S/N in opamp how low do i need?

[Tubis]

How low S/N ratio do a need to get acceptable noise levels if signal is boosted 12dB in front of lets say a Marshall 2203?
I have about 100units TL072 collecting dust in a box.
The datashheet says 18 nV/sqrt(Hz) which means nothing to me.
Is it to high?
Should a use an opamp with lover noise?
Could the noise level be reduced by parallelling the to amps in the TL072?

[Robert M. Martinelli]

Hi Tubis,

Noise figures for op-amps are expressed that way because ( thermal ) noise has a spectral density (it is present over the whole spectrum ), this means ( since the "sqrroot" term is the divider ), the biggest noise is expected to be found at low frequencies. Usually manufacturers test/measure an opamp NF by shorting the input to GND and measuring noise with a spectrum analyzer at the output. To state how much noise you can expect to find at the op-amp's output simply perform the calculation at the frequencies of your interest. This of course takes into account only the op-amp generated noise, not the noise from external circuitry, so the total noise will always be higher.

18 nV/(Sqrroot(Hz)) is a rather high NF. I would always try to keep noise as low as possible, and, since you seem to need a rather low gain, a single op-amp would suffice ( they ALWAYS have a better NF if compared with their "double" versions ).
I have successfully used NE5534s and OP27s ( less than 4 nV/(Sqrroot(Hz))) when a single opamp is needed, and NE5532s ( two NE5534s in the same package ) when I need a double opamp. The above mentioned single op-amps are pin compatible with 741s, the NE5532s are pin compatible with TL072s ( low power version of TL082s to be used in battery-powered devices ), TL082s, 4558s and the like.

Lately I have seen opamps from Analog Devices ( seems to me AD745 or something similar ) with a NF rated at 1 nv/(Sqrroot(Hz)), provided input impedance is kept within certain limits; I have never used such opamps, so I cannot say they keep what they promise. Maybe one day I' ll give'em a try, but until that day I' ll keep using the ones I mentioned above.

If noise is a concern, I would also recommend you to use metal film resistors, their performance, noise-wise is much better than that of standard carbon resistors.

Oops, I was forgetting....well...if noise had a predictable ( periodic ) behavior, paralleling op-amps and opposing phases would work, but noise has not a predictable behavior, so IMHO that's not going to work ( but you could try as well and come back to us with the results ).

Hope this helps

Best regards

Bob

[Tubis]

Thanks, and i always use metalfilm. I probably use the TLO72 for prototyping and then switch to other better ones to get the noise down. Putting a noisy booster in front of a mutiple gain stage tube amp is a noisegenerator waiting to happen

[Robert M. Martinelli]

Well, to get a good idea of the differences use a 4+4 DIP socket, this way experimenting with different op-amps ( and maintenance as well ) will be easier.

[Alex/Tubewonder]

Paralleling opamps/transistors/tubes to reduce input noise works:

Noise is reduced by SQRT(number of parallel devices). Paralleling 2 opamps in TL072 will give you ca 13nV/sqrt(Hz) equivalent input noise.

[Robert M. Martinelli]

Hi Alex, thanks, good to know! I never tried that, and I probably need to go back to school to read about noise theory

As to this specific application, I think he'd better use a 5534 or an OP27, as he would need a whole bunch of paralleled 072s to lower noise to the level of a single 5534

[Mike Sulzer]

There is more to it than noise voltage.

You have to be careful about noise performance. In general, devices are rated for both noise voltage and noise current. Tubes and FETs have insignificant noise current for what we do; they are all noise voltage. With bi-polar transistors, both can be significant. In general, bipolar has much lower noise voltage than FETs, but the noise current can make them worse in some situations.

Noise current is significant because it flows out of the input and develops a voltage across the source impedance. A guitar pickup has an impedance that changes with frequency, so figuring out the total noise when using an op amp with a bipolar input is not so easy.

So the 5534 might look a lot better than a FET input amp if you consider only noise voltage, but maybe not so much better if you consider the current.

By the way, reducing noise by paralleling devices is fine for those that have only significant noise voltage (Tubes, FETs). You can get more noise if you do this with a bipolar input.

[Mark Hammer]

Noise depends on more than just the op-amp itself. While there are considerably more op-amps to choose from now than there were in 1979, take a moment to read the article on bandwidth and noise in non-inverting amps from this scanned issue of DEVICE, a publication that Craig Anderton put out briefly: http://hammer.ampage.org/files/Device1-8.PDF .

One of the things you'll see is that, depending on the source resistance, some op-amps can be noisier or quieter than others. Well worth a look.

[Mike Sulzer]

Noise depends on more than just the op-amp itself. While there are considerably more op-amps to choose from now than there were in 1979, take a moment to read the article on bandwidth and noise in non-inverting amps from this scanned issue of DEVICE, a publication that Craig Anderton put out briefly: http://hammer.ampage.org/files/Device1-8.PDF .

One of the things you'll see is that, depending on the source resistance, some op-amps can be noisier or quieter than others. Well worth a look.

Yes, his data show that the op amps with the bipolar inputs are noisier with high source impedances than those with FET inputs, as you would expect from the noise voltage and current specs. However, he starts the discussion with something that appears to be incorrect: that using a large input resistor (to get the highest possible input impedance) might not lead to the best SNR. Actually, the noise from that resistor is shorted out by the source impedance of the device connected to the input. That is, the total noise level from impedances at the input is given by the paralleled impedances. Using a large resistor maximizes the signal; that is what counts.

[Robert M. Martinelli]

One of the things you'll see is that, depending on the source resistance, some op-amps can be noisier or quieter than others. Well worth a look.

Absolutely, this IS the key point - when I talked about the AD745, ( which is a BJT design ) I tried to make it clear that they guarantee a NF under 1nv/sqrroot(Hz) provided that the source impedance is kept below a certain value ( which I don't remember ) so in the end it's all about choosing the right op-amp for the application; when it comes to ( electro-magnetic ) guitar pickups, I have to say I have tried both FET and BJT op-amps in the past, and NE5534s and OP27s gave me the best results noise-wise ( much better than 4558s anyway ). It' s true that a pickup's impedance varies with frequency, OTH within the frequencies of our interest impedance resides in the KOhms realm, not Megs ( as we know at the resonant peak it equals resistance, becoming greater and greater as we move away from it ) and that's the reason why I think I got better results with the above mentioned op-amps.

[Mike Sulzer]

resonant peak it equals resistance, becoming greater and greater as we move away from it ).

You mean "lower magnitude of impedance as we move away from the resonance", right?

[Tubis]

In this Boss eq that i think is very noisy. Who is the bad boy?

Is it the input opamp(HA12017), the amplifing opamp(JRC4558) or the gyrators(TLO22) or all of the above.

And wat is the reason for upping highs at the input and then damp theme at the output(not in the picture).

[Steve Conner]

First of all, do you "think" it's noisy from reading about it, or do you own it and know that the noise level is unacceptable? Anything you put in front of a high-gain amp will increase the noise level to some extent.

Anyway, I think the noise comes mostly from the gyrators. The reason is that the output impedance of the gyrator op-amps begins to fall at high frequencies, as they run out of GBW. This causes the gain, and the noise gain, of the circuit to rise at high frequencies, as well as the desired boost/cut frequency, whenever any of the sliders is moved away from the centre position.

My guess is that the pre-emphasis (what you called "upping highs and damping it") is there to hide this effect.

My Ninja Toaster parametric EQ had the same problem, both in PSpice simulations and in practice. Changing the gyrator op-amp for the lowest-noise part I could find made a slight improvement. Since Boss used 6, a really low-noise part would have been outside the budget for a stompbox.

If you own the pedal, you can of course do this test: if the noise goes away with all sliders on 0dB, then it's the gyrators to blame.

[Mike Sulzer]

Anyway, I think the noise comes mostly from the gyrators.

I agree. This is an active inductor circuit used in many equalizers. That does not mean that it is the best, just that it uses one op amp. A better respected circuit uses two. For example, it can be found in this preview of Horowitz's Art of Electronics: http://books.google.com.pr/books?id=...um=6&ct=result. (Or if that link goes away, search for: active filter gyrator.)

[Mark Hammer]

TL022 op-amps are regularly used in circuits where low current draw is paramount (e.g., LFO circuits) and are generally not found in the audio path.

The low current draw is for two not particularly related purposes: 1) when you have a lot of op-amps and you want the battery to last for a while (note thatthe original MXR 6-band EQ lacked any jack for external power), 2) when you want the LFO to have low enough current requirements that it does not generate spikes on the supply lines also feeding the audio path (the famous "ticking" problem that accompanies many modulation pedals).

What particular advantage the TL022s give the Boss pedal (in comparison to other chip choices) I don't know. Clearly they have many more chips to choose from than, say, MXR did in 1976.