The first amp that I built, class A, of course had a extremely high slew rate due to no global feedback, and the sound quality of if it is nice and crisp (highs can be clearly head and are VERY fast).
However, with the class AB amp I just finished building, the slew rate was very slow at first (about 1.6V/us when applying a 20khz square wave for 10V at the output). You can clearly here this at the amp output as the sound sounded a lot more muted and sluggish than my class A.
With a few component value tweaks, I was able to increase the slew rate to about 3.3V/us but at the cost of some stability. Any further tweaks I do such as decreasing cdom, increasing the VAS or diff pair current sources, causes the amp to become unstable. I can also put the same values in spice and it's also unstable, so it looks the same in simulation land and on the real PCB.
Where are the common places you can increase slew rate while maintaining stability? Standard AB amp with CFP outputs.
I have the stability cap on the negative rail of the CFP stage, was 220pf, tried to reduce to 18p and was unstable so I had to put it back to 220pf.
However, with the class AB amp I just finished building, the slew rate was very slow at first (about 1.6V/us when applying a 20khz square wave for 10V at the output). You can clearly here this at the amp output as the sound sounded a lot more muted and sluggish than my class A.
With a few component value tweaks, I was able to increase the slew rate to about 3.3V/us but at the cost of some stability. Any further tweaks I do such as decreasing cdom, increasing the VAS or diff pair current sources, causes the amp to become unstable. I can also put the same values in spice and it's also unstable, so it looks the same in simulation land and on the real PCB.
Where are the common places you can increase slew rate while maintaining stability? Standard AB amp with CFP outputs.
I have the stability cap on the negative rail of the CFP stage, was 220pf, tried to reduce to 18p and was unstable so I had to put it back to 220pf.
3.3 V/µs with a bit of a following wind for a power amp, that doesn't sound too exciting indeed.
Please show an accurate schematic.
My guess would be that you may be using an undegenerated bipolar input stage, with resulting LTP current a bit on the low side. This then gets you into slew rate trouble with the Miller cap (things are not symmetrical, I assume?). Could it be that your output stage is a tad on the slow side?
Please show an accurate schematic.
My guess would be that you may be using an undegenerated bipolar input stage, with resulting LTP current a bit on the low side. This then gets you into slew rate trouble with the Miller cap (things are not symmetrical, I assume?). Could it be that your output stage is a tad on the slow side?
Don't have schematic on me, I apologize, but will post when I get home.
I don't have degeneration resistors at on the input pair, but do have them on the current source load.
I have tried increasing the current at the input stage, but that causes stability issues WITHOUT increasing the slew rate at all, so I don't think that is the issue. Increasing the VAS current does increase the slew rate considerability though, but that is at the cost of stability, and past a certain point it does nothing.
I don't believe that you will ever find that a high slew rate improves the sound of an amplifier. I tend to see a lot of talk about changing operational amplifiers and devices to increase slew rates. The problem is that the people involved are not fully aware that the fastest settling time obtainable from full 0dBU attenuation to nil with Redbook 16Bit/44.1kHz playback is 22uS, equivalent to a 0.85V/uS slew rate. For 24Bit/192kHz playback, it is 3.5V/uS and corresponds to a frequency about four octaves higher than the human hearing extends. Musical audio and even theatrical film audio does not make use of the full conversion rate because it would only be useful to recreate near-perfect square waves spanning the full >92dB dynamic range of digital audio, which are not components of any audio recording. Instead, we use a much smaller dynamic range in recorded audio and thus the rise time and slew rate of any recorded fast transients are reduced, too. Personally, if the amp sounds sluggish, I would be looking elsewhere.
Schematic:
Open loop response:
Step response (0 to 2V input step, 1ps rise time):
An externally hosted image should be here but it was not working when we last tested it.
Open loop response:
An externally hosted image should be here but it was not working when we last tested it.
Step response (0 to 2V input step, 1ps rise time):
An externally hosted image should be here but it was not working when we last tested it.
Hello Fusion.
Your stages are weak biased now.
To increase the slew rate you need to increase the dv/dt or I/C capability of the voltage amplifier stage .This means to increase the current which flows into the Miller capacitor and also to decrease this cap to mantain it stable.You should decrease R1 to 330 ohms to increase the current through the differential stage.ALso insert degeneration for Q2 ,an emitter resistor about 1kohm ,the same thing for Q3 .After this,you can decrease the C7 to less then 33pF .Doing this you will get a more than 60V/us slew rate .
The current through Q4 is very small in your schematic ~1 mA !You need to increase it to at least 5 mA .You can do this by decreasing R3 to 120-150 ohms .
What supply voltage you have choosed ?
Slew rate test is accurate when your outputs are full swing .
Your stages are weak biased now.
To increase the slew rate you need to increase the dv/dt or I/C capability of the voltage amplifier stage .This means to increase the current which flows into the Miller capacitor and also to decrease this cap to mantain it stable.You should decrease R1 to 330 ohms to increase the current through the differential stage.ALso insert degeneration for Q2 ,an emitter resistor about 1kohm ,the same thing for Q3 .After this,you can decrease the C7 to less then 33pF .Doing this you will get a more than 60V/us slew rate .
The current through Q4 is very small in your schematic ~1 mA !You need to increase it to at least 5 mA .You can do this by decreasing R3 to 120-150 ohms .
What supply voltage you have choosed ?
Slew rate test is accurate when your outputs are full swing .
Last edited:
Thanks, I'll give these suggestions a try and report back.
I can't really increase the current through Q4 too much unless I use some kind of vbe multiplier instead of the simple diodes due to it over biasing the output stage.
Nominal supply is around +/- 27V.
After tweaking values, adding the degeneration resistors, and changing the diodes to a vbe multiplier (can't have 5mA on Q4 with diodes and bias the output devices correctly), I can still only achieve about ~20V/us. Any suggestions or improvements? Or should I just go with 20V/us?
Schematic (ignore the open loop test module):
Open loop gain:
Schematic (ignore the open loop test module):
An externally hosted image should be here but it was not working when we last tested it.
Open loop gain:
An externally hosted image should be here but it was not working when we last tested it.
Believe it or not, its needed for stability. Bringing it down to ~100pf kills it.
20 V/µs sounds good enough to me. You need about 5 for 20 kHz at 100 W / 8 ohms, and I assume this won't be a super high power amp.
So you have both LTP and VAS currents at ~4 mA now? IMO you can ease up on the LTP a little, 1 mA usually is just fine. (Some amps run them at <100 µA when undegenerated in order to reduce input impedance distortion from nonlinear base current.) By contrast, 4 mA for the VAS seems a little tight - you don't have an EF3 output stage there. Anyway, the broadly similar ESP P3a doesn't use more than 5 mA either, but still I wouldn't expect any miracles into 4 ohm loads.
I have found in simulation that a Darlington VAS can introduce a fair bit of common-mode distortion that only goes away once you replace the buffer's emitter resistor by a current source - should be mostly a non-issue at 26 dB of gain though. This becomes more relevant in headphone amps.
What sort of output stage idle current are you getting? Optimum bias would seem to work out to 12 mA or so.
So you have both LTP and VAS currents at ~4 mA now? IMO you can ease up on the LTP a little, 1 mA usually is just fine. (Some amps run them at <100 µA when undegenerated in order to reduce input impedance distortion from nonlinear base current.) By contrast, 4 mA for the VAS seems a little tight - you don't have an EF3 output stage there. Anyway, the broadly similar ESP P3a doesn't use more than 5 mA either, but still I wouldn't expect any miracles into 4 ohm loads.
I have found in simulation that a Darlington VAS can introduce a fair bit of common-mode distortion that only goes away once you replace the buffer's emitter resistor by a current source - should be mostly a non-issue at 26 dB of gain though. This becomes more relevant in headphone amps.
What sort of output stage idle current are you getting? Optimum bias would seem to work out to 12 mA or so.
Last edited:
~25V rail, so for 8 ohm load about ~80W peak, ~55W rms.
I suppose I could bump up the VAS current a bit, I'll mess around with the LTP's current value a bit to see what difference it makes.
About 50mA, when I actually build the amp though I plan to dial the current down to get a specific Vre though (about 5-10mV).
Let's say you've got a very stiff supply and still manage 48 Vpp at the output with a bit of a following wind. (That's some fairly big Ifs.) That's 17 Vrms, or Vout²/Rload = 35 W into 8 ohms. Let the rails sag a bit, and you're closer to 25-30 W.
P3a is given with 60 W / 8 ohms typical when used with +/-35 V rails (that's +/- 35 V quiescent).
P3a is given with 60 W / 8 ohms typical when used with +/-35 V rails (that's +/- 35 V quiescent).
Fusion ,if you want a higher slew rate than 20V/us you need to decrease also the C8 from 150 pF to 68-39pF for example .Also it is better to connect the Miller cap not in base of Q9 but in emitter of Q2 .If you have issues with instability increase the differential emitter resistors to 1Kohm
I have tried to decrease it, but the lower it gets there a zero that causes peaking at high frequency that moves the gain near or past the 0db line when the phase is way past 180 degrees. I obviously don't know the exact mechanism at play here, but I simply could not get rid of this.
????
Is this a typo?
I have bias issues with the emitter resistors too high.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.