AndrewT said:
Yes I looked at the device parameters and saw that myself and then was a bit puzzled. It seems high output power requires the devices to be cool, yet by nature they are going to heat up in a situation like this, limiting the overall output they can provide.
It seems this amplifier isn't capable of delivering the quoted output power with anything other then huge heatsinks.
The temperature thing just seems to make more and more sense. If I play a musical track with heavy bass content, lets say a bass drum periodically plays every second and whilst this is going on there are other instruments adding to the mix.
Now if the bass drum is on its own, no clipping noise is detected, however if a large amount of other bass information bas preceded it, or say the bass drum has beaten 4 times per second in comparison to the usual 1 per second. The additional power and strain this has on the FETs causes them to momentarily heat up, at this point the next bass drum hit to happen AFTER the heavy bass usage, clips. But give it a second or two and there is no clipping any more. So in other words once the FETs have cooled back down, the amount of current required can pass and the clipping disappears.
So if I describe this again
Drum beat - no clip
Drum beat - no clip
Heavy bass passage + drumbeats - clipping
Drum beat - clipping
Drum beat - less clipping
Drum beat - no clip
AndrewT said:
I agree ! You need to do more involved measurements than are possible with a meter. Sorry but I think that's the bottom line here.
I still think the design is pushing the envelope for a dual pair of MOSFET's. I would have thought the double die type would have been more suitable with a 16 amp IDS rating. MOSFETS do inherently limit at high drain currents, it's not a design feature, it's to do with the physical properties.
This is not mentioned, however the picture Slone provides has them connected to the sink.
The other thing that keeps haunting me is that you are using an unmounted speaker as a load ( Have I got that right ). It's only meaningful to make measurements with a defined load- usually either 8 or in your case 4 ohms resistive. Which for 250 watts takes some doing, electric fire element in a bucket comes to mind (Really)
I can't do a lot about this at the moment, but I could probably rig something up if absolutely required.
It's decent I suppose. The main issue is there is between 10-30dB of gain through the bass due to the open baffle compensation, so the amps have to work hard in comparison to say the upper midrange and tweeter that have no power Eq at all.
Right just to clear this up.
8Arms into the 4 ohm load.
Am I correct in saying that this workload is shared by each rail, such that 4A is provided by the P channel and 4A is provided by the N channel?
If this is the case what current draw per rail at the amplifier supply should I see to allow this?
Haha, I was thinking of mentioning this, but decided not to, to help keep the atmosphere pleasant, after all it's not that hard for me type a sentence summing the problem up.
What I do not know currently, (but it seems realistic) is if the temperature is the limiting factor at the moment. There could be nothing wrong with anything except the FET temp. getting too high and thus the max output is curtailed. When playing bass heavy music (at just before clipping) for 30 mins the heatsinks are about as hot as they were with 0.15V bais per FET.
Mooly said:
I mentioned earlier if either of you cared to read it
that the rails did/do NOT droop by anything appreciable, we're talking a couple of volts per rail. Unless you were meaning something else.Alright I need more involved measurements, what do I need to do this?
And what capabilities do these things need to have?
I can connect the amplifier to the sound card and do a host of stuff with the ARTA suit. I can probably also get an oscilloscope program too, but as I said earlier the bandwidth would be horrible.
250W into 4r0 requires 44.7Vpk and 11.18Apk.
Vpk=sqrt[ P * Rload * 2].
Ipk=Vpk / Rload.
All of that 11.2Apk comes alternately from each supply rail.
It's the same argument when we were considering the power available when 3.3Apk was flowing to the load.
PS,
we are reading it, it being what data you are able to gather with just a DMM and no load resistor.
Vpk=sqrt[ P * Rload * 2].
Ipk=Vpk / Rload.
All of that 11.2Apk comes alternately from each supply rail.
It's the same argument when we were considering the power available when 3.3Apk was flowing to the load.
PS,
we are reading it, it being what data you are able to gather with just a DMM and no load resistor.
Hi,
an interesting experiment for you to try.
Switch your DMM to 200Vac and measure the DC voltage on each half of the PSU. Do this with the amp idling and normal bias. Do again with the bias turned right down to minimum. return the bias to normal and measure again when you are sending near maximum signal to the test load.
Tell us the three Vac readings and the three Vdc readings.
Then we'll try part 2 of the experiment
an interesting experiment for you to try.
Switch your DMM to 200Vac and measure the DC voltage on each half of the PSU. Do this with the amp idling and normal bias. Do again with the bias turned right down to minimum. return the bias to normal and measure again when you are sending near maximum signal to the test load.
Tell us the three Vac readings and the three Vdc readings.
Then we'll try part 2 of the experiment
Several posts back, it sounded like the amp was heating up and the current was going up all out of proportion to the signal level you described. It should have been blowing you out of the room at that point. It almost sounds like the amp is trying to drive a short circuit. Look for something loading it down- output network wired wrong, speaker problem, etc. IMO, this self correcting business with the FETs is a tangent. There is some fundemental wiring/hookup problem, or maybe a bad part.
AndrewT said:
Amp Idling with bias set.
+rail = 64.9
-rail = -64.9
It's latish in the night and as people are turning things off the rail voltages keep rising by a small amount, as the rails increases due to the lower demand. It's only small mind you, but each rail is always within 0.1V of each other. If I turn my 2kW heater on the rail voltages drop by almost a volt anyway I'll carry on.
Amp Idling with zero bias.
+ rail = a change of 0.3 volts
-rail = a change of 0.3 volts.
As the mains voltage is fluctuating I just measured the difference between nominally and zero biased.
Amp driving 55hz sine wave into paralleled XLS just before clipping.
Both lines dropped by 5V down to 61.5V, still 1.5 V higher then the 60V they are supposed to have.
If I didn't know the impedance of the loudspeaker and wanted to make a calculated guess would this be possible?
If I know the RMS voltage at the output and the current drawn by one of the amplifier rails? I suppose measuring the Irms current at the output would figure that out too.
Andrew mentions (I think) that the rail current = Ipeak on the output. If this is the case surely using Vrms on the output and Ipeak on one of the rails one could calculate the load being driven by the amplifier?
Or am I totally wrong on this?
If I know the RMS voltage at the output and the current drawn by one of the amplifier rails? I suppose measuring the Irms current at the output would figure that out too.
Andrew mentions (I think) that the rail current = Ipeak on the output. If this is the case surely using Vrms on the output and Ipeak on one of the rails one could calculate the load being driven by the amplifier?
Or am I totally wrong on this?
Conrad Hoffman said:
I measured the output voltage and output current fixed at 55hz.
0.27amps
1.04 Volts
V=IR
R=V/I = 3.85 ohms. Close enough to the 4 ohms I've always assumed it is.
I don't want to measure this at a higher voltage as people are sleeping >.<.
But it appears, at least from the output to the loudspeakers, that the amplifier is seeing the loudspeaker load as the load it's driving. Would this count out the "fundamental wiring/hookup problem"? or could something else internal be causing the issue.
Both amplifiers function in exactly the same way, so unless I've got two bad parts that seems a little unlikely.
Conrad Hoffman said:
I cannot give actual figures at the moment as its late at night. But...
When I first had the current being measured I slowly increased the volume whilst reading the current drain. The current increased slowly as the volume went up, there were no sudden steps once I went beyond a certain level.
One thing that I can say now though, is this.
I was applying an input signal from Audacity and was altering, ever so slightly, the input voltage (to the amplifier) by altering the amplitude of the generated sine wave.
Don't worry about the absolute figures I am just using this to illustrate a point.
If I put the signal through at level 0.8 there would be clipping almost instantly.
If I use a level of 0.7 there would be no clipping at all.
If I use a level of 0.76, after playing the signal for 2 seconds clipping starts.
If I use a level of 0.74, after playing now for say 10 seconds clipping begins to show up.
To me it seems like the FETs are heating up and are self limiting; the lower the drive level, that little bit longer it takes for them to reach the limiting temperature.
Now whether or not the temperature is actually correct for the power I am trying to throw around is another matter, which is why of course you asked if the rail power was proportional to the output power.
Well I can tell you from some testing I did earlier on that at 55hz, the aforementioned 3.85ohms, the clipping sound would start to appear after a few seconds when there was 25 volts exactly across the output terminals and at the same time, the negative rail had 3.98 amps running through it.
If I have this right and we are assuming say 4 ohms, as the loudspeaker voice coils will have heated up some (I have placed my finger on them after being driven like this for 10 mins or so and they are actually quite cool, but there is some heat there).
V=IR
I=V/R = 25/4 = 6.25amps rms.
If we turn this into a peak value we get 8.83amps which is more then double the value Andrew said it should be.
The only variable that could possibly change here is the loudspeaker or, if this is remaining pretty constant (this is why I want to measure the output current & voltage at the onset of clipping), could indicate a problem with the amplifier.
If the calculations so far are correct, with 3.98amps peak on the rails 3.98/sqrt of 2 = 2.81amps rms on the output. With 25 volts being achieved we're looking at a load impedance of
V=IR again
V/I = R = 25/2.81 = 8.9 ohms.
However am I right in assuming some of that input power is being dissipated as heat? If we assume 70% efficiency (figure from Slone's book) 30% of that 2.81 amps is going out as heat, leaving 1.967amps for the output meaning now....
25/1.967 = 12.7 ohms. This is obviously incorrect or there really is a huge problem with the amplifiers.
Another thought I had which seems completely random.
P=I^2R
On the output assuming 4ohms I'm getting (6.25^2)*4 = 156watts.
On the input we've got 4amps and 63 volts = P=VI = 252 watts. This is a peak value? 252/1.41 = 180 watts.
Assuming 70% efficiency. 180*0.7 = 125 watts.
So 156 total watts on the output with a 4 ohm load.
And 125 watts on the output after 70% amplifier efficiency, from the input current/voltage.
This is why I need to measure the current drawn on the output at the onset of clipping.
If we assume the loudspeaker impedance could in fact be a little bit higher....
Say it had reached 4.6 ohms, we'd now be at 136 watts.
These figures don't seem too far apart. And now you're going to shoot me telling me that although what I've calculated looks nice, it's wrong for X number of reasons
A further thought. We keep talking of 4 ohms. I suspect this load you are using is anything but.
I have been thinking about this and, in an inductive load (yours), the voltage and current are not in phase. The frequencies you are using are also at the "worst" value I suspect. 55Hz must be pretty near the speaker electrical resonance point I would guess
Again my maths fails me, I can visualise what's happening, but can't calculate it. In any case you need the values to put in to it all.
Put simply the current will either lead or lag the voltage across the load. The power in the outputs is the product of current flow (at any one instant) and the voltage at this point in time across the device. In a resistive load the current through the FET is increasing while the voltage across it is decreasing, as the output rises toward the rail voltage. With an inductive load the peak current is occuring either before or after minimum voltage occurs across the FET. The result is higher power dissipation in the outputs.
Anyone good at j notation and real and imaginery power
Now another thought, and I am sure you have not done this.
Years ago I was playing about with inductive filters in the rails to an Amp to see about reducing common mode noise on the rails, and tried a "filter " from an old VCR. I was gobsmacked at what happened. The filter coupled the two rails "magnetically" through its ferrite core, and at anything over about 1 Watt the outputs collapsed through the effect of the rails being modulated so violently.
I have been thinking about this and, in an inductive load (yours), the voltage and current are not in phase. The frequencies you are using are also at the "worst" value I suspect. 55Hz must be pretty near the speaker electrical resonance point I would guess
Again my maths fails me, I can visualise what's happening, but can't calculate it. In any case you need the values to put in to it all.
Put simply the current will either lead or lag the voltage across the load. The power in the outputs is the product of current flow (at any one instant) and the voltage at this point in time across the device. In a resistive load the current through the FET is increasing while the voltage across it is decreasing, as the output rises toward the rail voltage. With an inductive load the peak current is occuring either before or after minimum voltage occurs across the FET. The result is higher power dissipation in the outputs.
Anyone good at j notation and real and imaginery power
Now another thought, and I am sure you have not done this.
Years ago I was playing about with inductive filters in the rails to an Amp to see about reducing common mode noise on the rails, and tried a "filter " from an old VCR. I was gobsmacked at what happened. The filter coupled the two rails "magnetically" through its ferrite core, and at anything over about 1 Watt the outputs collapsed through the effect of the rails being modulated so violently.
Are you really putting these power levels into an actual physical speaker? Very few speakers can handle more than a dozen watts of continuous sine wave power for more than a few seconds to maybe a minute. The voice coils would simply burn up.
If everything is proportional and working as it should, this might indeed come down to getting the power out of the FETs. There is a limit to how much heat you can remove from smaller devices. IMO, very few silicone rubber pads are as good as mica and grease, though a few are better. Sometimes a copper spreader plate is used between a device and the heatsink- grease the device right to the copper, then isolate the larger copper area from the sink with mica or silicone pad. The only amplifiers I've ever had this sort of problem with were either going out of control on bias, or were oscillating. Try the AM radio trick- wave the antenna of an AM radio over the amp while tuned between stations. If you hear a signal that's obviously produced by the amp, you're oscillating!
If everything is proportional and working as it should, this might indeed come down to getting the power out of the FETs. There is a limit to how much heat you can remove from smaller devices. IMO, very few silicone rubber pads are as good as mica and grease, though a few are better. Sometimes a copper spreader plate is used between a device and the heatsink- grease the device right to the copper, then isolate the larger copper area from the sink with mica or silicone pad. The only amplifiers I've ever had this sort of problem with were either going out of control on bias, or were oscillating. Try the AM radio trick- wave the antenna of an AM radio over the amp while tuned between stations. If you hear a signal that's obviously produced by the amp, you're oscillating!
Yeah, it's a pair of peerless 830452 XLS subwoofer drivers, they are supposed to be good till 300 watts continuous each. As I mentioned before I have felt how hot the coils are, after testing the amplifier up to clipping and they are warm but by no means hot.
Considering that getting 250 watts into 4 ohms is pushing the devices, maybe it is just an issue with the heat, spread the job to three output pairs and the work load would be substantially decreased. However there are a couple of situations which just don't make sense to me with this, but I will leave it for now, in case something else crops up as the problem.
The pads I am using have a thermal resistance of 0.3 degrees/watt, that seems pretty good to me.
http://uk.farnell.com/681090/semico...g-antistatic/product.us0?sku=bergquist-k6-104
I'll have to find an AM radio for that -.-
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.