|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
65.19.76.104
In Reply to: RE: Wow, a 7KHz + 60 Hz IMD test? posted by Dave_K on August 16, 2015 at 12:11:16
Test software computes the result so there is no need to look at a display. However, if you want to, my software (Soundforge 10c) can scale the display window for any desired range, e.g.blow up the entire horizontal scale to go from 6900 to 7100 Hz. The FFT itself has more than enough frequency resolution with 65K points and Blackman Window.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Follow Ups:
Sorry, I should have explained what I was getting at yesterday but was running out the door.
The test software is reporting what it calls "intermodulation distortion + noise" using the SMPTE frequencies but if you look at the IMD+N level reported in the summary (0.0019%) it seems obvious that the software is neither using the SMPTE analysis method nor some other method of quantifying the IM sidebands. Or at least it's not doing it correctly. If it were, the result would be more like an order of magnitude lower.
Given the reported level of IMD+N (0.0019% = -94 dB), I suspect the software is including THD + IMD + noise + anything else not the signal. If I'm right, it's a useless test because all it's really measuring is the level of power supply hum in Archimago's test rig. In fact, I think 5 of the 8 metrics reported in the RightMark summary appear to be just measuring power supply hum: Noise level, Dynamic range, THD, IMD + N, IMD + N (swept).
But in order to know for sure whether I'm right or not, you would need to blow up the display to see what's going on near 7 KHz. Maybe you would see +/- 60 Hz sidebands at the -100dB level, but I doubt it. And I suspect the spectral leakage around the 7k peak in an FFT plot would obscure any lower level sidebands.
I think Archimago has a problem with power supply hum in his test setup which is above the level of any distortion products and it results in inaccurate metrics being reported by his software. And the only way to really separate the power supply related components from distortion products is via FFT analysis, in which case choosing a very low f1 for the IMD test isn't helpful.
Also, I really do think the choice of 7KHz and 60Hz frequencies is dumb. I honestly thought that the SMPTE IMD test was a legacy test that nobody used anymore, but after searching around I see that some people are still using it. The problem I see with it is twofold: First, the choice of a very low f1 will make it hard to interpret the result via FFT because low level sidebands could be buried in the spreading of the peak at f2. Second, choosing an f1 that coincides with the power supply frequency could potentially lead to confusing results in systems with a power supply hum problem. I think the DIN method with 8KHz + 250Hz is more useful, as is the HF intermodulation test.
If I were doing the tests, I would have used a wide variety of IM test signals. One needs to look at all the stages of signal processing and look at the effect likely defects and limitations have on the signal output by the system, choosing particular tests to catch each possible problem area.
Measurements are useful for engineering purposes when they are used interactively by an engineer. They are completely useless (if not dishonest) when used as marketing tools to non-technical customers. Even where they are directed at technical customers they are suspect, because the lack of interactivity makes it impossible to do hands-on testing to understand what is going on, which includes at least understanding the accuracy of the measurements and their significance. To use measurements to prove a "religious" point is the equivalent of counting angels on the head of a pin. (Especially as in this case some of the plots did show that operating system changes affected measurements.)
I was not commenting on the intelligence in the selection of IM test signals. However, now that the point has been raised, I suspect that there may be some merit to use of a low frequency test signal if one is looking to see low level non-linearities in a DAC's output. However, it might be even better to use a DC signal if there is a DC path from digital to analog to digital. This will show up a lot of converter artifacts as a change in noise floor. Sometimes one can infer this behavior by looking at published curves, e.g. DAC spec sheets.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Pretty much agree with all of that, particularly measuring being part of interactive, iterative process.
My impression of Archimago's blog is that he's just having fun experimenting like a kid with a chemistry set and not serious about digging into what's going on.
2 agendae points.......
1. He's angry that he can't afford higher performing equipment.
2. Therefore: he's creating a "straw man" to "prove" an objective position in a subjective field.
"Asylums with doors open wide,
Where people had paid to see inside,
For entertainment they watch his body twist
Behind his eyes he says, 'I still exist.'"
How do you know that the problem you create, apparently being gain calibrated to the 60Hz peak since it is by far the largest, doesn't totally confuse the algorithm and invalidate any IMD+N result? If nothing else you've limited the IMD test signal level to IMD artifact signal level ratio by having the gain set where the IMD test signal is lower than it could be under more ideal circumstances and so limited the resolution of any results that are attained.
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: