Pink Noise Generator

[loamobn]

I'm trying to build a pink noise generator, but am having a hard time amplifying the ~10nV peak-to-peak noise of a reverse-biased transistor to something that could drive a speaker and actually sound like pink noise (which pretty much sounds like white noise).

(The power supply is +/- 9V)
So my plan was to have a noise-generating stage: NPN with a floating collector and -9V base that has a DC biased emitter that is fed across a coupling capacitor. Simulations say that the output is about 10nV p-to-p.

Then, I fed that to an op amp with negative feedback, but the output (past another coupling capacitor) was only 100nV p-to-p!

What I want is a white noise output that is about 2V p-to-p, but I dont understand analog circuits well enough to confidently know how to do it.

I attached a diagram I lifted from the internet that is about the same thing as what I'm doing...I tried it, but it too produced a p-2-p V that was wayyy too low.

Thanks for any help, I'm really stumped

 

[TedA]

Ioamobn,

I think you had better wire up an actual transistor and resistor and see how much noise you get. It's pretty clear that your simulation is not accurately simulating the transistor used in your referenced design.

This is sort-of an off the books application for the transistor; the model for it likely does not attempt to model the noise in breakdown.

I think you will find that the noise from actual parts will be much higher than the 10nV you are getting in simulation. Another consideration is that the transistors are not controlled for this parameter in manufacture, so there may be much variation among different samples of the same part number. You may need a gain adjustment control earlier-on in your circuit.

If your intent is to stick to simulation, and not venture into reality, I suggest you look for a noise generator among the facilities provided by your simulator.

The BC548 is not a special part, any small signal silicon transistor might work ( in reality ).

It is fairly likely that all of your transistor models fail to account for this mode of operation. You might try various zener diode models. I'm sure that some zener models out there will account for noise, since this is a parameter that is important in normal operation, for a zener.

A final point is that unlike the virtual transistors in the simulation, the real ones will suffer damage from being operated this way, and should not be tossed back in the drawer with the virgin parts. The damage does not impair the noise generation function, but does degrade other transistor parameters.

If you should, in some circumstance, actually need to amplify a 10nV signal, you will find this quite challenging. The circuit noise will more than likely be much greater than your signal. The MC4558 is not a low noise part, but even the lowest noise devices are unlikely to be low enough.

Yes, pink noise resembles white noise in the same manner that pink paint resembles white. It is very obvious when you switch between pink and white noise over a speaker, but they do sound alike, when compared to other signal sources.

One lesson, here, is that simulation will only get you so far, and, in some cases, may give an entirely useless answer.

Ted

 

[(*steve*)]

the real ones will suffer damage from being operated this way.

Good points about simulation.

However I'm not sure the transistor will be damaged so long as the current through the reverse biased junction is kept low.

How low? Well that's the $64,000 question. I'd probably want to keep it under 100uA, but I have no real basis for that figure.

 

[TedA]

Steve,

Somewhere, there must be some published experimental data on how BE junction breakdown degrades transistors, but I have no idea where it might be found. My reference is the late Bob Pease, who mentioned this effect in his column more than once. As I recall, he said that the degradation began at quite low currents, and was cumulative over time. Usually the most important parameter that is degraded is the gain.

I suspect that the actual breakdown occurs in a very small region within the transistor structure, so the current density might be high, in spite of the total current being low.

Pease also explained why the collector-base junction develops a tiny forward bias voltage while the emitter-base junction is biased into reverse breakdown. Another effect unlikely to be implemented in the typical Spice transistor model.

Ole Bob also had a few interesting things to say about Spice.

Ted

 

[(*steve*)]

There are 2 obvious possibilities for the degradation of the BE junction.

The first is secondary breakdown. It can happen with forward or reverse (to avalanche) biased junctions. I'm sure you understand the process. Some regions are hotter than others and they have a lower voltage drop, so they attract more current, get hotter, have a lower voltage drop, ... eventually that part of the junction fails.

The other process is possibly the migration of ions across the transistor structure. Again, this is typically only seen with chips that operate at elevated temperatures.

I have been struggling to find failure modes that do not require significant energy (or a significant concentration of it) that would be exacerbated by reverse biasing a junction.

If you can find some published reference then I'd be very happy to read it.

 

[(*steve*)]

If you look here you will note that it is asserted that:

Reverse Bias. Some semiconductor devices are diode junction-based and are nominally rectifiers; however, the reverse-breakdown mode may be at a very low voltage, with a moderate reverse bias voltage causing immediate degradation and vastly accelerated failure. 5 V is a maximum reverse-bias voltage for typical LEDs, with some types having lower figures.

This seems to fit the BE reverse bias junction failure scenario, however LEDs are somewhat special in that the semiconductor is more sensitive to certain failures. However, on balance, it seems that these failure modes are unlikely to be related.

However, the breakdown of the junction (whether by zener or avalanche action) is in and of itself non-destructive.

Both of these breakdown processes are non-destructive and are reversible, so long as the amount of current flowing does not reach levels that cause the semiconductor material to overheat and cause thermal damage.

 

[loamobn]

Thanks for your help on the noise-generating portion of the circuit. Is the consensus, then, that it would be best to use a resistor? Or a zener? If there's a risk of degredation of the BJT I'd rather use something that would have a longer shelf life in this application.

It seems that either way, the p-to-p voltage of noise is going to be very low. Will an amplifier circuit be able to produce white noise (which can be filtered to get pink) on the other end? I really don't know the best way to make white noise: if there's any suggestion on that, however general, it would at least give me a jumping off point.

 

[(*steve*)]

A resistor is a very low noise element -- don't use that.

I'd go with the reverse biased BE junction. It will generate a much higher amplitude signal.

Don't concern yourself with life. In normal use a reverse biased junction may see a much lower impedance source of power which will destroy it rapidly. In this case the impedance is high and I consider the prospect of it failing soon is very low.

 

[alfa88]

You might want to check out one of my favorite sites.

http://www.musicfromouterspace.com/index.php?MAINTAB=HOME&VPW=1024&VPH=500

If I remember right I used a Zemer in place of a transistor for noise.

 

[(*steve*)]

*WOW* - he wants $15 for a transistor that has 30mV of noise on a reverse biased BE junction.

I want a piece if that action!

 

[alfa88]

Heh heh. I can't really say I've bought anything from him but he does have some nice pc boards, lots of schematics and masks.