My first discrete amp

Status
Not open for further replies.
Jump to Latest
#21
The C7 compensation cap seems pretty large at 1nF giving a low slew rate. If you reduce it you could increase R6 and R7 at the same time, do it right and gain bandwidth product (feedback level) can remain constant.

Here are some of my favourite amplifier design related sites,

http://www.synaesthesia.ca/
http://www.hardwareanalysis.com/content/article/1842/extrema-reference-class-a-diy-amplifier/
http://www.angelfire.com/ab3/mjramp/
http://www.cordellaudio.com/
http://users.ece.gatech.edu/~mleach/lowtim/

Also Doug Self's book if worth a read, a lot of practical info even if you don't like his point of view.
 
#22
i was going to say the same thing about the compensation cap..... but with a triple darlington the compensation cap will be rather large. also if you want to see a good paper on amp stability, look at the appendix about amp stability in Linear Technology's application note AN-47. it goes into all of the compensation/speed/stability issues in detail. the last appendix in there about Murphy's Law is pretty good too...
 
#23
Simulation results for Revision 0.5:

Frequency response:
FrequencyResponseRev05.PNG


Noise:
NoiseRev05.PNG


Poor Mans THD + Noise: 0.0793 %

After taking some of your advice here is revision 0.6:

The low pass filter values stays and so does the drivers. And I like to
overdo things so all the caps stays 🙂.

Schematic:
http://hem.bredband.net/b1438270/Amplifier/Rev06/AmplifierRev06.pdf

Frequency response:
FrequencyResponseRev06.PNG


Noise:
NoiseRev06.PNG


Poor Mans THD + Noise: 0.0157 %

Regarding the 100k resistor on the collector of the clip protection transistor Q7, do you thing its ok? I simulated clipping with 3V input and the maximum current through R14 was ~ 8mA.

With 1V input (and no clipping) the current through R14 is ~ 6mA.

I also tried lowering R12 but the distortion goes up and without it the results gets crazy. I also tried with 100R at the base of Q7 and R12 at 0R but still the results got crazy.

Oh and I wrote a matlab scrip that sort of calculates THD + Noise.
The script:
http://hem.bredband.net/b1438270/Amplifier/PoorMansTHDplusN.m

Hopefully this time all is ok so I can start drawing the pcb.
 
#26
Try doing and open-loop gain/phase response. To do that change the value of C1 to 0 (or some minimum value like .01 pF if the simulator won’t allow 0). Change C2 and C3 to 100,000 uf, and R11 to 0 ohms (if you're testing at the actual "ouput" node, change L1 to 0 as well). That will give some indication as to phase margin at unity gain.
 
#27
Hi,
try changing the ratio between C2 and C3.
either increase C3 to >=330uF
or
decrease C2 to <=2u2F

Try reducing VR17 to 500r to help resolution for setting the Ibias.

Try reducing R29 to somewhere between 2r2 and 4r7 5W

Add a resistor between Q8 collector and ground. This reduces the current in the EF when the amp is abused, i.e. stops it blowing up.

Try reducing C1 to 2n2F

Replace D1 & D2 with 1n4002 or 1n5402 or similar high current low cost diodes. May be use a pair of diode connected transistors.

Why have you chosen output devices for your driver stage?
Try MJE15034/5

Try adding a small cap from R8 and R18 to -ve supply.
or
splitting R8 & R18 and adding caps from junction of the split pair to -ve supply.
 
#29
Looking at the open-loop plot, with the closed loop gain you’re running (about 27 dB), it looks like your phase margin will be a bit over 80 degrees, which is pretty good (60 degrees is usually considered the minimum for a PA due to the reactive load it must drive).

Add a resistor between Q8 collector and ground.
Click to expand...

Good catch AndrewT, the power dissipation in Q8 can get fairly high during a clip (the current’s already limited by Q7). Looks like something around 430 ohms at 2 watts would work for a Q8 collector resistor.
 
#30
Revision 0.8:

Why have you chosen output devices for your driver stage? Try MJE15034/5
Click to expand...

The finished amplifier will look so massive with all those TO-264 devices 😎
And the drivers/output devices in question have really good specs. So why not?

Schematic:
http://hem.bredband.net/b1438270/Amplifier/Rev08/AmplifierRev08.pdf

Frequency response:
FrequencyResponseRev08.PNG


Open loop frequency response:
OpenLoopFrequencyResponseRev08.PNG


Noise:
NoiseRev08.PNG


Poor Mans THD + Noise:
0.0089 % @ 1 kHz
0.0058 % @ 10 kHz
 
#31
danfo098 said:
Revision 0.8:

And the drivers/output devices in question have really good specs. So why not?
Click to expand...

Layout for one. Two, it is not necessary to use a more expensive larger device with larger junction capacitances and lower ft for the purpose of driving the bases of the outputs. It is rather counter-productive. :smash:
 
#32
I have a plan for the layout so I don't think that will be a problem.

The MJL4281A/4302A have 5 MHz higher ft than the MJE15034/5 according to On Semiconductors datasheets and the junction capacitances I think I can live with.

And if I give in and change the drivers all originallity is lost and I at least want some of that 🙂.
 
#35
The MJL4281A/4302A have 5 MHz higher ft than the MJE15034/5 according to On Semiconductors datasheets and the junction capacitances I think I can live with.
Click to expand...

Hi

Ft is not a constant across the entire current range of operation. Looking at the datasheets for MJL4281 and MJE15034 and observing the GBP graphs of each, you can see that in the current range of operation when using the MJL4281 as the drivers, the Ft is way less than that of MJE15034. Unless you plan to run the MJL4281 outputs at a current greater than 4-5 amps, at which point their gain drops significantly because they aren't designed for it, the higher Ft transistor for the drivers at 100-300mA is actually the MJE15034. But then, don't let the facts convince you.:whazzat:
 
#38
The preliminary component layout is finished now. The board is 280x225mm and will be four layers.

The layer stack I'm going to use:

Layer 1: Signals.
Layer 2: Power.
Layer 3: Ground, will try to use the entire layer if possible.
Layer 4: Signals.

I also managed to squeeze in 16 more 1000uF caps in the power supply. 🙂

One thing that I'm not completley sure about is the footprint for the 2SA970/2SC2240, the datasheet that I have found isn't very specific so I might have turned Base/Emitter the wrong way?

I'm also having concerns about the close proximity of the transformers, that they will cause 50 Hz humming noise, what do you guys think about that?

Schematic rev 1.0 with preliminary component layout:
http://hem.bredband.net/b1438270/Amplifier/Rev10/AmplifierRev10.pdf

BOM for rev 1.0:
http://hem.bredband.net/b1438270/Amplifier/Rev10/AmplifierBOM.pdf
 
#39
You’re only using 15 VA transformers for your main supplies?? (jaw drop) Most guys use hundreds of VA. It’s also common to see power transistors mounted at the edge of a PCB for easier heat sink mounting. Q110 should be in a package that can easily mount to the heatsink also, such as the MJE340, for thermal tracking purposes. Maybe do some more reading before you pour money into PCBs at this point.
 
Status
Not open for further replies.
Top Bottom