Author Topic: Help with noisy analog circuit.  (Read 4156 times)

0 Members and 1 Guest are viewing this topic.

Offline Greg Robinson

  • Regular Contributor
  • *
  • Posts: 85
  • Country: au
Help with noisy analog circuit.
« on: August 31, 2015, 10:05:41 am »
Hi everyone.

I recently built the following circuit for use with guitar (hence the high input impedance). It has a gyrator and capacitor multiplier for variable frequency and cut level high and low pass shelving filters. It also includes a gain boost at the end of the circuit. While the circuit works as expected, it suffers from unacceptably high noise when the cut controls (VR1 and VR3) are turned down. As built, U1-4 are included on a single TL074, and U5 is half of a TL072. Not shown on this LTSpice schematic are local decoupling of 100n from both rails to ground and from rail to rail for each of the opamp chips.

As I understand it, these circuit topologies tend to be inherently noisy, but can be mitigated with proper selection of low-noise opamps. However, as I am self taught, there are some gaps in my knowledge, and I am uncertain of how to calculate noise in opamp circuits, and how to select appropriate devices for low noise.
If anyone could point me in the direction of a reference for calculating noise in opamp circuits, and how to select devices for low noise, or otherwise has comments or suggestions for my circuit, would be greatly appreciated.
 

Offline PChi

  • Frequent Contributor
  • **
  • Posts: 265
  • Country: gb
Re: Help with noisy analog circuit.
« Reply #1 on: August 31, 2015, 10:21:06 am »
If spending money is OK then buy 'Small Signal Audio Design' by Douglas Self at http://www.douglas-self.com/ampins/ampins.htm
Try the Texas Instruments, Linear Technology or Anaolg Devices web sites, http://www.analog.com/media/en/technical-documentation/application-notes/AN-358.pdf
R1 and R2 are very high values and my guess is the number one source. You could use an experimental approach and try isolating the stages to see which one is causing the most noise. Or reduce impedances and choose a lower noise part than the TL071.
 

Offline HackedFridgeMagnet

  • Super Contributor
  • ***
  • Posts: 1937
  • Country: au
Re: Help with noisy analog circuit.
« Reply #2 on: August 31, 2015, 11:46:02 am »
Not sure what you are trying to do.
Can you do a block diagram.

Don't like the voltage follower U4, I think that can push the op amp past it's limits. But then maybe that's how u make weird noises out of a guitar.
What do you mean by cut controls? Are they supposed to operate like faders to the U4?

What rail voltages are they?

 

Offline tggzzz

  • Super Contributor
  • ***
  • Posts: 9888
  • Country: gb
    • Having fun doing more, with less
Re: Help with noisy analog circuit.
« Reply #3 on: August 31, 2015, 01:53:40 pm »
It would help us - and probably you - if you drew a block diagram of the principal elements. Then you can annotate the inputs and outputs with expected voltages (min, max, and AC peak-to-peak).
There are lies, damned lies, statistics - and ADC/DAC specs.
Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".
Having fun doing more, with less
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 13743
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Help with noisy analog circuit.
« Reply #4 on: August 31, 2015, 03:06:28 pm »
Looks fine to me, what's a block diagram for..?

The problem with noise is you have relatively noisy amps at low signal levels.  So the nosie is a big fraction of the signal, and SNR sucks.  You also have a poor match for the subsequent op-amps.

Some thoughts:
- TL07x is fine for input -- you need the high impedance to avoid loading the source.  High impedance op-amps are always noisier (though there are amps available that are several times better than this one), and the ultimate limit is the source resistance anyway (which is 189 nV/rtHz as shown).
- You need gain at the input, and you need to reduce the impedance.  You have one of these -- the resistors after U1 are on the order of 10kohms, not megs, so the follower is doing its job.  But you need gain, too.  U1 should be wired for 5-10 gain, and this will push the signal level that much above the noise level of everything else.
- Since the impedance is lower in the rest of the circuit, TL07x is suboptimal.  If you have enough input gain to compensate, the SNR won't change much for noisy or quiet amps, so that's a priority.  If you don't want to use input gain (or can't, for some reason), it would be worth using lower noise, lower impedance (bipolar) amps for the rest, like LM833 or even LME49722.
- I'm not sure offhand how noisy the filters will be, but they look like the kind of thing that greatly amplifies small differences, for frequencies in the stop band.  Which invites noise there.  So doing these at the same signal level is probably a big sin.
- You've got all the gain at the end, which as you've noticed, is the wrong place for it. :)  It's also adjustable between "not much" and "a lot", which can be a bad idea because of the poor frequency response at high gain that dominant-pole op-amps suffer from.  (TL07x is modestly fast, so this probably doesn't affect anything here.)

Also for reduced noise, consider reduced bandwidth: add more input, intermediate, or output filtering (followed by a buffer for line driving capability), so that frequencies below the lowest and above the highest signal frequencies are attenuated.  For hi-fi audio, you don't need anything below 20Hz or above 20kHz; for guitar purposes, these can be tighter.  Bass guitar may still want 20Hz, but a shrill lead guitar might not need under 200Hz.  Neither will need much above 10kHz; for lowpass cuts below 10kHz, it will be more mellowed, which might also be desirable?

If you also want a distortion stage, you should probably do it with clipping diodes or something like that, rather than driving the final stage into saturation (which makes a huge signal level).  This keeps the amp happy (although I don't think TL07x suffers problems from that kind of treatment) and gives you more control over what kind of distortion you get, and when.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline macboy

  • Super Contributor
  • ***
  • Posts: 1954
  • Country: ca
Re: Help with noisy analog circuit.
« Reply #5 on: August 31, 2015, 04:21:02 pm »
The sub-circuit around U3 looks like it will produce a lot of noise. You have a high feedback resistance (looks like adjustable from 33k to ~1M) and a very small capacitor (some pF) to ground. I'd suggest trying a re-design with a smaller R and larger C (same RC). Doing so significantly reduced the impedance at the + input of U3 which will reduce input-referred output noise. Consider similar changes around U2 although with 330n at the input it doesn't seem as important.
 

Offline Greg Robinson

  • Regular Contributor
  • *
  • Posts: 85
  • Country: au
Re: Help with noisy analog circuit.
« Reply #6 on: August 31, 2015, 04:48:51 pm »
...snip...
Thanks for the links and suggested reading PChi, I should get around to purchasing some of Self's books, and I will definitely read the AD app note, it appears to be exactly what I am after.
Regarding R1 and R2, they will be in parallel with the source impedance of the passive magnetic guitar pickup ~10k. 1Meg is the traditional and expected input impedance for use with passive guitar pickups.

...snip...
Hi HackedFridgeMagnet. I thought I explained pretty well what I am trying to do in my first post, variable high and low pass shelving filters...
Ok, I have put together a block diagram, see attached.

Not sure what you're worried about or why you think U4 could be "pushed past it's limits". I am certainly not trying to get weird noises out of this circuit, it is intended to be free of distortion, only meant to provide equalisation of the signal (in the form of variable frequency and level of attenuation high and low pass shelving filters, this circuit does not even provide for boost of those frequencies). The circuit also allows for broadband gain, but I am not having problems with that part of the circuit.
The rail voltages are included on the schematic: +/-9V.

...snip...
Hi tggzzz,
As requested, I have attached a block diagram, annotated as you have suggested.

...snip...

Hi Tim,

Thanks for your detailed reply!
In hindsight, I probably should have truncated the circuit to the section that was giving me problems, and left out the rest, seems it's just caused confusion ;)
Yes, I should have made the input stage the one with gain, rookie error, however, that's not the area that is causing me trouble.
The only time that I have objectional levels of noise is when I turn down the cut controls (VR1 and VR3), this is why I believe that U2 and U3 (possibly U4?) are the area that I need to direct my attention. You are right, I should use lower noise bipolar opamps for these locations, as I don't require the high input impedance of the bifet TL07x in those positions.
I'm not trying to get any distortion out of this circuit, so won't be including any diode clippers. Just want to be able to vary boost between 0 and 40dB, this has proved to be a good range for varying pickup output levels for clean boost with 18V rails, and allows a nice level of overdrive for a 12AX7/ECC83 amplifier input.
I've got a bit of reading to do, I'll post an updated schematic once I digest everything.

...snip...
Hi Macboy,
Thanks for the input. You've raised some good points, I'll look into the changes you've suggested.
« Last Edit: August 31, 2015, 04:54:26 pm by Greg Robinson »
 

Offline tggzzz

  • Super Contributor
  • ***
  • Posts: 9888
  • Country: gb
    • Having fun doing more, with less
Re: Help with noisy analog circuit.
« Reply #7 on: August 31, 2015, 05:52:42 pm »
Thank you for listening to our questions/points, and dealing with them - all too often people don't!

It would help if you would note the op-amp output voltages you expect to see for different positions of the potentiometers.

I suggest you do some simulation of op-amp noise to see the contribution that the opamps are or aren't making. Opamps noise consists of two parts: a voltage source in series with one of the opamp inputs, and a current source between each input and ground. The values of the noise source can be found in the data sheet, with units V-per-root-Hz and I-per-root-Hz. Thus if the value is 5nV/sqrt(Hz) and the bandwidth of the opamp's circuit is 10,000Hz, then you should use a value of 500nV.

Yes, that is simplistic, but it will give you a first-order feeling for whether your guesses are right.
There are lies, damned lies, statistics - and ADC/DAC specs.
Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".
Having fun doing more, with less
 

Offline Zero999

  • Super Contributor
  • ***
  • Posts: 13044
  • Country: gb
  • 0999
Re: Help with noisy analog circuit.
« Reply #8 on: August 31, 2015, 10:22:17 pm »
Not sure what you're worried about or why you think U4 could be "pushed past it's limits". I am certainly not trying to get weird noises out of this circuit, it is intended to be free of distortion, only meant to provide equalisation of the signal (in the form of variable frequency and level of attenuation high and low pass shelving filters, this circuit does not even provide for boost of those frequencies). The circuit also allows for broadband gain, but I am not having problems with that part of the circuit.
I think he meant that U4 is driving too lower impedance load. The TL071 is not designed to drive 100R. You might be able to get away with it at very low signal levels but if the output is higher than a volt or so it will probably clip the signal.

See figure 6 on page 9 of the datasheet.
http://www.ti.com/lit/ds/symlink/tl072.pdf

U5 could be changed to non-inverting configuration which will have a much higher input impedance.

If cost/availability of parts is an issue you could replace that TL072 with the NE5532 (much lower noise but lower input impedance) and use it for the sections which are causing a problem.
http://www.ti.com/lit/ds/symlink/ne5532.pdf
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 6226
  • Country: de
Re: Help with noisy analog circuit.
« Reply #9 on: September 01, 2015, 08:35:06 am »
If signa levels alow it, it might be better to have the adjustable gain at the input. So signal levels are generally higher. It also saves one more OP.
As the circuit looks loke being battery powered, current consumption can also be a factor when choosing the OPs. So a NE5532 might not be to power hungry and lower power alternative might be good enough. The resistors at the input are not that bad as a noise source - in the usefull band and especially at higher frequencies, they are effectively paralled with the input cap. So the 180 nV/Sqrt(hz) are only below the frequencies of interest, or in open circuit case.

The part aboutd U3 looks quite noise.

To lokalize the noise source, LTspice has a good function, though not that easy to use. Let it calculate the noise. Clicking at the resistors / Ops it gives the noise from the individual sources.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf