TLE2071 to upgrade TL071 - Some Data

Application is a Trident style EQ. I started with the TL071 and TL072 combo (two of each) since that was the original opamp. All opamps bypassed with 0.1uF from each rail to ground.

I should first say I'm not really fond of the sound of the TL071 series, even when not driven hard. Switching out all the opamps to TLE2071 and TLE2072 sounds MUCH better to me.

Here is some data before and after (note there is heavy averaging going on). Don't read too much into the raw numbers, it's not a calibrated measurement, but it's useful for comparison purposes. There is a 1 kHz test tone at -6 dBFs, the measurement is zoomed in to see the low level signals.

First the EQs are hard bypassed, so you see the effect of the converter itself and also some of the 500 series power supply that might be coming in.

TL071 x2 and TL072 x2

TLE2071 x2 and TLE2072 x2

Harmonic distortion is improved, as is the noise floor. Though curiously the noise rises towards lower frequency.

-150dB? I take it you don't have any build-out resistors.

A bipolar op amp might give you even better results.

If you really need JFET, OPA2132 might also give you slightly better numbers.

But at that level I'm not sure it really matters.

I take it these plots are with the EQ out?

squarewave said:

-150dB? I take it you don't have any build-out resistors.

Click to expand...

That is just the spot noise. If the FFT display is per root Hz then you need to subtract 43dB from the 150dB to get the rms noise in a 20KHz bandwidth which makes it -107dBu (assuming the scale really is dBu) which is quite typical for a unity gain SS stage.

Cheers

Ian

Ok. Yeah. But I was just asking if there were build-out resistors. Pretty clearly there aren't in which case one might consider them seeing as how every modern piece gear that I've every looked at has between 33 and 100 ohms in series with the output. Yeah, the plot won't show the little differences in harmonics and LF noise of different op amps. But that's sort of the point too.

Output is 1646 driver which has 25 ohm resistors.  I double checked output noise with REW and numbers are similar.

squarewave said:

Ok. Yeah. But I was just asking if there were build-out resistors. Pretty clearly there aren't in which case one might consider them seeing as how every modern piece gear that I've every looked at has between 33 and 100 ohms in series with the output. Yeah, the plot won't show the little differences in harmonics and LF noise of different op amps. But that's sort of the point too.

Click to expand...

Are you suggesting build out resistors would affect the measured noise? Are you suggesting they should always be used, even internal to an EQ circuit? Or?

Cheers

Ian

ruffrecords said:

Are you suggesting build out resistors would affect the measured noise?

Click to expand...

What do you mean exactly by "measured noise"? I was just musing about the noise floor. Any series resistance is going to affect the noise floor. We don't know what 0dBFS is in john12ax7's measurements but to get a noise floor at -150dB the build out resistors have to be very small or 0dBFS is actually like +12dBu (which is very possible actually).

ruffrecords said:

Are you suggesting they should always be used, even internal to an EQ circuit? Or?

Click to expand...

Not sure what you mean by "internal". I was just pointing out that the test environment is not very typical. I'm of the opinion that measurements should be made with some realistic impedances. If you look at just about any modern circuit, there are build out resistors. The THAT 1646 is at the very low end at 25 ohms. The datasheet mentions this wrt accidentally plugging into phantom and using ferrite beads to keep RF out of it. So I'm not really advocating one way or another but if you do measurements with the source impedance at literally zero and no build out, yeah you'll see the differences between op amps but unless that's the principal objective, it would be more realistic to use 100 ohms source at least.

For the converter used 0 dBFs is +20 dBu. I'm really not sure what you are getting at regarding build out resistors.  As mentioned it was not meant to be calibrated data,  but rather a direct A/B of a real world situation of how you would actually use the EQ,  which in my case is a hardware insert.

squarewave said:

What do you mean exactly by "measured noise"? I was just musing about the noise floor. Any series resistance is going to affect the noise floor. We don't know what 0dBFS is in john12ax7's measurements but to get a noise floor at -150dB the build out resistors have to be very small or 0dBFS is actually like +12dBu (which is very possible actually).

Click to expand...

By measured noise I mean the noise shown in John's graphs. You seemed to be suggesting that with build out resistors the figures John got should be higher. However, the noise floor is not -150dB. As I mentioned before this FFT is most probably displayed as spot noise. To get the actual rms noise floor over the 20KHz bandwidth you have to integrate these readings over that range. If the plot is essentially flat then this means you need to add 43dB to the -150dB which gets you -107dB which is typical of many op amps at unity gain. A 150 ohm resistor has a Johnson noise of about -130dBu which is insignificant compared to -107dBu so you will not see them make any difference to the noise at the output of an op amp.

Cheers

Ian

ruffrecords said:

By measured noise I mean the noise shown in John's graphs. You seemed to be suggesting that with build out resistors the figures John got should be higher.

Click to expand...

Yes, that is exactly what I am suggesting saying. If the build out resistors were larger or if the source impedance was higher the noise floor would be higher yes and by noise floor I mean the "spot" noise line at -140 to -150. Looking at the noise floor is precisely what these plots are great for.

ruffrecords said:

However, the noise floor is not -150dB. As I mentioned before this FFT is most probably displayed as spot noise. To get the actual rms noise floor over the 20KHz bandwidth you have to integrate these readings over that range.

Click to expand...

If you're going to integrate and produce a datasheet number then great but what's the point of looking at the plot?

The nice thing about these FFT plots is that they show you what's going on at specific frequencies. You can see if harmonics are low-order of more 2nd vs 3rd or whatever. You can see if there's a little bump at 2kHz because of an SMPS throttling. That is the sort of thing that, with enough gain would present as a whistle noise (my MOTU Traveler phantom supply has this problem).

I'm not knocking anything going on here. I was just musing as to why the noise floor -150dB. So now we know that 0dBFS is +20dBu so that completely explains it. Low source impedance, small build-out resistors and +20dBu level is going to give you a -150dBFS floor.

squarewave said:

Yes, that is exactly what I am suggesting saying. If the build out resistors were larger or if the source impedance was higher the noise floor would be higher yes and by noise floor I mean the "spot" noise line at -140 to -150. Looking at the noise floor is precisely what these plots are great for.

Click to expand...

OK, understood but this is not true. As I mentioned, the noise from a 150 ohm resistor is much lower than the noise from a unity gain op amp (whether you look at the spot noise or the rms over 20KHz) so in this instance you will not see any noise contribution from a build out resistor

If you're going to integrate and produce a datasheet number then great but what's the point of looking at the plot?

Click to expand...

Because they give information that the rms value does not - like the presence of hum, distortion products, 1/f noise, spurious signals  for example. But this particular example is an op amp in an EQ circuit operating at unity gain. Under these circumstances you will not see and noise contribution due to build out resistors.

Cheers

Ian

ruffrecords said:

OK, understood but this is not true. As I mentioned, the noise from a 150 ohm resistor is much lower than the noise from a unity gain op amp (whether you look at the spot noise or the rms over 20KHz) so in this instance you will not see any noise contribution from a build out resistor

Click to expand...

I see what you're saying. The op amp is going to put out more noise than a little resistor. That's true.

So a TL071 is 18nV/√hz * √20000 = 2.5uV RMS whereas a 100 ohm resistor is 0.18uV RMS.

It's not a huge difference though. With a really quiet op amp and a little gain, it no longer holds true.

squarewave said:

I see what you're saying. The op amp is going to put out more noise than a little resistor. That's true.

So a TL071 is 18nV/√hz * √20000 = 2.5uV RMS whereas a 100 ohm resistor is 0.18uV RMS.

It's not a huge difference though. With a really quiet op amp and a little gain, it no longer holds true.

Click to expand...

I think it is a huge difference, - it is over 20dB.

Changing the gain of the op mp will increase the op amp output noise but the build out resistor noise will be unchanged so it will be even more insignificant compared to the op amp noise.

Cheers

Ian

Overall I don't think noise is much of a concern in line level applications.  The TL071 has higher noise, but it's not something that really jumped out audibly as a problem in use.

The main thing I noticed was the TLE2071 sounded a lot smoother.  Which might be partially due to significantly lower higher order distortion.

ruffrecords said:

I think it is a huge difference, - it is over 20dB.

Changing the gain of the op mp will increase the op amp output noise but the build out resistor noise will be unchanged so it will be even more insignificant compared to the op amp noise.

Click to expand...

Ok. Well it looks like you got me cornered on this one. I'm just going to step behind this curtain and ...

I will second the finding with the TLE2071(2). I replaced the ones controlling the HF and MF1 in my TAC Scorpion channel strips and the boosting the HF shelf is far more pleasant to the ear than the TL072. I haven't done a noise test but just doing some quick listening with program material revealed the apparent difference.

Thanks!

Paul

squarewave said:

Ok. Well it looks like you got me cornered on this one. I'm just going to step behind this curtain and ...

Click to expand...

LOL, don't worry about it. We are all here to learn. I still struggle with the intricacies of transformers. I know some good rules of thumb but the details of what goes on under the hood are still a bit of a mystery.

Cheers

Ian