then the "5V" number on the sheet is just a safety-margin down from 7V.

And again IIRC, it was risky to try this with a 9V battery, because the noise-power faded quickly as the battery drained. (Ah, but who uses battery today?) Perhaps more risky to also cleverly use the Vz as the bias for an amplifier? When done with 9V there was often a DC-block cap to the amplifier, which was biased conventionally (often just 1Meg C-E, grounded emitter).

Still a very clever low-parts design.

> Sounds just like the white noise on a "snow" screen on T.V.

That's just what it is, nearly. FM (TV) is de-emphasized (treble cut), this should be brighter. (And FM hiss is raw hiss through an FM demodulator, but I think the result is just random.) This plan lacks full bass response, a small difference (Q1 at 0.1mA and C9 at 1uFd gives 500Hz bass-cut). But over most of the "hiss zone" they are nearly pure random sources.

One thing to try, cost you a dime: raise C9 from 1uF to 10u and 100u. Bass-cut hiss is not natural, letting some rumble through may be more "realistic". Also there is 1/f noise, which gets greater at lower frequency. The audible effect is that the hiss is not constant from millisecond to millisecond.

And if you get deep (sub-Hz) 1/F noise boosted far enough to clip the amplifier path (which needs more than just Q1), you will get some "snap crackle and pop".

But the real problem with trying to raise non-Zener hiss up to "annoying" level is that you are likely to pick up all the buzz and even signal in the area. To a point, that may be good. But I think it will require obsessive shielding to be workable.

Change C3 0.1uFd to 1000pFd-500pFd. This shaves a trace more bass from the gitar, but also brings up the generous 1/f current-noise of your antique 1458. I have seen a phono preamp with a similar chip and too-small input blocking cap put out so much subsonic noise it tripped the "DC protect" relays in the speaker amp. This will also work with lusty 5532, but won't work with TL072 (no gate current, thus no current noise).

> The "noise" that is generated is a nice smooth white noise type sound and is pretty consistent.  Do you think there would be any benefit to changing the transistors to something like a 2N2222 or 2N3904 to get a different character to the noise?

No. This is raw semiconductor physics. As RG says, avalanche noise. '5088, '2222, '3904 are all essentially the same thing, sold in different sizes and proportions ('5088 may have a thinner base, real '2222 has a bigger die). Back in the late 1960s there were audible differences between -foundries- (some were cleaner than others), but not across devices from the same fab. Since the 1970s all fabs have got very-clean, and I doubt there is any audible difference.

Which does suggest one line of extended research. Get old crappy Germanium devices. Wire for gain and follow with lots more gain. Their dirty surfaces give lots of excess and 1/f noise. However we know from popular Ge FX pedals that even odd-lot Ge devices do not hiss excessively at guitar level. So you still need considerable gain.

Try the digital pseudo-random noise generators. Short registers produce not-very-random hiss. However they are repetitive: mine made a distinct thump every 2 seconds. On long room-tuning sessions, this was annoying. You may try 2 or 3 such hissers running on different clocks to try to "randomize" the loop artifacts. But this is getting complicated.

Look, this is the 21st century, and 2010 already. You can buy "message recorder" chips/modules for about what a handful of CMOS registers cost. They sell stuffed bears with these modules so grandkid can send a personal message to grandma. And they may turn up in post-holiday sales. Anyway, find a hiss you like, record it, loop playback. 30 seconds of time is probably enough to mask the repeat.

[R.G.]

> breakover voltage of the 2N5088 base-emitter seems to be about 4.5-5V from the specs I found

That may be a minimum?

It is, assuredly. The old standard was that base-emitters break at 6V, which is the reverse-BE spec on a 2N3904. I took that maximum as being a telegraph for saying "we'll ship anything that meets gain spec, even if the B-E is a little thin."

And again IIRC, it was risky to try this with a 9V battery, because the noise-power faded quickly as the battery drained. (Ah, but who uses battery today?)

It was - and it does. But who does? 

Perhaps more risky to also cleverly use the Vz as the bias for an amplifier? When done with 9V there was often a DC-block cap to the amplifier, which was biased conventionally (often just 1Meg C-E, grounded emitter).

Still a very clever low-parts design.

That's what I was after going voltage-mode instead of the more-common current mode circuit.

Try the digital pseudo-random noise generators. Short registers produce not-very-random hiss. However they are repetitive: mine made a distinct thump every 2 seconds. On long room-tuning sessions, this was annoying. You may try 2 or 3 such hissers running on different clocks to try to "randomize" the loop artifacts. But this is getting complicated.

Even simpler, get the PIC 12F509 or one of the PIC 10Fxxx chips. You can get these in 8-pin dip packages and program them for something in the 64-bit class of feedback shift register noise. A 31-bit feedback shift register repeats in 2.147 billion steps; if you set the step frequency to something out of the audio band for good verismillitude on wideband noise, say maybe 44kHz, the repeat would be something like 13.5 hours for a repeat. A 63 bit register (which the Math Devil made me calc out) would repeat once every 6.6 million years if your battery lasts that long.  Maybe we could use the smaller register length or higher rep rate.

These PICs are about $1.00. I would have done that, but then *-no one-* would ever have built one.    For some reason, no one here has tumbled to using PICs for logic or digital functions. I tried to push that for a while, but gave up when it was met with a deafening silence. The digital forum seems to have dropped into the DSP-only rut.

But yes, one eight pin dip and maybe a couple of resistors and caps could do a bang-up noise generator for cheap - and good.

[tiges_ tendres]

Even simpler, get the PIC 12F509 or one of the PIC 10Fxxx chips. You can get these in 8-pin dip packages and program them for something in the 64-bit class of feedback shift register noise. A 31-bit feedback shift register repeats in 2.147 billion steps; if you set the step frequency to something out of the audio band for good verismillitude on wideband noise, say maybe 44kHz, the repeat would be something like 13.5 hours for a repeat. A 63 bit register (which the Math Devil made me calc out) would repeat once every 6.6 million years if your battery lasts that long.  Maybe we could use the smaller register length or higher rep rate.

These PICs are about $1.00. I would have done that, but then *-no one-* would ever have built one.    For some reason, no one here has tumbled to using PICs for logic or digital functions. I tried to push that for a while, but gave up when it was met with a deafening silence. The digital forum seems to have dropped into the DSP-only rut.

But yes, one eight pin dip and maybe a couple of resistors and caps could do a bang-up noise generator for cheap - and good.

I think many folks on here are starting to pick up on PIC stuff now.  I think the draw back is the learning curve plus the cost of programming tools.  If you think about all of the resources available purely for guitar pedals, it seems odd that no one has offered the paint by numbers route for PIC programming yet.