Eric, thanks for joining
1. The Fets are very well matched. I have no problems with DC offset at all
2. I measure every part too
3. I'll try to set AC current gain back. But I've another higher power X-Aleph and it is also set to 69% and makes no problems
4. URail is +/- 18.4V
Bias is 3.3 Amps
Power dissipation per FET is around 10W
The stock XA60 uses 12 Fets per channel also, right?
Pass DIY Addict
Joined 2000
Paid Member
I just plugged some of the data you've provided into William's handy-dandy Aleph-X spreadsheet and some strange results pop up.
First, your input data for one monoblock: 19v rails, 3.1A bias, AC gain at 69%, 12 fets
Output data from the spreadsheet: 72w into 8ohms, 50w into 4 ohms, 9w dissipation per fet, and with the same level of heatsinking on my monoblocks (0.08c/w), a total thermal rise of 9c with a final chassis temperature of 32c (given 23c ambient). Not even body temperature.
My conclusion (though I may be wrong), is that you're not running the bias point high enough (currently less than 0.5A per fet) for the transistors to properly conduct, hence the distortion. This is a Class A design, it should run HOT! Yours is still well below body temperature (or it would be with my heat sinks). My guess is that your current configuration indicates a voltage drop somewhere near 0.17v across the source resistor for each fet. Essentially, your amp is current hungry (bias point is too low).
Here is a quick thing to try: crank up the bias point until you are measuring somewhere closer to 0.35v drop across your source resistors (assuming their value is 0.33 ohms) to your output mosfets. You want each mosfet to be dissipating closer to 1.0A. This should raise your bias point to about 6A and would be pushing about 20w through each output fet. At this point, your amp should begin getting warm and be on its way to getting hot.
If raising the bias point works (amp gets warm/hot, output waveforms look more clean), then I would recommend the following changes to have your amp still produce 70w into an 8ohm load:
1) Clip the power supply to the third mosfet on each transistor bank. This means your monoblock is now running with 8 instead of 12 fets.
2) Adjust the bias (v1 and v3) until you are measuring a voltage drop of 350mV across the source resistors to your output fets. This should provide a total bias setting of about 4.3A. Each fet should be dissipating about 20w (about 1.0A each) now.
3) Reset AC Current Gain back to 50%
4) You should be able to measure about 70w into 8 ohms and about 35w into 4 ohms.
Hope this helps!
Eric
First, your input data for one monoblock: 19v rails, 3.1A bias, AC gain at 69%, 12 fets
Output data from the spreadsheet: 72w into 8ohms, 50w into 4 ohms, 9w dissipation per fet, and with the same level of heatsinking on my monoblocks (0.08c/w), a total thermal rise of 9c with a final chassis temperature of 32c (given 23c ambient). Not even body temperature.
My conclusion (though I may be wrong), is that you're not running the bias point high enough (currently less than 0.5A per fet) for the transistors to properly conduct, hence the distortion. This is a Class A design, it should run HOT! Yours is still well below body temperature (or it would be with my heat sinks). My guess is that your current configuration indicates a voltage drop somewhere near 0.17v across the source resistor for each fet. Essentially, your amp is current hungry (bias point is too low).
Here is a quick thing to try: crank up the bias point until you are measuring somewhere closer to 0.35v drop across your source resistors (assuming their value is 0.33 ohms) to your output mosfets. You want each mosfet to be dissipating closer to 1.0A. This should raise your bias point to about 6A and would be pushing about 20w through each output fet. At this point, your amp should begin getting warm and be on its way to getting hot.
If raising the bias point works (amp gets warm/hot, output waveforms look more clean), then I would recommend the following changes to have your amp still produce 70w into an 8ohm load:
1) Clip the power supply to the third mosfet on each transistor bank. This means your monoblock is now running with 8 instead of 12 fets.
2) Adjust the bias (v1 and v3) until you are measuring a voltage drop of 350mV across the source resistors to your output fets. This should provide a total bias setting of about 4.3A. Each fet should be dissipating about 20w (about 1.0A each) now.
3) Reset AC Current Gain back to 50%
4) You should be able to measure about 70w into 8 ohms and about 35w into 4 ohms.
Hope this helps!
Eric
Last edited:
My heatsinks are only 0.175 k/W for each monoblock. It's a small case.
21°C above ambient.
I could crank up the bias to 5A. Means 32°C above ambient. Dissipation per FET then 15W.
I'll set the bias to 0.57V at the source resistor (it's 0.69 ohms). That will give app 5 amps bias.
read this pdf ;
two things :
1. try adding these 5V6 zeners
2. relax with volume ( if you need bigger amp , try with bigger one
)
now seriously - there is some excessive capacitance in system ( now when input CCS current issue is confirmed as non existent ) ..... that's why decrease in gain in frequency domain
my habit is to use one IRFP150 instead two smaller IRFPs ... ( 240 , 244)
twice power and current , but capacitance pretty much lower than with 2 smaller ones
two things :
1. try adding these 5V6 zeners
2. relax with volume ( if you need bigger amp , try with bigger one
)now seriously - there is some excessive capacitance in system ( now when input CCS current issue is confirmed as non existent ) ..... that's why decrease in gain in frequency domain
my habit is to use one IRFP150 instead two smaller IRFPs ... ( 240 , 244)
twice power and current , but capacitance pretty much lower than with 2 smaller ones
Pass DIY Addict
Joined 2000
Paid Member
This is your problem! 2.8V input signal is waaaaay too much. Most every amp will clip with a 2.0v RMS input.
Your scope and wave-form testing should take place with measured OUTPUT levels of 1.0vRMS and 10.0vRMS. Both of these will correspond to input voltage levels far below 2.8v. Output sine and square waves measuring 1.0v and 10.0v RMS should look very clean on your scope.
Regardless of the outcome, I would still run the amp with 50% AC gain and set up for only 8 output mosfets. Paralleling extra mosfets and increasing AC current gain are usually "tricks" for getting more output of your amp without overheating the mosfets. Since the power output target you've specified is not terribly demanding (70w into 8 ohms), I see no need to use 12 mosfets or increases in AC current gain.
Eric
Excessive C? So all unneeded C's out, right?
I removed C12 and C13 (220uF Panasonic FC)
I also removed C9 and C10 (3n3) lag cap
I don't use these caps in my bigger amp also.
Now it is much better.
Power out with 20kHz and 1% THD is 40W/8Ohms now.
Before it was 20W/8Ohms.
Power out with 1kHz and 1% THD is 65W/8Ohms now.
Before it was also 65W/8Ohms.
So I still have problems with higher frequencies but much better than before.
Last edited:
With these values they do look perfect.
I wanted to check the behavior in full power and to get full power I need 2.8V unbalanced in.
Didn't really nobody look into THD at full power with frequencies above 1kHz?
Last edited:
Pass DIY Addict
Joined 2000
Paid Member
I have no way to measure THD with my (borrowed) scope, so no, I haven't looked at THD before. I have noticed that high frequency output tends to distort more quickly than lower-frequency output for the same input level, though I haven't made a formal study of it yet. Perhaps I'll have a look at this the next time I have the scope hooked up...
Funny, I have measured maximum output into 4ohms and 8ohms lots of times, but I never actually measured the magnitude of the input signal that provides my maximum undistorted input signal.
Guess its time to get the scope out again soon...
I still think you'll find benefit to use fewer output transistors and drive them with greater power (20w or so power dissipation).
Funny, I have measured maximum output into 4ohms and 8ohms lots of times, but I never actually measured the magnitude of the input signal that provides my maximum undistorted input signal.
Guess its time to get the scope out again soon...
I still think you'll find benefit to use fewer output transistors and drive them with greater power (20w or so power dissipation).
Seems like my problem is more normality than a problem and I realized it only because I tried my new toy (HP8903B Audio Analyzer)
Try it. You'll be surprised
Many thanks to all that tried to help here!!!!
I'll try to experiment with fewer FET's
Pass DIY Addict
Joined 2000
Paid Member
That's quite the nice toy that you've acquired! I've been trying (to the extent possible) do things "on the cheap," so I've been borrowing equipment from the EE lab on campus.
Looks like I need to drag out the scope tonight or tomorrow to do some more learning. In reference to your removal of caps, I found square wave distortion when c9 and c10 were removed (detailed on my web page that you've been reading). I have never run the amp without C12 or C13...
Looks like I need to drag out the scope tonight or tomorrow to do some more learning. In reference to your removal of caps, I found square wave distortion when c9 and c10 were removed (detailed on my web page that you've been reading). I have never run the amp without C12 or C13...
I have a pair of 140W Aleph-X build with Karis boards. No C9/C10 and no C12/C13 and NO Problems
Pass DIY Addict
Joined 2000
Paid Member
Hmmm... I'll have to give it a go without C12 and C13 and see what differences show up in the measurements and audio performance.
Ok, after carrying the 70lb beast upstairs and waiting for it to warm up, I made some measurements of input and output voltages at the visible onset of clipping in the scope. Here is what I found:
1 kHz clip onset, 3.66v peak-to-peak input, 150w output into 4 ohms
5 kHz clip onset, 3.64v p-p input, 139w into 4 ohms
10 kHz clip onset, 3.05v p-p input, 96w into 4 ohms
15 kHz clip onset, 2.80v p-p input, 80w into 4 ohms
20 kHz clip onset, 2.56v p-p input, 65w into 4 ohms
These measurements were made with 22.4v rails and 9.4A bias current. Measured voltage drop across source resistors averages 0.515v. Final heat sink temp is about 55c.
I stand corrected, 2.8v input is not too high of a signal level. I thought for sure this was part of your problem. I did, however, duplicate your situation of reduced power output as input frequency increases.
Ok, after carrying the 70lb beast upstairs and waiting for it to warm up, I made some measurements of input and output voltages at the visible onset of clipping in the scope. Here is what I found:
1 kHz clip onset, 3.66v peak-to-peak input, 150w output into 4 ohms
5 kHz clip onset, 3.64v p-p input, 139w into 4 ohms
10 kHz clip onset, 3.05v p-p input, 96w into 4 ohms
15 kHz clip onset, 2.80v p-p input, 80w into 4 ohms
20 kHz clip onset, 2.56v p-p input, 65w into 4 ohms
These measurements were made with 22.4v rails and 9.4A bias current. Measured voltage drop across source resistors averages 0.515v. Final heat sink temp is about 55c.
I stand corrected, 2.8v input is not too high of a signal level. I thought for sure this was part of your problem. I did, however, duplicate your situation of reduced power output as input frequency increases.
Last edited:
Similar to my findings. App. half power with 20kHz than with 1kHz. From 20Hz to 1kHz nearly no change.
Without C9/C10, C12/C13 it got significantly better.
As I removed all 4 caps at one time I can't they which ones are "bad".
We will have to wait what you find when removing C12/C13 only. This way we will know.
I can understang gain changeing with frequency but why the input voltage, shouldnt this be constant? Input impeadance doesnt change so it doesnt make sense.