Music Electronics

This blog will no more be updated.

If you use a mac for your electronics design you have probably tried to install LTspice on your mac. Like me you’ll find that it’s quite different from the version that runs in windows! No menus and you have to be able to remember some syntax to make your simulations. This sucks. 

So now i just installed the windows version on my mac, and it works just fine!

Go give it a try, or ask me for the LTspice IV.app file to avoid having to make it your self. 

TMBT3906/TMBT3904 are bjt’s in SOT23 package and they seem to have quite ideal specs for low distortion amplifiers. They can only stand 50Vce, but often ultra low distortion calls for cascode stages anyway, so you can keep these safe from high voltages that way. 

As I’m struggling with the issues of current feedback in power amplifiers, it’s starting to dawn on me why Douglass Self is doing it the way he is. 

He really is achieving the needed performance with a remarkable simplicity and low cost. 

One obvious issue is that the current feedback amplifier will typically require a very low value resistor from the output to the negative input. - In the 100R to 2k2 order. This is because the open loop gain is heavily dependent on this resistor. The lower its value, the higher the open loop gain. And in a 200W amplifier you’ll end up having almost about one watt of power dissipation in your feedback resistors! This is pricey and might cause thermal distortion at lower frequencies. I presume the cheapest way to do it is to use a few resistors in series to create the desired value, but this is still very expensive with thin film or metal film resistors. 

Here is my latest circuit and what looks like a succesfull attempt at getting the open loop gain above 100dB. 

View larger

I’ve taken the output stage out of the equation because it really doesn’t have much to do with this part at all. So for practical tests I use an LME49610.

I’ll have to experiment a little with C3 and C4. They increase open loop gain, but not at lower frequencies, and they might introduce low frequency distortion as they can’t really be “large enough” - They’d have to be above 1000u to work well at low frequency, which is just not practical. 

I’ve also connected the frequency compensation capacitors back to the input rather than just to GND for some negative feedback on the gain stage. On final versions I might make room to populate both options. Which option works best might depend on the performance of the output stage. 

A current feedback has an input that consists of a unity gain buffer and two current mirrors in the configuration you see here: 

The buffer marked Buf1 has a quiescent current running in through +V terminal and out -V terminal. This current is mirrored so the same current runs from Q3 collector to Q4 collector. 

When +IN is higher than -IN, a current will flow out of the -IN terminal (this is the error current, hence the name current feedback). This current causes the current in terminal +V to be higher than the current in terminal -V. This in turn causes the upper current mirror to source more current than the lower one sinks. This difference in current is flowing to capacitor Cdom and so the voltage on Cdom rises and the output voltage follows and rises too. And so +IN > -IN = Vout rises, and vise versa. And so when you connect the output to the negative input you get a circuit that functions very much like a conventional opperational amplifier. 

It seems like a complex and expensive approach. But it’s probably a good circuit. And in its own right simpler and less convoluted to make than many of their other circuits. I’ve only ever build one ROG circuit, but I generally think who ever designs those circuits are full of interesting ideas. But I very much prefer that they now have chosen circuit that can be reliably replicated without having to individually trim the bias of J-fets that has more or less gone out of production. That seemed a little unnecessary. 

The LT1054 seems like an odd choice though. Isolated DC/DC converters with +/- outputs aren’t very expensive, and they have the benefit of eliminating the ground loops when using a wall wart for powering pedals. I think I would have gone for that option, especially for a DIY optimised circuit. 

Edit: 

I just checked, and sure enough, from my supplier, a Murata isolated +/- 12V 42mA DC/DC converter is cheaper than a LT1054! Go figure… I’d strongly recommend using a isolated converter. For any pedal, actually, regardless of wether a negative supply is needed, since it removes ground loops from using wall warts for powering it. 

PDF here

It’s a no budget solution, USB powered and it has a Baxandall volume. I made it because it bothers me that most computers headphone output has pretty decent performance when it’s unloaded and at max volume. And perform horribly at any other condition. So this allows the system volume to be turned to max (where it should be) and then you adjust the volume with an amplifier with adjustable gain, rather than with a noisy variable attenuator. 

My boss have decided that I’m going to use a different schematic capture and PCB layout software. 

If you’re not opting for one of the IR gate drivers that takes care of all this for you, here’s how to do it. 

You might for example want to use an iCoupler gate driver like ADuM3224: 

It has quite remarkable specs along with two fully isolated gate drivers, 43ns typical propagation delay, <5ns delay matching between the two fully isolated gate drivers! In other words, very ideal for class D. And not too expensive either. 

And this tiny bit of logic will take care of the dead time: 

This diagram is just conceptual, you’ll of course have to add the necessary driving circuitry, but I just added the mosfets for clarity.   

NEW!! 

Try a simulation of the circuit Click Here

This is great for when ever you need a tiny amount of power from a high voltage rail, cost optimisation or high reliability applications.

It will regulate to roughly the voltage of the zener. 

If you have tight tolerances for your supply voltage but still need the high input voltage of the discrete solution, just put a regular integrated linear regulator after it. 

Edit: I forgot that stability might not be trivial to any one who would need this circuit. This regulator of course need a miller cap on the upper most p-channel mosfet.

I’m developing a class D amplifier, and I have currently solved the problem of keeping the common mode output voltage zero at all times. I’ll write up on that later. Here’s something that i happen to find more interesting. It’s the general idea that a differential amplifier does not need to draw any ground current at all! 

This makes for much simpler and better pcb layouts since there is no need to deal with splitting the audio ground and keeping the switching ground the same. (yep that shit is actually possible. You make slits in the GND plane, and place ceramic capacitors across it. The switching currents will pass through the caps, but the LF audio signal will not!) 

Here in stead you make all decoupling between the two power rails and keep ground completely out of the picture on the output side. So the output filter capacitor (C3) is not connected from ground to output, it’s connected between outputs like you see here: