I studied Engineering in the late 60s early 70s when a computer was an air conditioned roomful of stuff (we had an IBM1620 at one University and a 360 on the other) and measuring instruments, think Hewlett Packard, Tektronix, Rhode & Schwarz, General Radio, etc. were impossibly expensive and generally available only at University Labs (student privileges I used and abused 😉 )
So how did we optimize designs lacking simulation (there was available only a crude early version of Spice, output was bedsheet size meters of zig zag folding line printer paper, oldtimers will remember that) and easily available metering?
Undisputed King of the Jungle was graphic design. (that´s what datasheet curves are for 😉 )
There was a method, I forgot details long ago for lack of use, which by measuring line to line distances (which were not uniform) along chosen loadline you could do a little Math and calculate a distortion percentage.
As in: "this preamp stage will have 5% distortion when putting out 10V RMS" or something like that.
Does anybody remember that?
jan didden?
(although you are probably too young for that).
So how did we optimize designs lacking simulation (there was available only a crude early version of Spice, output was bedsheet size meters of zig zag folding line printer paper, oldtimers will remember that) and easily available metering?
Undisputed King of the Jungle was graphic design. (that´s what datasheet curves are for 😉 )
There was a method, I forgot details long ago for lack of use, which by measuring line to line distances (which were not uniform) along chosen loadline you could do a little Math and calculate a distortion percentage.
As in: "this preamp stage will have 5% distortion when putting out 10V RMS" or something like that.
Does anybody remember that?
jan didden?
(although you are probably too young for that).
Well, I spent a whole summer internship writing code to do waterfall plots...
Today, a cheap $4 calculator does it better than the big sheets I was putting out of the plotter to the oohs! and aaahs! of the entire Fiber Physics Dept.
Things do change.
I don't want to measure another type of distortion. I want to a more comprehensive measurement of distortion. We have the ability to do that now, cheaply with programmable instruments... just press the button and off and it goes.
Who cares about the OP? He's long gone.
So, the guy took me to the equipment racks in the lab that controlled the nuclear accelerator and the detection units. This was '80. My senior thesis internship.
He points out to a PDP-9... and pulling out a 19" rack says" This is the 4K of core memory", pointing out the rows and rows of little ferrite cores. "We got this in '68, still works fine".
Then he moves over to a different chassis on a rack. "This here is a PDP-9A, we got it 6 months after the other one, it has 64K of memory, it's in chips". So he didn't bother to slide it out from the rack. "Still works fine".
Then he shows me my desk and points out a Commodore PC.... with a 6502 ( or maybe it was a Z80, I forget) and 64KB. It had more power than the other two, two or three magnitudes more power!
They were measuring nuclear resonance with those machines!
I still think no one has improved on FORTRAN to do numerical analysis. I love its simple ability to specify the precision of a value. Sure, Matlab is fancy with all of its libraries... but, nothing like Hollerith...
Last edited:
TonyEE,
You said: "I want to a more comprehensive measurement of distortion. We have the ability to do that now, cheaply with programmable instruments... just press the button and off and it goes."
Tell us the details of your more comprehensive measurement of distortion, please.
Which equipment does that?
What are the parameters, units, etc. of such a measurement?
Thanks!
You said: "I want to a more comprehensive measurement of distortion. We have the ability to do that now, cheaply with programmable instruments... just press the button and off and it goes."
Tell us the details of your more comprehensive measurement of distortion, please.
Which equipment does that?
What are the parameters, units, etc. of such a measurement?
Thanks!
Bimo,
77V peak at 20kHz from a solid state amplifier into an 8 Ohm load resistor at only 0.000107% distortion is very impressive.
But put 77V peak at 20kHz into a tweeter, that would almost certainly destroy it, unless it is a very short burst.
The 2nd harmonic is at 40kHz, bats can hear that, I can not hear 20kHz, and I certainly can not hear 40kHz.
60kHz 3rd harmonic anybody?
77V peak at 20kHz from a solid state amplifier into an 8 Ohm load resistor at only 0.000107% distortion is very impressive.
But put 77V peak at 20kHz into a tweeter, that would almost certainly destroy it, unless it is a very short burst.
The 2nd harmonic is at 40kHz, bats can hear that, I can not hear 20kHz, and I certainly can not hear 40kHz.
60kHz 3rd harmonic anybody?
There is cool experiment related to this. Take a tweeter / driver in your hand, feed high frequencies to it, put it behind your head. Its weird how the high frequencies are perceived more there for some reason. Noticed this when I was outside at silent night time and weird sound came somewhere and it was very hard to locate. Turned out to be a fuse box / electronics box on a shed wall which emitted some buzzing sound. Further away I could hear it only when it was behind me and sound disappeared when facing at it, weird stuff.
I mean, even if hearing tapers of the sound highs could still be perceived somehow, I guess. But, sound emitted by loudspeakers and coming from behind has already travelled quite some time and got attenuated in the way so aubility of >20kHz coming from loudspeakers is probably quite low.
I mean, even if hearing tapers of the sound highs could still be perceived somehow, I guess. But, sound emitted by loudspeakers and coming from behind has already travelled quite some time and got attenuated in the way so aubility of >20kHz coming from loudspeakers is probably quite low.
tmuikku,
Damping factor?
Take an amplifier of 0.001 Ohms output impedance, speaker cable of 0.001 Ohms DCR, a high pass woofer inductor of 0.001 Ohms DCR.
That is a total of 0.003 Ohms.
Take a typical "8" Ohm woofer that has 6.000 Ohms voice coil DCR.
The total is 6.003 Ohms.
Super cooled voice coils anybody? . . . Not for me.
Damping factor?
Take an amplifier of 0.001 Ohms output impedance, speaker cable of 0.001 Ohms DCR, a high pass woofer inductor of 0.001 Ohms DCR.
That is a total of 0.003 Ohms.
Take a typical "8" Ohm woofer that has 6.000 Ohms voice coil DCR.
The total is 6.003 Ohms.
Super cooled voice coils anybody? . . . Not for me.
Definitely not too young for that!
I do remember it, but not sure where. Best bet is Norman Crowhurst in his series, but I loaned them out so can't look it up.
Now that I think about it, definitely Crowhurst.
Jan
Thanks.
Way back then, "good abbundant" tech literature here, including Tubes of course, was by Philips Eindhoven, think late 40´s to mid 60´s stuff, doubly so because we had FAPESA, our own Philips plant manufacturing everything from lightbulbs (the original market they cornered) to Tubes and Transistors, think our own Miniwatt or Valvo or Mullard branch.
So they were "The (Audio) Bible" for us (with good reason) and that method was often mentioned there.
But I lost track and accessibility (although maybe those books are still at the University Library, who knows?),I asked here because somebody else might remember the graphical method to calculate distortion.
Thanks for answering.
PS: Norman Crowhurst was known here but much later.
PS2:all that Industrial goodness was dismantled and destroyed starting in 1976, when a Banks backed military coup browbeat us into "The (financial) Market" and "Globalization".
We paid 30000 dead for the "privilege" of losing 90% of our Industry.
Only a fistful of greybeard diehards like yours truly continue the fight, and not much future left.
Oh well.
Way back then, "good abbundant" tech literature here, including Tubes of course, was by Philips Eindhoven, think late 40´s to mid 60´s stuff, doubly so because we had FAPESA, our own Philips plant manufacturing everything from lightbulbs (the original market they cornered) to Tubes and Transistors, think our own Miniwatt or Valvo or Mullard branch.
So they were "The (Audio) Bible" for us (with good reason) and that method was often mentioned there.
But I lost track and accessibility (although maybe those books are still at the University Library, who knows?),I asked here because somebody else might remember the graphical method to calculate distortion.
Thanks for answering.
PS: Norman Crowhurst was known here but much later.
PS2:all that Industrial goodness was dismantled and destroyed starting in 1976, when a Banks backed military coup browbeat us into "The (financial) Market" and "Globalization".
We paid 30000 dead for the "privilege" of losing 90% of our Industry.
Only a fistful of greybeard diehards like yours truly continue the fight, and not much future left.
Oh well.
At that time (mid-60's) I worked at Philips in Eindhoven, how's that for dating me.
On the distortion calculation:
1 - work out the pos output from a selected neg grid value. Say with 1V neg grid drive the anode voltage rises A volts. Do the same with pos grid drive; say with 1V pos drive, Va drops B volts.
Then distortion D = 0.5*(|A-B|)/(A+B).
Jan
On the distortion calculation:
1 - work out the pos output from a selected neg grid value. Say with 1V neg grid drive the anode voltage rises A volts. Do the same with pos grid drive; say with 1V pos drive, Va drops B volts.
Then distortion D = 0.5*(|A-B|)/(A+B).
Jan
IMHO, I've explained it quite well... but I will repeat it, again!
Sweep the input fundamental frequency from 20Hz to 20Khz. At each point, do a frequency scan from 20Hz to 20Khz for the harmonic distortion components.
The current standard measurement does this with a 1Khz input signal and it also does a sweep from 20Hz to 20Khz to find the TOTAL harmonic distortion... it does not produce a frequency plot during the sweep... the "waterfall" display.
tonyEE,
Have you ever seen and used the Tektronix real time spectrum analyzers that covered from 10Hz to way above audio frequencies?
They had a waterfall display.
Just run a very slow sweep on the signal generator, at the low frequency portion of the 20Hz to 20kHz (you can not sweep quickly when using a 1Hz or 3 Hz, or 10Hz resolution bandwidth filter, the amplitude will be very much attenuated, and also smeared.
At the higher portion of the 20Hz to 20kHz frequency sweep, you can adjust the real time spectrum analyzer resolution bandwidth to be much wider, so the sweep speed can be much faster.
The key to maximum sweep speed is the rise time of the required resolution bandwidth that gives enough resolution between the fundamental and the harmonics.
I recommend you get your hands on one of these older Tektronix products.
Good Luck getting your hands on one of those, nobody that has these obsolete products are going to sell them, they are too valuable for such measurments.
Too little, too late.
In case you are wondering, I ran these analyzers, and gave support around the world to those who had them.
Have you ever seen and used the Tektronix real time spectrum analyzers that covered from 10Hz to way above audio frequencies?
They had a waterfall display.
Just run a very slow sweep on the signal generator, at the low frequency portion of the 20Hz to 20kHz (you can not sweep quickly when using a 1Hz or 3 Hz, or 10Hz resolution bandwidth filter, the amplitude will be very much attenuated, and also smeared.
At the higher portion of the 20Hz to 20kHz frequency sweep, you can adjust the real time spectrum analyzer resolution bandwidth to be much wider, so the sweep speed can be much faster.
The key to maximum sweep speed is the rise time of the required resolution bandwidth that gives enough resolution between the fundamental and the harmonics.
I recommend you get your hands on one of these older Tektronix products.
Good Luck getting your hands on one of those, nobody that has these obsolete products are going to sell them, they are too valuable for such measurments.
Too little, too late.
In case you are wondering, I ran these analyzers, and gave support around the world to those who had them.
LOL.... do you remember those plotting CRTs that Tektroniks used to have back in the late 70s? That's the one I programmed, with FORTRAN, to create waterfull plots. And then, I hooked it up to one of the big paper plotters.
They also made awesome scopes, until HP came out with the 1980B programmable scope and killed everybody, including all of those analog signal analyzers...
Tektronix produced the T4002 computer terminal. It had a Flood Gun Island Storage CRT, that had a non storage Compose Line at the bottom of the CRT phosphors.
It had a multifunction unit on the front, with tri-state color lighted clear buttons, and an alpha numeric keyboard. The complete assembly was called the "keyboard".
During my 1969 student summer hire, during the engineering phase of the T4002, my engineer boss and I designed the keyboard tester.
It was quite an odd looking kludge, but it worked.
Later, I found out that it was used in production to test all of the production keyboards.
It had a multifunction unit on the front, with tri-state color lighted clear buttons, and an alpha numeric keyboard. The complete assembly was called the "keyboard".
During my 1969 student summer hire, during the engineering phase of the T4002, my engineer boss and I designed the keyboard tester.
It was quite an odd looking kludge, but it worked.
Later, I found out that it was used in production to test all of the production keyboards.
Yeah but this is what is sold in HiFI marketing isn't it?
While it might be true that high damping factor is related / byproduct of reduced distortion in amplifier and would make valid sales argument as such for the amplifier the marketing material leaves out that side effect of this is that driver / loudspeaker is now allowed to manifest its distortion fully and system distortion could even increase compared to some other amplifier.
All that matters is what becomes audible, right? Its left for the consumer to understand whats going on in their systems and how to reduce system distortion, how amplifier affects loudspeaker and how loudspeaker affects the amplifier, they are both connected through same circuit. Purifi paper is kind of eye opener on this for the driver distortion: if driver load impedance is not taken care of (in crossover) to reduce driver distortion then it would be better to have "worse" amplifier, cabling, which would worsen their performance some but increase loudspeaker performance, reduce distortion made by loudspeaker by increasing impedance in the circuit, which would probably net less distortion emitted by the system.
Well, I've got no idea about magnitude of these various distortion mechanisms but since THD reading is bigger for loudspeakers than for amplifiers I would assume reducing distortion in loudspeaker would be higher priority than reducing amplifier distortion. Although, Geddes has showed THD is meaningless number as it does not correlate with perceived sound quality. Well, anyway, something to be considered and tested in our own systems.
edit. if taking the topic as philosophical question and try to relate that to our own systems, then it would make sense to figure out what in the THD umbrella is most offensive sounding and try eliminate that. Iterate until there is no meaningful difference anymore. I start by checking out what the Purifi paper reads and continue from there 🙂 Yep, I don't know what my system distortion performance is and which aspects of it is more audible than something else, yet.
Last edited:
Things have changed incredibly and keep accelerating.
When I started on this career, we'd have ten hardware guys for each "software" (*) guy, and we ALL knew how to read schematics. The hardware guys wore clip on ties, the SW guys, rebels as we are, wore shirts with NO ties. True rebels, huh? ;-) But we all wore pocket protectors.
In my latest job, we're using a Xylinx FPGA assembly that contains an SoC with four A7s and two R5s, with the busses, DMA, cache, etc.... AND a real, programmable FPGA for doing the IO devices. All of them.
About five years ago, this would have required a very large FPGA ( to do the cores ) or two chips (SOC and IO FPGA).
I went to the lab and took a look at the hardware... not only is the board tiny, but the Xylinx is on a mezzanine card so that it could be upgraded in the future. The "mother board" only contains the IO connectors and some power supply filtering.
I recall that 20 years ago, an SBC with quad PPC board and quad PCI mezzanines was 12 inches long and six inches high. We were doing SMP and we thought we were the most awesome (I was getting 98% bus utilization over the PCI busses).
Then in the last 10 years, I've found myself in SoC design teams with as many as 28 ARM cores, with DMAs, switch matrices, As, Rs, Ms, etc, etc... and those go... yep.. INTO YOUR PHONE!
Not to mention the 32TB PCI-e SSD prototypes I had on my desk three years ago. They are up to 64TB now.
Then, I take a look at my recently built, and superb sounding, Aleph 2s... Oh well, no one said my Samsung's DAC and headphone amps were meant to sound really good, that's why I use an external USB DAC/headphone device with my Samsung.
BTW, that little USB DAC is also a marvel of design though.
(*) we call it firmware nowadays.
Last edited:
D
Deleted member 375592
The loudspeaker non-linear distortion measurements - this is exactly what I've been doing in the last 3 months. Oversimplifying, I am using (a version of ) Least Squares to calculate the linear part of the measured signal, and then take the difference between measured and linear (residual). If anyone is interested, please read the attached .pdf (5400 words and ~100 figures), and I will be very grateful for your comments.
I am quite sure THD is nearly useless... but I do not know a better way to express loudspeaker quality in one number.
I am quite sure THD is nearly useless... but I do not know a better way to express loudspeaker quality in one number.
What exactly is LTI distortion?
I guess the cone driver (circular membrane being pushed pistonicall) having higher 3rd order harmonics could be because of constructive interference at frequencies where the wavelength is a multiple of the radiating cone radius (minus the center cap). I suppose this is a type of resonance because, as you note, there is no air resistance at the 3rd (and higher harmonics) since part of the cone is moving in and other parts are moving out for a total zero pressure differential within the cavity of the cabinet.
How about cones in an open baffle? On second guess, they might not be any different for the harmonics.
A planar is constructed differently so it may not have such interference, huh?
I need to read this a 3rd time.... too late now.
Thanks.