N-Channel mosfet amplifier with servo bias
I'm going to built a new amp whith what i have in my drawers so i did the attached schematic but i have many questions before drawing a pcb because i'm not sure of what i did.
I want to keep bias current constant so i measure it by calculating the difference of the total current trough the output transistors and the current in the speaker. A slow integrator of difference adjusts the current generators Q1 and Q3 .
I want to use only one quad amp-op so i drawed a single phase rectifier for the speaker current and substract it twice to the total current.
When A4 output is 0V, i choose R6 and R11 to have only half of bias current, so the current grows slowly when starting the amplifier.
I'd like to have your comments about amplifier section also (does it work ?)
Can you simulate it with spice ?
Any comment is welcome
Your bias circuit will most likely not work. It's really hard to filtering out the real bias component of the total current.
Sliconix had an auto bias mosfet amp in their application notes .... somewhere. Can someone find it?
I began to draw a sample and hold that sample voltage on R23 R24 when volage is about 0V at the output but it should be able to sample/hold in 10ns , more, V and I are not in phase because a speaker is not a resistance !! Then i thought this way "could" work........i'm very interrested in the application you speak about.
Maybe measuring the energies dissipated by R22,23,24 could be an other way ?
I had an idea of an auto bias circuit:
Between the gates, place a microcontroller with a temperature sensor. The output of the MCU is PWM or a DAC. Vgs vs temperature is measured and stored in the MCU. Variyng temp => Varying Vgs
The thing was to let this MCU be floating and be driven directly from ther gate drive voltage.
This idea never left my brain....and got into the computer or lab.
I found it, but the article is copyrighted so I can't post the schematic without permission (I emailed Siliconix). Anyway, the article is by Bill Roehr and it may have originally been an AES preprint. It's dated April 1982 and titled "The Autobias Amplifier." Shown are 25 and 50 watt circuits. These are all single-supply, bipolar input, n-channel MOSFET output amps with a big output cap. Brrrr!
Peranders : sure we can do nice things whith a microcontroller, i'd preffer to keep it for a direct numeric input amp with class D output....(hope to find a project one day in the digital forum) my desire is not to "compensate" (i could do a curve with ampop, zeners,res, and a temperature sensor, depending of the Vgs caracteristic) but to control with closed loop, just for fun and experimentation.
In my bipolar amp, i see clearly with the scope, the bias current passing through the emitter resistances. This current never gets lower than the minimum value, so i had another idea and will post it soon.
SY : I thing i found the application you are speaking about :
Single supply and all AC coupled......I don't understand all, Vgs decreases non linear with Id....I have to take time to analyse.
Tks for your posts
I did a new servobias schematic (see attached)
The idea is to compare the current crossing R22 and R23 to a minimum value witch is the image of the bias current. If it goes under this value, a slow integrator will increase the bias current by changing the voltages at Q1 and Q3 emitters.
And if it goes greater ? no problem, a negative current is continuously injected in the input of the integratorso the output of A3+D1 must be 100mw, corresponding to an error of 0.5mA of the bias current.
With this method, there is no need to measure an AC current at the output, all what is usefull is that the continuous part of the current does not go under the choosen value.
(sorry for syntax errors)
You had better use a very fast...FET or FET hybrid op amp for A1. The difference in what is needed to drive this quasi complementary pair output is going to be different for each device. Some very high frequency harmonics are going to be needed here to cancel those created by the bottom IRFP460, acting as a voltage amplifier, or common source. The top one is a source follower or current amplifier. If complementary operation is what you are looking for in this circuit, you should use complementary N-ch & P-ch. However, if you want my opinion, you should rebias this circuit as a class A amplifier and use the top IRFP460 as the output transistor as a source follower(as is now), and use the bottom IRFP460 as a current source, biasing Ids bias through the top one. Direct couple to speaker as is currently shown, I bet you will get better overall results without so much herky-jerky.:rolleyes:
May its not as power efficient as class AB complementary or quasi-complementary, but quality would be better...of course you wouldn't need a servo bias circuit.
Bias ideas and comments
Some years back Linear Technology was kind enough to send me some samples of one of their new 8 pin function blocks, the LT1166. This IC is designed for power output stage automatic bias. Although it is not directly suitable in the circuit outlined in this tread it could possibly be adapted to provide a good portion of the required functionality. They had a nice 16 page brochure that may be available on their web site as a application note that gives a great deal of detail on the LT1166.
In general any type of automatic bias adjustment acn add substantially more complexity to a power amplifier circuit with potential negative benefits. With a well designed bias network that monitors heat sink temperature and without doing any kind of sampling the output stage current the output stage current can be kept constant within 1.5% over a temperature range of 110 to 190 degrees Fahrenheit. Such a method can do better than this over the desired operating temperature range of 135 to 155 Fahrenheit, (145 or so being preferred for maximum stability). The bias circuit I used in my 1R1 amplifier that Per-Anders is placing on his web site has the ability to do this. The drawback of using Bias control networks such the one I used in that amplifier is that it is sensitive to construction of the heat sinks and thus it is optimized for a specific heat sink design, and also affected by the placement of the devices on the heat sink.
My feeling is that Iím generally against using any type of auto bias circuit that adjust bias by measuring the voltage drop across resistors placed in series with the output devices. I have a gut feeling that such current sensing method look better on paper than they work in actual practice when applied to audio power amplifiers. Such methods been tried by quite a number of manufacturers and appear to all become abandoned after a period of time. This of course does not mean that we donít experiment with auto bias and other method to stabilize bias settings.
If a isolated method of controlling bias is required then a very useful circuit of for making a dual opto isolator very linear with an op amp was published in Electronic Design quite a number of years back. It provides a good voltage in versus a current output curve. It used one section of the dual opto for driving a isolated load and the other half of the opto for providing feed back to the op amp. The two led light sources being in series. As I recall the LED current was also part of the feedback loop to the op amp. One of the inputs to the op amp excepted the input drive voltage. I had used it in the past for some experiments and it gave excellent results. Thus such a circuit should be kept in mind when the need for controlling current or bias and isolation is required.
|All times are GMT. The time now is 07:27 AM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2015 DragonByte Technologies Ltd.
Copyright ©1999-2015 diyAudio