Thank you Carlos, for your inspiration... In my testing I noticed that the circuit actually added quite abit of 2nd harmonic distoriton at high frequencies. Perhaps this is why the Harpsichords "sparkled"!
I do feel as if your view on audio lacks some objectivity. But I will not judge. I like having you around, as do I Lineup and pretty much any DIY'er.
Thank you Paul, your information on this bias scheme is useful and appreciated.
When first designing a circuit with this output I drive it between the emitters of the bias transistors as you show. Simulating this gave me a THD of around .001% at 2V pk-pk, 1KHZ. Now that you say this, I wonder why I didn't just stick with that. Thank you!
I will think about this some more. I was convinced this circuit wasn't suitable for class B operation but apparently it can be done. However a 1k resistor will limit the amount of output power.
Much appreciated everyone!
- keantoken
I do feel as if your view on audio lacks some objectivity. But I will not judge. I like having you around, as do I Lineup and pretty much any DIY'er.
PaulBysouth said:
Thank you Paul, your information on this bias scheme is useful and appreciated.
When first designing a circuit with this output I drive it between the emitters of the bias transistors as you show. Simulating this gave me a THD of around .001% at 2V pk-pk, 1KHZ. Now that you say this, I wonder why I didn't just stick with that. Thank you!
I will think about this some more. I was convinced this circuit wasn't suitable for class B operation but apparently it can be done. However a 1k resistor will limit the amount of output power.
Much appreciated everyone!
- keantoken
Keantoken,
Just to be clear, the frequency compensation resistors and capacitors go (referring to your circuit in Post #35) ...
- resistor between the emitter of Q1 and the base of Q3 (let's call it Rb), and
- a capacitor between the base of Q3 and the base of Q10 (let's call it Ca),
- similarly for the lower half (Q2, Q4, Q11).
As you are current driving the collectors of Q3,Q4 use a value for Rb in the range 100 to 1kohms. A low value will cause current limiting, a high value will allow operation into Class B (with an increase in distortion). Ca will probably be around 100pF.
If you were to drive the circuit between the emitters of Q3 and Q4, then Rb needs to be larger to prevent excess base current into Q3,Q4 (typically 1k or larger). Large Rb values will allow the output to go into Class B.
Adding Rb will also increase the bias current in Q1,Q2 (due to the voltage drop across Ra caused by the base current). So you may need to increase R5,R8 to 12 ohms.
I'm planning to build a variation of the Nelson Pass F5 with an Allison biased output stage and lateral Mosfets, but its about 3 or 4 amplifiers down the list so it won't be till late 2009.
Paul Bysouth Feb, 2009.
Just to be clear, the frequency compensation resistors and capacitors go (referring to your circuit in Post #35) ...
- resistor between the emitter of Q1 and the base of Q3 (let's call it Rb), and
- a capacitor between the base of Q3 and the base of Q10 (let's call it Ca),
- similarly for the lower half (Q2, Q4, Q11).
As you are current driving the collectors of Q3,Q4 use a value for Rb in the range 100 to 1kohms. A low value will cause current limiting, a high value will allow operation into Class B (with an increase in distortion). Ca will probably be around 100pF.
If you were to drive the circuit between the emitters of Q3 and Q4, then Rb needs to be larger to prevent excess base current into Q3,Q4 (typically 1k or larger). Large Rb values will allow the output to go into Class B.
Adding Rb will also increase the bias current in Q1,Q2 (due to the voltage drop across Ra caused by the base current). So you may need to increase R5,R8 to 12 ohms.
I'm planning to build a variation of the Nelson Pass F5 with an Allison biased output stage and lateral Mosfets, but its about 3 or 4 amplifiers down the list so it won't be till late 2009.
Paul Bysouth Feb, 2009.
Thank you for more info Paul, I discovered something myself about base current however which I detail below.
Okay, ever since I started this amplifier, something's been going wrong.
I finally figured it out!
It was always trippin'!
Producing random noise. Figured it out!
I'm surprised no one has pointed this out to me. Referring to the attachment in post 35, there is only about .6V across Q4's collector. In my simulations and prototypes, B-C voltage is so low that it shuts off the B-C junction and huge base current ensues. In the schematic below is my current fix. We can also just delete the diodes and connect Q6 and Q5's emitters to the collectors of Q1 and Q2. Pretty much the same thing.
Now it is completely silent and stable and sounds fine, although it still doesn't have the sparkle it did a few tests ago. I suspect that the 2nd harmonic distortion was a result of my asymmetrical layout, with parasitics distributed unevenly. Now everything is symmetrical. However, that might not be it. Before, the LTP was horrendously unbalanced. that could most definitely be the source of H2.
Anyways, right now I know there is some kind of unpleasant distortion, even though the sound is mostly transparent.
I am having a bad problem right now. My oscilloscope only shows signs of slight oscillation, but the heatsinks warm up fast. Plus, I'm getting high off the fumes.
My o'scope is only a tek 561B and the diff. amp only has bandwidth of 300KHz. So I do have a problem.
Not sure how to fix this, but I will try the compensation caps that Paul suggests and see what happens.
- keantoken
Okay, ever since I started this amplifier, something's been going wrong.
I finally figured it out!
It was always trippin'!
Producing random noise. Figured it out!
I'm surprised no one has pointed this out to me. Referring to the attachment in post 35, there is only about .6V across Q4's collector. In my simulations and prototypes, B-C voltage is so low that it shuts off the B-C junction and huge base current ensues. In the schematic below is my current fix. We can also just delete the diodes and connect Q6 and Q5's emitters to the collectors of Q1 and Q2. Pretty much the same thing.
Now it is completely silent and stable and sounds fine, although it still doesn't have the sparkle it did a few tests ago. I suspect that the 2nd harmonic distortion was a result of my asymmetrical layout, with parasitics distributed unevenly. Now everything is symmetrical. However, that might not be it. Before, the LTP was horrendously unbalanced. that could most definitely be the source of H2.
Anyways, right now I know there is some kind of unpleasant distortion, even though the sound is mostly transparent.
I am having a bad problem right now. My oscilloscope only shows signs of slight oscillation, but the heatsinks warm up fast. Plus, I'm getting high off the fumes.
My o'scope is only a tek 561B and the diff. amp only has bandwidth of 300KHz. So I do have a problem.
Not sure how to fix this, but I will try the compensation caps that Paul suggests and see what happens.
- keantoken
Attachments
AndrewT said:
If I understand correctly then the Vbe's of the bias transistors will always be almost equal if we don't take into account the effect of Vcb, which you didn't mention.
The proportionality of the bias resistors' currents will be somewhat nonlinear because of transistor nonlinearities, while the bias transistors are adjusting the quiescent current in realtime.
I don't understand how Vbe can be different for the bias transistors.
- keantoken
Have you ever played with the Rush Cascode? It is the same circuit and concept as these biasing transistors.
I don't think the problem you are describing is the problem I am having. The problem I once had, as far as I am aware, is that the Vcb of the bias transistors became so low that they turned off during output swing. I have fixed that problem with the diodes in the schematic I posted.
I may actually be fooling myself and the heatsinks are as warm as they always have been. I need larger heatsinks but that will have to wait.
Could you describe why you think the Vbe's of the bias transistors will be unequal?
- keantoken
I don't think the problem you are describing is the problem I am having. The problem I once had, as far as I am aware, is that the Vcb of the bias transistors became so low that they turned off during output swing. I have fixed that problem with the diodes in the schematic I posted.
I may actually be fooling myself and the heatsinks are as warm as they always have been. I need larger heatsinks but that will have to wait.
Could you describe why you think the Vbe's of the bias transistors will be unequal?
- keantoken
All those folks are very strange, well, we cannot trow too much stones.....
when our own roof is made of glass.... there are others more strange then those exotic guys into our community... i am one of them with my pictures.... but as i work hard and use to cooperate with others, some folks tolerates me.
We have dogs too Andy....... or Dog picture at least.... well... dog is a very good thing... a special friend we can have.
Lumba is faceless.... maybe his dream was to be a computer terminal or something alike... computers are becaming very clever those last days.... it is strange that Lumba has no picture... i found him a very sensistive guy despite all darkness and cold weather of his country he seems warm.
Yes... i am strange too... ahahahha folk i am.
Be happy.... live problems will still trying to invade our shield...but they hate guys that laugh.... smile folks...it is a problem vaporizer.
More ridiculous is a corn can with this brand.... Olé!.. imagine all that imagination to decide for a name... imagine an amplifier with this name.... Olé Amplifier....ahahahahahah!
Yes... a clever marketing strategy...no one will forget such a ridiculous name.,, good to fast things, scooter, bicicles, sport cars, football, and human phisical competionos...not Corn!
regards,
Carlos
when our own roof is made of glass.... there are others more strange then those exotic guys into our community... i am one of them with my pictures.... but as i work hard and use to cooperate with others, some folks tolerates me.
We have dogs too Andy....... or Dog picture at least.... well... dog is a very good thing... a special friend we can have.
Lumba is faceless.... maybe his dream was to be a computer terminal or something alike... computers are becaming very clever those last days.... it is strange that Lumba has no picture... i found him a very sensistive guy despite all darkness and cold weather of his country he seems warm.
Yes... i am strange too... ahahahha folk i am.
Be happy.... live problems will still trying to invade our shield...but they hate guys that laugh.... smile folks...it is a problem vaporizer.
More ridiculous is a corn can with this brand.... Olé!.. imagine all that imagination to decide for a name... imagine an amplifier with this name.... Olé Amplifier....ahahahahahah!
Yes... a clever marketing strategy...no one will forget such a ridiculous name.,, good to fast things, scooter, bicicles, sport cars, football, and human phisical competionos...not Corn!
regards,
Carlos
Attachments
The blue one, the amplifier at the picture's rigth side
Is the Dx Amplifier.
It is there because does not produce fatigue when playing... and goes playing low volume all day long... switched on 5 hours morning and off 10 PM.... i do not feel tired listening.
It is simple... diferential, voltage amplifier and output alike a darlington.... no CCS, no current sink... no mirror... very simple... audio level does not interfere with bias... it remains stable into 100 miliamperes... supply well filtered, input rail voltage stable and all transistors conducting having VBE around 600 milivolts.
I have others that plays better treble...better sound...but this one does not produce fatigue...i do not feel tired.
I think, because simple, despite do not measure so good as others, sounds nice....not better than others.... more confortable, more friendly, more warm than many others.
Because simple.
Carlos
Is the Dx Amplifier.
It is there because does not produce fatigue when playing... and goes playing low volume all day long... switched on 5 hours morning and off 10 PM.... i do not feel tired listening.
It is simple... diferential, voltage amplifier and output alike a darlington.... no CCS, no current sink... no mirror... very simple... audio level does not interfere with bias... it remains stable into 100 miliamperes... supply well filtered, input rail voltage stable and all transistors conducting having VBE around 600 milivolts.
I have others that plays better treble...better sound...but this one does not produce fatigue...i do not feel tired.
I think, because simple, despite do not measure so good as others, sounds nice....not better than others.... more confortable, more friendly, more warm than many others.
Because simple.
Carlos
Attachments
keanton,
I have a few comments on the last circuit you posted.
I would replace the Rush cascode with a more conventional Vbe multiplier. While the Rush circuit is useful in some applications, a Vbe multiplier is better suited to maintain the proper bias for a class AB amplifier. You can also eliminate D1 and D2.
You need to put emitter resistors in each individual transistor of the output devices. Delete R5 and R6. The purpose of the emitter resistors is to help the paralleled devices share the current. In the real word (vice simulation) these devices will not be perfectly matched and they may not share the current very well. BTW, these resistors are a form of negative feedback. In general, negative feedback reduces the sensitivity of the circuit to variations in the parameters of the active devices.
What are R2 and C5 for?
Rick
I have a few comments on the last circuit you posted.
I would replace the Rush cascode with a more conventional Vbe multiplier. While the Rush circuit is useful in some applications, a Vbe multiplier is better suited to maintain the proper bias for a class AB amplifier. You can also eliminate D1 and D2.
You need to put emitter resistors in each individual transistor of the output devices. Delete R5 and R6. The purpose of the emitter resistors is to help the paralleled devices share the current. In the real word (vice simulation) these devices will not be perfectly matched and they may not share the current very well. BTW, these resistors are a form of negative feedback. In general, negative feedback reduces the sensitivity of the circuit to variations in the parameters of the active devices.
What are R2 and C5 for?
Rick
sawreyrw said:
I can eliminate D1 and D2 without doing away with the Allison multiplier, if I connect the emitters of Q6 and Q5 to the bases of Q3 and Q4. Personally I still want to stay with the Allison (or Rush) multiplier as it makes the circuit unique.
Emitter resistors will be added when I build the official circuit, but I haven't been able to use them in real life because I don't have any.
R2 and C5 help the Allison circuit if you're not driving it from the emitters. It causes the CCS to act as a push-pull counterpart to side-step the relatively high output impedance of the Allison circuit. It doesn't improve distortion much at small signals but it helps when it operates at its max output.
The best way to drive the Allison is between the emitters, and this offers the lowest distortion.
I have discovered a good technique for reducing the size of R5 and R6, and so deleting them should not be necessary. This idea is in the attached schematic, which I will test out next, if I am able to. I would solder this instead of breadboarding, but I don't trust myself with solder yet. Breadboarding is proving annoying and confusing at the least. I suppose I will get used to it.
- keantoken
Attachments
The Allison multiplier is not dead!
Sander Sassen's ExtremA uses it, apparently to practical effect.
Official schematic here:
http://www.hardwareanalysis.com/content/image/11997/
Sassen's thread here:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=96853
With this I think the Allison multiplier is perfectly fine to use in a class A amplifier.
- keantoken
Sander Sassen's ExtremA uses it, apparently to practical effect.
Official schematic here:
http://www.hardwareanalysis.com/content/image/11997/
Sassen's thread here:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=96853
With this I think the Allison multiplier is perfectly fine to use in a class A amplifier.
- keantoken
I got the output stage to my latest schematic working.
It seems that the voltage across the bias resistor keeps going up by 100mV every time I turn the thing on.
Everything is normal, however.
EDIT: I just built the LTP portion. Wow, it put out a pure 1.5MHz sinewave! None of my circuits have looked that good whilst in oscillation!
I was only able to detect this with my oscilloscope's highest sweeprate setting. This probably means I did something wrong building it, because if it was unstable enough to oscillate I would expect a higher frequency.
EDIT2: I just now plugged it in to test it out anyways while it was oscillating. It sounds really clean and smooth, sooooooo much better than my previous circuits (it seems I have constructed a fine oscillator indeed - oscillation does not modulate with the audio!). This is with Enya however. Once it cools down again I will test it with some electronic stuff.
There is still weird artefacts, probably a B-C junction shutting off somewhere in the Allison or associated circuitry. I can probably fix this with a diode.
- keantoken
It seems that the voltage across the bias resistor keeps going up by 100mV every time I turn the thing on.
Everything is normal, however.
EDIT: I just built the LTP portion. Wow, it put out a pure 1.5MHz sinewave! None of my circuits have looked that good whilst in oscillation!
I was only able to detect this with my oscilloscope's highest sweeprate setting. This probably means I did something wrong building it, because if it was unstable enough to oscillate I would expect a higher frequency.
EDIT2: I just now plugged it in to test it out anyways while it was oscillating. It sounds really clean and smooth, sooooooo much better than my previous circuits (it seems I have constructed a fine oscillator indeed - oscillation does not modulate with the audio!). This is with Enya however. Once it cools down again I will test it with some electronic stuff.
There is still weird artefacts, probably a B-C junction shutting off somewhere in the Allison or associated circuitry. I can probably fix this with a diode.
- keantoken
Okay, I've got it to stop oscillating. Listened to it and it sounds decent, will listen more.
I'm in dire need of larger heatsinks, I will try to drill holes in a good one I have with a hand drill but no guarantees. :/
I would describe the sound as very relaxed, not very bright.
Whereas my old design was very unstable and very hard to get working and had strange problems, this one has given me no trouble.
- keantoken
I'm in dire need of larger heatsinks, I will try to drill holes in a good one I have with a hand drill but no guarantees. :/
I would describe the sound as very relaxed, not very bright.
Whereas my old design was very unstable and very hard to get working and had strange problems, this one has given me no trouble.
- keantoken
I didn't notice any difference when it was oscillating, from when it was not.
I managed to drill two holes in a heatsink I have and so it can handle longer listening times.
I'll say now that it doesn't sound as clear as older versions. Perhaps this has to do with something in the LTP. We shall see. There seem to be plenty of unpleasant harmonics. Hopefully it's just glitching.
I don't really have any good reference CDs, unfortunately.
I suspect bias current is a bit high. I shall switch to 2.5 ohm bias resistors and see if I can run it cooler.
- keantoken
I managed to drill two holes in a heatsink I have and so it can handle longer listening times.
I'll say now that it doesn't sound as clear as older versions. Perhaps this has to do with something in the LTP. We shall see. There seem to be plenty of unpleasant harmonics. Hopefully it's just glitching.
I don't really have any good reference CDs, unfortunately.
I suspect bias current is a bit high. I shall switch to 2.5 ohm bias resistors and see if I can run it cooler.
- keantoken
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.