Remember that the definition of class A is that each output side conducts the full 360 degrees of the signal cycle. In the past, designers have build amps that keep the devices on during their off cycle by keeping them conducting 10 or 20mA DC bias. Sometimes this is called non-switching class A - the devices do never switch off.
But if you look at the power supply and return current for class A, you'll see a replica of the signal, a sine with a sine signal. With the non-switching class A, you'll see half-wave, pulse like supply and return current as with class (A)B. Those cause a lot of high harmonics spray on the supply lines, which is not the case with true class A.
So non-switching class A for me does not merit the class A name because a) the devices do not conduct the signal for the full 360 degrees, and b) they have very different supply currents making it hard to keep the amp 'clean'.
Now square or cube law amps do conduct the full 360 degrees signal cycle. The power supply currents do not exactly follow the signal (they look more like a skewed sine wave with sine wave signals) but they are close and for me that is enough to call it class A.
Jan
As for these other points, these are not really different from other amps I'd think. You can lower the distortion by using feedback - in fact Ian proposed inclusion of an opamp in the forward path for higher ol gain that can be used for feedback. Lots of parallel output devices and lower Re values will also help, just like with other amps.
And the bias setting sensitivity depends basically on the same sort of factors as any other amp.
This is all important but is not specific to square or cube class A.
Jan
Another question for me would be, how does the HF IMD performance of this amp compare to more traditional designs? I expect the IMD at high power and frequency (maybe 19+20kHz) would be pretty bad, as the output stage will be generating big 3rd or 5th order products and relying on global feedback to get rid of them.
Of course the audibility of this is another question as nobody listens to 19+20kHz sine waves at 100W, so maybe it is a case of trading crossover distortion at low levels for IMD at high levels, rearranging the shortcomings into places where they won't be audible.
If there are IMD test results in the issue of Linear Audio I promise to buy it
Of course the audibility of this is another question as nobody listens to 19+20kHz sine waves at 100W, so maybe it is a case of trading crossover distortion at low levels for IMD at high levels, rearranging the shortcomings into places where they won't be audible.
If there are IMD test results in the issue of Linear Audio I promise to buy it
Last edited:
Another question for me would be, how does the HF IMD performance of this amp compare to more traditional designs? I expect the IMD at high power and frequency (maybe 19+20kHz) would be pretty bad, as the output stage will be generating big 3rd or 5th order products and relying on global feedback to get rid of them.
Of course the audibility of this is another question as nobody listens to 19+20kHz sine waves at 100W, so maybe it is a case of trading crossover distortion at low levels for IMD at high levels, rearranging the shortcomings into places where they won't be audible.
If there are IMD test results in the issue of Linear Audio I promise to buy it
Ian has prepared an extensive reply to your questions which I will be posting at my site, but for the impatient readers I'll post it here in it's original form.
I think you'll appreciate it.
Jan
To better prepare myself for my build I have been studying the schematic and associated articles (book and web links). I have a question regarding the 5 CMOS Inverters (U1-5).
In the Linear Audio article Mr. Hegglun discusses the CMOS inverters used in the differential input stage. “Each inverter is actually three inverters connected in parallel since there are 6 inverters in a package. Any unused inverters upset the operation of the used inverters so they all have to be used.”
Just one inverter pair is shown in the schematic (probably for simplicity) but a quick look at the pcb show that all 6 inverters in the package are used as described above. 3 used in U1-A and 3 used in U1-B.
U3-A on the pcb reflects all 6 inverters are wired in parallel.
However, I am confused with how U2, 4, & 5 are shown on the schematic and as wired on the pcb. The schematic shows that the gate(input) and the drain(output) are shorted together. The pcb also reflects that all six inverters in the package have the gates and drains all tied together.
I have a general understanding of how a CMOS hex inverter works. But I don’t see how this configuration works. I have looked online for examples using this configuration without luck. I have to assume that this is correct. So can someone please explain this to me or point me to some reading material that explains this configuration.
Here is my interpretation of how U2,4, & 5 are wired on the pcb:
In the Linear Audio article Mr. Hegglun discusses the CMOS inverters used in the differential input stage. “Each inverter is actually three inverters connected in parallel since there are 6 inverters in a package. Any unused inverters upset the operation of the used inverters so they all have to be used.”
Just one inverter pair is shown in the schematic (probably for simplicity) but a quick look at the pcb show that all 6 inverters in the package are used as described above. 3 used in U1-A and 3 used in U1-B.
U3-A on the pcb reflects all 6 inverters are wired in parallel.
However, I am confused with how U2, 4, & 5 are shown on the schematic and as wired on the pcb. The schematic shows that the gate(input) and the drain(output) are shorted together. The pcb also reflects that all six inverters in the package have the gates and drains all tied together.
I have a general understanding of how a CMOS hex inverter works. But I don’t see how this configuration works. I have looked online for examples using this configuration without luck. I have to assume that this is correct. So can someone please explain this to me or point me to some reading material that explains this configuration.
Here is my interpretation of how U2,4, & 5 are wired on the pcb:
An externally hosted image should be here but it was not working when we last tested it.
I certainly do appreciate it!
Thanks very much for publishing my question and to Ian for replying. I think I basically agree with Ian that lowering the order of the output stage's nonlinearities is more important than reducing the magnitude of them. I've always been skeptical of the CFP output stage, that conceals a small, sharp gain wiggle in a transfer function that's otherwise nice and flat. The local feedback in the Sziklai pair sharpens up the knee, and to me that is the very last thing you want to do. Doug Self raised this point in his book, but he left the question open as to whether a small, sharp discontinuity was preferable to a larger but smoother one.It was news to me that a cube-law characteristic in the crossover region gives a harmonic series falling off faster than the usual BJT crossover glitch. I suppose this makes sense because the cube law is polynomial but the BJT glitch is exponential. An exponential is an infinite polynomial series, so it generates every harmonic, whereas a cubic function just generates the 3rd.
Another thing that fascinates me about this business is that a lot of the work has been done in simulation, but the simulated MOSFETs and BJTs behave very differently to the real ones at low currents. The plots of MOSFET crossover regions in Self's book are all wrong. Real MOSFETs give much better results. Bob Cordell went some way to rectifying this with his tutorials on making your own SPICE models.
Think out of the box. These are not inverters - these are N- and P-JFET pairs as shown in the schematic.
What's a FET with gate connected to source?
Jan
Last edited:
I hope this is alright to post here. This was listed in the references of his article in Linear Audio Volume 8. Available here: http://perso.orange.fr/francis.audio2/AmpHegglun.doc
Send me an MP with your email address.
The schematics at Francis Brooke's site comes from my collection of the british magazines Electronics World and Wireless World. It a was wonderful magazine but slowly and surely lost the plot at the beginning of this century.
Thanks to Jan Didden, we now have Linear Audio where the authors have plenty of room to publish very detailed articles.