Re: LT1166 vs Vbe multiplier
That's not particular smart. You better wait until I have finished my design.
BTW, at the moment I'm busy with a new auto bias circuit that has far less impact on the THD. It's fully symmetrical and conceptually less complex than my previous autobias thingie or the LT1166
syn08 said:Who knows, maybe I already did?
That's not particular smart. You better wait until I have finished my design.
BTW, at the moment I'm busy with a new auto bias circuit that has far less impact on the THD. It's fully symmetrical and conceptually less complex than my previous autobias thingie or the LT1166
Re: Re: Re: LT1166 vs Vbe multiplier
Not bad at all. Probably a testament to the distortion reduction capabilities of your output stage. When you went -180 mV, what did the bias current decrease down to?
I don't see MOSFET Vgs tempco mentioned much at all. When I was researching for my MOSFET power amplifier, I just measured it myself.
In my MOSFET amp, that is why I resorted to mounting one of the HEC error differencing transistors (which also act as Vbe multipliers) on the heat sink.
Cheers,
Bob
Edmond Stuart said:
Not bad at all. Probably a testament to the distortion reduction capabilities of your output stage. When you went -180 mV, what did the bias current decrease down to?
I don't see MOSFET Vgs tempco mentioned much at all. When I was researching for my MOSFET power amplifier, I just measured it myself.
In my MOSFET amp, that is why I resorted to mounting one of the HEC error differencing transistors (which also act as Vbe multipliers) on the heat sink.
Cheers,
Bob
Mosfet as temp sensor
It's from my QRV01 headphone amp.
They can have up up -10 mV/deg C. I used a BS170 as temp sensing element and it worked quite good.Edmond Stuart said:
It's from my QRV01 headphone amp.
Re: Re: Re: Re: LT1166 vs Vbe multiplier
Hi Bob,
Iq dropped from 150mA to 43mA.
Whether the not bad results are due to specific properties of my OPS, I'm not sure. Please notice that THD figures are only relative. Besides, my OPS is not that special. I think it's more due to the wider class-A region of MOSFETs compared to BJTs.
I've read figures between 4 to 7mV/K, so let's stick on 5.5mV/K for the moment.
Each of your differencing transistors multiplies the Vbe by a factor a two (correct me if I'm wrong). So, if you mount only one of them on the heat sink, you get a tempco compensation of only ~2*2.2mV/K, while the total Vgs tempco is 2*5mV/K. This is about two times more. Do I have overlooked something?
BTW, I'm still fiddling with auto bias circuits, one the holy grails in audio. If one gets it truly right (without adding any distortion), that would mean a major step forward.
Cheers,
Edmond.
Bob Cordell said:
Hi Bob,
Iq dropped from 150mA to 43mA.
Whether the not bad results are due to specific properties of my OPS, I'm not sure. Please notice that THD figures are only relative. Besides, my OPS is not that special.
I think it's more due to the wider class-A region of MOSFETs compared to BJTs.I've read figures between 4 to 7mV/K, so let's stick on 5.5mV/K for the moment.
Each of your differencing transistors multiplies the Vbe by a factor a two (correct me if I'm wrong). So, if you mount only one of them on the heat sink, you get a tempco compensation of only ~2*2.2mV/K, while the total Vgs tempco is 2*5mV/K. This is about two times more. Do I have overlooked something?
BTW, I'm still fiddling with auto bias circuits, one the holy grails in audio. If one gets it truly right (without adding any distortion), that would mean a major step forward.
Cheers,
Edmond.
Re: Re: Re: Re: Re: LT1166 vs Vbe multiplier
The total bias spread for a pair of HEXFETs is on the order of 8V, of which half is generated by one of the differencing transistors, so if I recall correctly, mounting just one of those transistors on the heatsink provides substantially more than a net 2.2 mV/C. Indeed, the PAIR of MOSFETs together need on the order of 11 mV/C for compensation. When using a Vbe multiplier in this way, the ratio of Vgs_on to Vbe_on comes into play.
Cheers,
Bob
Edmond Stuart said:
The total bias spread for a pair of HEXFETs is on the order of 8V, of which half is generated by one of the differencing transistors, so if I recall correctly, mounting just one of those transistors on the heatsink provides substantially more than a net 2.2 mV/C. Indeed, the PAIR of MOSFETs together need on the order of 11 mV/C for compensation. When using a Vbe multiplier in this way, the ratio of Vgs_on to Vbe_on comes into play.
Cheers,
Bob
Re: LT1166 vs Vbe multiplier
Hi Bob,
I've just simulated you output stage at two different temperatures and got other results. With all BJTs held at the same temperature, I got a total temp. compensation of 5mV/K (between the two gates), far less than expected, by you and by me. If times allows, would you have a closer look at this case?
Cheers,
Edmond.
Bob Cordell said:
Hi Bob,
I've just simulated you output stage at two different temperatures and got other results. With all BJTs held at the same temperature, I got a total temp. compensation of 5mV/K (between the two gates), far less than expected, by you and by me. If times allows, would you have a closer look at this case?
Cheers,
Edmond.
Re: What about in real life?
Hi Per-Anders,
I don't know, and it is not my intention to build a complete HEC output stage, just to figure out how large the tempco of the bias generator is.
Normally, we use a calculator for this kind work. Sane people first think, then calculate and simulate, then build and test, not the other way around.
Cheers,
Edmond.
peranders said:
Hi Per-Anders,
I don't know, and it is not my intention to build a complete HEC output stage, just to figure out how large the tempco of the bias generator is.
Normally, we use a calculator for this kind work. Sane people first think, then calculate and simulate, then build and test, not the other way around.
Cheers,
Edmond.
Re: Re: What about in real life?
Some sane people first simulate, then think (as much as simulation results allow), and that's all about it. Calculate, build, test is for wusses.
For a FQP19N20 and FQP12P20 pair, Bob's HEC has a measured tempco of about 8.5+/-1mV/C depending on the pair's Vth spreading.
Which is the reason why in the PGP amp I decided to put both trannies on the heatsink, in case it is build with vertical MOSFETs.
Edmond Stuart said:
Some sane people first simulate, then think (as much as simulation results allow), and that's all about it. Calculate, build, test is for wusses.
For a FQP19N20 and FQP12P20 pair, Bob's HEC has a measured tempco of about 8.5+/-1mV/C depending on the pair's Vth spreading.
Which is the reason why in the PGP amp I decided to put both trannies on the heatsink, in case it is build with vertical MOSFETs.
Re: Re: Re: Re: What about in real life?
Sorry, I shoud add that two trannies on the heatsink matched that tempco very well (within the measurement precision, which is obviously less than you are able to simulate). So it's indeed about 2*2.2mV per tranny. The bias was rock solid at around 100mA per device.
Edmond Stuart said:
Sorry, I shoud add that two trannies on the heatsink matched that tempco very well (within the measurement precision, which is obviously less than you are able to simulate). So it's indeed about 2*2.2mV per tranny. The bias was rock solid at around 100mA per device.
Re: Re: Re: Re: Re: What about in real life?
The problem is that my simulator tells me (or lies to me) that the total tempco of the bias generator is only 5mV/K, while it should be 2*2*2.2=8.8mV/K. I've checked and and double checked it, used even different trannies. But it remains 5mV/K.
Perhaps you can check the HEC on your simulator.
syn08 said:
The problem is that my simulator tells me (or lies to me) that the total tempco of the bias generator is only 5mV/K, while it should be 2*2*2.2=8.8mV/K. I've checked and and double checked it, used even different trannies. But it remains 5mV/K.
Perhaps you can check the HEC on your simulator.
