Aha, here is a picture inside from the manufacturers website. I will drop them a note and see if they have any ideas.
http://www.kempelektroniks.nl/index.php?item=1
http://www.kempelektroniks.nl/index.php?item=1
All of these Class D amps generate large amounts of EMI. Doesn't matter who makes them. You have to be aware of that when you connect anything to them.
I have friends who build products with similar units. All by themselves, they pass the emissions test. Connect anything to them, especially the speaker, and they fail. Consider any wire connected to them to be an effective antenna.
Jocko
I have friends who build products with similar units. All by themselves, they pass the emissions test. Connect anything to them, especially the speaker, and they fail. Consider any wire connected to them to be an effective antenna.
Jocko
I can't vouch for my esteemed competitors, but you should definitely try UcD then. Not only do we measure the amplifiers with all cables installed (3metres, and fanning out in all directions), we target (and manage) EMI performance better than 20dB below agency limits.Jocko Homo said:
The modules are also used in audio sets containing tuners. The acid test there is quite different, and much worse.
Such sets have only a piece of wire hanging out the back for an FM antenna, running in parallel with the speaker cables at a few inches distance. So we drape the cables straight down to the floor, along with the FM antennq.
The spec: FM reception on a weak station should not worsen noticeably when the amp is operating. More specifically, we check the field strength required to have 28dB SNR from the FM tuner (=very bad reception). Then we turn on the amp. If the signal strength needs to be upped more than 3dB the amp is back to the drawing board.
It is very easy to make a circuit that complies with agency limits when it has no cables coming in or out. How could it get any emissions out otherwise? Cables are an integral part of any valid EMC test setup.
Some of your competitors design their product for use in powered loudspeakers, where the speaker leads are fairly short.
And then someone gets the idea to make a stand-alone amp with them..............and looooong speaker cables.
Jocko
And then someone gets the idea to make a stand-alone amp with them..............and looooong speaker cables.
Jocko
Yeah I know who you mean. They use a full bridge with only a differential-mode filter. The common mode voltage potentially spikes from rail to rail. In reality it's a bit better luckily and they're only 10dB over the limit (at least the 250W model I tested).
You are positive that yours is better...............?
How high does your spectrum analyser go??
(Mine goes to 18 GHz, no need to go that high. 80 MHz or so will work.)
Jocko
How high does your spectrum analyser go??
(Mine goes to 18 GHz, no need to go that high. 80 MHz or so will work.)
Jocko
Jocko Homo said:
I measure up to 1GHz but that's pro forma because really there's nothing coming out of the noise floor above 150MHz.
I have spent some time reading through all of this thread, and it's links to the outside world.
The Kemp line conditioner does indeed seem to be fully passive, and thus does not have an internal switching freq that can interfere with ZAPpulse (Like also mentioned by Charles).
Every ZAPpulse module is noise tested before being shipped, by
amplifying the output signal 50 times, and listening to the
noise floor. Any whine, crackle etc. would turn up rather clear in this test.
A single ZAPpulse module by itself can not make a whining noise because of the modulator design, so an interference signal must come from the outside. This is also confirmed by Ed's taking out the interconnects and finding the noise gone.
So what can be the problem? Either a RF noise is coming in on the interconnect signal wires or on the interconnect GND wires.
To find out, just use a bit of systematics:
To remove the latter connect the GND wire directly to the center of the PSU (goes for any kind of power amp BTW). Not to ZAPpulse's input GND.
We have also seen a (disturbingly large) number of people who connect pin 1 and GND of an RCA receptor to the - (MINUS) input of ZAPpulse. The is WRONG ....! It should of course hook up to GND. Of course this is not the case here.
So hook up pin 1 to the center of the power supply (GND).
Pin 2 to + input and
Pin 3 to - input.
And also make sure that your intercoonect cable
is the IEC standard with the above signal standards (not all are).
Futher to see if the noise is induced on the cable or coming from inside the preamp, simply disconnect the OTHER end of the interconnect cable. If the whine is still there, it's induced into the cable from an outside source.
Bruno: Yes we have had that exact same test here for years. With an FM tuner and a ZAPpulse module. We also call it our workshop radio 😀
🙂
The Kemp line conditioner does indeed seem to be fully passive, and thus does not have an internal switching freq that can interfere with ZAPpulse (Like also mentioned by Charles).
Every ZAPpulse module is noise tested before being shipped, by
amplifying the output signal 50 times, and listening to the
noise floor. Any whine, crackle etc. would turn up rather clear in this test.
A single ZAPpulse module by itself can not make a whining noise because of the modulator design, so an interference signal must come from the outside. This is also confirmed by Ed's taking out the interconnects and finding the noise gone.
So what can be the problem? Either a RF noise is coming in on the interconnect signal wires or on the interconnect GND wires.
To find out, just use a bit of systematics:
To remove the latter connect the GND wire directly to the center of the PSU (goes for any kind of power amp BTW). Not to ZAPpulse's input GND.
We have also seen a (disturbingly large) number of people who connect pin 1 and GND of an RCA receptor to the - (MINUS) input of ZAPpulse. The is WRONG ....! It should of course hook up to GND. Of course this is not the case here.
So hook up pin 1 to the center of the power supply (GND).
Pin 2 to + input and
Pin 3 to - input.
And also make sure that your intercoonect cable
is the IEC standard with the above signal standards (not all are).
Futher to see if the noise is induced on the cable or coming from inside the preamp, simply disconnect the OTHER end of the interconnect cable. If the whine is still there, it's induced into the cable from an outside source.
Bruno: Yes we have had that exact same test here for years. With an FM tuner and a ZAPpulse module. We also call it our workshop radio 😀
🙂
Thanks Lars, I appreciate the response.
I have dropped KE a quick question about the internals of this thing, and we will see if they can offer any other insights
Since I first wrote I have tried a couple of things, first using heavily shielded mains cables from the KE thing to the Zappulse - whine comes back. Then switch back to normal mains cable into the wall (same wall socket as the KE is plugged into, ie same earth and ring main) - whine goes away, noise floor drops.
However, I have sometimes managed to get the whine to come back by waving the mains cable near the "top" of the KE, ie over what would be the center of the two big torroids inside. Right now it's not doing it, but it was doing it everytime earlier...
Finally, as you saw, I tried leaving the Zappulse plugged into the KE, and removing the interconnects at the amp end (I'm nervous about blowing the modules up by disconnecting the other end...). This also made the whine go away.
It's going to be a short while before I can dismantle the device again and try to change the grounding on the inlet socket. I have to get on with the rest of the active system since I am now quite a long way behind schedule. However, perhaps the above info gives something of a clue?
By the way, another note to builders who saw my first picture. I have now added a 2mm thick aluminium sheet under the Zappulse modules and it's perhaps a little smaller than A5. It's glued with thermal epoxy to the steel base (stuff you stick on heatsinks with), and now the top of the case is a little more bearable, but still extremely hot...
I think that builders who are putting together a stack of these things need to think about better cooling than just using the bottom of the case? It's now not hot enough to burn you, but it's still very "hot" to the touch. I'm a bit worried about putting several of these in a stack and then sticking them in a fairly closed equipment rack, so I might need to rethink the cooling on the next pair of amps.
Thanks again
I have dropped KE a quick question about the internals of this thing, and we will see if they can offer any other insights
Since I first wrote I have tried a couple of things, first using heavily shielded mains cables from the KE thing to the Zappulse - whine comes back. Then switch back to normal mains cable into the wall (same wall socket as the KE is plugged into, ie same earth and ring main) - whine goes away, noise floor drops.
However, I have sometimes managed to get the whine to come back by waving the mains cable near the "top" of the KE, ie over what would be the center of the two big torroids inside. Right now it's not doing it, but it was doing it everytime earlier...
Finally, as you saw, I tried leaving the Zappulse plugged into the KE, and removing the interconnects at the amp end (I'm nervous about blowing the modules up by disconnecting the other end...). This also made the whine go away.
It's going to be a short while before I can dismantle the device again and try to change the grounding on the inlet socket. I have to get on with the rest of the active system since I am now quite a long way behind schedule. However, perhaps the above info gives something of a clue?
By the way, another note to builders who saw my first picture. I have now added a 2mm thick aluminium sheet under the Zappulse modules and it's perhaps a little smaller than A5. It's glued with thermal epoxy to the steel base (stuff you stick on heatsinks with), and now the top of the case is a little more bearable, but still extremely hot...
I think that builders who are putting together a stack of these things need to think about better cooling than just using the bottom of the case? It's now not hot enough to burn you, but it's still very "hot" to the touch. I'm a bit worried about putting several of these in a stack and then sticking them in a fairly closed equipment rack, so I might need to rethink the cooling on the next pair of amps.
Thanks again
I think the heat issue is that you are playing a massive amount of HF energy out of your amp at the moment, probably at full level. That explains the poor high frequency resolution. I had this with a Tripath when I connected balancing transformers to the inputs and forgot to load the winding [putting a resistor across it]. Result- thin sound, and amp output in saturation, using power like it was going out of business. These things won't get hot unless they are running flat out for weeks...sounds like something similar is happening to your amp.
That's a really interesting idea. Any suggestions on how to debug my issue though?
It's getting really hot even without a signal from the source for example (but interconects plugged in)
There is no HF whine coming from the speakers right now
Topside of the steel case, above the amp modules, is about 50C. Amp has been running for several hours now at normal quite lowish listening levels. Does this sound a little high?
Bottom of the case is slightly cooler (which makes sense)
It's getting really hot even without a signal from the source for example (but interconects plugged in)
There is no HF whine coming from the speakers right now
Topside of the steel case, above the amp modules, is about 50C. Amp has been running for several hours now at normal quite lowish listening levels. Does this sound a little high?
Bottom of the case is slightly cooler (which makes sense)
I stand corrected
They do get quite warm. I have hooked mine up, and even with no signal, heat is produced. Mine are bolted to a 3mm aluminium plate.
And, of little consolation I know, but they are totally silent, even with unshielded interconnects.
They do get quite warm. I have hooked mine up, and even with no signal, heat is produced. Mine are bolted to a 3mm aluminium plate.
And, of little consolation I know, but they are totally silent, even with unshielded interconnects.
I note the cookbook at lcaudio.com states:
"Cooling requirements depends on the
applied supply voltage, as most of the
heat is generated in idle mode, with no
load."
Elsewhere - it states:
"Asimple rule says that at no signal, a
ZAPpulse loses some 9.5 Watts power
to capacities in the MOSFET's, and at
full power it loses another 9.5 Watts to
serial resistance. So at 500 W out you
only have to dissipate around 19 Watts
to the heat sinks."
I imagine it's some kind of U curve, where in normal volume levels it stays the coolest.
20watts, at idle, for 2 modules could feel warm, but it's not like a regular amp that gets much hotter the more you drive it.
"Cooling requirements depends on the
applied supply voltage, as most of the
heat is generated in idle mode, with no
load."
Elsewhere - it states:
"Asimple rule says that at no signal, a
ZAPpulse loses some 9.5 Watts power
to capacities in the MOSFET's, and at
full power it loses another 9.5 Watts to
serial resistance. So at 500 W out you
only have to dissipate around 19 Watts
to the heat sinks."
I imagine it's some kind of U curve, where in normal volume levels it stays the coolest.
20watts, at idle, for 2 modules could feel warm, but it's not like a regular amp that gets much hotter the more you drive it.
Indeed, this is how I understood it to work. I have 63V going into the modules (quite a lot). They do have the external power source option hooked in from an external 20v feed, which is supposed to drop the idle requirements by a few watts.
As I understand it, a class D amp with only two states, ie +ve and -ve has to generate a zero voltage by oscillating between those two states. So at idle it burns basically as much juice as at full chat. However, I believe that there are some class D designs which have a three state output where no output is also an option, but I don't know which are which of the commercial designs (if you see what I mean).
Curiously, through the top of my case, one of the amps feels a lot hotter than the other (it's that right hand channel which is much hotter). However, I will have to take the lid off to see it this is real or just an abberation of the way they are installed, one of them not being as well in contact with the aluminium plate, or something else spurious like that.
However, I'm still most keen to find out what I can do to improve the resistance of these amps to external junk on the earth wire (which looks like my problem here). Can anyone suggest a process to go through?
Thanks
As I understand it, a class D amp with only two states, ie +ve and -ve has to generate a zero voltage by oscillating between those two states. So at idle it burns basically as much juice as at full chat. However, I believe that there are some class D designs which have a three state output where no output is also an option, but I don't know which are which of the commercial designs (if you see what I mean).
Curiously, through the top of my case, one of the amps feels a lot hotter than the other (it's that right hand channel which is much hotter). However, I will have to take the lid off to see it this is real or just an abberation of the way they are installed, one of them not being as well in contact with the aluminium plate, or something else spurious like that.
However, I'm still most keen to find out what I can do to improve the resistance of these amps to external junk on the earth wire (which looks like my problem here). Can anyone suggest a process to go through?
Thanks
I am thinking what if the KE's internal coils and capacitors form a tank circuit, that resonates at the frequency of another hooked up SMPS, or is simply triggered to make resonance peaks by the charge peaks of the ZAPpulse amplifier? This could generate the effect you describe.
For cooler operation, why not simply hook up the external Gate Driver circuit, that you have on your Predator PSU. This should effectively cut the heat dissipation in half 😉
For cooler operation, why not simply hook up the external Gate Driver circuit, that you have on your Predator PSU. This should effectively cut the heat dissipation in half 😉
Lars Clausen said:
I'm not quite sure what a tank circuit is, but I guess that you are perhaps on the right track here? In my testing there was only one other item plugged into the KE, the Meridian 502 preamp. I would have removed that as well, but didn't have a long enough flex to read the wall, and didn't have time to go find one! I can retest I guess.
Yep, already done that. That was what I meant at the top of my prev message about the 20v external feed.
I have raised the case up a bit by another half inch so that there is some more airflow underneath, and it seems to be a little cooler today. Top of the case is still very warm to the touch, but then since the bottom of the case is hot, and the only way it's going to get out is convecting up, then this is no real suprise. I don't know how how 2x 5W of dissipation should be, but above the modules it's physically hot to the touch 40-50C I guess (it's a 1U case).
I think somehow I may now need to add some more cooling holes to the case. Since I added the aluminium heat spreader below the modules I am now blocking the cooling slots in that position which would have let some air in. There is still some airflow obviously from the left of the case, and plenty of space above, but probably little air is coming straight up and through past the modules themselves
Idle dissipation and 3-level amps
At idle, a class D amplifier generally dissipates between 1% and 2% of the maximum power level it was designed to handle (in terms of MOSFET ratings). If LC's modules are designed for 500W (given due cooling), a dissipation at idle of 5W to 9W is perfectly normal.
-
Three-level class D amps also switch all the time, but the two legs switch in phase so differentially it looks like the output is "between the rails".
The inability to achieve soft-switching in three-level mode (unless 2 single-ended output filters are used) offsets the potential efficiency advandage brought forward by the reduction in switching frequency, and introduces class-B like distortion at low power levels.
Three-level amplifiers are not normally commercially made except by one company ("If it isn't ****** it isn't digital" was/is their slogan) who have introduced even more complications in the scheme- to further detriment.
At idle, a class D amplifier generally dissipates between 1% and 2% of the maximum power level it was designed to handle (in terms of MOSFET ratings). If LC's modules are designed for 500W (given due cooling), a dissipation at idle of 5W to 9W is perfectly normal.
-
Three-level class D amps also switch all the time, but the two legs switch in phase so differentially it looks like the output is "between the rails".
The inability to achieve soft-switching in three-level mode (unless 2 single-ended output filters are used) offsets the potential efficiency advandage brought forward by the reduction in switching frequency, and introduces class-B like distortion at low power levels.
Three-level amplifiers are not normally commercially made except by one company ("If it isn't ****** it isn't digital" was/is their slogan) who have introduced even more complications in the scheme- to further detriment.
ewildgoose said:
That's cool because I had the same thought after reading about the power cord being passed over it induces the whine.
A tank circuit is a resonator, it oscillates, sometimes it needs a bit of a kick to get it going but not always. I think Lars is theorizing the switching of the ZAPpulse could be hitting near its resonant frequency causing it to oscillate or peak. Hmm... that'd have to be one of the most interesting ground loops..
I guess it's a pretty simple test to find that out though, remove it, all noise should be cured.
- Status
- Not open for further replies.