By all means, hence my insistence on absolute minimal Y caps.FuriousD said:
I could be wrong but I can't see why we'd need an Y cap if the transformer is double-shielded (single turn foil shield) and the two shields start and end at the same "angle" around the center leg (so no vertical e core with the primary side coming out of one hole and the secondary out of the other!!!)
There would be absolutely no voltage capacitively coupled across them so there would be no voltage source between pri and sec that needs to be short-circuited.
Adding another shield as first and last layers would prevent any capacitive coupling through the ferrite.
Using foil windings and well-thought-out sequencing of the windings should actually get the same effect without explicit electrostatic shields (=make sure the ends that "see" eachother are HF cold).
Bruno Putzeys said:
Hi Bruno,
It seems you're quite right, cf:
http://www.fairchildsemi.com/an/AN/AN-4140.pdf
Extract: "Figure 14 shows a shielding example, which allows the
removal of the Y-capacitor that is commonly used to reduce
common mode EMI."
Y-capacitors implies that the circuit has an earth connection, so I will assume you are not referring to a non earthed, double insulated product. Double shielding the transformer will intercept much of the power circuit switching noise current being injected into chassis/earth ground, but not all. If the switching devices (mosfets or diodes) are mounted to chassis connected heatsinks, they can generate almost as much noise as an unshielded/unbalanced transformer. Even if they are floating, they and their circuit traces still will have some stray capacitive coupling to chassis (earth). Also, the transformer shield itself will never be perfect (may yield only a 20 to 40dB reduction), so EMI filtering most likely will still be necessary.
Regards -- analog(spiceman)
analogspiceman said:
It all makes sense unfortunately.
Indeed a non earthed design (with a two-wire power cord, like DVD players and such) would definitely ease things!
Well maybe, but only if one never connected them electrically to anything else, 'cause then the EMI gremlins would be right back working their common mode mischief.
The signal path would be out the line cord, 'round through earth, back through the audio cable and down the input of the DVD player's sound amp.
A well designed device is immune to a certain amount of EMI. Amplifiers that pick up minute amounts of common mode EMI are liable to be the same high quality amps that spontaneously become radio receivers if the speaker cable's too long. Colon Products Incorporated.
I've picked up a bad habit somewhere of also referring to a cap connected between pri and sec in a non-earthed product as "y" capacitors. Apparently the fairchild app note does the same. It is probably not entirely correct terminology though.analogspiceman said:
I was talking about caps placed thus in order to short-circuit the capacitively coubled voltage standing across pri and sec of the transformer.
Since we're concerned with reasonable powers (500W) and are simply going for the lowest EMI we can get, capacitive coupling to chassis can be avoided by using pcb mounted heat sinks that return capacitively coupled currents to local "ground" and using ground planes on the pcb.
Of course the transformer shield is not perfect, but if we can get by with, say 220pf across it I believe the smps would not be an impediment to good sound quality.
analogspiceman said:
Well maybe, but only if one never connected them electrically to anything else, 'cause then the EMI gremlins would be right back working their common mode mischief.
The signal path would be out the line cord, 'round through earth, back through the audio cable and down the input of the DVD player's sound amp.
Amplifiers that pick up minute amounts of common mode EMI are liable to be the same high quality amps that spontaneously become radio receivers if the speaker cable's too long. Colon Products Incorporated.
You mean Excorporated 🙂
Not sure I follow you, if the non-earthed product's audio output is floating and connected to a loudspeaker only, and its differential audio input is connected to a source via a reasonably high impedance, how can it transmit conducted common mode noise to the power line?
Radiated noise is another matter of course, but measures can be taken against that.
Assume that the conducted EMI exists. It must flow in a circuit. If it goes out the line cord, it must return either through the speaker cables, the audio input cables, or capacitive coupling from the amplifier's isolated chassis to nearby grounded metal. Remember, a connection that blocks common mode signals at audio frequencies may readily conduct them at EMI frequencies. In fact, unless specific measures have been taken to prevent it, this is usually the case.
Regards -- analog(spiceman)
Hi Bruno, your habits are perfectly civilized, I'm sure.
As far as I know, the term "y-capacitor" only implies that the part has been blessed
by the Safety Agencies as a "fail proof" or fail safe part in situations where its failure would otherwise expose the user to dangerous line voltages. That would certainly include across the transformer applications, although there they would require a special Y1 type part with a double blessing (equivalent to two normal Y2 caps in series). Y1 capacitors are called for wherever double insulation is required - such as when a product is designed for use with a two wire line cord. However, without an earth ground to shunt them away, such a capacitor may send line hum currents through the amplifiers input cables. This can lead to audible buzz problems (especially if the ac power connection has hot and neutral reversed and the amplifier uses single ended phono jack style inputs). Properly configured balanced inputs (such as present on your UcD amps
) may knock this down by 40dB or so. Of course, unless it's inaudible, this is still a problem.Regards -- analog(spiceman)
It seems you may be right about the unfamiliarity factor, at least among the audio diy-er community. And I agree with your statements, although my understanding of the terminology is a little different. To my experience, EMC (electromagnetic compatibility ) encompasses design and testing to limits, both for emissions and susceptibility (EMI is electromagnetic interference). The general idea of compatibility is that the acceptable levels of electrical noise going out an electronic device is well under the disruptive noise tolerance level a neighboring electronic device. As I (sort of) recall the safety factor is set to at least a couple of orders of magnitude.
Hasn't this general concept always been part of the CE, FCC, VDE, etc. approval process for consumer equipment (including audio)? Maybe I'm mistaken, but that is why I was incredulous when FuriousD reported:
If this isn't, as I expect, just a little marketing mumbo-jumbo or golden ear religious hysteria then that Krell amp has serious, inexcusable design shortcomings. EMC standards have been in place for decades. What happens to Krell audio quality when somebody in the next room uses their cell phone or someone in the basement shop runs an electric drill or someone in the study boots up their computer? Something doesn't add up.Well, agreed that a (double blind) listening test should be the ultimate measure of sound quality, but why should wide current loop bandwidth imply greater susceptibility?
Good suggestions, but I think you forgot the part about splitting the output windings and putting the diodes in the middle (okay, who knows what this would achieve?).
Regards -- analog(spiceman)
Hi analog,
you are correct, proper expression is of course EMI(ElectroMagnetic Interference) immunitiy. As I see it EMC makes sure that your device does not put garbage on the outside, while EMI immunity makes sure that you are not affected by the outside garbage.
One correction to FuriousD. Capacitors across power lines are commonly refered as X capacitors. There is no legal limit to the size of the X capacitors, (think of power factor correction capacitor banks on industrial lines) so developer of any equipment that is affected by X capacitor should start redesigning. One effect that X capacitor might have is to equalize common mode interference on both lines.
Regarding audio susceptibility of PFC circuit, here is quick opinion which might also be wrong. Any garbage on input voltage is directly transfered to the current refernce. Input current to output voltage has a single pole transfer function, while voltage loop transfer function has additional pole positioned very low. That means that voltage loop is not able to completely attenuate higher frequency garbage present on the input. Situation becomes worse with load feedforward which further distorts input voltage and therefore influences other power supplies working on the same line. I currently do not have time for a detailed analysis, but if your LTspice model is able to perform AC analysis, it would be easy to plot Vin to Vout transfer function. Even more interesting would be to use two PFC models connected to the line model and plot current feedforward to opposite output voltage transfer function.
Best regards,
Jaka Racman
you are correct, proper expression is of course EMI(ElectroMagnetic Interference) immunitiy. As I see it EMC makes sure that your device does not put garbage on the outside, while EMI immunity makes sure that you are not affected by the outside garbage.
One correction to FuriousD. Capacitors across power lines are commonly refered as X capacitors. There is no legal limit to the size of the X capacitors, (think of power factor correction capacitor banks on industrial lines) so developer of any equipment that is affected by X capacitor should start redesigning. One effect that X capacitor might have is to equalize common mode interference on both lines.
Regarding audio susceptibility of PFC circuit, here is quick opinion which might also be wrong. Any garbage on input voltage is directly transfered to the current refernce. Input current to output voltage has a single pole transfer function, while voltage loop transfer function has additional pole positioned very low. That means that voltage loop is not able to completely attenuate higher frequency garbage present on the input. Situation becomes worse with load feedforward which further distorts input voltage and therefore influences other power supplies working on the same line. I currently do not have time for a detailed analysis, but if your LTspice model is able to perform AC analysis, it would be easy to plot Vin to Vout transfer function. Even more interesting would be to use two PFC models connected to the line model and plot current feedforward to opposite output voltage transfer function.
Best regards,
Jaka Racman
The term ElectroMagnetic Compatibility entails *both* immunity and emission. EMC implies that equipmentJaka Racman said:
*is not allowed to radiate above a certain gabarith AND
*should not suffer any adverse effects from externally applied fields up to a certain gabarith.
Between the two, the emission and the susceptibility limits, there is a margin. If both are fulfilled, you can be certain that two devices complying with the same set of EMC rules will not interfere with eachother (ie are Compatible).
(Of course, you know all this - only explaining the terminology)
You're right about the input feedforward effect. However, this effect is much less on PFCs than on, say, 50Hz linear supplies where (at least during conduction) coupling is 100%. For this reason I'm not very concerned about this being a serious issue for audio quality.
analogspiceman said:
Hi Analog,
Agreed, I have looked at the effect of 1uF, 2uF, 4uF of capacitance across the mains and yes it does distort the mains waveshape, but not anywhere near as much as my wife turning the Hoover on in the next room does! I agree, the Krells are appauling if they are bothered by this. (Aside: I'm pretty sure that they're 50Hz txfmr, brdige + caps, no linear reg stage and by the sounds of it very little PSU rejection on the amplifier. This is not so uncommon; I have a linear based class-A far eastern amp on which I can hear these changes to the mains)
One way however to deal with this (In an "I'm squeaky clean" type way and this works on my far eastern class-A linear amp) is to make sure that the line filter used with the SMPS uses a common mode choke (CMK) first. i.e. Line In - CMK - X Cap - CMK - Y cap to Chassis. This way I can fit as much X Cap as I like (tested up to 4uF) after the common mode choke and it doesn't bother the Class-A linear amp.
F.
PS. Bruno, your description of the EMC terminology may be correct in principle, but EM compatibilty only requires that your product keeps working. It does not specify that it works to the best of it's ability. It would still meet the EMC directives if it sounded "dog rough" as long as it didn't switch itself off.
FuriousD said:
I believe that the EMC directive states: - “continues performing as specified” – so if you where to specify in the users guide that the units performance degrades or even fails under the influence of EMC then this is OK? – not sure about the customers confidence levels though...
Or at least this was the interpretation of the directive by a large EMC test house in HK… but I see the same test house “Test” the emission levels of CD / DVD players without audio / video leads connected to the units!
European laws require the equipment to maintain given SNR (40dB re 50mW output power) when subjected to a specified field, so that is slightly better than only requiring functionality.FuriousD said:
IMO the honest way would be to check the equipment performs to its full published spec, but I doubt such a thing will ever become law. Krell would lobby against it 😀
Me personally I like the idea mentioned by Jaka within post # 50. You still have the advantage of using a smaller transformer than needed with an ordinary "linear" PSU but you can move the safety responsibility to the manufacturer of the transformer.
Another topology that crossed my mind is an "ordinary" PFC, followed by an unregulated switching supply. I assume the latter could be made "stiff" enough for audio use.
Regards
Charles
Another topology that crossed my mind is an "ordinary" PFC, followed by an unregulated switching supply. I assume the latter could be made "stiff" enough for audio use.
Regards
Charles
That's how it's "usually" (in a relative sense) done, and this was the basis for analog's first simulation.phase_accurate said:
Bruno Putzeys said:
Bruno,
Whats the advantage of using an unregulated switching supply; apart from simplicity of course.
F.
Two advantages in fact: simpler than a regulated SMPS and smaller than a linear PSU.
With a good PFC in front of it we even have some preregulation.
Regards
Charles
With a good PFC in front of it we even have some preregulation.
Regards
Charles
- Status
- Not open for further replies.