Some of you remember when CD players came out.
1. Analog Filters:
Many early CD players used Analog Filters, (Real LC filters Only). There were No digital filters in those models.
The square wave test signals had ringing right after the voltage step, the ringing died down in time across the square wave flat section.
All ringing was gone before the next transition (voltage step).
2. Digital Filters:
Then along came CD players that used digital filtering . . .
The square wave test signals had "pre" ringing before the voltage step (transition), and ringing right after the voltage step; but the center of the square wave flat section's ringing was much lower amplitude.
In layman's terms, the digital filter "predicts" when the voltage step is coming, so it rings before the voltage step (transition).
. . . Do Not believe this over-simplified explanation, instead the effect is due to the digital filter's particular phase / time response which differs according to the frequencie(s) of the signal. A good fast rise square wave has frequency components all the way from DC to far in excess of the 22,000 CD audio band.
A digital function generator that uses digital filtering has exactly the same effect.
I believe that is what you are seeing from your DAC.
3. 16 bits, or 14 bits:
Then there were early CD players that did not have 16 bit DACs, they only had 14 bit DACs, ($aved Money), they threw away the last 2 bits that came off the 44.1k 16 bit stream (44.1k x 2 [L and R channel]).
Less resolution, but many did Not notice the difference; except at the ca$h register.
4. Two DACs, or one DAC:
An early CD player had only one DAC. That DAC ran at 88.2k.
It was used in an interleaved mode: 44.1k times per second it spit out the Left Channel, and 44.1 times per second it spit out the Right Channel.
There was a switch that connected the interleaved output first to the Left Channel analog circuitry, than to the Right Channel analog circuitry.
Many did not notice the difference.
Because:
The interleaved time error was not very much . . . just turn your right ear 1/16 inch forward, and your left ear 1/16 inch backward with respect to the perfectly placed speakers, and you sitting exactly in the sweet spot (no cheating, all 3 measurements need to be within (1/16 inch)/3, = 1/200 inch or you did not compensate properly).
5. Ringing:
A. Sometimes a 'ringing' square wave indicates an oscillation.
B. Sometimes a 'ringing' square wave indicates a phase disturbance, or amplitude versus frequency disturbance, or both (thes do Not indicate an oscillation).
Learn to tell the difference between A. and B., or you may loose sleep thinking it is A. when it actually is B.
It does not make sense to be 'disturbed' if it is Not an oscillation.
6. Sometimes, what we do not know, just might not hurt our listening sessions.
"Enjoy the Music"
1. Analog Filters:
Many early CD players used Analog Filters, (Real LC filters Only). There were No digital filters in those models.
The square wave test signals had ringing right after the voltage step, the ringing died down in time across the square wave flat section.
All ringing was gone before the next transition (voltage step).
2. Digital Filters:
Then along came CD players that used digital filtering . . .
The square wave test signals had "pre" ringing before the voltage step (transition), and ringing right after the voltage step; but the center of the square wave flat section's ringing was much lower amplitude.
In layman's terms, the digital filter "predicts" when the voltage step is coming, so it rings before the voltage step (transition).
. . . Do Not believe this over-simplified explanation, instead the effect is due to the digital filter's particular phase / time response which differs according to the frequencie(s) of the signal. A good fast rise square wave has frequency components all the way from DC to far in excess of the 22,000 CD audio band.
A digital function generator that uses digital filtering has exactly the same effect.
I believe that is what you are seeing from your DAC.
3. 16 bits, or 14 bits:
Then there were early CD players that did not have 16 bit DACs, they only had 14 bit DACs, ($aved Money), they threw away the last 2 bits that came off the 44.1k 16 bit stream (44.1k x 2 [L and R channel]).
Less resolution, but many did Not notice the difference; except at the ca$h register.
4. Two DACs, or one DAC:
An early CD player had only one DAC. That DAC ran at 88.2k.
It was used in an interleaved mode: 44.1k times per second it spit out the Left Channel, and 44.1 times per second it spit out the Right Channel.
There was a switch that connected the interleaved output first to the Left Channel analog circuitry, than to the Right Channel analog circuitry.
Many did not notice the difference.
Because:
The interleaved time error was not very much . . . just turn your right ear 1/16 inch forward, and your left ear 1/16 inch backward with respect to the perfectly placed speakers, and you sitting exactly in the sweet spot (no cheating, all 3 measurements need to be within (1/16 inch)/3, = 1/200 inch or you did not compensate properly).
5. Ringing:
A. Sometimes a 'ringing' square wave indicates an oscillation.
B. Sometimes a 'ringing' square wave indicates a phase disturbance, or amplitude versus frequency disturbance, or both (thes do Not indicate an oscillation).
Learn to tell the difference between A. and B., or you may loose sleep thinking it is A. when it actually is B.
It does not make sense to be 'disturbed' if it is Not an oscillation.
6. Sometimes, what we do not know, just might not hurt our listening sessions.
"Enjoy the Music"
Last edited:
To sum up what I read here my amp seems fine. Unfortunately the DAC has its ringing. To really test what is going on with my amp I would have to invest in a proper signal generator. The wave overshoot we see when connecting to my speakers might be from the speakers is what I read here.
I might also try a Zobel network on my speaker just for the sake of seeing what that would do to the wave function.
One more thing: I want to learn more about amp design and the stuff we talk about here, ideally from a source like a book. Does anyone have suggestions?
I might also try a Zobel network on my speaker just for the sake of seeing what that would do to the wave function.
One more thing: I want to learn more about amp design and the stuff we talk about here, ideally from a source like a book. Does anyone have suggestions?
The problem - your expectation is wrong.
From my my experience of this kind of ringing, it may spice up the highs a little, and that can be a good or bad thing depending on your speakers
There are books on amplifiers, such as those by Doug Self and or Bob Cordell. However, they both presume you have enough basic electronics to start at a certain level. Don't know how much you know at this point.
Also a must-have reference book is, 'The Art of Electronics.' Again, depending how how much basics you have at this point it may or may not be easy to understand.
Thing is, electronics is a huge subject area. Just audio is a huge subject area. You have to find the right starting material that fits were you are now, then take it from there.
You might try going to: https://sound-au.com/articles/index.htm ...then scroll down the section with beginner articles and see if you already know all that stuff, or if it might be a good place to start.
That said, I don't necessarily agree with Rod Elliot on everything. IMHO he is a little outside is area of expertise with some of what is on his philosophy page.
Last edited:
The ringing is also a result of limiting the bandwidth of the square wave to fs/2 before the square wave is sampled. There is a low-pass filter before the ADC in addition to the one after the DAC.
In the case of square waves generated by software, the software omits harmonics above fs/2. If you plot the data values on the disc, you will see that the "square wave" is already not square.
Ed
In the case of square waves generated by software, the software omits harmonics above fs/2. If you plot the data values on the disc, you will see that the "square wave" is already not square.
Ed
Yarach,
Your Post # 23 has a 1kHz square wave, picture #2.
The "ringing" on that is the digital filtering effect inside the signal source electronics.
Your amplifier schematic shows that there is No digital filter in your amplifier.
So, it will not distort a proper accurate 1kHz square wave to look like picture # 2.
Your Post # 23 has a 1kHz square wave, picture #2.
The "ringing" on that is the digital filtering effect inside the signal source electronics.
Your amplifier schematic shows that there is No digital filter in your amplifier.
So, it will not distort a proper accurate 1kHz square wave to look like picture # 2.
EdGr,
I have a test CD. It has a 100Hz square wave.
All my CD players are digitally filtered, so they all show pre-transition ringing, then smaller ringing all the way to the other pre-transition ringing.
Since the CD has a 44.1kHz sample rate, the ringing is at 22.05kHz (yes, as you said, sampling rate/2).
CD players have DACs.
The digital filter(s) in the CD players are before the CD DAC.
CD recorders have ADCs.
A recording device, like a CD recorder, must have an Analog Low Pass Filter before the ADC (Analog to Digital Convertor).
That analog low pass filter cuts off signal frequencies that are higher than the 44.1k/2 rate. It prevents Aliasing of any/all excessively high frequency signal(s) which would fold back into the audible range of frequencies (adding bad sound into the recording).
A single pole 6dB/octave analog filter does not cause any 'ringing'.
Higher order analog filters often cause 'ringing'
The 'ringing' that is being talked to here, is not an oscillation.
Examples:
Many Fires require Fuel, Heat, and Oxygen. Fire!
Many Oscillations require Gain, Phase delay or advance, and feedback. Oscillation!
That is not ringing, that is oscillation.
Amplifiers have gain, phase delay, and feedback. If those numbers are bad, we can have one of two things:
1. An Oscillator instead of an amplifier, once oscillation starts, it never ends, until the power is turned off.
2. What some call "ringing", it is a transient oscillation, it is not continuous. Most often this is referred to as amplifier Instability.
Neither of these two is the same thing as the "ringing" of a digital filter.
I have a test CD. It has a 100Hz square wave.
All my CD players are digitally filtered, so they all show pre-transition ringing, then smaller ringing all the way to the other pre-transition ringing.
Since the CD has a 44.1kHz sample rate, the ringing is at 22.05kHz (yes, as you said, sampling rate/2).
CD players have DACs.
The digital filter(s) in the CD players are before the CD DAC.
CD recorders have ADCs.
A recording device, like a CD recorder, must have an Analog Low Pass Filter before the ADC (Analog to Digital Convertor).
That analog low pass filter cuts off signal frequencies that are higher than the 44.1k/2 rate. It prevents Aliasing of any/all excessively high frequency signal(s) which would fold back into the audible range of frequencies (adding bad sound into the recording).
A single pole 6dB/octave analog filter does not cause any 'ringing'.
Higher order analog filters often cause 'ringing'
The 'ringing' that is being talked to here, is not an oscillation.
Examples:
Many Fires require Fuel, Heat, and Oxygen. Fire!
Many Oscillations require Gain, Phase delay or advance, and feedback. Oscillation!
That is not ringing, that is oscillation.
Amplifiers have gain, phase delay, and feedback. If those numbers are bad, we can have one of two things:
1. An Oscillator instead of an amplifier, once oscillation starts, it never ends, until the power is turned off.
2. What some call "ringing", it is a transient oscillation, it is not continuous. Most often this is referred to as amplifier Instability.
Neither of these two is the same thing as the "ringing" of a digital filter.
Last edited:
6A3sUMMER - We are saying the same thing. The neat thing about software is that square-wave harmonics above fs/2 don't need to generated at all. Just add the harmonics below fs/2 ("additive synthesis"). This is the perfect filter.
Ed
Ed
Surprised no "standard speaker test load" (who likes listening to continuous tones..?) has been proposed, for two and three way system emulation.
I mean you could take a straightforward commercial crossover, put resistors where the speakers go and voila! But then "who's" crossover would be regarded as a "standard".
Having such a design that could be referenced repeatedly (sticky?) would certainly advance the state of DIY audio. Otherwise "MY amp and MY speakers sound great - WHY?" is no more better answered than you got lucky - which happens.
When the phase of the current is leading / lagging that of the voltage a few times as frequency goes across the audio range, I've got to believe some amp designs handle that "situation" better than others. I've no idea what one would look for when driving a standard real speaker load, but I'd bet there's something to see, with the analysis tools available in, say, REW.
I mean you could take a straightforward commercial crossover, put resistors where the speakers go and voila! But then "who's" crossover would be regarded as a "standard".
Having such a design that could be referenced repeatedly (sticky?) would certainly advance the state of DIY audio. Otherwise "MY amp and MY speakers sound great - WHY?" is no more better answered than you got lucky - which happens.
When the phase of the current is leading / lagging that of the voltage a few times as frequency goes across the audio range, I've got to believe some amp designs handle that "situation" better than others. I've no idea what one would look for when driving a standard real speaker load, but I'd bet there's something to see, with the analysis tools available in, say, REW.
Maybe its not practical represent all speakers according to one standard model?
Perhaps better than a (non-inductive) resistor - the popular 1st order choice. Unless it's everyone's (who designs amplifiers) "secret sauce", then I'd understand.
But one would think there'd be a legitimate "Can your amp drive this?" thread, showing a topology arrangement of passive elements (presenting an impedance which isnt ridiculous even though as I understand, some speakers are) that could illuminate, by measurement, amplifier shortcomings and where they excel better than a resistor.
There are at least 2 threads in Tubes/Valves that give schematics of speaker load simulators.
They are different in complexity, to do different simulations.
Always start with a standard non inductive load resistor (4, 8, 16, etc. as appropriate for the amplifier's secondary output tap).
If an amplifier does not have good stability, good frequency response, good distortion performance, and moderate damping factor . . .
Then it certainly will not be good with a large number of different speaker models.
A few years ago I used a Rhode & Schwarz Vector Network Analyzer (VNA) to test the impedance and phase angle from 10Hz to 20kHz of lots of speaker models.
Then, I connected the same speakers to an amplifier output, and drove the ampilifier with the VNA signal output, and measured the total system's phase and frequency response, 10Ha to 20kHz, from the amplifier input to loudspeaker terminals.
I also did the same test for a non-inductive 8 Ohm resistor, to test the amplifier phase and frequency response to a resistive load.
Then I compared the two measurements, by looking at the pairs of phase and frequency response curves.
. . . Very revealing.
They are different in complexity, to do different simulations.
Always start with a standard non inductive load resistor (4, 8, 16, etc. as appropriate for the amplifier's secondary output tap).
If an amplifier does not have good stability, good frequency response, good distortion performance, and moderate damping factor . . .
Then it certainly will not be good with a large number of different speaker models.
A few years ago I used a Rhode & Schwarz Vector Network Analyzer (VNA) to test the impedance and phase angle from 10Hz to 20kHz of lots of speaker models.
Then, I connected the same speakers to an amplifier output, and drove the ampilifier with the VNA signal output, and measured the total system's phase and frequency response, 10Ha to 20kHz, from the amplifier input to loudspeaker terminals.
I also did the same test for a non-inductive 8 Ohm resistor, to test the amplifier phase and frequency response to a resistive load.
Then I compared the two measurements, by looking at the pairs of phase and frequency response curves.
. . . Very revealing.
EdGr,
If you want to test the stability and frequency response of an amplifier to signals that are above 22.05kHz, you need a fast rise square wave.
Whether it is digitally generated, or analog, it needs the fast rise at the up and down transistions.
The approximate bandwidth of a well behaved amplifier is:
0.35/rise time, or for the other direction . . . 0.35/fall time; where rise time and fall time are defined as the time between 10% to the 90%, or 90% to 10% voltage points.
Some amplifiers do not have identical rise time and fall time.
Example:
An amplifier has a rise time of 6us, and a fall time of 7us.
0.35/6us = 58.3 kHz
0.35/7us = 50.0 kHz
The amplifier has about 54kHz bandwidth (-3dB).
To verify this, you need a square wave generator with a rise time / fall time of 0.5us or better.
Of course if you have a sine wave generator that goes from 10Hz to 100kHz (with equal amplitude),
you can do spot frequency response tests to find the -3dB lower and upper bandwidths of your amplifier.
Lots more work than the quick fast rise square wave test.
For some of us, we will test both ways.
Trust but Verify.
All my life I have been waiting to meet the "perfect" filter.
Analog Filter days . . . refer to the definitive filter bible, written by a Russian named Zverev, in a book published by Westinghouse.
8 x 11 inches rectangular, and 3 inches thick.
I used that book and another engineer's software that was written for the Intel 8087 Co-processor, to design the Tektronix 2782 and 2784 helical filters, -3dB bandwidth of 10MHz centered at 525MHz.
Digital Filters were invented, and they had some of the same tradeoffs as Analog filters; but they also added some new tradeoffs.
Some applications and systems require both analog and digital filters for optimum performance.
Some of you will filter out most of what I said above. :^)
I often filter out things myself, especially to long disertations and/or ramblings like I wrote above.
If you want to test the stability and frequency response of an amplifier to signals that are above 22.05kHz, you need a fast rise square wave.
Whether it is digitally generated, or analog, it needs the fast rise at the up and down transistions.
The approximate bandwidth of a well behaved amplifier is:
0.35/rise time, or for the other direction . . . 0.35/fall time; where rise time and fall time are defined as the time between 10% to the 90%, or 90% to 10% voltage points.
Some amplifiers do not have identical rise time and fall time.
Example:
An amplifier has a rise time of 6us, and a fall time of 7us.
0.35/6us = 58.3 kHz
0.35/7us = 50.0 kHz
The amplifier has about 54kHz bandwidth (-3dB).
To verify this, you need a square wave generator with a rise time / fall time of 0.5us or better.
Of course if you have a sine wave generator that goes from 10Hz to 100kHz (with equal amplitude),
you can do spot frequency response tests to find the -3dB lower and upper bandwidths of your amplifier.
Lots more work than the quick fast rise square wave test.
For some of us, we will test both ways.
Trust but Verify.
All my life I have been waiting to meet the "perfect" filter.
Analog Filter days . . . refer to the definitive filter bible, written by a Russian named Zverev, in a book published by Westinghouse.
8 x 11 inches rectangular, and 3 inches thick.
I used that book and another engineer's software that was written for the Intel 8087 Co-processor, to design the Tektronix 2782 and 2784 helical filters, -3dB bandwidth of 10MHz centered at 525MHz.
Digital Filters were invented, and they had some of the same tradeoffs as Analog filters; but they also added some new tradeoffs.
Some applications and systems require both analog and digital filters for optimum performance.
Some of you will filter out most of what I said above. :^)
I often filter out things myself, especially to long disertations and/or ramblings like I wrote above.
Last edited:
I am probably one of the few hobbyists who have designed and built signal generators, both the hardware kind and the software kind. 🙂
Ed
Ed
THat is a great source to start! I already have knowledge about how electornics work in the sense that I know energy transmutationand it's workings. I know what resistiance is and why the excitation of atoms causes heat and in extremer cases light emission and that kind of stuff. THerefore I know why a tube "amplifies" while in reality we only modulate the ammount of excitation of another loop of atoms by using a lesser excited modulated loop as the "control valve".
What I mostly lack is what causes oscilaltons, distortion, phase shifting and that kind of sstuff.
I found an article about sqaure waves and what we are seeing thanks to Markw4.
It exactly explain as why we are seeing said ringing in the squarewave from the DAC.
https://sound-au.com/articles/squarewave.htm#s1
It exactly explain as why we are seeing said ringing in the squarewave from the DAC.
https://sound-au.com/articles/squarewave.htm#s1
Yes. I agree with your description @wg_ski .
This is also my prefered shape of square waves for a single-ended tube amplifier. It is already a very good performance for such kind of amp, very difficult to rival at those frequencies in SE, given that the FB loop is circa 5.5dB only. This usually gives a fatigue-less, deep and crisp sound...
The quality of the output transformer proves to be very fine finally - these are unknown Chino amp production OEM, bought 2nd hand on a craiglist for almost nothing "because it is Chinese", from a Guy who replaced them by Hammond Standard Series 125GSE (see below)...
Effectively, on my schematic, the presence of the 1nF 500V MicaAg capacitors in the FB loop and accross screen and plate taps of the OPT corrects the treble region against a little leading edge spike and attenuates ringing, while maintaining a correct bandwidth though.
And that's true, it's a little flaw from the amp's Chino OPT, which can't be corrected further without a treble bandwidth increased sacrifice, the square waves from my Siglent SDG-810 generator being absolutely perfects.
But honestly, the result could have been much worse ! 😉
T
Constructing a physical model of a typical 2-way bookshelf speaker was the subject of this old Stereophile magazine article:
https://www.stereophile.com/reference/60
Using this physical model is safer than testing with a real speaker, and easier on the ears too. 🙂
For simulation purposes, I created a SPICE model of this "speaker" from the schematic on page 2 of the article. This lets me run SPICE simulations and then bench test an amplifier circuit, using a physical speaker model that corresponds to the SPICE model. Is this what you were thinking?