@ ostripper
Ok, Thanx for the info
@ Alan0354
I admire your thirst for knowledge, & self taught dedication Your careers have proved that qualifications are not always, All That
Don't let anyone bully you, or put you off !
My advice would be, to learn to chill as well as learning & tinkering
"Re"garding Re & bias etc, have you seen my thread & the ideas & comments etc in there http://www.diyaudio.com/forums/solid-state/277005-oliver-offset-optimiser.html
Ok, Thanx for the info
@ Alan0354
I admire your thirst for knowledge, & self taught dedication
Your careers have proved that qualifications are not always, All That Don't let anyone bully you, or put you off !
My advice would be, to learn to chill as well as learning & tinkering
"Re"garding Re & bias etc, have you seen my thread & the ideas & comments etc in there http://www.diyaudio.com/forums/solid-state/277005-oliver-offset-optimiser.html
If you are going to use digital correction of OPS bias I think the smarter way to go is to use standard bias spreader techniques to do a first order compensation and then the uC to do the second order. Just use the I/O lines directly to control the spreader transistor base. With 4 or 5 lines you can get 8 to 16 correction steps around the 1st order comp from the spreader. Importantly, you can make the whole thing self calibrating save for the initial standard bias spreader adjustment.
I touch very briefly on this on a document on my website covering thermal comp schemes.
I touch very briefly on this on a document on my website covering thermal comp schemes.
Could you elaborate?
Maybe some pseudo code?
What is it supposed to do when the amp is making a noise?
What to do when the noise stops?
Does it need to know when the amp is supposed to be making a noise?
If so, how does it know?
As I said, I've an analogue method to do what you propose but I'm not sure it's what I want.
Thanks
At those levels of distortion, I don't expect to see much difference with BAxxx but you probably will with 1n4148 & the like.
The effect is greatest at HF
I meant using the LTspice transistor.
Hm.mmm! It appears even Linear have stopped selling them?
No, I was just throwing out some thoughts since it's on the subject. I personally doesn't even want to follow as I don't even think it's necessary. It's one thing to do it because you can, it's another thing to actually put in the effort to actually try to implement it.
I since think about it more, you can actually simplify to the point not needing the uP and go pure analog. Just use individual closed loop servo to do what I describe. Just have a pot to trim a universal reference that everyone follows. Then each of the individual servo circuit of each transistor follow the reference voltage and adjust the voltage across the Re by adjusting the current through the base stop resistor.
Idea is very simple but still it's a lot of circuits. I don't like it, but idea has it's way to intruding into my mind and disturb my tv program!!!
Regarding to the noise and signal, my thinking is make the time constant very long for the adjustment and just "integrate" out all the noise and signals. As long as all the signal and noise average out to zero, it is not going to affect the outcome.
But as I said, I am not going to go deep into it. If it works, it works. If it doesn't, then oh well!!! Don't take it too serious. I am just stirring up trouble here to keep you guys awake!!!
I am drowning on my own amp and hogging is the least of my concern for the moment. I just found out I made a bubu big time on the anti clipping on the VAS and it's not easy to fix. So you are on your own.
Last edited:
currnt hogging prevention
Have seen some interresting ideas about balancing currents in paralleled devices:
Transformer circuit for balancing current flow through parallel semiconductors
Parallel transistor amplifiers
Negative feedback circuit for parallel-connected thermionic amplifiers
However there is also a patent for more than a pair of OPTs (with lots of tranformers).
Have seen some interresting ideas about balancing currents in paralleled devices:
Transformer circuit for balancing current flow through parallel semiconductors
Parallel transistor amplifiers
Negative feedback circuit for parallel-connected thermionic amplifiers
However there is also a patent for more than a pair of OPTs (with lots of tranformers).
The most difficult issue here which is glossed over is how to know what the actual Ibias is. Yes, you compare it to a ref and adjust the bias spreader accordingly by closing the loop - that's simple.
But how do you extract the actual value of the Ibias in the first place? ALL of the systems I have seen have ultimately had to focus (and sometimes fail) on this! Except in class A amps with their full audio signal across the Re - it's much simpler there; see for instance Nelson Pass' solutions for that situation.
Jan
There is a detailed explanation and discussion of Phase Intermodulation Distortion (PIM) in an AES Convention Paper over 30 years ago, and that paper is available on my web site.
In a nutshell, any signal-dependent gain in the amplifier will cause the closed loop bandwidth of the amplifier to move as a function of the signal. The closed loop phase lag through the amplifier created by the closed-loop bandwidth rolloff will change with signal, and thuse there is phase distortion. This is no different than the distortions of differential gain and phase referred to in the analog video world.
The HF loop gain can be modulated by the signal in numerous ways, but amplitude nonlinearity (gm) in the input stage and variation of the total Miller compensation capacitance with signal are two of them. Of course, such nonlinearities will also necessarily create THD and conventional IM. Thus, if your THD-20 is very low, your PIM must also be very low. This is the part that Otala overlooked in suggesting that it is a new kind of distortion that does not show up on conventional measurements. Note that I also showed PIM measurements in my MOSFET power amplifier paper, also available on my web site.
It is unfortunate that the datasheets for the 1N4148 and 1N4149 do not show curves of capacitance versus reverse bias voltage. It is well known that this capacitance goes down significantly with increased reverse bias. The capacitances at 0V bias for these diodes is 4pF and 2pf, respectively. The capacitance can be expected to go down to parhaps 1pF or 0.5pF under high reverse bias. A good LTspice model will show you this.
If you have a 3pF signal dependent variation in capacitance in parallel with a 30pF Miller compensation capacitor, you can expect the ULGF to vary on the order of +/- 5% with signal swing.
Cheers,
Bob
This article doesn't address current hogging, but it does provide a way to automatically set the bias of the output stage:
Linear Audio | your tech audio resource
Linear Audio | your tech audio resource
you don't have to "extract" idle current from the peaks - the "problem" is setting/controlling it when both Q are conducting
when one is on and the other off there's no problem with thermal runaway
so you can measure with needed resolution near output current zero crossings
you should also do full range measurements to have the information for thermal modeling for feedforward too
but needing gain ranging on the current measurement is not a big deal
digital control allows conditionals, looking where and when with the needed gain, resolution
when one is on and the other off there's no problem with thermal runaway
so you can measure with needed resolution near output current zero crossings
you should also do full range measurements to have the information for thermal modeling for feedforward too
but needing gain ranging on the current measurement is not a big deal
digital control allows conditionals, looking where and when with the needed gain, resolution
Sorry guys, it just dawn on me when I was in bed last night that my "simple" idea only works on Class A amplifier. In AB, the voltage across the Re is 0 almost half the time. Can't do that!!! Sorry.
This 5 minute idea always have a way to get complicated. My last thinking is you are going to need to have a separate uP at ground level monitor the input and tell the uP floating at the OUTPUT that there is no signal and go ahead and do calibration of the bias current................
I really need to get out of this topic. As I said from the very beginning, I won't do it, it's too complicated. This is just talking for the sake of talking. Curiosity just got the best of me yesterday. I am not going to involve in this anymore, I made a big bubu in my anti clipping circuit and I cannot think of a graceful way to fix it.
So this is really on your own. I am out.
Attached is the image of my OPS on the heatsink. Finally I got the 0.22, burn in the 0.22 and put it all in. I measured the resistors and put the two that is like 2.5% lower value at the two ends and put the two transistors that has beta 1% higher than the ones in the middle at the two ends. Vbe of all the PNP and NPN are matched to at least 1%.
The temperature of the transistors at the two end has better cooling, so I make it work harder to heat it up slightly more and for more accurate temperature sensing. All in all, I don't think I need all those fancy uP stuffs to avoid hogging. I am a true believer of "less is more". Don't get stuck in all the fancy ways of controlling the bias current.
Sorry.
This 5 minute idea always have a way to get complicated. My last thinking is you are going to need to have a separate uP at ground level monitor the input and tell the uP floating at the OUTPUT that there is no signal and go ahead and do calibration of the bias current................
I really need to get out of this topic. As I said from the very beginning, I won't do it, it's too complicated. This is just talking for the sake of talking. Curiosity just got the best of me yesterday. I am not going to involve in this anymore, I made a big bubu in my anti clipping circuit and I cannot think of a graceful way to fix it.
So this is really on your own. I am out.
Attached is the image of my OPS on the heatsink. Finally I got the 0.22, burn in the 0.22 and put it all in. I measured the resistors and put the two that is like 2.5% lower value at the two ends and put the two transistors that has beta 1% higher than the ones in the middle at the two ends. Vbe of all the PNP and NPN are matched to at least 1%.
The temperature of the transistors at the two end has better cooling, so I make it work harder to heat it up slightly more and for more accurate temperature sensing. All in all, I don't think I need all those fancy uP stuffs to avoid hogging. I am a true believer of "less is more". Don't get stuck in all the fancy ways of controlling the bias current.
Sorry.
Attachments
Last edited:
Hello Bob
I am away from my archives at present, but from what I recall of the investigations we did into ThermalTraks some years ago (2012?) this was really not the case.The metal header is connected to the heatsink through a relatively large area of thermal pad and the diode is much nearer heatsink temperature than junction temperature. I did some measurements on p540-542 of Audio Power amplifier Design 6th edition, which demonstrated this.
As I say, I am away from my notes, but I suspect that the difference here is that you are using a rather higher thermal resistance between device and heatsink than I am.
I wholly agree. A bit of foam lagging reduces the thermal lag, as you might
say. I have tried this and it does work.
As I see it the single bias generator (I cannot like the phrase leg-, I mean bias-, spreader; sounds most inelegant) eliminates most of the possibly positive feedback on all the output devices. After that the differences between devices & Re's would have to be pretty gross to allow any instability.
And I will repeat, I have only ever seen thermal runaway once in my life.
More importantly, does controlling the bias to the extreme improves the sound?
How much circuit it takes to really control the bias? If that does not improve the sound significantly, no body will want to buy it even if it works. Amps has been around like forever, they last forever. who in the right mind will spend more money, making the pcb bigger, much more complicated to do nothing of importance.
Only in DIY that people gone to the end of the world just to say he/she did it.
How much circuit it takes to really control the bias? If that does not improve the sound significantly, no body will want to buy it even if it works. Amps has been around like forever, they last forever. who in the right mind will spend more money, making the pcb bigger, much more complicated to do nothing of importance.
Only in DIY that people gone to the end of the world just to say he/she did it.
I can't look it up at the moment, but I would have thought that heat basically travels at the speed of sound in the material, so some, possibly tiny, change could be detected immediately. Of course the heating up of the material on the way is another matter altogether, and gives the usual exponential curve.
Is there a physicist to hand?
You sure about that? If that's the case, we won't be talking about this at all. then there should be no issue even the sensing device is bolted next to the big transistor on the heat sink as the change is instantaneously. You use a cloth to hold a cup of hot water, takes a long time to slowly feeling the heat. I think the biggest problem is the phase lag where there is a delay from the heat source to the detector that can cause instabilty.
I do agree current hogging is not a big issue if people are careful. Hell, amps have been around like forever, they last and last. I don't see controlling the bias improve the sound in any way, if it's not broken, don't fix it.
Last edited:
"Low value (Re = 0.1) emitter resistor to minimise class AB distortion." --- D.Self.
IIRC way back in the Wireless World era there was an article showing distortion lower with lower Re.
Was there a null value of Re for harmonic distortion. ?? 2H?
View attachment Lower distortion with low Re.pdf
THx-RNMarsh