White noise not being amplified

[LeatherNeck]

Cannot get white noise to come through this op amp. I have read that I will need to use dual supplies in this design.
1. Do I need to use dual supply?
2. Can this op amp pass white noise?

 

[FvM]

You either need dual supply or a DC bias for U1 non-inverting input.
Furthermore Q1 CB diode junction doesn't work as a noise source because it has high breakdown voltage. Use EB diode instead.

 

[betwixt]

A dual supply is preferable but it should work as it is. The problem is the output of the first stage is going to be almost at ground potential and it will then be reduced by the pot before entering the second stage. Raising the voltage to mid rail, or, using a split supply overcomes that problem. Is there a reason for using 1K feedback resistors when the input resistance of the op-amp is in the Giga Ohm realm. Much higher values should be adequate and will reduce loading and current consumption.

Brian.

 

[LeatherNeck]

You either need dual supply or a DC bias for U1 non-inverting input.
Furthermore Q1 CB diode junction doesn't work as a noise source because it has high breakdown voltage. Use EB diode instead.

Oh yes, sorry, I actually did use the EB junction. That was the first problem I found with the circuit. So you think maybe 100K feedback resistors?

 

[LeatherNeck]

Ok I’ll try higher values and provide a dc bias for the non inverting input. This was not my design, but then again, I’m not a designer so I’m careful when critiquing another engineer’s design. But the original designer is not here anymore. Ok thanks.

 

[betwixt]

Assuming you can live with a response roll off at very low frequencies, add a capacitor at the ground end of R34 (value depends on LF roll off but I would suggest 47uF) then add a 330K resistor from pin 3 of U1 to +5V. Personally, I would increase the feedback resistors, all of them to at least 100K.

Brian.

 

[Audioguru]

The second opamp also has its inputs biased at 0V then its output will not be able to produce the negative swings of the input signal. Its (+) input also must be biased at half the supply voltage.

 

[LeatherNeck]

Assuming you can live with a response roll off at very low frequencies, add a capacitor at the ground end of R34 (value depends on LF roll off but I would suggest 47uF) then add a 330K resistor from pin 3 of U1 to +5V. Personally, I would increase the feedback resistors, all of them to at least 100K.

Brian.

Ok, I changed the 330k to 100k voltage divider and increased the feedback resistors to 100k and 10k with a 10uF cap to ground on R34. It is actually amplifying the signal now. There is 100mV of signal out now. Should I keep trying to increase the gain of the stage? White noise is different right? Just because I increase the gain, doesn’t mean I’ll get what I expect. May some more amplitude but different frequency content? Frequency content seems a bit different since I increase the gain to 10, well 10 plus one right. 1 + Rf\Rin.

- - - Updated - - -

Oh yes, thanks. That is true. I believe I will have to add the cap on Rin to ground also
To separate the ac response from the DC. But thank you for the response.

 

[betwixt]

I'm glad it's working.
The problem was that although the op-amp can handle rail-to-rail outputs, you had no bias on the input stage so the output was sitting at, or almost at, zero DC which didn't leave headroom for any amplified signal. By moving the output voltage nearer to mid supply voltage, it can freely swing it's output higher and lower without distortion.

Increasing the gain will have an effect on the output frequency range. This is a fact of life when you use low power op-amps. The choice of LT6003 isn't a good one if you need more amplitude, that device has a very small gain-bandwidth product. Basically, that means the more gain you set, the narrower the bandwidth it can manage. If you look at the data sheet for the LT6003 you will see it can only manage about 3KHz with no gain and even less when you amplify with it. The gain of 2 you have set in the schematic means it can only manage up to about 2.5KHz, well inside the audio range and that explains the change in sound you observe. There is no fix for this except to use a device with bigger GBW product.

Brian.

 

[LeatherNeck]

Ok that all makes perfect sense to me. I did look at the data sheet and, yes very small GBP. So the intent of the designer was to use a 35 watt halogen light to simulate the flame inside a gas or oil burner. Using this white noise signal summed with a DC level to mimic flame flicker. Another engineer says he would like ten percent flicker voltage. Max dc output To drive light is about 12volts. So I would like to achieve somewhere close to 1.2 volts noise to sum with the drive voltage. If I use the same components in the second amplifier stage will there still be more bandwidth loss?

 

[betwixt]

Wow! that is probably the most complicated and least efficient way to do that!

The response of a halogen lamp is very slow, the filament is incapable of rapid brightness changes so a frequency response of less than 10Hz would be more than adequate. Controlling high current in that way is extremely inefficient though and you will probably dissipate more heat in the control circuit than in the lamp. A far simpler solution, using probably fewer than 10 components is to generate random numbers and use them to control the on/off ratio of a PWM generator. Basically, instead of controlling the voltage to the lamp as an analog value, you rapidly (> 100 times per second) vary the time that full power is applied. The visual effect is identical but because you are hard switching the power on or off, there are very few losses in the control circuit.

I use a similar PWM technique to control 20W LED lighting to give a flicker effect and all the current is switched through a T0-92 transistor and it runs almost cold. I also built a multi-channel version that drives lots of red, yellow and white LEDs to simulate a flame in a log fire, it is actually very realistic and people are hesitant to touch it in case they get burned when in reality it is stone cold.

Brian.

 

[LeatherNeck]

Yes, I am familiar with this technique. I have done that with the PWM output of an Arduino micro to modulate intensity of an LED. You are correct, the output of our circuit has a big power transistor mounted on a sizable heat sink.

 

[LeatherNeck]

Also, do you think I can get that low of a frequency out of the lt6003 amps used in this circuit. I feel they have set me up for failure. OMG.

 

[KlausST]

Hi,

the low frequencies are not the problem. Especially not those low frequencies that make the bulb flicker.

Klaus

 

[FvM]

Before designing a simulator, I would measure the flame flicker spectrum with a photodiode and FFT analyzer. I would be surprised if the spectrum is similar to white noise, more likely it's something like pink noise (1/f density distribution).

 

[LeatherNeck]

Ok, I can do that. I believe we have a scope that can do an FFT. It would be good to know what my frequency distribution is.

Thanks.