0.2 Gnat Farts... I like your perspective...
The electrostatic shield (aka belly band) should, in theory, catch any high-frequency noise present on the mains. I doubt it makes a huge difference, but on the other hand, it's not much of a cost adder, so why not? My power supply board will have a terminal where you can connect the shield, if your transformer has one.
~Tom
Hi peufeu, I think I'm about 75% in agreement with you.
However, some people prefer to use both soft-recovery diodes AND snubbers. Thanks to the soft-recovery diodes, their live circuit has very, very little ringing (or none), so there's nothing to see on a scope when you attempt to dial-in the snubber. Your procedure fails for these people. The obvious workaround is to make two passes using two sets of diodes; first, hard-recovery diodes ("bad diodes") that ring like crazy, letting you dial-in the snubber resistance. Second, unsolder the bad diodes and solder in the final, soft-recovery, good diodes. But this is clumsy.
Another slight advantage of dialling in the snubber resistor NOT on the live circuit is, you don't need to be anywhere near the lethal AC mains. This is comforting for people who are (rightly) fearful of working near live AC circuits. They can get perfectly useful and perfectly correct bellringer-derived snubbers with just a 9V "transistor radio" battery, which is far less frightening because it is far less dangerous.
To use the highpass idea, I suggest it's best to start with a calculation of the attenuation you need at the mains frequency, 50 Hertz. If the secondary is V volts peak, and the ringing amplitude is (let's just say) 0.4 volts peak to peak, then I suggest that you want an attenuation of at least (V/0.4) at the mains frequency, 50 Hz.
For example, the Modulus recommended transformer (22VAC RMS) has V=31 volts peak, so I suggest an attenuation of at least (31/0.4) = 78X at 50Hz. You could round that up to 100X, meaning that the highpass filter should be designed to have a corner frequency of 5 kHz (or greater). This would pass tomchr's 67 kHz oscillatory ringing waveform quite nicely.
Build the highpass with a large R (>50K) and a small C. You want to be sure that the highpass R is so big that it does not contribute any meaningful amount of damping; you don't want the measurement apparatus to have any effect upon the signal being measured!
Last edited:
The GOSS band can be in addition to the electrostatic shield. See Standard Range Toroidal Transformers: CM0225220: 225VA 230v to 2x20v for an example (and the type I am considering once I decide how many channels per case.
Could be you are not generally expected to want both though...
Could be you are not generally expected to want both though...
The problem is that the "2/10ths of a gnats fart" is often where the action is - the difference between Ye Olde Boring audio system, and "Wow!!" ... which is why some spend time playing in that area ...
At those prices, heck yes, I'd certainly buy both.
Remember, you spend the money in this household-fiscal-year, but you get to enjoy the results for several more h-f-y's at zero incremental cost! For the price of a very crappy bottle of Freixnet sparkling wine at a Chili's chain-restaurant in Midlothian, Texas - - - you could buy not one but two peace-of-mind emblems that prove to you (but only suggest to others) that you have sincerely Tried Your Damndest to get rid of hum and noise. Toss in a couple of ferrite beads and/or Common Mode Chokes and I'll start calling you "John Curl" as long as you're still buying the beers.
I disagree with the notion that the ear can reliably discern detail that cannot be resolved by modern test gear. With my gear at home, I can easily measure down to -140 dBV without even trying. Getting below that requires a couple of mouse clicks to dial up the averaging. At work, I routinely measure signals with 165 dB of dynamic range (+10 dBm carrier, -165 dBc noise floor), again, without trying very hard. The instrument is barely breaking a sweat at that point. The human ear-brain combo for even a trained person, has a hard time discerning signals 40-50 dB down from the strongest signal. Some people claim to be able to discern signals 60 dB down. That's still a far cry from the dynamic range of modern test gear.
I do agree that if you train yourself to listen for certain detail, you can more clearly articulate what you do and don't like about a certain setup. That was reflected in the Harman-Kardon study (Ooh! Actual data! Concept!) linked to by twest820 a while back. However, the auditory working memory (phonological loop) is only about 30 seconds long. For a good A/B comparison, the sound samples really need to be confined dramatically in duration.
We should also not forget about the various cognitive biases. It turns out, that if you prep the brain with an expectation of what to hear, you can in fact hear things in noisy signals that just aren't there. Michael Shermer has a brilliant and highly entertaining example with Stairway to Heaven played forwards and backwards in his TED talk: TED: Why People Believe Strange Things. The entire video is worth watching. It is highly entertaining. However, if you can't afford to take 13 minutes of your life to watch it, at least watch the last few minutes starting at 8:50 for the auditory illusion example.
The human mind is a wonderful thing...
~Tom
I do agree that if you train yourself to listen for certain detail, you can more clearly articulate what you do and don't like about a certain setup. That was reflected in the Harman-Kardon study (Ooh! Actual data! Concept!) linked to by twest820 a while back. However, the auditory working memory (phonological loop) is only about 30 seconds long. For a good A/B comparison, the sound samples really need to be confined dramatically in duration.
We should also not forget about the various cognitive biases. It turns out, that if you prep the brain with an expectation of what to hear, you can in fact hear things in noisy signals that just aren't there. Michael Shermer has a brilliant and highly entertaining example with Stairway to Heaven played forwards and backwards in his TED talk: TED: Why People Believe Strange Things. The entire video is worth watching. It is highly entertaining. However, if you can't afford to take 13 minutes of your life to watch it, at least watch the last few minutes starting at 8:50 for the auditory illusion example.
The human mind is a wonderful thing...
~Tom
Last edited:
What the human hearing does an admirable job of doing, is picking up transitory distortion artifacts; momentary glitches in the integrity of the playback. Which is exactly what MP3 and its ilk is all about - deliberately distorting the signal so that it can be compressed more; and all audio cognoscenti agree that MP3 is terrible playback, - it must be easy to hear that it's distorting, !! So, either human hearing can pick up distortion, or it can't - doesn't work having it both ways ...
Yes, the best bit rates in the compression make it much harder; but people train themselves - just like I mentioned - to use certain patterns of sound, types of music, to catch these algorithms out.
And the same process is used to catch out faulty playback in normal audio replay - you hear the system misbehaving under certain conditions, and then endeavour to track down the cause and effect relationship
Test gear would have a terrible time trying to detect MP3 distortion, but it's certainly there - because people can hear it happening - and all the time, they claim. The key thing being, that MP3 "distortion" is not "simple"; it's highly complex ... but human hearing can pick it without difficulty.
- it must be easy to hear that it's distorting, !! So, either human hearing can pick up distortion, or it can't - doesn't work having it both ways ...Yes, the best bit rates in the compression make it much harder; but people train themselves - just like I mentioned - to use certain patterns of sound, types of music, to catch these algorithms out.
And the same process is used to catch out faulty playback in normal audio replay - you hear the system misbehaving under certain conditions, and then endeavour to track down the cause and effect relationship
Test gear would have a terrible time trying to detect MP3 distortion, but it's certainly there - because people can hear it happening - and all the time, they claim
. The key thing being, that MP3 "distortion" is not "simple"; it's highly complex ... but human hearing can pick it without difficulty.
Last edited:
Previous scope shots are with BYV27-150 fast/soft recovery diodes which generate a very small spikes (invisible without highpass filter). The highpass should be set way too high (10k+100pF) so it does not act as a snubber itself...
tomchr said:
Nice gear
I've always wondered about this objectivist/subjectivist debate, tring to think about tests that could measure things that don't appear with traditional steady-state measurements...
- With FFT, all random signals appear more or less as a raise in the noise floor ; many signals have a "white noise" spectrum, for example a series of randomly placed spikes, or regular spikes of random polarity, appear as white noise
- THD measurement do not measure transient phenomena (like thermal) ; some time ago I I had fun plotting the Gm versus current versus time of an amplifier (using my soundcard) : the test signal was a low frequency high amplitude signal plus a high frequency very low amplitude signal, and a low resistance load. I used IQ detection on the HF at the output (in software) : the detected HF amplitude gives Gm (Y axis), the LF amplitude at a specific time gives "DC" current (X axis) ; plot that and you get a wing diagram, which evolved from underbiased to overbiased over the course of a few LF periods on a cold amplifier. It's basically intermodulation, but in real-time, without FFT.
I would specify the GOSS band, and the Faraday shield, and ask them to put on 10% more primary turns so the core runs at a lower flux density. But be aware this will reduce the VA rating by 10% for a given core size.
I also disagree with this. I think test gear is far more sensitive than the ear. What it doesn't do is assess the subjective impact of the distortions that it's measuring. Most audiophiles seem to believe that tiny measured aberrations map to huge subjective changes, so they end up obsessing over picoseconds of jitter and deci-gnat-farts of distortion, simply because they can be measured. I think it is the other way around. Measured aberrations have to be huge (by the standards of modern test gear) before they map to any subjective change at all.
You have obviously never tried measuring MP3 distortion. It's huge and very obvious, both in frequency domain and time domain measurements. Try MP3 compressing a 1kHz square wave, uncompressing it again, and comparing the waveform to the original.
As you probably know, the MP3 codec uses a psychoacoustic model to estimate what parts of the signal you wouldn't miss, and then throws these parts away. Unsurprisingly this causes obvious changes in both the spectrum and time series.
Last edited:
Sadly I have some common mode chokes. Was given them as left overs from a 'you don't want to know' project. 15A chokes so OTT for anything I could need, but hey free chokes and belief in better. Belt and braces, and extra belt.
should I ever end up in your neck of the woods beer would be a good plan.
A mains transformer electrostatic shield should connect to the screening enclose with a very low inductance lead.0.2 Gnat Farts... I like your perspective...
The electrostatic shield (aka belly band) should, in theory, catch any high-frequency noise present on the mains. I doubt it makes a huge difference, but on the other hand, it's not much of a cost adder, so why not? My power supply board will have a terminal where you can connect the shield, if your transformer has one.
~Tom
It should not be a long lead.
It should not connect to any PCBs.
It should not connect to any audio circuits.
or,
am I wrong again?
That pair of amplifiers were binned, a short while later.
That is why I did not reveal the name of the Designer.
But the experiment proved that the capacitor exacerbated the problem.
The capacitor made the quality of the supply worse, i.e. further from clean DC.
H.Ott tells us that the interference must be given a route to exit from the primary to the "Earth", that big capacitor under our feet.
The electrostatic shield must connect to Earth with the lowest impedance so that the interference sees maximum attenuation on it's way to the audio circuits.
The electrostatic shield must connect to Earth with the lowest impedance so that the interference sees maximum attenuation on it's way to the audio circuits.
A Faraday shield in a transformer is an extension of the Faraday cage formed by the chassis. You are trying to make it so the primary is outside the cage and the secondaries are inside. So ideally the Faraday shield should be connected to the chassis by a short low impedance path.
However, the Faraday shield unavoidably has some inductance of its own (a hole in it if you like) as it can't be allowed to form a shorted turn. The Faraday shields in the Airlink toroids are made by wrapping multiple turns of insulated copper foil, so they probably have far more inductance than the connecting lead and won't do much at really high frequencies.
However, the Faraday shield unavoidably has some inductance of its own (a hole in it if you like) as it can't be allowed to form a shorted turn. The Faraday shields in the Airlink toroids are made by wrapping multiple turns of insulated copper foil, so they probably have far more inductance than the connecting lead and won't do much at really high frequencies.
Tom did state that the value of the cap could make things worse when it is too high. Do you recall the value you used and if it was higher than the default value Tom suggested? Also, to address the point of a resistor being necessary to snub the ringing, don't some caps have a parasitic resistance within them?AndrewT said:
Last edited:
It was a wima 0.1uF 100V MKS4, precisely what the amplifier Designer specified.
Probably too low an esr to be an effective snubber.
According to Tomchr, using a too large cap moves the frequency even lower into the passband of the amplifier and thus the amplifier is more capable of attenuating the ripple on the supply rails.
Probably too low an esr to be an effective snubber.
According to Tomchr, using a too large cap moves the frequency even lower into the passband of the amplifier and thus the amplifier is more capable of attenuating the ripple on the supply rails.
Last edited:
It should connect to the system ground through a low-impedance path.
An experiment can't prove anything. It can show support for a theory or fail to show support for a theory. You removed the cap, something happened. As far as I read your post, you never found the root cause. Maybe the root cause was a cold solder joint somewhere and you just happened to tweak it just right when you pulled the cap from the supply, which made it look like the cap was the issue.
Three types of snubbers are being discussed:
C only: Cap across the transformer winding. In my view (and Blencowe's also), this, while not optimal, is the best universal solution. The value of the capacitance is not critical. 100 nF seems like a good one.
C or RC across each diode: If optimized, this can be a good solution. The challenge is that it needs to be optimized in every build as the transformer and wiring is different. The risk with this type of snubber is that you can actually make the ringing worse.
C||RC across the transformer winding: I don't think the C is needed. An RC across the winding should be sufficient, but it does need to be optimized and the component values will vary depending on the exact transformer chosen.
The capacitor does have some series resistance, but for a good film cap, this tends to be around 1-10 mΩ. That's not enough to dampen the ringing.
I still maintain that a properly designed amplifier should be immune to power supply non-idealities. The Modulus-86 is a good example of such an amplifier.
~Tom