I’m thinking of designing a microprocessor calibrated amp. The idea is to have a microprocessor calibrate the initial gain stage to bias & balance the drive to the output transistor section. This would compensate for drift due to age & temperature of the components. It would also super refine the output bias keeping the offset voltage to the speaker to a ridiculous 0v, +/- 0.00001v.
Comments…
Comments…
sounds like an interesting project... just picking up a cheap microcontroller, a digital potentiometer and making a way to measure the bias, then compensate with the digital potentiometer. What sensor are you considering using?
Maxim has some interesting sensors:
http://para.maxim-ic.com/compare.asp?Fam=Sens_Cond&Tree=Sensors&HP=Sensors.cfm
--
Brian
Maxim has some interesting sensors:
http://para.maxim-ic.com/compare.asp?Fam=Sens_Cond&Tree=Sensors&HP=Sensors.cfm
--
Brian
This was (in a cruder form) done for a tube amp some years ago, the Audionics BA150. The key thing is, unless you're running in Wallpaper Peeling Marshmallow Roasting True Class A mode, inhibiting the circuit when there's music running through and holding the bias points until there's a second or two of silence to recalibrate.
Well it sounds like a PIC would be perfect. They have ADCs onboard. Could you use that to measure the bias? As for programming them, I use a language called JAL. It's fantastic! I have used Dallas / Maxim pots with them with perfect results in about 5 minutes!
You could even display the output variables such as DC content and Bais on a small LCD (or have an output at the back) with the spare IOs!
Sounds interesting. I was looking at doing a similar project for my final year project for uni at one point.
Gaz
You could even display the output variables such as DC content and Bais on a small LCD (or have an output at the back) with the spare IOs!
Sounds interesting. I was looking at doing a similar project for my final year project for uni at one point.
Gaz
Im planning on using a 24 bit audio codec. 24 bit ADC for reading the drive to the final emmiter follower transistors, 24 bit DAC to drive the Vactrols which adjust the Bias current & DC balance offset.
The calibration will be real time, even with music. I'll probably calibrate once every 5 seconds, moving the 24 bit DACs no more than 1 or 2 steps for every other second. With some internal flash memory, bias settings will be remembered & tracked.
PICs typically draw too much current & make too much electrical noise. I'll end up using a TI MSP430.
Dac pots suck royally, for audio purity, only Vactrols will do. They virtually mimic a fixed resistor. When used properly, they should outperform a potentiometer since there is no sliding mechanical contact. I’m even making a pre-amp using a few of them.
The calibration will be real time, even with music. I'll probably calibrate once every 5 seconds, moving the 24 bit DACs no more than 1 or 2 steps for every other second. With some internal flash memory, bias settings will be remembered & tracked.
PICs typically draw too much current & make too much electrical noise. I'll end up using a TI MSP430.
Dac pots suck royally, for audio purity, only Vactrols will do. They virtually mimic a fixed resistor. When used properly, they should outperform a potentiometer since there is no sliding mechanical contact. I’m even making a pre-amp using a few of them.
Automatic bias adjustment and DC-offset is a nice idea. I'd also add something which monitors the current through the output stage and trips an output relay if its too high. This would probably have to be done with hardware, a flip-flop or something like that.
You may want to consider making this as universal as possible so it can be fitted to existing amplifiers.
Finally I can suggest using an AVR uC. They have a sleep mode which shutsdown pretty much everything so they should generate far less noise. As far as I know they are much better than PICs.
You may want to consider making this as universal as possible so it can be fitted to existing amplifiers.
Finally I can suggest using an AVR uC. They have a sleep mode which shutsdown pretty much everything so they should generate far less noise. As far as I know they are much better than PICs.
Re: 10uV?
Yes I am.
The TI MSP430 is the cheapest at 1$ & will operate under 1mA at 3.3v @ 32KHz, or, around a mA at 4MHz & in sleep mode, it will operate under 1 uA.
Using 6 interface wires, I'll use 2 TLV320AIC23, $2.85 each, which will give me 4x 24bit ADCs, 4x 24bit DACs.
Inputs from amp will be put through a 0.1Hz low pass RC filter & R/R voltage divider, passed through an TLO64, then to the 4 ADCs.
The controler board will operate right from the rail supplies of the amp.
Dave said:
Yes I am.
jackinnj said:
The TI MSP430 is the cheapest at 1$ & will operate under 1mA at 3.3v @ 32KHz, or, around a mA at 4MHz & in sleep mode, it will operate under 1 uA.
jackinnj said:
Using 6 interface wires, I'll use 2 TLV320AIC23, $2.85 each, which will give me 4x 24bit ADCs, 4x 24bit DACs.
Inputs from amp will be put through a 0.1Hz low pass RC filter & R/R voltage divider, passed through an TLO64, then to the 4 ADCs.
The controler board will operate right from the rail supplies of the amp.
kevin gilmore said:
Kevin, I'm very interested in your automatic bias. Do you have a schematic, maybe? If it's large send me a private message.
I started a thread in the subject:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=3881
I had also an idea of perfect bias adjustment especially with MOSFET's.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.