Understanding which combination of measured devices will fair better with an inbetween.

[Trdat]

I have a basic understanding of measurements but was hoping someone will chime in and tell me more about what would fair better? And also if anyone has experience in comparing the crossover in a AV to a active crossover?

Option 1. A Marantz AV7005 into a hypex NC502MP amp with dual subwoofers using the internal crossover and XLR outs. I know the Marantz AV7005 has not been measured but for the discussion lets take the AV7705 result of 75 SINAD.

Option 2. Is to use a well measuring DAC like the SMSL SU-8 with a SINAD of 110 and put an active crossover in between the DAC and Hypex NC502MP amplifier such as the Marchand XM44 which has a SINAD of 74.

Now how much will the marchand crossover degrade the sound in option 2? By how many db of SINAD? I am presuming I have to minus something from something but not sure what. This will give me at least on a SINAD comparison which one will fair better.

And lastly do we know anything about an AV crossover vs an active one?

Please some detail and an explanation would be apprecaited for the process of learning and understanding the peculiarities' of SINAD.

 

[Blumlein 88]

Impossible to say without more details on measured performance. Which is why I don't like SINAD. SINAD is the ratio of THD+Noise.

So one device could be SINAD of -74 db with a very low noise floor and 3rd harmonic distortion which equals -74 db which dominates the SINAD number. Another device could be a bit noisy with THD of less than -100db , but enough noise to cover it up and result in SINAD -74 db.

What you really want to know in your case is harmonic distortion levels of the first few harmonics and the level of noise. Not a combined figure for those.

As for your question, the crossover in an AVR or AV preamp is an active crossover. Whether it is better than a Marchand or not depends upon the relative roll off rates and quality of the circuit. Probably in most circumstances the one in the AV product will do just as well.

 

[dougi]

another option could be to put a digital xover before the DAC that goes to the Hypex and get a second DAC for the subs. A miniDSP nandigi has optical and coax inputs and coax outputs. Do the xover in there, room correct the subs and send coax outs to DACs. No SINAD issues to worry about. I do this for some active speakers, using Topping E30 DACs.

 

[Trdat]

Impossible to say without more details on measured performance. Which is why I don't like SINAD. SINAD is the ratio of THD+Noise.

I remember you have mentioned this before Blumlien. And it also has been stated by Amir(if ai am not mistaken) that SINAD is not the be all and end all of measurements, which I am aware of. I thought Ill take that as a starting point then compare the other measurments.

What you really want to know in your case is harmonic distortion levels of the first few harmonics and the level of noise. Not a combined figure for those.

So your saying to compare noise and distortion seperately? Can you point me to the direction of the measurement of the first few harmonics...?(Ill take a look into it) The marchand crossover doesn't have all the typical measurements.

As for your question, the crossover in an AVR or AV preamp is an active crossover. Whether it is better than a Marchand or not depends upon the relative roll off rates and quality of the circuit. Probably in most circumstances the one in the AV product will do just as well.

Thanks for this.

 

[Blumlein 88]

Okay look here, recent review. Notice how high the 2nd and 3rd harmonic are. The noise level isn't super, but much lower than the distortion, so the distortion limits your SINAD. Notice how there are other higher order harmonics, but much lower in level. This is pretty typical, 2nd and 3rd are usually the highest with others low enough not to matter much. If these were lower, like -100 db they are of no concern.

In this device the highest harmonic is -111 db in the 3rd harmonic. All others are lower. You'll never hear any of this distortion. While the noise level isn't horrid, it is surprisingly high in this case and over the entire 20 khz band adds up enough that it is higher than the distortion levels. The channel in red being -103 db which sets your SINAD level. All the harmonic distortion is much lower than this.

So when you asked is a setup with 75 db SINAD better than one with 74 db SINAD well it depends upon more than just this one number.

If 74 db is just a single harmonic maybe the 3rd at -74 db and the noise is low and other distortion is low then you'll probably never hear any problem from any of it. If on the other hand, all distortion is very low and noise level over the whole band is -74 db, that level of noise with your system turned up loud might be audible. Or suppose you have some 60 hz related hum at this level setting SINAD. It too turned up is going to be audible. Which means it isn't as good a result even though SINAD for both situations is pretty much the same.

Amir's reason for using SINAD has to do with looking at the very best. To get SINAD numbers better than -110 db obviously both noise and distortion must be very low. Any gear designed to have such numbers or better is sure to be blameless on both noise and distortion and idle tones and power supply noise bleed thru etc. All of that must be handled in an excellent manner.

It is in the lesser SINAD levels you have to be careful. Middling SINAD results can be okay, very good, or not good enough depending upon the particulars. I think it is the view of many, that with so much gear having great SINAD numbers and some of it not very expensive why would you bother with something scoring merely okay SINAD numbers? It might perceptibly be just as good, but why take the chance, just get the top notch engineered gear with top notch test results and have no worries. Reward those who make truly great gear.

 

[Trdat]

another option could be to put a digital xover before the DAC that goes to the Hypex and get a second DAC for the subs. A has optical and coax inputs and coax outputs. Do the xover in there, room correct the subs and send coax outs to DACs. No SINAD issues to worry about. I do this for some active speakers, using Topping E30 DACs.

I already have a triamp active speaker set up, the proposed system is my second system which I would like to use for home theatre duties as well as music listening hence why I want to use what I got and keep my money for upgrading my larger more pristine system.

 

[Trdat]

Okay look here, recent review. Notice how high the 2nd and 3rd harmonic are. The noise level isn't super, but much lower than the distortion, so the distortion limits your SINAD. Notice how there are other higher order harmonics, but much lower in level. This is pretty typical, 2nd and 3rd are usually the highest with others low enough not to matter much. If these were lower, like -100 db they are of no concern.

Starting to see how distortion and noise depending on its level plays into the whole thing over just SINAD itself. And how distortion limits the SINAD, obviously to me this means that we are getting a horrible SINAD score due to the distortion when the noise isn't that bad. I hope I have understood correctly... But why does distortion limit SINAD can noise limit the distortion for SINAD?

In this device the highest harmonic is -111 db in the 3rd harmonic. All others are lower. You'll never hear any of this distortion. While the noise level isn't horrid, it is surprisingly high in this case and over the entire 20 khz band adds up enough that it is higher than the distortion levels. The channel in red being -103 db which sets your SINAD level. All the harmonic distortion is much lower than this.

So am I understanding correctly that in this second device the noise is pretty bad and although the distortion is inaudible you get a high SINAD result which works out to be worse than the first device?

So when you asked is a setup with 75 db SINAD better than one with 74 db SINAD well it depends upon more than just this one number.

What I really was asking is how an inbetween device effects the performance of well measuring equipment compared to a lower SINAD system but your reply is absolute gold Blumlien really appreaciate it. Learning much more about measuremnts and this probably can help me with so many grey areas of understanding measurements and at least gives me the skill in chosing between SINAD numbers more confidently.

If 74 db is just a single harmonic maybe the 3rd at -74 db and the noise is low and other distortion is low then you'll probably never hear any problem from any of it. If on the other hand, all distortion is very low and noise level over the whole band is -74 db, that level of noise with your system turned up loud might be audible. Or suppose you have some 60 hz related hum at this level setting SINAD. It too turned up is going to be audible. Which means it isn't as good a result even though SINAD for both situations is pretty much the same.

This thing is there is probably so many variablilties in distortion/noise and certain harmonics that can and can't be audible but like I said it gives me a base to dig deeper into this. I will try and look into the particular devices.

Amir's reason for using SINAD has to do with looking at the very best. To get SINAD numbers better than -110 db obviously both noise and distortion must be very low. Any gear designed to have such numbers or better is sure to be blameless on both noise and distortion and idle tones and power supply noise bleed thru etc. All of that must be handled in an excellent manner.

This is also great to know!

It is in the lesser SINAD levels you have to be careful. Middling SINAD results can be okay, very good, or not good enough depending upon the particulars. I think it is the view of many, that with so much gear having great SINAD numbers and some of it not very expensive why would you bother with something scoring merely okay SINAD numbers? It might perceptibly be just as good, but why take the chance, just get the top notch engineered gear with top notch test results and have no worries. Reward those who make truly great gear.

Agree, I bought my marcahnd crossover well before ASR and it works well just curious how it would effect the overall SINAD score but I will look into it further.

Also, Blumlien I have a post under your Marantz AV7701 measurement it would be greatly appreciated if you could also one day in your time get to that question. A discussion there will also catapult me from darkness to basic competence.

 

[Trdat]

Also can you tell me where I am looking at the noise in the measurements?

 

[Blumlein 88]

Also can you tell me where I am looking at the noise in the measurements?

I answered your question in the 7701 thread.

On noise, Amir doesn't measure just noise, he usually shows SNR or signal to noise ratio, and also dynamic range which is similar, but not the exact same thing. So with SNR you have a max signal and you have how much noise is left if you notch out the test signal.

If you had something with a SNR of 85 db, it could also have SINAD of about -85 db if all distortion is lower than -85 db which is possible and happens sometimes. But when you look at FFT's of such a thing, you will see the spikes of the harmonic distortion lower than -85 db and the noise floor might look like it is down around -124 db. How can this be if noise is -85 db? The answer is in understanding what an FFT graph is showing.

Noise has to have a specified bandwidth. In audio this will often be 20 khz. If I connect a wide bandwidth multimeter to a -85 db white noise source over 20 khz I'd get the reading of -85 db relative to full scale. What happens if I use some filters and split this into two bands? 0-10 khz and 10khz-20 khz. What will my meter read for each band? It will read -88 db. Let me suppose I add more filtering. Splitting it into 4 bands 0-5khz, 5-10khz, 10-15 khz and 15-20khz. If I measure each of my filtered bands I now get a reading of -91 db in each band.

This is what an FFT is doing. It is splitting the total 20 khz band into many much smaller bands so the noise level in each band (called bins in an FFT) is very low which lets us see signals below the noise floor clearly. The larger the number of bins the smaller the bandwidth and the smaller the noise level in each bin. If I used a 32,768 bin FFT, on -85 db white noise, then each bin would have about -124 db in it. So if there is a fixed tone from harmonic distortion at -110 db I'll see it on the FFT even though the noise level for the device over the full 20 khz band is -85 db. As the FFT will have reduced the noise in each bin it is showing to -124 db.

Does this seem clear and make sense to you?

 

[Blumlein 88]

https://www.audiosciencereview.com/...-audio-sy-dap2002-review-dsp-amplifier.20114/

Here is an excellent example of an amp where noise is much higher than distortion. -72 db noise and -95 db distortion. So SINAD is -72 db.

If noise and distortion numbers were reversed this would actually be a much better amp for music. This much noise is probably going to be heard when you turn your system well up in volume.

 

[Trdat]

On noise, Amir doesn't measure just noise, he usually shows SNR or signal to noise ratio, and also dynamic range which is similar, but not the exact same thing. So with SNR you have a max signal and you have how much noise is left if you notch out the test signal.

Okay, but is it something I can read on the results of Amir's SINAD graph? I can't seem to discern the noise from the 2nd and 3rd harmonic distortion you explained. I'm going to hazard a basic guess and say the noise is the DBA level at which the "frequency sine wave sweep" (whatever it's called) starts from the left hand side on the DBA graph?

If you had something with a SNR of 85 db, it could also have SINAD of about -85 db if all distortion is lower than -85 db which is possible and happens sometimes. But when you look at FFT's of such a thing, you will see the spikes of the harmonic distortion lower than -85 db and the noise floor might look like it is down around -124 db. How can this be if noise is -85 db? The answer is in understanding what an FFT graph is showing.

Noise has to have a specified bandwidth. In audio this will often be 20 khz. If I connect a wide bandwidth multimeter to a -85 db white noise source over 20 khz I'd get the reading of -85 db relative to full scale. What happens if I use some filters and split this into two bands? 0-10 khz and 10khz-20 khz. What will my meter read for each band? It will read -88 db. Let me suppose I add more filtering. Splitting it into 4 bands 0-5khz, 5-10khz, 10-15 khz and 15-20khz. If I measure each of my filtered bands I now get a reading of -91 db in each band.

This is what an FFT is doing. It is splitting the total 20 khz band into many much smaller bands so the noise level in each band (called bins in an FFT) is very low which lets us see signals below the noise floor clearly. The larger the number of bins the smaller the bandwidth and the smaller the noise level in each bin. If I used a 32,768 bin FFT, on -85 db white noise, then each bin would have about -124 db in it. So if there is a fixed tone from harmonic distortion at -110 db I'll see it on the FFT even though the noise level for the device over the full 20 khz band is -85 db. As the FFT will have reduced the noise in each bin it is showing to -124 db.

Does this seem clear and make sense to you?

This is slightly more technical but it will register sooner or later, I get the jist of it. This information has given me a lot more insight to measurements.

 

[BDWoody]

This is slightly more technical but it will register sooner or later, I get the jist of it. This information has given me a lot more insight to measurements.

You have gotten some of the best explanations I've seen. When people actually want to learn about this stuff, the people here are amazingly knowledgeable and helpful. Kudos to @Blumlein 88 , and to you for actually making the effort to understand.

 

[Trdat]

You have gotten some of the best explanations I've seen. When people actually want to learn about this stuff, the people here are amazingly knowledgeable and helpful. Kudos to @Blumlein 88 , and to you for actually making the effort to understand.

And I am very thankfull for that, it is effort and it is greatly appreciated.

BDwoody, yeh I use all these threads as a reference reading them over and over again. Things come later or with some other reading, I come back and just instantaneously pick it up. I have always wondered how far someone can learn in Audio from being self taught, without a tutor or a formal basic electrical engineering course. I am pushing the boundaries, I don't expect to design amps or fully understand schematics but if I can get the above down pat its an accomplishment thanks to many on this forum who put effort in answering threads.

Do you think maybe some sort of categorized system for some of these threads that ouze information and educational material can be made? Maybe a list somewhere with headings. I know we can always use a search funtion but its just an idea.

 

[BDWoody]

Do you think maybe some sort of categorized system for some of these threads that ouze information and educational material can be made?

There is the huge reference library here, but there are definitely some threads that seem to get some of these more technical but critical questions answered better than others.

Short an index, that's one of the big benefits of being a regular around here. I'm like you, in that my background has nothing to do with any of this. What I've learned, I've learned from the people here over the last couple of years a spoonful at a time.

 

[Blumlein 88]

Okay, but is it something I can read on the results of Amir's SINAD graph? I can't seem to discern the noise from the 2nd and 3rd harmonic distortion you explained. I'm going to hazard a basic guess and say the noise is the DBA level at which the "frequency sine wave sweep" (whatever it's called) starts from the left hand side on the DBA graph?





This is slightly more technical but it will register sooner or later, I get the jist of it. This information has given me a lot more insight to measurements.

Amir gives dynamic range which is not much different than SNR and gives you an idea of the noise floor. A dynamic range of 110 db means noise is -110 db from max output level.

As for looking at the FFT for his SINAD you have to do some simple math and it is only approximate. You may have noticed Amir or others talk about FFT gain. That is the lowering by 3 db each time you double the FFT bins. I believe on the SINAD charts Amir uses an FFT size which gives 36 db FFT gain. What does that mean? Look at this chart below. I've added a yellow line estimating the average level of the fuzzy noise floor in this graph. I've also marked the 2nd and 3rd harmonic spikes sticking out from the noise floor. You have 36 db of FFT gain pushing the noise in the graph down. So -108+36 db FFT gain = -72 db noise level over the entire 20 khz band roughly. And indeed Amir noted in this review noise level set SINAD because it was so high. And his SINAD measured 71 db and 72 db for each channel. See the next example below.

In this second example noise is much lower. I added a yellow line estimating it as -150 db which with the 36 db gain would make me think noise is -114 db. Indeed the dynamic range measured about -114 db. The third harmonic which I've marked is higher than the other harmonics you see sticking up out of the grassy noise floor. It is about -112 db. So in this case SINAD is set almost equally by noise and 3rd harmonic. And you see his SINAD rating is about -111 db which is pretty close to what -112 and -114 db would add up to.

So you can eyeball the average noise floor level in the SINAD FFT and reduce it by 36 db to get a reasonable estimate of noise level.

 

[Trdat]

Things are much clearer now the only thing is the FFT bins that are a little difficult to totally comprehend(your math of minusing 36 is enough for now to see the noise level with Amir's measurements and compare which I have understood well) but I don't think that is of paramount importance to understand the overall concept. I do have a few more questions though, simple ones that will clear a few things up, if you don't mind.

1. Amir's test is a 1khz tone. So is performing the test only at this frequency representantive of the whole bandwidth when it comes to distortion?

2. You mention that everything under about 100db is generally inaudible. I understand that is very technical and open for debate but is the this the general consensus with noise and distortion?(at least what we agree here on ASR)

3. You also mentioned a few times that noise is worse than distortion depending on the distortion, especially if the distortion is the 3rd harmonic and if under 100db? Again, is that the general consensus that noise with a similar result is more audible and more worrying than distortion?

4. Lastly, as the harmonics continue say 2nd, 3rd, 4th are they less important than the previous? So 2nd harmonics is the most audible and the one we should be looking out for.

 

[Blumlein 88]

Things are much clearer now the only thing is the FFT bins that are a little difficult to totally comprehend(your math of minusing 36 is enough for now to see the noise level with Amir's measurements and compare which I have understood well) but I don't think that is of paramount importance to understand the overall concept. I do have a few more questions though, simple ones that will clear a few things up, if you don't mind.

I am afraid you may get a lot of maybe and might be answers here. These are all very good questions you ask.

1. Amir's test is a 1khz tone. So is performing the test only at this frequency representantive of the whole bandwidth when it comes to distortion?

1 khz is something of a standard to test. The first few harmonics occur where our hearing is most sensitive. Our hearing is less sensitive to distortion at low frequencies, and at higher frequencies the harmonics are too high to hear. Good devices don't change distortion over frequency very much, but not all devices are good. So sometimes distortion rises at higher frequencies.

2. You mention that everything under about 100db is generally inaudible. I understand that is very technical and open for debate but is the this the general consensus with noise and distortion?(at least what we agree here on ASR)

Another gray area. Some might prefer -110 db, some might say you can get away with even -80 db. One reason for these gray areas is we can come up with particular listening tests where we can probe the limits of hearing distortion or noise. When we are listening to more complex sounds from music our ability to hear it is much reduced. 1% distortion on music is likely fine and it might not matter until more like 3% with some music. 1% is merely -40 db.

Noise is a little clearer to answer. If you turn your system up as loud as you'll ever listen do you hear noise? If not you are okay. Depending upon how quiet your room is, how sensitive your speakers, that might be -80 db or it might be -100 db. If you get noise down around -100 db you are very unlikely to hear it, and if you barely did in quiet portions almost any music at all would mask it.

Also remember noise can be things like hum. 60,120, 180 hz from power supply hum doesn't have to be very loud until it ruins quieter portions of music.

3. You also mentioned a few times that noise is worse than distortion depending on the distortion, especially if the distortion is the 3rd harmonic and if under 100db? Again, is that the general consensus that noise with a similar result is more audible and more worrying than distortion?

Harmonic distortion only occurs during a signal to be distorted. So it can be masked by that signal. Especially 2nd and 3rd harmonics. Noise on the other hand can be constant and therefore isn't only going to occur when there is other signal to mask it. Without a signal it is simply there to be heard. Portions of music will cover it, but quiet parts of recordings or in between there is nothing to mask constant noise.

4. Lastly, as the harmonics continue say 2nd, 3rd, 4th are they less important than the previous? So 2nd harmonics is the most audible and the one we should be looking out for.

All the harmonics depends upon levels. Typically after the 3rd higher harmonic distortion is at lower and lower levels. There are exceptions, and upper harmonics can be more noticeable if they aren't low because being further from the main signal they are masked less.

Reading this wikipedia entry on masking is probably worthwhile.
https://en.wikipedia.org/wiki/Auditory_masking

Masking can be rather complex with forward and backward masking which varies with level and frequency and how long it lasts. Generally you'd strive to keep all harmonic distortion so low it doesn't need any masking to be inaudible.

All these are debatable areas. If you wanted a near indisputable set of guidelines, you want all noise, all distortion and all spurious noise to be -120 db below full output or lower. That would be right up against state of the art (SOTA) limits of what is physically possible. OTOH, most of the time, nearly all of the time, some specifications can be loosened and have no consequences you can hear. If it is inaudible, making it more inaudible doesn't make any difference. There is a big price difference in a 100% blameless audio system and one that is effectively so 98% of the time.

Transducers are the bottleneck to performance. On the playback end we are talking speakers and headphones. The frequency response and distortion levels put a limit on how good the sound can be in terms of fidelity.

 

[Trdat]

I am afraid you may get a lot of maybe and might be answers here. These are all very good questions you ask. 1 khz is something of a standard to test. The first few harmonics occur where our hearing is most sensitive. Our hearing is less sensitive to distortion at low frequencies, and at higher frequencies the harmonics are too high to hear. Good devices don't change distortion over frequency very much, but not all devices are good. So sometimes distortion rises at higher frequencies.
Another gray area. Some might prefer -110 db, some might say you can get away with even -80 db. One reason for these gray areas is we can come up with particular listening tests where we can probe the limits of hearing distortion or noise. When we are listening to more complex sounds from music our ability to hear it is much reduced. 1% distortion on music is likely fine and it might not matter until more like 3% with some music. 1% is merely -40 db.

Masking can be rather complex with forward and backward masking which varies with level and frequency and how long it lasts. Generally you'd strive to keep all harmonic distortion so low it doesn't need any masking to be inaudible.

Transducers are the bottleneck to performance. On the playback end we are talking speakers and headphones. The frequency response and distortion levels put a limit on how good the sound can be in terms of fidelity.

As hard as we try you can't find information like the above.

One more question though, I was reading through the SoundandVision measurements and they give you a distortion reading of say .0037% and I was trying to work out what that was. Now that you mention that -40db is 1%, is there a graph that shows me what percentage is at what minus DB so I can compare various tests that might have a percentage vs db. Of course if this is the same? And what type of distortion is this just to confirm? Is it THD+N?

Lastly in the soundandvision tests its' mentioned 100milivolts, what is that in vrms?

I think this will be the last of it for now. Again, simple questions that you can't really find solid info on and an explanation especially one from you take us deep. So thanks again.

 

[Blumlein 88]

Online calculator for % to db distortion.
http://www.sengpielaudio.com/calculator-thd.htm

1%= -40 db
.1%= -60 db
.01%= -80 db.
.001%= -100 db

SINAD is the obverse of thd+n. One is given in minus db and the other usually in %, but really the same measurement.

db is just a handy way to express large ratios. So it can represent noise only, thd only or both.

100 millivolts is .1 volts.