Valuable input! Thank you. I built a BJT version of this amp as well. They measure about the same, but subjectively I like the BJT version slightly better. Might very well be confirmation bias…
The big cap on the VBE and the high value gate resistors are mainly derived from looking at other designs. I’ll reduce them the next time I’m poking around.
First I've heard that ridiculous claim. Learn from the masters like Self and Cordell, or basically learn about analog electronics and have fun building stuff. In your circuit the current mirror pretty much eliminates 2nd order distortion due to asymmetrical currents in the pair as well as increasing the transconductance by a factor of two.
Hi
To me "grainy" sound is the sound of swirling your fingers through sand. It sounds "particulate" and like white noise that is maybe filtered? In certain cap positions in preamps, a ceramic will have that grain where a plastic won't.
Current mirrors and current sources/sinks should use fast BJTs as they are part of the signal path.
Regarding clipping with the EF-enhanced VAS: Using same-sex EF and VAS BJTs tends to have the sticking problem. I was curious about the inverted EF enhancement using an NPN with resistor to ground, and that seemed to clip cleaner. In front-ends without current-mirrors, the collector resistors for the diff must be reduced in value. This probably reduces DC gain, but most circuits have tons anyway. Lower-gain and feedbackless designs clip clean, so it is with high-gain designs that care is needed for clean clipping. Baker clamps and other diode connections are oten added, too.
To me "grainy" sound is the sound of swirling your fingers through sand. It sounds "particulate" and like white noise that is maybe filtered? In certain cap positions in preamps, a ceramic will have that grain where a plastic won't.
Current mirrors and current sources/sinks should use fast BJTs as they are part of the signal path.
Regarding clipping with the EF-enhanced VAS: Using same-sex EF and VAS BJTs tends to have the sticking problem. I was curious about the inverted EF enhancement using an NPN with resistor to ground, and that seemed to clip cleaner. In front-ends without current-mirrors, the collector resistors for the diff must be reduced in value. This probably reduces DC gain, but most circuits have tons anyway. Lower-gain and feedbackless designs clip clean, so it is with high-gain designs that care is needed for clean clipping. Baker clamps and other diode connections are oten added, too.
Small signal bipolar transistors are fast (from an audio point of view)! Well, some positions are more demanding like the VAS, but 20kHz isn't a challenge really.
Hi
My point about fast devices in current sources relates to the number of circuits you see that use slothy high-voltage parts instead of more capable ones. There are also tons of statements that "the current source isn't part of the audio path"... well... it is. What do you think creates the other half of the signal in a VAS like the common circuit has.
In the 2-BJT current source the control BJT is low-voltage and is usually way fast and the high-voltage one may be quite a bit slower. This was more of a problem in the past when it was difficult to have speed, high-voltage and good gain all in one device.
You have to look past 20kHz if you want good THD20. Self always uses Miller compensation or 2-pole, but with quite an early roll-off. You see the THD rise quickly at 10kHz for all of his amps. If you use alternmate compensation, the relationship between bandwidth and speed is broken and THD20 (20kHz) can be greatly improved.
The current mirror eliminates a critical resistor in the diff amp and improves performance as you have seen in your sim. You have a proper current source in the tail of the diff and this allows the circuit to bias properly at pretty low rail voltages - good for testing and for using the identical circuit at many different supply voltages for different applications.
You might want to add a buffer stage to drive the mosfet gates as this will improve the slew rate and usually reduces THD. It might require a tweak of the compensation values.
There is a lot of debate regarding THD numbers and musicalty. Electronic distortions are unnatural to our ear, yet you can have a high-THD low-gain or feedbackless circuit that sounds "nice" to you. usually it will be nice with open and airy music but not with a busy midrange. THD in these circuits tends to be low-order. When you get into high-gain circuits where the feedback factor is larger, there will be more higher-order harmonics and your ear is incredibly sensitive to these.
"Intermodulation" is when two signals interfere with each other. IM distortion (IMD or just IM) is much more harsh to our ear as the harmonics created by it are not musically related to the signal, it is just noise that fatigues our brains as we try to filter it out or make sense of it. generally, anything that reduces THD20 also reduces IM.
My point about fast devices in current sources relates to the number of circuits you see that use slothy high-voltage parts instead of more capable ones. There are also tons of statements that "the current source isn't part of the audio path"... well... it is. What do you think creates the other half of the signal in a VAS like the common circuit has.
In the 2-BJT current source the control BJT is low-voltage and is usually way fast and the high-voltage one may be quite a bit slower. This was more of a problem in the past when it was difficult to have speed, high-voltage and good gain all in one device.
You have to look past 20kHz if you want good THD20. Self always uses Miller compensation or 2-pole, but with quite an early roll-off. You see the THD rise quickly at 10kHz for all of his amps. If you use alternmate compensation, the relationship between bandwidth and speed is broken and THD20 (20kHz) can be greatly improved.
The current mirror eliminates a critical resistor in the diff amp and improves performance as you have seen in your sim. You have a proper current source in the tail of the diff and this allows the circuit to bias properly at pretty low rail voltages - good for testing and for using the identical circuit at many different supply voltages for different applications.
You might want to add a buffer stage to drive the mosfet gates as this will improve the slew rate and usually reduces THD. It might require a tweak of the compensation values.
There is a lot of debate regarding THD numbers and musicalty. Electronic distortions are unnatural to our ear, yet you can have a high-THD low-gain or feedbackless circuit that sounds "nice" to you. usually it will be nice with open and airy music but not with a busy midrange. THD in these circuits tends to be low-order. When you get into high-gain circuits where the feedback factor is larger, there will be more higher-order harmonics and your ear is incredibly sensitive to these.
"Intermodulation" is when two signals interfere with each other. IM distortion (IMD or just IM) is much more harsh to our ear as the harmonics created by it are not musically related to the signal, it is just noise that fatigues our brains as we try to filter it out or make sense of it. generally, anything that reduces THD20 also reduces IM.
Often the best answer to the question "What would happen if I did X?" is to reply "do X and record your results!"
In this case, the amplifier designed and built by @njswede is delightfully easy to try both ways: with, and without, current mirrors. Schematic fragment below.
Simply remove transistors Q3 and Q4, and solder a pair of wires into the holes for pins 1 and 3. Effectively turning the transistors into short-circuits from emitter to collector. Then remove resistor R3 and replace it with a suitable resistance which operates the input pair of transistors, and the subsequent Voltage Amplifier Stage, at the desired bias voltages and currents. Done! Now listen to the amplifier and decide for yourself, is the sound different? Is it suddenly "less grainy"? Did removing the current mirror, improve your enjoyment?
_
In this case, the amplifier designed and built by @njswede is delightfully easy to try both ways: with, and without, current mirrors. Schematic fragment below.
Simply remove transistors Q3 and Q4, and solder a pair of wires into the holes for pins 1 and 3. Effectively turning the transistors into short-circuits from emitter to collector. Then remove resistor R3 and replace it with a suitable resistance which operates the input pair of transistors, and the subsequent Voltage Amplifier Stage, at the desired bias voltages and currents. Done! Now listen to the amplifier and decide for yourself, is the sound different? Is it suddenly "less grainy"? Did removing the current mirror, improve your enjoyment?
_
Attachments
Hi
Just a clarification to my post regarding high-gain amps and their higher-order distortion harmonics. Because the feedback loop is so powerful, the level of these high harmonics will be vanishingly low, yet our hearing perceives them which might explain why some people do not like "the sound" of amps with high amounts of feedback.
I hope that makes sense?
Just a clarification to my post regarding high-gain amps and their higher-order distortion harmonics. Because the feedback loop is so powerful, the level of these high harmonics will be vanishingly low, yet our hearing perceives them which might explain why some people do not like "the sound" of amps with high amounts of feedback.
I hope that makes sense?
High levels of feedback require attention to stability in general and particularly what happens at clipping - it is easy to get oscillation, hanging and phase inversion.
Get that right and what you are left with is the lack of distortions that some people like as an effect.
Get that right and what you are left with is the lack of distortions that some people like as an effect.
I already test with and without CM as LTPs load years ago.
The answer is YES, CM make the sound hear more heavy/bold while without it everything become more relax, chill, neutral.
The price must pay that is i lost a lot of zeros THD. Um, who care?! Music for my ears.
I remember that Naim NAP, they are also didn't use CM as LTP load.
The answer is YES, CM make the sound hear more heavy/bold while without it everything become more relax, chill, neutral.
The price must pay that is i lost a lot of zeros THD. Um, who care?! Music for my ears.
I remember that Naim NAP, they are also didn't use CM as LTP load.
Attachments
What specifically about the CMs makes the sound “heavy/bold”? And what does that even mean?
That's for a whole amp, not a single transistor. Amps can be slow because they have to be stable in the face of multiple poles in the response, a single BJT doesn't need compensation to make it stable usually!
Just done a little simulation with this circuit:
Measuring the distortion in the current from the collector of Q2 at several frequencies:
1kHz -71dB 0.028%
10kHz -71dB 0.028%
100kHz. -70.5dB 0.03%
1MHz -59.5dB 0.1%
3MHz -50dB 0.3%
Clearly the current mirroring holds up well into the 100's of kHz range, with some drop in performance at 1MHz but not massive.
But in normal use such a current mirror is inside a global feedback loop and likely out-distorted by the output stage anyway, so we are basically limited in overall performance by output stage and open-loop gain whether a current mirror is used or not.
Measuring the distortion in the current from the collector of Q2 at several frequencies:
1kHz -71dB 0.028%
10kHz -71dB 0.028%
100kHz. -70.5dB 0.03%
1MHz -59.5dB 0.1%
3MHz -50dB 0.3%
Clearly the current mirroring holds up well into the 100's of kHz range, with some drop in performance at 1MHz but not massive.
But in normal use such a current mirror is inside a global feedback loop and likely out-distorted by the output stage anyway, so we are basically limited in overall performance by output stage and open-loop gain whether a current mirror is used or not.
@Mark Tillotson I’m in my 50s so my ears max out at about 12kHz, so I think I’m good. 🙂 A THD of 0.03% at 100kHz is still really good. Keep in mind that the second harmonic is at 200kHz for a 100kHz tone, which is very far outside the audible range.
I have no idea exactly what caused that. The theory of CMs pointed out that they have large dynamic impedance, doubling Gm, increase OLG.
However, my answer is enough for your question. It is not really ''detrimental effect'', just one of the ways to make your amp as you want.