I think Andrew is right when he mentions sliding bias schemes, although of course, there's a myriad of them out there, and some are probably very good.
However, I still prefer the classic approach, let the output stage run at a fair bias current, say 100-130 mA per device, and you will be hearing a smoother sound without the full class A penalty. Simpler is somehow always better.
However, I still prefer the classic approach, let the output stage run at a fair bias current, say 100-130 mA per device, and you will be hearing a smoother sound without the full class A penalty. Simpler is somehow always better.
I've been thinking about this dynamic bias stuff lately and fooling around on the simulator, thought maybe I was on to something, and it was quite funny that an almost identical schematic appeared in the last issue of Linear Audio.
Anyway, I have measured the voltage on my speakers which have reasonable efficiency (~90dB)... 1W is already pretty loud... so with a standard AB output stage biased a bit hot, it will most likely spend lots of time in class A anyway.
Also with these schemes it is easy to introduce more gremlins that it is supposed to remove...
Anyway, I have measured the voltage on my speakers which have reasonable efficiency (~90dB)... 1W is already pretty loud... so with a standard AB output stage biased a bit hot, it will most likely spend lots of time in class A anyway.
Also with these schemes it is easy to introduce more gremlins that it is supposed to remove...
We looked at this sometime ago . Without me paying much attention the use of triples seemed to offer the better way of getting the optimum amount of sensible class A power . Especially as in normal triples there would be spare driving current which would drive 6 output devices per channel .
The Quad 303 is happy at between 5 and 10 mA standing current . It was explained to me that it was a defect of the design that it has to sit so low . I must try 130 mA to see where it goes . My feeling is Dvv is right and and it will be enough to mimic class A . More so if multiple devices . Perhaps less good if somewhere between 10 to 100 mA perhaps ? Douglas Self calls this the gain doubling region . If will happen at 130 mA . The excuse being it is already loud so ears will be the source of distortion more than the amp ?
Coincidentally 100 mA is ideal bias for Exicon MOSFET's . Above that and the Ron is too high . It is easy to hear the looseness of the bass if too much used . Naturally there is no Ron with bipolar transistors . The opposite in fact . 100 mA is said to be the tipping point where Ron incenses in Audio MOS FET's .
Dvv . I notice you use 0R47 ( I think so ) . This almost puts the devices in MOSFET country . Self says 0R1 . 0R22 seems a safe choice .
I noticed class D protection works by Drain to Source voltage in theoretical papers . This could be used in class B MOSFET . Less so in bipolar as the losses are small . Even so a comparator might be able to do it .
The Quad should tolerate 130 mA for long enough to do some tests . I hope to hear the same sound with more low level colour . The Quad has out of this world measurements for a 1967 amp . The very low bias I think does something mildly like MP3 . What it doesn't do it doesn't show without reference to better amps . I have been running the Hypex Ucd 180 at home both on Hypex SMPS and conventional PSU . I have to say it is close run with the Quad and perhaps the Quad is better ? I listened to an old Arcron DC 300 the other day . Whilst it was more filtered than Quad or Hypex it seemed rather wonderful . The Hypex sounds different on the PSU's and measure slightly nicer on conventional . It is not all that way and Hypex should be commended for how well the SMPS works . I would rather like to try the IR IRS 2092 . On paper it looks good . This has made me think a Quad 303 with 6 outputs might be my cup of tea .
I wonder if anyone has used light bulbs as emitter resistors ? Might be OK as it would be in the feedback loop .
The Quad 303 is happy at between 5 and 10 mA standing current . It was explained to me that it was a defect of the design that it has to sit so low . I must try 130 mA to see where it goes . My feeling is Dvv is right and and it will be enough to mimic class A . More so if multiple devices . Perhaps less good if somewhere between 10 to 100 mA perhaps ? Douglas Self calls this the gain doubling region . If will happen at 130 mA . The excuse being it is already loud so ears will be the source of distortion more than the amp ?
Coincidentally 100 mA is ideal bias for Exicon MOSFET's . Above that and the Ron is too high . It is easy to hear the looseness of the bass if too much used . Naturally there is no Ron with bipolar transistors . The opposite in fact . 100 mA is said to be the tipping point where Ron incenses in Audio MOS FET's .
Dvv . I notice you use 0R47 ( I think so ) . This almost puts the devices in MOSFET country . Self says 0R1 . 0R22 seems a safe choice .
I noticed class D protection works by Drain to Source voltage in theoretical papers . This could be used in class B MOSFET . Less so in bipolar as the losses are small . Even so a comparator might be able to do it .
The Quad should tolerate 130 mA for long enough to do some tests . I hope to hear the same sound with more low level colour . The Quad has out of this world measurements for a 1967 amp . The very low bias I think does something mildly like MP3 . What it doesn't do it doesn't show without reference to better amps . I have been running the Hypex Ucd 180 at home both on Hypex SMPS and conventional PSU . I have to say it is close run with the Quad and perhaps the Quad is better ? I listened to an old Arcron DC 300 the other day . Whilst it was more filtered than Quad or Hypex it seemed rather wonderful . The Hypex sounds different on the PSU's and measure slightly nicer on conventional . It is not all that way and Hypex should be commended for how well the SMPS works . I would rather like to try the IR IRS 2092 . On paper it looks good . This has made me think a Quad 303 with 6 outputs might be my cup of tea .
I wonder if anyone has used light bulbs as emitter resistors ? Might be OK as it would be in the feedback loop .
That's the whole idea, Pierre, to spend the first watt, which we use in well over 90% of the rime in our rooms, with reasonably efficient speakers (mine do 92 dB/2.83V/1m) buried in class A anway.
If one errs with buasing, it's always better to have more than less. Much more gentle.
Ah, Sir Pearson riseth from the dead! Good to see you again, Nige.
No, not 0.47, but 0.27 Ohms. I have NEVER used 0.47, the worst I have done was 0.33 Ohms.
I honestly don't know what kind of biasing I'd have to use with 0.1 Ohms, off hand probably something like 200 mA per device or so. That's a bit too much, methinks.
You know I am a sworn follower of the Bart Locanthi Triple Output Stage Club, simply because it generally tends to sound better controlled to me, and with that kind of gain, 130 mA i quite enough.
I don't see how idle bias changes Rds on on power mosfets in a way that matters. Rds on is the on resistance when the fet is turned on hard, a forward bias value in the datasheet. Idle bias may heat up the fet which would increase the Rds on but you are only at that condition when hard clipping. I'm not interested in listening to amplifiers that are clipping. If the discussion is class D then there is no idle bias. What did I miss?
Demian, I was referring exclusively to BJT power devices. I do not work with MOSFETs, as I do not particularly like their sound, they lack oomph! for my taste. Hence, my experience with them is exactly nil.
However, this may change, as UK (Nigel) and Germany (Thorsten) are joint in converting me into a believer.
One of the key problems was availability - for me, it was as poor as a church mouse. Well, that seems to be going away rapidly, and consequently you will see more Japanese devices in my work, just received a package full of 2SC2240/2SA970 and 2SC3503/2SA1381 as a trial run and it went without a hitch,
However, this may change, as UK (Nigel) and Germany (Thorsten) are joint in converting me into a believer.
One of the key problems was availability - for me, it was as poor as a church mouse. Well, that seems to be going away rapidly, and consequently you will see more Japanese devices in my work, just received a package full of 2SC2240/2SA970 and 2SC3503/2SA1381 as a trial run and it went without a hitch,
amp 101 questions continue..............
take a simple differential amplifier, two NPNs, dual supply, collector load resistors and tail resistor, entirely conventional, I want an expression for gain at the first collector for a differential input. This would seem to be the simple ratio of collector resistance to tail resistance, however this neglects the impedance seen at the top of the tail looking towards the second emitter. Since the tail resistance is quite high and emitter 2 appears as a load in parallel with it the effect of the second emitter on gain would seem to be great. In his tutorial Leach (http://users.ece.gatech.edu/~mleach/ece4435/tutorial.pdf) shows a derivation of the transfer function that is a little over my head, in it on page 2 he states that the tail resistor is not included in gain calculation as it is not an AC source. Similarly Cordell presents on page 30 of his book a formula for gain that omits the tail resistance. Why is the tail resistance ignored?
take a simple differential amplifier, two NPNs, dual supply, collector load resistors and tail resistor, entirely conventional, I want an expression for gain at the first collector for a differential input. This would seem to be the simple ratio of collector resistance to tail resistance, however this neglects the impedance seen at the top of the tail looking towards the second emitter. Since the tail resistance is quite high and emitter 2 appears as a load in parallel with it the effect of the second emitter on gain would seem to be great. In his tutorial Leach (http://users.ece.gatech.edu/~mleach/ece4435/tutorial.pdf) shows a derivation of the transfer function that is a little over my head, in it on page 2 he states that the tail resistor is not included in gain calculation as it is not an AC source. Similarly Cordell presents on page 30 of his book a formula for gain that omits the tail resistance. Why is the tail resistance ignored?
Because the calculation here is concerned with differential gain, when the bases of the diff pair see a differential voltage. In this case the tail resistor has no voltage variation across it.
Yes.
Say your LTP has voltages V1 and V2 on the bases...
If you want to do the math for it, the much simpler way is to do two calculations,
- one for Vcommon = (V1+V2)/2, this will give your the sum of output currents
- one for Vdifferential = V1-V2, this will give your the difference of output currents
Say your LTP has voltages V1 and V2 on the bases...
If you want to do the math for it, the much simpler way is to do two calculations,
- one for Vcommon = (V1+V2)/2, this will give your the sum of output currents
- one for Vdifferential = V1-V2, this will give your the difference of output currents
I guess D is in Crimea ..
You mean on Crimea, it is a penninsula, you know?
No, of course I'm not there, I'm right at home when I'm not in some medical lab being bled to death by vampires posing as doctors. The odd thing is, so far, all my results were litarelly text book examples of how they should be for a geezer my age, including sugar. The biggie results come in on Monday, but I suspect they will be about the same. Which suggests my ailing legs are victims of something catching and by default viral transient, once you nail down the cure. And that isn't easy any more, what with the mutated viruses and everything.
Right now, all I'm taking is a French mixture pill of potassium, manganese and B1, B2, B6 and B12 vitamins, and it's doing me a hell of a lot of good.
I shy away from politics, but I must say I am enjoying the Crimean affair with great pleasure. I am enjoying watching the West squirm as it's cooking in its own recipe, this is Kosovo all over again, same but different players, and the boot is now on the other foot.
Crimean is a done deed, there's NOTHING that West can, or is willing to do about it, all that's left is maybe a deal regarding the eastern half of Ukraine, which is 70% populated by ethnic orthodox Russians. MAYBE they can salvage something there. Just maybe, but time is quickly running out.
And it's about to get better, much better. Now the Scots vote on independence, and if they don't make it this time, the next time they will, mark my words. The Basques in Spain (an in France), as well as the Catalonians, come soon enough, very possibly followed by the Corsicans in France, who have wanted to seceeded for over 50 years now.
America was warned but as usual, didn't liosten, once you open the Pandora's box of seccession, there is no end and there's NOTHING you can do about it. It's truly like a domino effect, the only questions is nor is there a next, but who's next? Tyrollians, who want to leave Italy and join Austria? Moldova, which is ready to burst at its seams, as over 50% of them want to join Romania? Eastern Macedonia, mostly populated by ethnic Albanians? How long do you think it will take ethnic Kurds, and we're talking about literally millions of people (about 20 million all told) here, will want to seceede from Turkey, Iran and partly Syria and finally have a state of their own? Just to add spice to the Syrain problem, as if it wasn't already complex enough.
And, as usual, the stupid Europeans will be footing the bill. Russia has a viable alternative, as China will buy all the gas and oil Russia is willing to sell it, but the Europeans, and among them especially the Dutch ($75 billion worth of gas and oil from Russia) and the Germans (40% of their gas comes from Russia, $54 billion worth), have no alternative just now, the earliest solution is estimated at best to be 5 years away if all goes well, not counting the construction time for brand new facilities.
The only thing lacking on the world scene is simple, old fashoned sanity.
On the electronics front, I'm still mopping up from your project. I've tried several suggestions, but was not very happy with any of them. But I have pushed that basic circuit further, evolved it, so to speak, and right now, I'm still testing for the key differences.
The key difference is that the FET buffered second stage is still a differetial stage with a current mirror loading, but it now has an active CCS rather than plain passive resistors. Plus a few odd bits here and there, small fry, but I think worth the time and trouble. We'll see.
The key difference is that the FET buffered second stage is still a differetial stage with a current mirror loading, but it now has an active CCS rather than plain passive resistors. Plus a few odd bits here and there, small fry, but I think worth the time and trouble. We'll see.
- Status
- Not open for further replies.