REW and Testing Electronics - what interface do you like?

[John Mulcahy]

Is it because the mesurement BW increases with frequency?

A log sweep has a pink spectrum, so its energy falls at 10 dB/decade as frequency increases.

 

[JLM1948]

A log sweep has a pink spectrum, so its energy falls at 10 dB/decade as frequency increases.

OK. Same thing, said differently.

 

[jschwender]

I use an M2 for recoording huitar/base, and for measurement. Both scenarios have specific requirements. I don't do microphone recording, but that is where the M2 is really good at. The distortion is low the dynamic range is pretty good, and i think you don't find a bettter device unless you have significantly larger budget. The unbalanced inputs on this device are significantly worse: the distortions are about 20 dB higher, dominated by even order harmonics. The cause is, that they actually don't have unbalanced inputs, and turn the balanced into unbalanced by shorting (-) input to GND. IMHO this is a tinkering solution. This way the M2 produces those distortions, not even in the input buffers, but in the following gain setting ampifier, a THAT 6263. To improve that, i use a pre-amplifier that does proper unbalanced-balanced operation, and this signal is fed into the balanced inputs of the M2. The result is, that even for unbalanced signals like from a guitar or amplifier measurement distortions are far better. The output path of the M2 has very low distortions, due to a pretty good DAC inside (ES9820PRO). The internal monitoring loop is done in a kind of stopgap solution, it changes the load on opamps inside, the way that the level setting of the phones output has influence on the distortions of the monitoring output. Not so nice. In my opinion the level meter display is a nice and useful thing, not only for recording. All potentiometers are digital, not analog. They operate in 3 dB steps, which makes it impossible to precisely tune to a certain level. This is not a disadvantage for music recording, and for measurement it may be a disadvantage in some cases. On the other hand, settings are precisely reproducible, which is very good in some measurement applications.

 

[JLM1948]

The unbalanced inputs on this device are significantly worse: the distortions are about 20 dB higher, dominated by even order harmonics. The cause is, that they actually don't have unbalanced inputs, and turn the balanced into unbalanced by shorting (-) input to GND. IMHO this is a tinkering solution.

I beg to differ. It's common practice, even in high-end and instrumentation equipment.

This way the M2 produces those distortions, not even in the input buffers, but in the following gain setting ampifier, a THAT 6263. To improve that, i use a pre-amplifier that does proper unbalanced-balanced operation, and this signal is fed into the balanced inputs of the M2.

I understand that as you connect the "balanced "output of whatever your source is to an unbalanced input. If the output stage is not floating, you are shorting the inverting leg to ground, which causes the increase in distortion. That's a very common problem (known as the "Tascam problem"). See

Balanced output w/ MC33079

Is there any reason why an MC33079 would be a bad idea in a balanced output stage? Want to use them cos they have been lieing around. Something like the attached schematic but with DC isolation caps on output and input. Also, R3 and R6, why no pf caps for filtering ? Would MC33079 need some ?

groupdiy.com

The result is, that even for unbalanced signals like from a guitar or amplifier measurement distortions are far better.

The guitar is floating and the amp has an unbalanced output, so there is no cold leg shorted to ground.

 

[jschwender]

The guitar plug is a mono plug and it causes the (-) channel of the balanced socket to ground. From an engineering standpoint it is working, which does not mean it is the intended use case or optimum.
Perhaps I have been misleading: my external preamp has an unbalanced input where guitar goes in, and a balanced output, which i plug into the balanced input of the M2. I am not shorting any output, and within the M2 the whole signal path is balanced. A balanced signal path has a high common mode suppression, but only if both signals are exactly inversely phased. If you just short one channel, this is no more the case. This way your balanced signal path transmits unbalanced signal. By definition this is not the intended use case, this is just a fallback solution. On a properly balanced signal path non-linearities of even order do not cause harmonics!. If you use it unbalanced, the same non-linearities cause harmonics.

 

[JLM1948]

The guitar plug is a mono plug and it causes the (-) channel of the balanced socket to ground. From an engineering standpoint it is working, which does not mean it is the intended use case or optimum.

Intended or not, it's a perfectly valid connection.

Perhaps I have been misleading: my external preamp has an unbalanced input where guitar goes in, and a balanced output, which i plug into the balanced input of the M2. I am not shorting any output,

Then why do you say that there is distortion when using an unbalanced source?

and within the M2 the whole signal path is balanced.

I really doubt the whole signal path is balanced in the interface. It is certainly between the input buffer and the A/D converter, because that's how ESS have designed it, but for the rest, certainly not.

A balanced signal path has a high common mode suppression, but only if both signals are exactly inversely phased.

That is plainly untrue. Common Mode rejection works even if only one leg is active, as long as the Wheatstone bridge constituted by the source and load impedances are balanced.

If you just short one channel, this is no more the case. This way your balanced signal path transmits unbalanced signal.

I agree that shorting one leg breaks the balance of the Wheatstone bridge. You're mixing up balanced signal and balanced connection.

By definition this is not the intended use case, this is just a fallback solution.

Says who?

On a properly balanced signal path non-linearities of even order do not cause harmonics!. If you use it unbalanced, the same non-linearities cause harmonics.

Distortion is not related to common-mode, neither to balanced signal transmission.

You are entitled to your opinion, but it is not backed up by EE science.

 

[jschwender]

Intended or not, it's a perfectly valid connection.
You are free to consider any level of distortion on a connection as valid. My general approach is: getting lower distortion is better than having higher distortion. I am not saying it is relevant for everyone and everywhere. If i do guitar recordings, i actually don't mind 2nd harmonics at -99 dB, that is still perfect. As for this application, this whole consideration is totally academic. But for doing measurements, it may matter.

Then why do you say that there is distortion when using an unbalanced source?
1. Because i can measure it. And everyone can verify that, the test is easy.
2. Because signal theory predicts it.

I really doubt the whole signal path is balanced in the interface. It is certainly between the input buffer and the A/D converter, because that's how ESS have designed it, but for the rest, certainly not.
Why do you doubt that? What do you mean by »the rest«??? Please go and have a look into the M2! I did: Signal path is balanced from input jack up to the ADC and from the DAC to the outputs. Even the USB bus uses balanced signal transmission ;–) And the power supply is kind of balanced, it is +5V and -5V, but that is not part of the signal path. One exception is the phones output, this one is unbalanced.

That is plainly untrue. Common Mode rejection works even if only one leg is active, as long as the Wheatstone bridge constituted by the source and load impedances are balanced.
This named condition is exactly not met by feeding a balanced signal into a balanced input by shorting the second leg. Consequently, your falsifying verdit is incorrect.

I agree that shorting one leg breaks the balance of the Wheatstone bridge. You're mixing up balanced signal and balanced connection.

Says who?

Distortion is not related to common-mode, neither to balanced signal transmission.

You are entitled to your opinion, but it is not backed up by EE science.

I disagree.

 

[JLM1948]

You: "You are free to consider any level of distortion on a connection as valid. "
This is definitely what I said. I said applying an unbalanced signal to a balanced input introduces no distortion, as long as the performance of said input is adequate. I'm not telling that distortion does not need to be limited.

Mehen why do you say that there is distortion when using an unbalanced source?
You:. Because i can measure it.
Me: Then your measuremlent setup is faulty.
You: And everyone can verify that, the test is easy.
Me: I have verified many times that it does not.
You: Because signal theory predicts it.
Me: What signal theory? Distortion is due to non-linearities. They can happen in unbalanced and balanced circuits as well.
Now, if you short one leg of a symmetrical output, the signal is shorted to ground, which results in clipped signal and parasitic currents that generate distortion. Note that I don't write "balanced", because balanced refers to equal impedance of the legs and high common-mode impedance. An output stage that uses an inverter to provide a polarity-reversed output is NOT a balanced output, beacuse the CM impedance is near-zero..

Me: That is plainly untrue. Common Mode rejection works even if only one leg is active, as long as the Wheatstone bridge constituted by the source and load impedances are balanced.
You: This named condition is exactly not met by feeding a balanced signal into a balanced input by shorting the second leg. Consequently, your falsifying verdit is incorrect.
Me: You don't understand. I didn't say the second leg is shorted in that case.You don't get the difference between a leg that is impedance balanced without signal and one that s also impedance balanced but also carries signal.
CMRR is a matter of impedance. Period. It has nothing to do with whatever level is present or not.

 

[jschwender]

Got it! The definition of the term balanced has caused misunderstanding. Please note that i am not a native english speaker, and i used the term "balanced" because for me it seems widely synonymously used for symmetric. Also in the manual of the M2 they only talk about "balanced" and "unbalanced", the term symmetric does not occur in it. Therefore i was sure to use that term properly. Of course, by your definition the signal paths inside the M2 are not balanced, they are symmetric. Knowing that gives me a different view on you comments above, now i notice your rich experience and in-depth knowledge! Thank you, i learned a lesson.

What i refer to is symmetric signal routing. And my point is that symmetric signal transmission cancels out all even non-linearities, (if they are equal in both lines). And this is exactly what i measured, and can prove in theory. This is what happens if you run a sine signal over a device with a nonlinear transfer function:

If you now make the second signal Y2 zero (feed asymmetric signal and short the second input line to gnd) then the distortion does not cancel out as there remains a term with cos(2omegat), the 2nd harmonic. QED. In the M2, this is very noticable. In my Steinberg device this is also the case.

 

[JLM1948]

If you now make the second signal Y2 zero (feed asymmetric signal and short the second input line to gnd) then the distortion does not cancel out as there remains a term with cos(2omegat), the 2nd harmonic. QED. In the M2, this is very noticable. In my Steinberg device this is also the case.

Your demonstration is well known , but it applies generally to push-pull stages, where most of teh distortion is caused by the output devices, not by the drivers.
In the case of balanced (or symmetric) input or output stages, the situation is different; there is one input signal that is split and polarity reversed, so whatever distortion is in the signal is polarity reversed, as well as the useful signal; as a result the signal is doubled and distortion too. Of course, the individual distortions of the output devices could partially cancel, or not...
If that is what you experience with your setup, it shows that one of the elements (or both) has a hidden defect.
I know that many active balanced output circuits are not really very good when one leg is shorted. Actually teh only that work satisfactorily atr the THAT1606 and 1646. The competitors from LT (SSM) and TI are a notch below in terms of performance. Those that use standard opamps are even worse.

 

[jschwender]

In case if the M2 the THAT 6263 causes significant distortion on high levels and one leg shortened on the input jack. The preamps that feed it are pretty good in terms of distortion. I don't think it is caused by a defect, as the data sheet shows even for the symmetric case and high signal levels pretty poor distortion curves.
I would say my demonstration applies to any symmetric signal processing device that has non-linearities. Even harmonics cancel out, odd harmonics don't. I am not talking about harmonics that are inversed an fed in, that is something different.

 

[JLM1948]

In case if the M2 the THAT 6263 causes significant distortion on high levels and one leg shortened on the input jack. The preamps that feed it are pretty good in terms of distortion. I don't think it is caused by a defect, as the data sheet shows even for the symmetric case and high signal levels pretty poor distortion curves.

I have looked at the documents on the MOTU website and the 6263 datasheet. I couldn't find anything related to distortion being different when one leg of the input is grounded.

I would say my demonstration applies to any symmetric signal processing device that has non-linearities. Even harmonics cancel out, odd harmonics don't.

It is true for output stages that have non-linearities concentrated in the output devices, but for inputs, the non-linearities are so much lower I doubt it is significant.

I have looked at your attachment. I don't understand the graphs. What is what? Is there one that shows the difference between balanced and unbalanced?

 

[jschwender]

I have looked at the documents on the MOTU website and the 6263 datasheet. I couldn't find anything related to distortion being different when one leg of the input is grounded.
That is right, id did not say that. The Motu manual does not specify if the given specifiactions are valid for symmetric or asymmetric, at least not explicitly. The THAT data sheet does not mention asymmetric operation at all, therefore i believe that this is not intended use. But what the data sheet tells is that the distortion above gain settings of 0 dB is increasing strongly, and my measurements confirm this.
It is true for output stages that have non-linearities concentrated in the output devices, but for inputs, the non-linearities are so much lower I doubt it is significant.

I have looked at your attachment. I don't understand the graphs. What is what? Is there one that shows the difference between balanced and unbalanced?

In fact this is only the asymmetric measurement, with variation of input level and gain setting. What it shows is that distortion rises with gain setting. Low distortion values are only reached with gain setting of <0 dB, or for levels of -20dBFS. Again, this is asymmetric operation.
This one shows the symmetric use case. Note that this is a loopback, so distortion is the sum of out+in distortion. The assumption is that the output distortion is lower than the input distortion which i verified only for a single point of operation. But even though, the loopback distortion is far lower compared to the asymmetric case above, and the 2nd harmonic is less dominant.

 

[JLM1948]

In fact this is only the asymmetric measurement, with variation of input level and gain setting. What it shows is that distortion rises with gain setting. Low distortion values are only reached with gain setting of <0 dB, or for levels of -20dBFS. Again, this is asymmetric operation.

I don't disagree with that; indeed it's quite common to note that the analog stages have distortion increasuing with level.
What I would like to see is the differnce in distortion, all else being equal, when the connection is balanced vs. when one leg is grounded.

 

[jschwender]

Different device, similar effect. I used the M2 to generate a signal of FS level and fed that into the Steinberg UR22. One time asymmetric only to the (+) leg, one time to the (-) leg and one time as symmetric signal. 2nd harmonic is much lower with sym input.

 

[JLM1948]

Different device, similar effect. I used the M2 to generate a signal of FS level and fed that into the Steinberg UR22. One time asymmetric only to the (+) leg, one time to the (-) leg and one time as symmetric signal. 2nd harmonic is much lower with sym input.

I think it just demonstrates that the M2's output stage does not perform well when one leg is shorted. That is quite common actually. That's the biggest challenge designers of electronically balanced output stages are faced with.
THAT and TI are honest with this issue, their datasheets show different THD for sinle-ended (one leg shorted) and differential mode. AD/SSM do not mention that.
Most half-discrete implementations (two or three opamp Cross-Coupled Output Stage) are much worse in this respect.
In comparison, electronically balanecd input stages are much easier to optimize in this respect. They present other challenges, though, like CMRR optimization.