gl said:
Thanks Graeme.
Are you referring to AC coupling of the FE to output stage when you mention Tazzz' take? I'm unconvinced of us mainly because it would be difficult to control the DC conditions in the output stage. How do you think this is achieved?
Thanks for the tip about VR2 - I would have fallen into this hole!
I'm already building but progress is slow. I have all the parts except the Dual FETs but I'm still pursuing the latter for now.
Ian.
GRollins said:
If Aleph is taken to mean the modulated current source then this is not an Aleph. I'm using the "A" loosely to denote an X amplifier that is operated mostly in PP class A. The Aleph is SE class A and the X is class AB so really this is an amalgam of both. Really I should start a new AX thread I suspect.
Ian.
Hello!
Have corrected the mistakes in my schematic!
In the heat of the moment I had forgotten to put in the right values for the input stage. Now you will get a gain of 26dB and an input impedance of 15k unbal. / 30k bal.
The relay switching circuit will follow shortly! But this will not be a problem and easy to handle.
With +/- 32V supply rails and a bias currrent of 2,5A per bridge half (5A total), it should reach 100W@8Ohm pure PP classA or 200W@4Ohm classAB.
But be warned, I still haven't build this amplifier and the bias circuit needs some attention and work.
Regards
Dirk
Have corrected the mistakes in my schematic!
In the heat of the moment I had forgotten to put in the right values for the input stage. Now you will get a gain of 26dB and an input impedance of 15k unbal. / 30k bal.
The relay switching circuit will follow shortly! But this will not be a problem and easy to handle.
With +/- 32V supply rails and a bias currrent of 2,5A per bridge half (5A total), it should reach 100W@8Ohm pure PP classA or 200W@4Ohm classAB.
But be warned, I still haven't build this amplifier and the bias circuit needs some attention and work.
Regards
Dirk
noisefree said:
I think the thermistor and the resistors should only be in series with the pot for setting the bias voltage and not in series with the whole bias generator as it is now.
The additional capacitor should then be coupled across them. It is told in the digital forum that the tl431 performs its very best when decoupled both across it in its entierly and also over the feedback resistor itself.
But how do you account for those extra 20% ?
Hi Tazz,
the used thermistor as temperatur sensor is a NTC element.
When you want to put a thermistor in series with the pot (between reference pin and cathode of the TL431) you have to use a PTC.
A second possibility for the NTC (NTC network) is between reference pin and anode.
The resistor in series with the reference element is needed to get a lower voltage bias when the whole thing is only biased by the 10k resistors in stand by mode (instead of the much higher front end bias).
Can you please give me the exact thread and post for the elcap / TL431 information?
Have interest to read it...
Have not find an idea for 20% extra swing without seperat front end voltage.... :-(
Regards
Dirk
the used thermistor as temperatur sensor is a NTC element.
When you want to put a thermistor in series with the pot (between reference pin and cathode of the TL431) you have to use a PTC.
A second possibility for the NTC (NTC network) is between reference pin and anode.
The resistor in series with the reference element is needed to get a lower voltage bias when the whole thing is only biased by the 10k resistors in stand by mode (instead of the much higher front end bias).
Can you please give me the exact thread and post for the elcap / TL431 information?
Have interest to read it...
Have not find an idea for 20% extra swing without seperat front end voltage.... :-(
Regards
Dirk
Me again....
think I need more sleep - make to much mistakes...
...have confused anode and cathode....
Here are the two corrected sentences:
When you want to put a thermistor in series with the pot (between reference pin and anode of the TL431) you have to use a PTC.
A second possibility for the NTC (NTC network) is between reference pin and cathode.
Sorry and......
.......good night!
Tzzzzzzzzzz.........
think I need more sleep - make to much mistakes...
...have confused anode and cathode....
Here are the two corrected sentences:
When you want to put a thermistor in series with the pot (between reference pin and anode of the TL431) you have to use a PTC.
A second possibility for the NTC (NTC network) is between reference pin and cathode.
Sorry and......
.......good night!
Tzzzzzzzzzz.........
The resistor is intentionally in series with the whole bias generator for the reasons Dirk mentions. The NTC thermistor is used to thermally compensate this resistor, not the TL431 which is thermally stable anyway. You are right that this, and probably the whole bias generator, should be bypassed with a capacitor though.
Dirk's circuit has no bootstrapping and hence cannot swing close to the rail voltages. The circuit I posted earlier has some bootstrapping that increases the maximum signal swing but will still be some way off the rail voltage. If there is another 'trick' here nobody has offered any suggestions yet...
Ian.
Dirk's circuit has no bootstrapping and hence cannot swing close to the rail voltages. The circuit I posted earlier has some bootstrapping that increases the maximum signal swing but will still be some way off the rail voltage. If there is another 'trick' here nobody has offered any suggestions yet...
Ian.
Ian Macmillan said:
I assume the thermistor is there to thermaly compensate the amps outputstage during operation by adjusting the voltage setting on the tl431?
It seems to me anyway that this could be accomplished by mounting this net in or around the 2.5v reference internal to the tl431 instead. And choose the values so as to not to turn it on without the bias current runing through the UGS module. Then the bias servo would show a steady 100mR impedance over a large range of currents/ frequencies.
And a smaller stand by voltage could still be generated from the 10K resistors and the resistors determining the voltage across the tl431 while it is operational.
Is the reason you don’t like the resistor/thermistor network in series with TL431 bias generator due to the increase in AC (as well as DC) impedance and the need to bypass with a capacitor to reduce this effect? Or is the motivation behind wanting to place the resistor/thermistor network in the TL431 reference loop to increase the effectiveness of the thermistor? Either way I agree that this is a perfectly reasonable way to do things if the intent is to thermally compensate the output stage by altering the bias.
The intent in both Dirk’s and my circuit is a little different. We are looking for an effective bias voltage that is lowest when the amplifier is in standby, i.e. the FE is powered down, increasing to its maximum when the amplifier is switched out of standby to help speed the warm up period. As the temperature increases, the desire is to reduce the bias current to the normal working value. Direct compensation of the output stage temperature, as might be achieved by mounting the thermistor on the output heatsink, is not the idea.
I can’t see how to achieve these aims with the thermistor in the TL431 reference loop, particularly in terms of the relatively abrupt behaviour required when switching in and out of standby. If we rely only on the thermistor, the bias current will remain at a high value when exiting standby until the chassis, and hence the thermistor, cools sufficiently. I don’t feel this is particularly desirable, hence the insertion of the series resistor.
No doubt there are better ways of achieving this but so far they haven’t occurred to me. All suggestions for better alternatives are very welcome
Ian.
The intent in both Dirk’s and my circuit is a little different. We are looking for an effective bias voltage that is lowest when the amplifier is in standby, i.e. the FE is powered down, increasing to its maximum when the amplifier is switched out of standby to help speed the warm up period. As the temperature increases, the desire is to reduce the bias current to the normal working value. Direct compensation of the output stage temperature, as might be achieved by mounting the thermistor on the output heatsink, is not the idea.
I can’t see how to achieve these aims with the thermistor in the TL431 reference loop, particularly in terms of the relatively abrupt behaviour required when switching in and out of standby. If we rely only on the thermistor, the bias current will remain at a high value when exiting standby until the chassis, and hence the thermistor, cools sufficiently. I don’t feel this is particularly desirable, hence the insertion of the series resistor.
No doubt there are better ways of achieving this but so far they haven’t occurred to me. All suggestions for better alternatives are very welcome
Ian.
My understanding of thermistors used in such locations is that the major response is to the ambient temperature inside the case. Since (in a class A amp) the source of heat inside the case comes almost entirely from the output section, I would therefore conclude that this thermistor is present to provide coarse bias regulation for said output section. I don't think things are more complicated than this.
Graeme
Graeme
IMO the the thermistor circuitry would have been the last thing tweeked in the final design of the amp. You can't finalize the response curve until the circuitry is installed in the case, so at least one or two or the resistors would be present to allow fine tuning of the thermistor response "in place".
For a DIY effort, that presumably would have different casework, you would have different circuit values around the thermistor or perhaps no thermistor at all. Once again I would maintain that this whole subject is getting more attention than it deserves.
Graeme
For a DIY effort, that presumably would have different casework, you would have different circuit values around the thermistor or perhaps no thermistor at all. Once again I would maintain that this whole subject is getting more attention than it deserves.
Graeme
Graeme,
I agree with you on most counts with the possible exception of this getting more attention than it deserves. It deserves attention only because it may be the key to how to handle the standby feature (if this is important to you). Standby may be useful in helping eliminate thumps from the speaker due to the SE bias.
Ian.
I agree with you on most counts with the possible exception of this getting more attention than it deserves. It deserves attention only because it may be the key to how to handle the standby feature (if this is important to you). Standby may be useful in helping eliminate thumps from the speaker due to the SE bias.
Ian.
When you switch off the front end in stand by mode, you have to leave the putput stage in a small conducting state. There are SE-bias resistors which are connected between output and negativ supply voltage!
You have to blow enough currrent into the output stage for not seeing the negativ rail voltage at the outputs. This is be done with the 10k resistors which are connected between the gates and the rail voltages. Alternatively you can put a switch between bias resistors and negativ supply and switch them off when using the stand by mode.
Dirk
You have to blow enough currrent into the output stage for not seeing the negativ rail voltage at the outputs. This is be done with the 10k resistors which are connected between the gates and the rail voltages. Alternatively you can put a switch between bias resistors and negativ supply and switch them off when using the stand by mode.
Dirk
To the voltage bias circuit:
Yes, the dimensions of the parts will be individuell and depends on location of the thermistor and other facts. Before you start implementing such a thermistor compensating network, I would do three measurements with the complete amplifier (and only a normal voltage reference as bias):
- Which bias voltage is needed to get full / maximum bias current when amplifier is cold.
- Which bias voltage is needed to get full / maximum bias current when amplifier is hot.
- Which bias voltage is needed to get minimum bias current and 0V at the outputs. (think this should be measured when the amp is hot, because then it doesn't matter that the voltage bias will rise a bit when it is cooling down)
With this three individuell values and a choosen thermistor it should be possible to calculate the series resistor and the resistor network.
When it is implemented you can correct the bias a bit with the trimmer.
Dirk
Yes, the dimensions of the parts will be individuell and depends on location of the thermistor and other facts. Before you start implementing such a thermistor compensating network, I would do three measurements with the complete amplifier (and only a normal voltage reference as bias):
- Which bias voltage is needed to get full / maximum bias current when amplifier is cold.
- Which bias voltage is needed to get full / maximum bias current when amplifier is hot.
- Which bias voltage is needed to get minimum bias current and 0V at the outputs. (think this should be measured when the amp is hot, because then it doesn't matter that the voltage bias will rise a bit when it is cooling down)
With this three individuell values and a choosen thermistor it should be possible to calculate the series resistor and the resistor network.
When it is implemented you can correct the bias a bit with the trimmer.
Dirk
gl said:
It probably will not of itself in the sense if both halfs of the 'bridge' are balanced then any offset should be common mode and hence will not be seen by the speaker. I was just concerned that everything will need to both balance well throughout the startup period as the PS caps charge as well as steady state.
Ignore me
Ian.
noisefree said:
As per my previous post, the negative supply due to the SE bias is not a problem for the speaker as it will see it as a common mode signal. That said, I believe I recall Nelson saying that he bleeds additional current into the output stage on standby to reduce the common mode voltage to close to zero. Just turning the PP devices on will not do it for the relatively small standby consumption we are looking for. I don't know, but I would assume this is done by making the 10k resistor to the +ve supply of slightly smaller value. Perhaps using a parallel resistor? This would allow another, similar resistor to be switched in with the FE supply to balance things in normal operation if desired.
Ian.
Ian,
yes, that's clear, the speaker would not see the negative supply voltage.
My answer was short and imprecise (difficult for me to come to the point in english!).
It is the working absolute offset (generated by switching on and off and maybe different on both bridge sides) which can result in moments with to much differential offset.
Also it is handsome to get 0V absolute offset in both operating methods (stand by and power).
But I have to disagree with you about the answer of Mr. Pass:
He explained that the implementation of the 10k resitors (between gate and supply) would result in outputs which float at about -4V ABSOLUTE offset. Bleeding enough current into the system would rise it close to 0V. (Post number is 1451 but better to start with 1445 in this thread)
I don't know if I have interprete it in the right way, but it seems to me that only this two tools (10k resistors and high enough bias voltage in stand by) would do the "self balance" between N- and P-mosfets.
Other point:
Have thought about the thermistor network. It believe it is easier to calculate this network when you use a single resistor in series with the TL431 and put the thermistor-resistor-network between cathode and reference pin. Then the series resistor can be calculated independently of the thermistor network.
Regards
Dirk
yes, that's clear, the speaker would not see the negative supply voltage.
My answer was short and imprecise (difficult for me to come to the point in english!).
It is the working absolute offset (generated by switching on and off and maybe different on both bridge sides) which can result in moments with to much differential offset.
Also it is handsome to get 0V absolute offset in both operating methods (stand by and power).
But I have to disagree with you about the answer of Mr. Pass:
He explained that the implementation of the 10k resitors (between gate and supply) would result in outputs which float at about -4V ABSOLUTE offset. Bleeding enough current into the system would rise it close to 0V. (Post number is 1451 but better to start with 1445 in this thread)
I don't know if I have interprete it in the right way, but it seems to me that only this two tools (10k resistors and high enough bias voltage in stand by) would do the "self balance" between N- and P-mosfets.
Other point:
Have thought about the thermistor network. It believe it is easier to calculate this network when you use a single resistor in series with the TL431 and put the thermistor-resistor-network between cathode and reference pin. Then the series resistor can be calculated independently of the thermistor network.
Regards
Dirk