No? Then can you give an example of when you have intentionally gone with a "worse" number because it sounded better?
se
On the subject of distortion, it might be interesting to put forth a little of what we know about distortion measurement, and its history.
First, there was harmonic distortion measurement. Usually, the residual was a bit high, as quality oscillators were difficult to build, but it worked OK. Around 1950, SMPTE IM distortion measurement was put forth. It appeared to have a lower inherent residual, and more sensitive than harmonic distortion for a given static nonlinearity. For tube equipment, it was even better, as it tended to ferret out substandard output transformers. For some reason, (at least lost to me) this measurement was better for the film industry and this is why the SMPTE sponsored this distortion measurement. Heathkit made a really nice unit, cheaply, that could be further modified to get as low as .005% distortion, residual. I used this for about a decade.
Harmonic distortion was still very important for analog tape recorder measurement, but NOT THD. Instead we used wave analyzers that measured separate harmonics. Usually, only 2nd and 3rd harmonic were important in magnetic tape, but it was important to separate them, as 2'nd harmonic implied a problem with the tape recording system. 3'rd was very predictable for a given brand of tape, and almost always the same.
In the middle 1970's, we all changed to THD, because HP and Sound Technology came out with a vast improvement in oscillator design and convenience in measuring harmonic distortion. This gave us reasonable measurements from 10-100,000 Hz, and is still popular today. SMPTE IM was also available in the same test set, so measurements could be easily compared. Usually IM gave slightly bigger numbers with a low frequency static nonlinearity, but harmonic was more interesting above 1KHz or so.
Sometimes, we would bring out our old fashioned wave analyzers and look at the residual of the THD or IM analyzer. This is always enlightening, more than the actual measured amount of distortion, in most tests.
Today, we tend to use some sort of spectrum analyzer, either as a piece of test equipment or as a program in a computer. This is a convenient way to measure, but not necessarily better than the older methods, and sometimes obscures distortion measurements with a great deal of extraneous 'stuff' like hum, etc. Next, I will talk about more exotic tests that we have found interesting and useful, that don't necessarily track the harmonic or IM tests.
First, there was harmonic distortion measurement. Usually, the residual was a bit high, as quality oscillators were difficult to build, but it worked OK. Around 1950, SMPTE IM distortion measurement was put forth. It appeared to have a lower inherent residual, and more sensitive than harmonic distortion for a given static nonlinearity. For tube equipment, it was even better, as it tended to ferret out substandard output transformers. For some reason, (at least lost to me) this measurement was better for the film industry and this is why the SMPTE sponsored this distortion measurement. Heathkit made a really nice unit, cheaply, that could be further modified to get as low as .005% distortion, residual. I used this for about a decade.
Harmonic distortion was still very important for analog tape recorder measurement, but NOT THD. Instead we used wave analyzers that measured separate harmonics. Usually, only 2nd and 3rd harmonic were important in magnetic tape, but it was important to separate them, as 2'nd harmonic implied a problem with the tape recording system. 3'rd was very predictable for a given brand of tape, and almost always the same.
In the middle 1970's, we all changed to THD, because HP and Sound Technology came out with a vast improvement in oscillator design and convenience in measuring harmonic distortion. This gave us reasonable measurements from 10-100,000 Hz, and is still popular today. SMPTE IM was also available in the same test set, so measurements could be easily compared. Usually IM gave slightly bigger numbers with a low frequency static nonlinearity, but harmonic was more interesting above 1KHz or so.
Sometimes, we would bring out our old fashioned wave analyzers and look at the residual of the THD or IM analyzer. This is always enlightening, more than the actual measured amount of distortion, in most tests.
Today, we tend to use some sort of spectrum analyzer, either as a piece of test equipment or as a program in a computer. This is a convenient way to measure, but not necessarily better than the older methods, and sometimes obscures distortion measurements with a great deal of extraneous 'stuff' like hum, etc. Next, I will talk about more exotic tests that we have found interesting and useful, that don't necessarily track the harmonic or IM tests.
Last edited:
Last Friday, I rolled off some of the high frequencies and left in a bass bump on a theater sound system to make it more musical. Not as flat as it could be, but everyone who listened agreed on the changes. Common practice. Flat sounded bad and would be unreliable.
At this point, I think it useful to mention the so called, Hirata Test. This was a test devised by Dr. Hirata, and it measures something rather different than harmonic or IM distortion. It is difficult to tell you exactly what it measures, BUT it does measure differences and they tend to track with subjective impressions of the human ear.
The Hirata test signal is composed of some step signals that are asymmetrical in the way they are added together. Most of the time we think of a 'staircase' of being a series of ascending and descending steps. In this case, the ascending staircase is somewhat different than the descending staircase. You might say ascending might be double tall, and the descending might be double wide, but only single tall. It all averages out from a DC point of view, but NOT in the short term.
There is ample explanation and theory in the AES and other literature from about 1980, by Dr. Hirata, but it is the results of this test that I found interesting. First of all, most EARLY solid state amps measured lousy in this test, BUT most good tube amps like the Marantz 7, etc measured very good. Why, I would be hard pressed to prove.
However, I tried the Hirata test with a circuit of my own. At this time, I used the typical differential fet input, second stage differential with a current mirror, so popular here and elsewhere. In order to get the LOWEST measured IM and harmonic distortion, I added an AC balance control, consisting of an electrolytic capacitor and a potentiometer to the standard circuit. To my amazement, adding the AC balance control LOWERED the IM distortion, BUT RAISED the Hirata distortion to a significant degree. Afterward, I decided to remove the AC balance control, in favor of lower Hirata distortion, therefore getting increased IM distortion in the bargain. This is but one example where tradeoffs are necessary in audio design.
The Hirata test signal is composed of some step signals that are asymmetrical in the way they are added together. Most of the time we think of a 'staircase' of being a series of ascending and descending steps. In this case, the ascending staircase is somewhat different than the descending staircase. You might say ascending might be double tall, and the descending might be double wide, but only single tall. It all averages out from a DC point of view, but NOT in the short term.
There is ample explanation and theory in the AES and other literature from about 1980, by Dr. Hirata, but it is the results of this test that I found interesting. First of all, most EARLY solid state amps measured lousy in this test, BUT most good tube amps like the Marantz 7, etc measured very good. Why, I would be hard pressed to prove.
However, I tried the Hirata test with a circuit of my own. At this time, I used the typical differential fet input, second stage differential with a current mirror, so popular here and elsewhere. In order to get the LOWEST measured IM and harmonic distortion, I added an AC balance control, consisting of an electrolytic capacitor and a potentiometer to the standard circuit. To my amazement, adding the AC balance control LOWERED the IM distortion, BUT RAISED the Hirata distortion to a significant degree. Afterward, I decided to remove the AC balance control, in favor of lower Hirata distortion, therefore getting increased IM distortion in the bargain. This is but one example where tradeoffs are necessary in audio design.
Last edited:
John this is interesting info. I need to read up on Hirata.
One of the reasons for the relative impopularity of the Hirata test may be that its diffficult to express in a number. That's the advantage of THD: everybody associates 0.01 as better than 0.05. Of course, without more info on the harmonic composition it's anybodies guess which one sounds better, but that question often remains unasked, isn't it.
jd
I found this:
audiobytz - biweekly e-newsletter
P.S. hallo Andre ! i see you are visitting this thread. I have not forgotten your CD and will send it to you after the High End in Munich.
audiobytz - biweekly e-newsletter
P.S. hallo Andre ! i see you are visitting this thread. I have not forgotten your CD and will send it to you after the High End in Munich.
Yes I know Jan, I just wanted to emphasize this.
I often hear that some methods are 'difficult to evaluate or express', and THD or CCIF IMD give numbers easily
OK. It's like politics, really. It's not so important whether the message is correct, as long as it is a simple one!
jd
Sure, however let me start by saying that out of context and repeated second hand its really not correct. And I have the info second hand. I needed some so I can optimize the interface for the server component of the system.
The correct description is that it doesn't used standard LSI functions like an AES receiver or DAC. Its all done in FPGA's large programmable chips that can be made to implement pretty much any digital function, including the digital parts of a DAC. The conversion from digital to analog is a mixture of the FPGA and external discrete components and the output is a discrete amplifier. Based on this you could say no standard IC's (although FPGA's are becoming more commonplace).
Thanks for your clarification - you're using FPGAs which most definitely are 'off the shelf' parts. Scott, you were right
John,
I read again the Hirata paper, it's indeed a smart way to get information about the TYPE of non-linearity of an amp. The plots are revealing.
It seems more a tool for a designer than something that can be used to spec an amp to the buying public.
I wonder though if there is anyone in the industry who actually uses this method. Do you know any?
jd
I read again the Hirata paper, it's indeed a smart way to get information about the TYPE of non-linearity of an amp. The plots are revealing.
It seems more a tool for a designer than something that can be used to spec an amp to the buying public.
I wonder though if there is anyone in the industry who actually uses this method. Do you know any?
jd
- Status
- Not open for further replies.