What do you look for in the distortion spectrum?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hello all.

In my different threads regarding my designs, I have read "well I like this way better, because there's less 7th harmonics (or somesuch)".

This is not a thread to debate Distortion vs. Ear, I simply want to know what you prefer to see in the distortion spectrum so I have some other ideas to aim for when I get tired of simply trying for lowest THD. I expect differing views and encourage them.

Thank you,
- keantoken
 
i usually look for lowest distortion in a spectrum display. some forms of distortion cause even harmonics and some cause odd harmonics, so if you have high odd harmonics, the causes are different from even harmonics, and the mechanisms that cause one type or the other is what i concentrate on correcting. it is also a good idea to check your signal source first to know what it looks like, and record the source's harmonic content by harmonic and relative amplitude. that way, you aren't chasing "ghosts" when you are "tweaking" your design. a spectrum analyzer will also show the hum and noise content as the amplifier is operating with signal. that's not something a noisemeter can do.

things like "high 7th harmonic" don't really happen. usually if one odd harmonic is high, so are the other odd harmonics, same goes for the even harmonics.
 
Hmm.

Surely we can hook the spectrum analyzer with a diff. amp and only look at the difference between the input and output.

BTW, as a side note for anyone interested, I ordered a 3A9 diff. amp plugin and a 3B4 timebase for my vintage Tek 561B, along with a Fluke 8800A, an S/D datapulse Versatester and an old HP signal generator.

So if and after I get done calibrating, I'll have a far more versatile test rig... Now I need to work on a power supply. O_O

Anyways, as I said, all opinions are welcome.

- keantoken
 
keantoken said:
Hmm.

Surely we can hook the spectrum analyzer with a diff. amp and only look at the difference between the input and output.

actually that's what a distortion analyzer does.
keantoken said:


BTW, as a side note for anyone interested, I ordered a 3A9 diff. amp plugin and a 3B4 timebase for my vintage Tek 561B, along with a Fluke 8800A, an S/D datapulse Versatester and an old HP signal generator.

So if and after I get done calibrating, I'll have a far more versatile test rig... Now I need to work on a power supply. O_O

Anyways, as I said, all opinions are welcome.

- keantoken
i used to have a dual diff amp plugin for my 541. unfortunately it got damaged (bent switch shafts) during a move, so all i was left with is the 2 channel module. if you need to calibrate the horizontal timebase, let me know, there are a few ways of doing it correctly that don't require expensive timebases. the ramp linearity i think is the difficult one since there are a few adjustments that interact. (edit...) you live in TX, so you should have a stable WWV signal to zero-beat a 10Mhz oscillator against, which if i remember correctly should give you a 10Mhz timebase that's accurate enough for calibrating that vintage scope. in the Army, we had a timebase that was sync'ed to WWVB and was accurate to 1 part in 10E-11 (10E-12 on really good days), but for audio work that's not really necessary.

as far as the vertical alignment, that's usually simpler, requiring accurate voltage sources.
 
My rule about non-linear distortions:
The closer to zero power, the closer to zero non-linear distortions, the lower is their order. However, it does not mean that on higher currents they are allowed to be huge with wide specter, because they cause intermodulations.
Hence, the choice of topologies: less of active elements working linearly, plus feedback to sweep the garbage under the rug on higher powers.

And, of course, thinking of distortions I keep in mind everything I have the control of: from mics to speakers. The system made of optimal subsystems can never be optimal -- the main axiom I follow.
 
unclejed613 said:


actually that's what a distortion analyzer does.

i used to have a dual diff amp plugin for my 541. unfortunately it got damaged (bent switch shafts) during a move, so all i was left with is the 2 channel module. if you need to calibrate the horizontal timebase, let me know, there are a few ways of doing it correctly that don't require expensive timebases. the ramp linearity i think is the difficult one since there are a few adjustments that interact. (edit...) you live in TX, so you should have a stable WWV signal to zero-beat a 10Mhz oscillator against, which if i remember correctly should give you a 10Mhz timebase that's accurate enough for calibrating that vintage scope. in the Army, we had a timebase that was sync'ed to WWVB and was accurate to 1 part in 10E-11 (10E-12 on really good days), but for audio work that's not really necessary.

as far as the vertical alignment, that's usually simpler, requiring accurate voltage sources.

Hmm. I'll have to make a crystal radio or something so I can "tune in". As far as voltage sources, I might have a problem there. I think I have a TL431 sitting on the floor somewhere :cannotbe:. I can use a pot to make it into a shunt reg. I'll have to see if that Fluke is accurate enough to adjust it. I can have my own poor-man's DC reference Shunt reg... I wonder if I can hook it up to the scope and see the noise? The 3A9 has 10uV/div sensitivity so it's possible...

And I've always been fascinated with the idea of making miniature ovens to keep important components temperature regulated. I would really love to make myself a bunch of calibration devices and references...

Anyways.

Wavebourn said:
My rule about non-linear distortions:
The closer to zero power, the closer to zero non-linear distortions, the lower is their order. However, it does not mean that on higher currents they are allowed to be huge with wide specter, because they cause intermodulations.
Hence, the choice of topologies: less of active elements working linearly, plus feedback to sweep the garbage under the rug on higher powers.

And, of course, thinking of distortions I keep in mind everything I have the control of: from mics to speakers. The system made of optimal subsystems can never be optimal -- the main axiom I follow.

Interesting axiom. I will think about it. Theoretically, a perfect system must act as one. As far as speakers and Mics, I have pretty much nothing of value, except perhaps my Boston Acoustics A40 enclosures, which don't have the original speakers in them. This is one of the reasons I want to hurry up and make myself an amp. But until I get some kind of soldering rig, I think I'm stuck.

- keantoken
 
And if there was, then phase wouldn't be flat.
And if there was, then dispersion wouldn't be even.
And if there was, something else surely still be wrong...

Its mostly about musical errors vs non-musical errors.
Here, THD is usually your friend. IMD usually your enemy.
Assume things will occasionally go outside the envelope,
and the loudspeaker wasn't put in your way to help.

You got to look at the speaker and output stage as a
unified system that have to fail elegantly together, and
fake good like it was intentional. If they are trippin over
each other's toes, it can be a trainwreck.
 
kenpeter said:
it can be a trainwreck.

believe it or not, there's a guitar amp by that name....... :rofl:



actually for the WWV thing, you should use a good stable receiver, and a good antenna (preferably a dipole 15 meters long with the antenna oriented perpendicular to a line towards Ft Collins CO). you want to zero beat the oscillator while the station is giving stable reception with very little or no fading, and during the portion of the hour that has no test tones ( HH:45 to HH:00 if i remember correctly). a 10.00000Mhz cpu clock oscillator should be pretty close, and i think you can fine tune it by adjusting the 5.00V power supply voltage between 4.8 and 5.2V (adjust VERY slowly and allow the oscillator to thermally stabilize after each adjustment of voltage. you might think about using a 10-turn pot to adjust the 5V regulator. if you want to "ovenize" the oscillator for thermal stability, you can pack it in antistatic styrofoam or wrap it up in antistatic bubble wrap. the oscillator output should be terminated with a resistor (look in the data sheet for the correct value and then capacitively coupled to the receiver through a "gimmick" capacitor (two insulated wires twisted together). you may or may not get a tone from the oscillator interfering with the WWV signal (you may just hear the WWV signal begin fading in and out which is a subsonic difference in frequency) as you get close to completely zero beating the signal, the WWV signal will fade slower and slower, until you get to the point where either the WWV signal stops fading altogether, or it fades out and stays faded out. this is zero beat, and now the oscillator is operating at almost exactly the frequency of WWV, WWV's received frequency stability is something like 10E-5 to 10E-8, depending on reception conditions. zero beating an oscillator should be good for 10E-2 to 10E-3, which is plenty for a vintage oscope
 
i was looking at the cal procedure for the 561B. horizontal accuracy is supposed to be +/- 0.5 minor divisions on the screen. that's 1% or 10E-2. so the WWV/XTAL oscillator method is accurate enough, even if the oscillator is 100khz off. if you get the oscillator to within 10hz, then your oscillator has an accuracy within 10E-7 and that would also allow for ionospheric doppler shifts and fading.
 
unclejed613 said:

actually for the WWV thing, you should use a good stable receiver, and a good antenna (preferably a dipole 15 meters long with the antenna oriented perpendicular to a line towards Ft Collins CO). you want to zero beat the oscillator while the station is giving stable reception with very little or no fading, and during the portion of the hour that has no test tones ( HH:45 to HH:00 if i remember correctly). a 10.00000Mhz cpu clock oscillator should be pretty close, and i think you can fine tune it by adjusting the 5.00V power supply voltage between 4.8 and 5.2V (adjust VERY slowly and allow the oscillator to thermally stabilize after each adjustment of voltage. you might think about using a 10-turn pot to adjust the 5V regulator. if you want to "ovenize" the oscillator for thermal stability, you can pack it in antistatic styrofoam or wrap it up in antistatic bubble wrap. the oscillator output should be terminated with a resistor (look in the data sheet for the correct value and then capacitively coupled to the receiver through a "gimmick" capacitor (two insulated wires twisted together). you may or may not get a tone from the oscillator interfering with the WWV signal (you may just hear the WWV signal begin fading in and out which is a subsonic difference in frequency) as you get close to completely zero beating the signal, the WWV signal will fade slower and slower, until you get to the point where either the WWV signal stops fading altogether, or it fades out and stays faded out. this is zero beat, and now the oscillator is operating at almost exactly the frequency of WWV, WWV's received frequency stability is something like 10E-5 to 10E-8, depending on reception conditions. zero beating an oscillator should be good for 10E-2 to 10E-3, which is plenty for a vintage oscope

Thanks for the info. Whenever I get my hand on a good oscillator I will do this.

- keantoken
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.