Very nice illustration of the effects of varying bias! Maybe some additional info: you showed several residual distortion waveforms. In general, you'd want a straight line, which would mean zero distortion (no residual after removing the fundamental).
Some of the residual waveforms looked pretty much like a sime wave. That means the residual is a single frequency, most probably only 2nd or 3rd. You could verify that by checking the horizontal time scale and convert to frequency.
The other thing with the 'weird' waveform, that's when you go so far down in bias that you get serious cross over distortion, which is a whole slew of even and odd harmonics. And all those harmonics together look anything but a sinewave! The 'weird' one you show btw is giving the 'scope problems to trigger so you see several traces running through each other. If you would trigger on the fundamental of the generator it would be stable.
jan
Some of the residual waveforms looked pretty much like a sime wave. That means the residual is a single frequency, most probably only 2nd or 3rd. You could verify that by checking the horizontal time scale and convert to frequency.
The other thing with the 'weird' waveform, that's when you go so far down in bias that you get serious cross over distortion, which is a whole slew of even and odd harmonics. And all those harmonics together look anything but a sinewave! The 'weird' one you show btw is giving the 'scope problems to trigger so you see several traces running through each other. If you would trigger on the fundamental of the generator it would be stable.
jan
Aha! Thank you!!
I knew somebody would explain what was occurring if I posted that photo. I will adjust the trigger and see if I can get a 'clear' trace of the heavily-distorted residual.
Yes it looks like you have a Tek scope, at least 2-channel. If you input the fundamental from the generator into the 2nd channel and set the trigger to ch 2 it should be rock-steady.
If you have a 1-ch scope it should have a separate trigger input.
This type of signal is hard for trigger circuits.
jan
If you have a 1-ch scope it should have a separate trigger input.
This type of signal is hard for trigger circuits.
jan
Yes that'll work; it's doesn't really matter which is which, but having the fundamental on one channel gives you two advantages: you can trigger on the nice fundamental signal which is pretty stable, and you immediately see which harmonic is dominant if it is in the same picture as the fundamental.
If you see a nice sine-wave as residual you know that that represents a single dominant harmonic. If, in contrast, you would have equal 2nd and 3rd, the residual would look like a combination of two sines and depending on the relative phase of the harmonics look like a distorted sine or a distorted triangle. All this corresponds to the FFT of course.
The 'weird' wave you showed consists of many harmonics and is typical for cross-over distortion. Cross-over distortion is caused by sharp switching between the output devices which means many, especially higher order, harmonics (which is also clear from the FFT). The higher-order ones make up the sharp edges.
jan
This is extremely interesting!
Me and my homeboy will build the F5 soon and as good as the F5 seems to be in its original form one wonder how to implement it the best way possible. As I will build my own chassis; Would it be a good idea to go over the top with cooling and to up the bias >600 mV?
Me and my homeboy will build the F5 soon and as good as the F5 seems to be in its original form one wonder how to implement it the best way possible. As I will build my own chassis; Would it be a good idea to go over the top with cooling and to up the bias >600 mV?
Even in you have amazing cooling you will run into wattage limits on the transistors.
I will take a few more measurements and post them, but when I was playing around with it the 600mV bias point seems to be the sweet spot - it took much more bias for small benefits in reduced distortion.
Jan.Didden - I did trigger from the fundamental as you suggested, and it worked exactly as you said! Thanks!
I will take a few more measurements and post them, but when I was playing around with it the 600mV bias point seems to be the sweet spot - it took much more bias for small benefits in reduced distortion.
Jan.Didden - I did trigger from the fundamental as you suggested, and it worked exactly as you said! Thanks!
That's a great idea! I love it, and would really like to see what it would do for the distortion figures. Unfortunately, I have yet to get a variac...
I would also like to do some investigation into that -- from what little I have played around with this, my gut says fewer devices with more bias. But there will be a 'sweet spot' point as well.
That's a good question for Nelson... I wonder if he's reading this...
THere is no doubt that it would lower distortion as a result of higher bias, but with more pairs, you have fets operating well within their SOA, with plenty of current headroom to spare. ON top of that, as a result of the cumulative heat buildup on the sink, their temp is higher due to external factors rather thatn internal heat. I wonder if this does not help in both the area of conduction, like we have seen in the discussions about the diodes i the F5T.
The result is very informative, thanks!
Now for amps that using lots of out pairs, I remember reading somewhere NP said the total bias matters, not individual FET. Like X350.5 I measure only 106mA bias per FET, total should be 1A though. But I do hear distortion sometimes, when very hot (even smaller bias because of temperature rises when air-con isn't on and room temperature is about 30C), and start to worry if small bias per FET is the primary cause.
Now for amps that using lots of out pairs, I remember reading somewhere NP said the total bias matters, not individual FET. Like X350.5 I measure only 106mA bias per FET, total should be 1A though. But I do hear distortion sometimes, when very hot (even smaller bias because of temperature rises when air-con isn't on and room temperature is about 30C), and start to worry if small bias per FET is the primary cause.
My gut also says a single mosfet at one bias current will sound 'nicer' than two parallel mosfets at half the bias current each (but same total bias current). Mustn't this be the case, as that single transistor is operating under much different 'conditions' compared to a parallel pair? (And also, to realize any theoretical lower distortion from higher total bias, the pairs in parallel must be well matched to each other?)I would also like to do some investigation into that -- from what little I have played around with this, my gut says fewer devices with more bias. But there will be a 'sweet spot' point as well.
That's a good question for Nelson... I wonder if he's reading this...
But apparently for mosfets there is no biasing sweet spot, i.e. the distortion goes down and down until the transistor fails from heat (excess current). Seems to me that for us DIYers the practical 'sweet spot' is balancing the minimum number of devices in parallel to supply sufficient current needed for an intended power level into an intended loudspeaker load. So it's all about designing the amp for the load...
Attachments
I still wonder if the advantage of multiple mosfets is more about total current capacity and DF. From what I have read, Andersonix seems to be correct that with increasing current, both THD and higher order harmonics drop. THe only exception I have seen is in lateral mosfets. I believe this app note shows that at some point, the laterals THD plateaus and begins to get subtley worse. I assume this is related to Yfs ceiling, which is not present at reasonable bias levels on the other fets tested.
http://www.ti.com/lit/an/snaa045a/snaa045a.pdf
http://www.ti.com/lit/an/snaa045a/snaa045a.pdf
Why doesn't Nelson use more outputs in his FW amps? PL is another thing, they are high power commercial products, built to work with any speakers at any conditions.
I think that ideally you want to have a pair at a higher bias. If you can't provide them enough cooling, stability, etc, you start paralleling them, paying price of matching, capacitance change and so on. My speculation, though...
I think that ideally you want to have a pair at a higher bias. If you can't provide them enough cooling, stability, etc, you start paralleling them, paying price of matching, capacitance change and so on. My speculation, though...
Multiple parallel mosfets gives better reliability (longer life) as well. There's also a personal comfort level to be considered: Lots of heat from high biasing can take some getting used to!
BTW, that Fig 3 Distortion vs Watts graph is from NP's 'Leaving Class A' article, and as bias current goes up, distortion goes down and into the noise floor, so there's not much point in going too crazy with the bias unless the PS and 'everything else' is also very 'quiet.' More bias -> more PS ripple... on and on...
BTW, that Fig 3 Distortion vs Watts graph is from NP's 'Leaving Class A' article, and as bias current goes up, distortion goes down and into the noise floor, so there's not much point in going too crazy with the bias unless the PS and 'everything else' is also very 'quiet.' More bias -> more PS ripple... on and on...
I think in the right instance he has - for example there is no reason why the F4 must have 3 pair of output fets. There is, however, no doubt that more fets let you pass more current more easily.
The question that I hope to determine is if less distortion just plain sounds better... It might not, depending on the ratio of 2nd harmonic to 3rd harmonic, as Nelson alludes to in a few different places.
Also, I need to hook up the Aleph and do a similar series of bias vs. distortion points like above, as the Aleph's distortion has some interesting things in it as the bias is changed. That's my next thing to document when I have the chance to.
(Emphasis mine)
I think that is exactly it... Make the amp that you need/want. Someone who likes conventional 3-way speakers is going to have very different requirements than a guy with Edgarhorns.
So try a few different amps, figure out which one you like the best, and go for it!
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.