My first discrete amp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hello!

This is my first post here and my first discrete amplifier also.
I have built a couple of Gainclones in the past but now I thought it
was time for a "real" amplifier ;).

The constant current sources are set to 5mA but I'm not sure if that's ok?

Thoughts, comments, questions, all feedback will be appreiciated.

Thanks.
 

Attachments

  • amplifier.zip
    64.7 KB · Views: 469
looks pretty simple and straightforward.....
you have a good opportunity to thermally couple the input transistors. put a sil-pad between them and bolt them together.

you might want to run the VAS current source at 10 or 20ma, maybe higher

you might want to beef up those drivers (Q11, Q12) though... with a beta of 25 on the output devices, you'll be pushing those drivers hard......
 
I changed the drivers to pre-drivers instead and increased the VAS constant current to 25mA. I also added an extra 10R in parallel with the output inductor and a couple of protective diodes in parallel with the output drivers.

Feel free to give me more feedback, I'll start working on the PSU for now.
 

Attachments

  • amplifier.zip
    69.1 KB · Views: 127
danfo098 said:

Hello :)

Thoughts, comments, questions, all feedback will be appreiciated.
Thanks.

Looks like a reasonable circuit to start with. I'd like to add to what has been suggested. It is a good idea for the dc resistance seen by the bases of Q3 and Q4 to be equal. This is because as the temperatures of Q3 and Q4 drift, so will their betas and their base currents. If the dc resistance is different on each base, then the base voltages will drift apart causing the output offset voltage to drift. Does this make sense?

IOW, the value of R1 and R4 should be equal. If you make R4=10k you can make R5=330 for a similar closed loop gain.

Regarding C2. The base currents of Q3 and Q4 will cause a small voltage drop across R1 and R4 which will cause their bases to be at a slightly negative voltage. So if you use a polarized cap for C2, the positive pin of C2 should go to GND. BTW, my experience is that a non-polarized electrolytic will sound a little better here, if you have one. 10uF is ok but a little on the low side - 50uF would be better.

Brian
 
your compensation cap should be adjusted for the increased miller capacitance from adding the intermediate drivers. that will have to be found by first running the sim (which is what i'm guessing you did to get the original value), and again you might have to change it a bit when the circuit is built. remember, if you use a sim to design a circuit, it's pretty much a "ballpark" model. there will always be things that have to be changed with real components. i've actually done pretty well by sketching a circuit out on a piece of paper and building it, and using a spice program helps me save money in smoked components ("there's 10A flowing through a 2N2222?????"). but even without a spice program, you can do pretty well with "off the cuff" designs as long as you have an idea why some things are done the way they are. i'll have to dig out some old circuit design books i have, but if you run into a bandwidth problem with that diff amp, there are ways to fix that that are ultra simple that just seem to have been forgotten.
 
With no kind of limiting on Q7 & Q8 you may end up blowing both devices with a positive clip. It’s also a good idea to have a low pass RC filter on the input to minimize parasitic oscillation caused by stray feedback from the output. Some people use 50KHz as a rule of thumb for this filter, though there is no standard. You’re also running a huge gm in the LTP, loop oscillation is almost assured. It’d be a good idea to put some 100 ohm resistors in series with Q3 and Q4’s emitters to reduce the gm to manageable levels.
 
With no kind of limiting on Q7 & Q8 you may end up blowing both devices with a positive clip. It’s also a good idea to have a low pass RC filter on the input to minimize parasitic oscillation caused by stray feedback from the output. Some people use 50KHz as a rule of thumb for this filter, though there is no standard. You’re also running a huge gm in the LTP, loop oscillation is almost assured. It’d be a good idea to put some 100 ohm resistors in series with Q3 and Q4’s emitters to reduce the gm to manageable levels.

Right on the money... what we have is a 50khz oscillator (attached)your exact schema - LT4,
emitter degeneration is your friend :D use it often.
the OPL gain is 160+ db.. also 20+ ma for the VAS is
a little heavyhanded.
next post I will get it to work with a few friendly, cheap resistors.
OS
 

Attachments

  • oscillator.gif
    oscillator.gif
    15.6 KB · Views: 1,094
I'm starting to lean towards using small signal devices now, I just can't decide. I'm going to have to sleep on it. I chose a pair that had nice specs, they're in the schematic down below.

And regarding the 100pF Cdom/compensation cap I haven't done any simulations I just chose 100pF because most schematics I've looked at usually have that value.

I also added a low pass filter with cutoff at 48kHz and som extra resistors that kaos suggested.

Revision 0.4:
http://hem.bredband.net/b1438270/Amplifier/AmplifierRev04.pdf
 
I'm starting to lean towards using small signal devices now, I just can't decide.

yes , much faster are the 2240/970's (as yoda says) about
dead even with the 1845/992 's that I use. keep the
larger devices for your VAS (BD's or MJE340/50) and Vbe.
BTW .. phillip's BD139/40 are discontinued devices..:bawling:
here is the version 3 with the gain changed (RE's) and
VAS at 8ma...(attached)

Edit: for such a small amp the EF'ed VAS would create a square
wave (too much gain) even without it i had to increase your global feedback (4.7k) to keep from clip at .7V input.. that why i removed
it. On the "big boys" (+-60 or above) this extra gain is ok as
you have to produce a VERY big voltage swing..

OS
 

Attachments

  • works now.gif
    works now.gif
    51 KB · Views: 1,242
danfo098 said:
I'm starting to lean towards using small signal devices now, I just can't decide. I'm going to have to sleep on it. I chose a pair that had nice specs, they're in the schematic down below.

And regarding the 100pF Cdom/compensation cap I haven't done any simulations I just chose 100pF because most schematics I've looked at usually have that value.

I also added a low pass filter with cutoff at 48kHz and som extra resistors that kaos suggested.

Revision 0.4:
http://hem.bredband.net/b1438270/Amplifier/AmplifierRev04.pdf

that's ok.... as you learn WHY things work the way they do, you'll have a better understanding HOW to select the right component for the job. everybody is here to learn, and part of that is "bouncing" ideas off of each other..... i've seen compensation caps all the way from 5 or 10pf all the way up to 1000pf it just depends on the devices used and the diff amp and VAS currents, and the miller capacitance of the output stage. these all have an effect on what that capacitor should be. some people even add an integrating cap in the feedback loop, where others don't like to have reactive components in the feedback loop (i don't like reactive components in the feedback loop, but i'll use a small integrating cap if it's absolutely necessary).
 
Here we go again :)

I found some transistors that i liked even better, they are in the schematic. This time I have done some simulations also mainly to se that it doesn't oscillate and that the crossover distortion is ok.

Revision 0.5 schematic:
http://hem.bredband.net/b1438270/Amplifier/Rev05/AmplifierRev05.pdf

If you want to do som simulation on this amplifier I'm including spice models and schematic.

LT Spice IV schematic:
http://hem.bredband.net/b1438270/Amplifier/Rev05/AmplifierRev05.asc

Spice models that I used:
http://hem.bredband.net/b1438270/Amplifier/Rev05/standard.bjt

Simulation screenshot:
Spicesim.PNG


I got the spice models from ON Semiconductors webpage for all models except for the 2SA970/SC2240 which I found on this forum.

Simulations on my todo list:
THD (don't know how to do that yet though...)
Frequency response
And maybe some more that you will tell me about

Thats it for now, time to sleep.
 
Personal suggestions :)

The 2SA970/2SC2240 works well. Also good performers are Fairchild's KSA992/KSC1845 pair (clones of the equivalent 2SA/2SC devices). For a basic amp like this, even the "old" 2N5401/2N5551 or BC546/556 work great too.

Use small signal devices for the mirror as well. Thermally couple them.

Q5 could stay as MJE340 even though a small signal device should be plenty. It's reference Q6 should be a small signal device, or use an LED which is my personal favorite.

Q7 is fine, Q8 should be an MJE350 or BD140. Q10 fine (or BD139), Q11 again, a small signal transistor or LED. If you want to push the boat out here, look at KSA1220/KSC2690 from Fairchild or 2SA1209/2SC2911 from Sanyo.

For Q14/15 look at MJE15032/33 or MJE15034/5 from OnSemi.

Don't overdo it with the charge suckout caps across the resistors. A good quality 1uF polyester cap across the drivers Q14/15 resistor is all you need here. Likewise, dont overdo it with the capacitors across the VBE multiplier - either a 10uF electrolytic and a 100nF polyester, or just a good quality 1uF polyester.

Q9 - small signal devices are better, but i prefer to use a TO-126 package simply because it's handier to mount. BD139 or MJE340 work ok. If youre going to use a small signal, drill a TO-92 sized hole in your heatsink and stuff the body in there with thermal grease.

edit: BTW OnSemi's models for BD139/140 are crap. Get Fairchild's ones
 
You still have a self destruct problem with Q7 & Q8 using the 2SA970s if the amp clips hard positive. Under a positive clipping condition what happens is that Q3 will come on hard injecting approximately 6 mA into the base of Q7, turning it on hard. Q7 coming on hard will inject a large current into the base of Q8 and R13, essentially 30V – 2Vbe over 100 ohms, or approximately 288 mA, destroying Q8 (the 2SA970’s are only rated for 20 mA base current, 100 mA collector) and probably taking Q7 along with it. To resolve this issue place another PNP device between the +30 V rail and Q7’s base. Connect the base of this new device to the junction of R13 and Q8 (you may need a 100 ohm base stopper resistor, hard to say). Change Q8 to a heftier device that can take a 60 mA base current (the MJE350 will work as one possible option, as jaycee recommends, for Q8, there are many that can be used there though).

The R1/C1 combo of 220 ohms and 10 nF is a pretty heavy load for a pre-amp to drive, most probably increasing HF distortion in the pre-amp. Maybe make an adjustment in values there.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.