F5 power amplifier

Moderator
Joined 2002
Paid Member
@jacco - reason is, if I increase the bias, I may not get additional power output if the amplifier has already reached its voltage swing limits. However if there is some swing to spare, I could probably get a little more output. I have 8 ohm speakers that are a gentle load, but very very low efficiency. Let's just say each watt counts :)
 
Someone asked me about an unbalanced F5 with Toshiba MOSFETs.

Well, I have not built one, and I am honoured that you would ask me instead of asking Nelson.
But if I were to build one (unbalanced), I would do the following :

1. Stick to 0R47 source resistors
2. Keep the drain resistors of the JFETs the same as the original circuit.
3. Reduce the JFET bias (by using an Idss of say 7mA) to account for the lower Vgs of the Toshibas (around 2.xxV for 1.3A).
4. If you want to play the gain balancing game, then use 5 ohm more on the source of the 2SJ74s, and scale UP the feedback resistor for the negative rail half according (so that closed loop gain for top and bottom halves are the same).


Hope it's clear,
Patrick
 
Ex-Moderator R.I.P.
Joined 2005
F5 is specifically designed to drive relatively efficient 8ohm speakers, but takes on any kind of load, thus 2ohm stable

Honestly, I will only use my F5 during the cold winter time
Wouldnt dream of using it on hot summer days
Like the tube guys with heavy 845 PP amps
They usually have what they call "summer amps"

Here we have a season fore just about everything
 
I don't have any evidence that midrange magic is a
function of matching. Perfect matching reduces 2nd
harmonic. Those SE tube designs with the coveted
midrange are supposed to have more than their share of
2nd harmonic.


:cool:

:D

the whole amplifier will run 15Cdegrees hotter when your day time temperatures are @ 38degC than when it is set up for daytime temperatures of 23degC.

If the Ts=50degC at set up then it will run @ 65degC when Ta=38degC.
From there you will find that Tc and Tjmax will both be 15Cdegrees hotter as well.

The thermistors will trim back the hot running current slightly, that is why they are there, but they cannot temperature compensate for big changes in Ta.


At what temp will the thermistors start to reduce current ? Is it possible just shut the amplifier PS down instead of any current restrictions ?
 
...At what temp will the thermistors start to reduce current ?
They do that all the time - whenever the ambient temperature changes, they change their resistance, when rasistance falls the Vgs of MOSFETs gets lower and Id through MOSFETs decreases. Google NTC thermistors - some thermistors are PTC but we don't want them here ;)

...Is it possible just shut the amplifier PS down instead of any current restrictions ?

Yes it is (with thermo-switch), but why would you want to do that? F5 has built-in over-current protection.
 
Member
Joined 2003
Paid Member
Yes, you can leave both current protection and temp-comp out. Without temp-comp it takes longer to properly adjust. Some say that without the extra "stuff", that it sounds better, I wouldn't know, I built it this way and it sounded great.
 

Attachments

  • F5-no tempco-no current limit.JPG
    F5-no tempco-no current limit.JPG
    30 KB · Views: 928
Moderator
Joined 2002
Paid Member
Right. It's the same temperature, or even slightly higher. I wish the bias was a little more stable, the temperature in my room varies quite a bit (10 degrees C) through the day. I will try moving the thermistors closer to the transistors. It is pretty much certain I can't run without them.
 
Does anybody else see the amp run cooler with a load connected and music playing, or is it just me?

Hi sangram
It's not your imagination. Class A amps draw a constant amount of power and when no music is playing this is dissipated as heat. When playing music some of the power is used and less heat is produced. Not as concise and elegantly put as AndrewT's equation, but less cryptic. Class A motto: "Play loud and be cool!":cool:
 
Moderator
Joined 2002
Paid Member
I guess I'm being a pain in the butt, but this is what happens when I'm adjusting the bias.

An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.


These shots are taken a few seconds apart, the bias and the offset both drift up and down, plus there's a lot of hysteresis in the adjustment. Meaning once I move the pot, there is an immediate response, then after some time the bias tries to go back in the opposite direction, with the offset continuing on its journey wherever it was going, after some time things settle down with very little movement, but it's very tough to reach a precise adjustment.

Question for the fine gentlemen reading this: what would you term as acceptable for the drift of the bias/offset? +/- 20mv/10mV, +/- 10mV/5mV, +/-5mV/1mV respectively? I was shooting for the last set but can't quite seem to nail it with that precision just yet...

Or should I even worry about it as long as it's within parameters of +/-50mV offset? I don't like more than 10mV offset, but maybe that's just OCD kicking in.

Thanks for your help.
 
One thing to be mindful of -- as the resistors heat, their impedance changes! You have a lot of current through the feedback, gain and source (drain) resistors so the resistance is all affected -- the 3W Panasonics are spec'd with a tempco of 300ppm, but for the 0.47 ohm I've found it to be much worse (I am using a calibrated HP3456a DVM in 4 wire mode.) The 2W are 350ppm. I measure the current through each channel with a fused dmm. I've built 8 of these amps now and usually find that the optimal bias is a bit lower than spec'd in the article. I am using heat sinks with a thermal impedance well within the bounds suggested by NP.
 
Moderator
Joined 2002
Paid Member
@Jack: Thanks for that. I am trying to bias it near the 600mV point in the article but yes, it's a bit tricky. Do you think it'll help to go to the 50ppm Mills Audio resistors? they're spendy, but maybe I'll do it for my next build. My heatsinks are also just within the thermal spec - I'm probably doing 22-24 degrees above ambient.

@Zen Mod: Thanks. I don't think it is oscillating (I don't have a scope but the sinks run at normal temp) as the temps are within control - but yes, I do have flyouts from the PCB to the transistors, maybe I'll rework those tomorrow.

Just to clarify, my offset fluctuates between 0.590 and 0.610mV, a 20mV range. The offset is usually between 3 and -3mV, but sometimes hops to a larger number (20mV or so) before dropping back again. It's almost as if the transistors are fighting each other to maintain bias, and then as the FET grows hotter and reduces bias, the other one starts to take over.

My pics are terrible (and so is the build strategy) but here they are:

An externally hosted image should be here but it was not working when we last tested it.


I finally had to rip the FETs off the board because I could not stick to the original plan of opposed heatsinks. They are now mounted with stranded 15AWG copper wire for the Drain and Source, and single-strand wire for the Gate. So obviously it does not look as pretty anymore: (

An externally hosted image should be here but it was not working when we last tested it.


This is a shot of the finished amp - it sounds amazing so I don't mind how it looks... when it's retired in summer I'll look at a chassis concept.

An externally hosted image should be here but it was not working when we last tested it.


An externally hosted image should be here but it was not working when we last tested it.


No hint of grain or harshness, just pure music. It was obviously created for rock - I finally have an amp that makes my vintage rock recordings sound good again :) So no complaints really, but was worried that it might be me doing something wrong to get unstable bias - if, indeed it really is that unstable - I've never built a discrete before so don't have a basis to compare, and chipamps always give me pretty steady offset readings (though I can't actually measure the bias).

I have yet to mount the thermistors close to the MOSFETs as I'm waiting on some Arctic Alumina. I'd be happy to hear any other ideas or suggestions. Once again, the amp sounds great and I'm probably just being a little obsessive. But that's probably because it's just my first discrete build, and the first amp I've ever heard that sounds so good. I just don't want to be sitting and checking offset every two days - though I'm willing to do that if that's what it will take:)