I have tested the output stage in isolation initially after your previous reply, to finally find out that one of the output transistors had failed shorting collector-emitter. That helped!
Regarding testing without the output stages: I started off with what Self calls a small-signal class-A output stage (constant current) in his book (got the 6th edition). That helped already with getting some confidence that the input stage and VAS are at least doing something right.
Now that I got a nice present from @steveu in the form of a working circuit (.asc file), I'll continue from there with different driver devices.
Thanks!
@de Ocampo Thanks for the suggestion about the bias generator.
The basics work, and after seeing that it generates the correct range of voltage, I started to play around with letting it run at higher power for a few minutes.
Clearly the thermal compensation was not in place yet at all: I did not put the sensor transistor close to one of the drivers.
Then I put some wires on the sensor transistor so that I could put the two closer together. Nothing proper yet, but this already increases the thermal following of the bias generator.
I will share my circuit a bit later after doing some more testing, and after actually trying with BD139/140 drivers (think I'll need to order those...)
Thanks and cheers
The basics work, and after seeing that it generates the correct range of voltage, I started to play around with letting it run at higher power for a few minutes.
Clearly the thermal compensation was not in place yet at all: I did not put the sensor transistor close to one of the drivers.
Then I put some wires on the sensor transistor so that I could put the two closer together. Nothing proper yet, but this already increases the thermal following of the bias generator.
I will share my circuit a bit later after doing some more testing, and after actually trying with BD139/140 drivers (think I'll need to order those...)
Thanks and cheers
LPT and low level signal path stages should be as fast as is practical, ie ~300MHz. This is because every bit of phase lag is a threat to stability and must be hidden by lowering the dominant pole (a larger VAS capacitor) and reducing the slew rate. Yes, there are complicated multi-pole compensation schemes, but the foundation of stability is always a dominant pole that reduces the loop gain below 0dB before the poles of the LPT etc can add phase shift. When you are a math whiz with lots of experience, then you can push the boundaries of feedback compensation, but beginners should stick to the basics. Have a look at the phase margin simulations like the LTC examples:
C:\Users\yourname\Documents\LTspiceXVII\examples\Educational\
audioamp.asc
LoopGain.asc
LoopGain2.asc
C:\Users\yourname\Documents\LTspiceXVII\examples\Educational\
audioamp.asc
LoopGain.asc
LoopGain2.asc
Hi all,
First of all: Merry Christmas to those that celebrate it! Happy Humpday to the rest 🙂
Apologies for not posting any updates sooner. I have been busy with LTSpice, but also the actual circuit of course.
Before diving into details: I was able to acquire a pair of Boston A25 speakers second hand. But I did not want to hook them up right away, first I needed some DC offset protection. That's exactly what I built, including an output relay. Then it was time to play some music. Boy, those speakers sound just fine! So in short: the amplifier works. But I'm not done yet, of course.
Now for some things I did:
First thing was to replace the drivers with BD139 / 140. That basically solved the HF oscillations I saw in the simulation.
Then I started looking into impact of the LTP current (while keeping transconductance approximately constant by simulating the input stage in isolation) and VAS current on distortion. This resulted in selecting approx. 4 mA input tail current with 100 Ohm degeneration (giving about 8.8 mS transconductance), and 9 mA VAS current.
Then I started re-reading about setting the HF gain for stability. Just by assuming that 30 dB of NFB factor at 20 kHz should be stable (took this value from Self's book) and having a closed-loop gain of about 26 dB, the open-loop gain should be about 56 dB at 20 kHz. With a transconductance of 8.8 mS, I set Cdom to 100 pF.
I was also curious to actually see the open-loop gain in the circuit. Thus I connected my QA403's output to the amplifier's input, and buffered the input stage's inputs to feed them to the input of the QA403. Setting the 0 dBr to the output level (19 dBV output for -7 dBV input), it resulted in attached graph. It's not as smooth as in Self's book, but I was very happy to see at least some resemblance, and to see that open-loop gain at 20 kHz is indeed approx. 56 dB.
Last week I also created an output coil, but still need to include it in the amplifier. It should be around 2 uH. Before using it, I'd like to measure its actual inductance and think about an LC tank for that with a measured C (measure C by feeding a square wave to an RC and get the time constant from that). If anyone has any good tips on how to measure the L, or confirm that this is the way to go, that would be greatly appreciated.
Lastly, I learned to work a bit with the .four statement in LTSpice. After struggling with my simulation and by applying reduction, I discovered that the big trouble maker (after replacing the driver devices) was my simulated power supply. Cutting that out, I started getting reasonable (at least for me and my first amplifier, that is 🙂) distortion values. Then the next biggest impact on distortion I found is the input capacitor! Completely bypassing it in simulation gave the best results, but increasing it from 10 uF to 100 uF gave almost the same results as removing it. That is not a realistic thing, and I want to stick to the 10 uF WIMA MKS2 or MKS4 capacitors I have. Any thoughts on that?
When increasing the input capacitor to 100 uF in simulation the distortion based on the .four measurements for 1 kHz are: Partial Harmonic Distortion: 0.000647%, see attached log.
Next steps are to include the output coil (with 10 Ohm, 3 Watt damping resistor) and measure the THD with / without output relay into an 8 Ohm resistive load at different levels.
Also the current limiting circuitry still awaits implementation in the actual circuit.
Currently I have the Omron G2R-1A-E relay, and it works. But I am in doubt about its ability to disconnect in case of gross overload situations... For that reason I've just now ordered two of the Amplimo RL-24's. Those should have little influence on signal quality, while having the ability to disconnect larger DC currents.
When all of this is done and I'm happy with the result, I'll start working on the power supply. Have all the components waiting for me, so no dull moments this Christmas Holiday for me ;-)
Also looking forward to start designing a PCB for the amplifier. Still wondering if it makes sense to have it all on one big PCB (supply, amplifier, output protection) or that I will create separate boards for those three. Where current limiting will still be part of the amplifier of course.
Finally things are getting somewhere.
Sorry for the long read. I guess I was a bit hyped when I started writing this 🙂
Thanks for all the help so far!
/Paul
First of all: Merry Christmas to those that celebrate it! Happy Humpday to the rest 🙂
Apologies for not posting any updates sooner. I have been busy with LTSpice, but also the actual circuit of course.
Before diving into details: I was able to acquire a pair of Boston A25 speakers second hand. But I did not want to hook them up right away, first I needed some DC offset protection. That's exactly what I built, including an output relay. Then it was time to play some music. Boy, those speakers sound just fine! So in short: the amplifier works. But I'm not done yet, of course.
Now for some things I did:
First thing was to replace the drivers with BD139 / 140. That basically solved the HF oscillations I saw in the simulation.
Then I started looking into impact of the LTP current (while keeping transconductance approximately constant by simulating the input stage in isolation) and VAS current on distortion. This resulted in selecting approx. 4 mA input tail current with 100 Ohm degeneration (giving about 8.8 mS transconductance), and 9 mA VAS current.
Then I started re-reading about setting the HF gain for stability. Just by assuming that 30 dB of NFB factor at 20 kHz should be stable (took this value from Self's book) and having a closed-loop gain of about 26 dB, the open-loop gain should be about 56 dB at 20 kHz. With a transconductance of 8.8 mS, I set Cdom to 100 pF.
I was also curious to actually see the open-loop gain in the circuit. Thus I connected my QA403's output to the amplifier's input, and buffered the input stage's inputs to feed them to the input of the QA403. Setting the 0 dBr to the output level (19 dBV output for -7 dBV input), it resulted in attached graph. It's not as smooth as in Self's book, but I was very happy to see at least some resemblance, and to see that open-loop gain at 20 kHz is indeed approx. 56 dB.
Last week I also created an output coil, but still need to include it in the amplifier. It should be around 2 uH. Before using it, I'd like to measure its actual inductance and think about an LC tank for that with a measured C (measure C by feeding a square wave to an RC and get the time constant from that). If anyone has any good tips on how to measure the L, or confirm that this is the way to go, that would be greatly appreciated.
Lastly, I learned to work a bit with the .four statement in LTSpice. After struggling with my simulation and by applying reduction, I discovered that the big trouble maker (after replacing the driver devices) was my simulated power supply. Cutting that out, I started getting reasonable (at least for me and my first amplifier, that is 🙂) distortion values. Then the next biggest impact on distortion I found is the input capacitor! Completely bypassing it in simulation gave the best results, but increasing it from 10 uF to 100 uF gave almost the same results as removing it. That is not a realistic thing, and I want to stick to the 10 uF WIMA MKS2 or MKS4 capacitors I have. Any thoughts on that?
When increasing the input capacitor to 100 uF in simulation the distortion based on the .four measurements for 1 kHz are: Partial Harmonic Distortion: 0.000647%, see attached log.
Next steps are to include the output coil (with 10 Ohm, 3 Watt damping resistor) and measure the THD with / without output relay into an 8 Ohm resistive load at different levels.
Also the current limiting circuitry still awaits implementation in the actual circuit.
Currently I have the Omron G2R-1A-E relay, and it works. But I am in doubt about its ability to disconnect in case of gross overload situations... For that reason I've just now ordered two of the Amplimo RL-24's. Those should have little influence on signal quality, while having the ability to disconnect larger DC currents.
When all of this is done and I'm happy with the result, I'll start working on the power supply. Have all the components waiting for me, so no dull moments this Christmas Holiday for me ;-)
Also looking forward to start designing a PCB for the amplifier. Still wondering if it makes sense to have it all on one big PCB (supply, amplifier, output protection) or that I will create separate boards for those three. Where current limiting will still be part of the amplifier of course.
Finally things are getting somewhere.
Sorry for the long read. I guess I was a bit hyped when I started writing this 🙂
Thanks for all the help so far!
/Paul
Just measured THD to be 0.01418% at 1 kHz, at 8 Watt into 8 Ohm resistive load.
Measured this at the NFB take-off point, before the output coil and relay (but with those present).
My guess would be that my circuit and its ground is too messy to get much better results right now.
Measured this at the NFB take-off point, before the output coil and relay (but with those present).
My guess would be that my circuit and its ground is too messy to get much better results right now.
Congratulations on good results.
100 uf as an input capacitor to a common emitter input transistor is rediculous. HD below 0.2% is rediculous. I cannot hear above that on speakers, and my speakers have HD 25 db down 60hz to 12000 hz at 5 watts. I tried once my CS800s amp in the full differential mode with no input capacitor and direct outputs, versus an Apex AX6 with capacitors in and out. Same wattage out, less than 1. The CS800s was worse because the cables picked up some hum. Back to RCA connector coax, with the input capacitor.
Peavey uses 15 turns 16 gauge solid wire wound around a china marker or AA battery, for output inductor. They call that 800 uH.
The inductor measuring meters all come from ***** and are very likely garbage. There is a B&K for $700, and the $120 CH150 from an oriental source I do trust disappeared the month I had the money saved for it. It should be possible to measure inductance by measuring phase shift of voltage across a series resistor from the voltage across the inductor on a 2 channel scope.
I do not trust any AC rated relay to break 1000s of amps of short circuit current. Unless salvaged from an amp with a good reputation, like Peavey, Crown, or QSC. Once I have the amp hulk, why not fix it instead of building something? The M-2600 I am listening to now has no DC on output protection, but I like the low wattage (70/ch) and no fan. I tried 2500 uF capacitor series output as DC protection Saturday, and it sounded okay, but the sound stopped after 30 minutes. Probably the thermal switch opened. Okay after I put the wire only cable back on, about a 4 minute break. Apparently 800 uH is not enough to prevent oscillation running into a 2500 uF load. Will try to build a R J Keene nfet protection circuit next. Two series low resistance nfets series the output S to S. I have a dozen .08 ohm nfets and 12 APV1122 gate drivers. Lack the 7 v bipolar breakover device (diac), the flip flop to remember that a fault occurred until power off reset, the red/green LED to indicate fault, the DIP sockets.
100 uf as an input capacitor to a common emitter input transistor is rediculous. HD below 0.2% is rediculous. I cannot hear above that on speakers, and my speakers have HD 25 db down 60hz to 12000 hz at 5 watts. I tried once my CS800s amp in the full differential mode with no input capacitor and direct outputs, versus an Apex AX6 with capacitors in and out. Same wattage out, less than 1. The CS800s was worse because the cables picked up some hum. Back to RCA connector coax, with the input capacitor.
Peavey uses 15 turns 16 gauge solid wire wound around a china marker or AA battery, for output inductor. They call that 800 uH.
The inductor measuring meters all come from ***** and are very likely garbage. There is a B&K for $700, and the $120 CH150 from an oriental source I do trust disappeared the month I had the money saved for it. It should be possible to measure inductance by measuring phase shift of voltage across a series resistor from the voltage across the inductor on a 2 channel scope.
I do not trust any AC rated relay to break 1000s of amps of short circuit current. Unless salvaged from an amp with a good reputation, like Peavey, Crown, or QSC. Once I have the amp hulk, why not fix it instead of building something? The M-2600 I am listening to now has no DC on output protection, but I like the low wattage (70/ch) and no fan. I tried 2500 uF capacitor series output as DC protection Saturday, and it sounded okay, but the sound stopped after 30 minutes. Probably the thermal switch opened. Okay after I put the wire only cable back on, about a 4 minute break. Apparently 800 uH is not enough to prevent oscillation running into a 2500 uF load. Will try to build a R J Keene nfet protection circuit next. Two series low resistance nfets series the output S to S. I have a dozen .08 ohm nfets and 12 APV1122 gate drivers. Lack the 7 v bipolar breakover device (diac), the flip flop to remember that a fault occurred until power off reset, the red/green LED to indicate fault, the DIP sockets.
Last edited: