Yet another boring Bipolar amp design!

[ceharden]

Hi there,

I've finally got round to starting designing an amp for myself rather than just fixing other peoples when the blow them up!

It's probably a bit early to start posting schematics but I've reached a milestone in that I've got a simulation of my own initial ideas working in Spice. There are many bits missing from the circuit to make it work (stable) in real life and the transistor choice is limited to what's in the parts library.

But here it is:

Amplifier schematic

The aim is to start with a design delivering maybe 200W into a 4R load from +/- 50V-ish rails. The prototype will depend on the 'doner' amplifier which I will be ripping the output stage out of and keeping the power supply and heatsink!

It's a fairly basic LTP input stage with current source and mirror into a class A driver with current source. The biasing is done with diodes in the simulation as I intend to use the new OnSemi Thermaltrak output devices. Output is a standard Siklai driver into what will be the NJL3281/1302 pair but currently simulated as MJ15022/23. Obviously the real amp will use two or more in parallel.

I'll post more as the project progresses but this is very much intended as a learning project before moving onto bigger designs. A big hurdle will be the PCB layout but one step at a time as they say!

Thanks for reading,

Chris

[AKSA]

Nice amp.

Flip Q16 vertically, and put a quality 1uF cap across the 100R resistor between the pre-driver emitters.

Cheers,

Hugh

[ceharden]

Thanks Hugh, also had one of the output devices upside down.

Schematic updated.

Really looking forward to prototyping this. Wonder if I could get away with Veroboard (stripboard) if I was careful?

Need to add a few small caps to limit the HF response, where is the best place to do it? Some schematics I've seen suggest from the output of the class A VAS to the inverting input of the LTP?

[AKSA]

Chris,

Try starting out with 22pF across the collector/base of Q5 (lag compensation).

Consider also around 10pF from emitter of Q10 to base of Q2 (phase lead).

You will have a much more accurate picture of the values if you analyse the circuit with LTSpice, now in Version IV, reputably far superior to Version III.

Veroboard is fine for a prototype.

You can manipulate stability for your chosen Lag Comp cap by raising or lowering the OLG with subtle changes to the LTP degeneration, R2/R3.

You may also need to drop one of the bias diodes, leaving three, and place a variable resistor in series with all three, as the pre-driver in the CFP will have lowish Vbe, and you will probably need around 2.40 volts bias to fire it up to say 60mA quiescent. Aim for 30mV across the 0.47 ohm output emitter resistors.

Hope this helps,

Cheers,

Hugh

[ceharden]

Funny you should mention the bias diode, just realised the quiescent current through the output stage had got a bit high! Something I haven't worked out yet is exactly how to use the diodes in the Thermaltrak packages. If I've got 4 devices (2 per side) but only need 3 diodes, do I just not connect one, or parallel it with another?

Thanks for the hints with the compensation. I understand most bits about how amplifiers work but I'm not really up on the theory and maths about why they do!

Can you recommend any other transistors for the input LTP? The BF423's only have a HFE of 50 or so which is a bit low for that part of the amp. I've seen 2SA1085's used in that position which are minimum HFE 250.

I'm thinking that 0.47R for the output emitters is a bit high but maybe I should start there with the prototype.

My other area of concern at the moment is the dissipation in Q8 and Q10. With 7mA of bias through that stage, Q8,10 are dissipating over 300mW which is quite close to their rating.

Cheers,

Chris

[AndrewT]

with R11=100r I would expect <6.5mA passing the CCS, maybe even 6mA.
Use To126 devices if you want to keep the near 300mW dissipation. They can do upto 5W to 10W with good characterics to suit VAS duty. Look at 2sa1360/c3423.
With the output triple, you have adequate current gain for any severe 4ohm load and VAS current could be reduced slightly 3 to 4mA? and still work adequately. This would allow To92 devices to be used. E-line can dissipate a touch more and cool better with a flag on them.

BTW,
Roender's 3pair (3281/1302) output stage would be an excellent model for a 200W into 4ohm amp driven from +-50Vdc. Your higher value of Re will allow it to run cooler and adopt a smaller heatsink (0.5 to 0.6C/W).

[ceharden]

Thanks for your comments AndrewT

I'm getting about 6.8mA through the CCS, but I've looked at the specs of the BF423 again and they have a maximum dissipation of 625mW. The highest dissipation of any transistor is 340mW. If I reduce R11 to 120R the idle current is just under 6mA and dissipation down to 280mW. Might be ok with the TO92 devices unless supply rails end up being higher. Really should measure the 'doner' amp's supply.

Ultimately I'm going to be moving to surface mount (this is part of a self-teaching project!) so that will require another search for parts....

I'm trying to simulate the amp with a load but it's going unstable in the simulation! Didn't think spice was that clever! Or it could be something wierd in my settings.

[CBS240]

Hi

I bet you could get away with less than 6mA of VAS bias. The base of the pre-driver shouldn't require very much current. I wonder if R11 is a bit too high. IMO, I would try to run the bias of Q5 at least about 1mA. A lot of small signal BJT's loose their gain and bandwidth with currents too much lower than 1mA. Well, assuming the beta of Q10 isn’t unreasonably low I could see about 700uA bias for Q5, might not be too bad.......

Good and simple.

Some superficial simple minor improvements might be to use J-fets for the input transistors. Then you might be able to use a smaller input cap, maybe polypropylene. A Wilson mirror instead of a simple mirror to load the LTP is a simple but worthwhile improvement. Hugh gives good advice about the compensation scheme, but it will be more useful to apply in the real circuit because PCB layout may affect this aspect of the circuit. So when you build your veroboard PCB, keep in mind that you will be changing these values to fine tune the compensation. Hope you can get it to work.

[MJL21193]

Originally posted by ceharden
I haven't worked out yet is exactly how to use the diodes in the Thermaltrak packages. If I've got 4 devices (2 per side) but only need 3 diodes, do I just not connect one, or parallel it with another?

Can you recommend any other transistors for the input LTP?

My other area of concern at the moment is the dissipation in Q8 and Q10.

Hi,
If you are looking for ~60mA idle current, 4 of the NJL diodes in series are needed. Like Hugh said, put a pot in series with them to adjust. I know from experience that the simulator will not give you the correct drop using the N4001's. Here's a better model to make a suitable diode:
.MODEL NJLXXXXD d
+ IS=5.5e-008 RS=0.65 N=2 TT=5.1937e-008 CJO=2.08272e-011 VJ=1.5+/n M=0.653713 EG=1.3 XTI=2.33354 BV=400 IBV=1e-005

You need a 100V+ device for the input. I have been looking for higher voltage inputs and found a couple of Panasonic units that are reasonable gain and 150Vce. 2SA1123 and 2SC2631. Should be easily available.

As others have said, lower your VAS current, then you will not need to worry about dissipation. BTW, good for this position is the 2SA1381 and 2SC3503. Current production and pretty cheap. Also good drivers are 2SC4793 ans 2SA1837. These are Toshiba and cheap but very good.
Goodluck!

[Nico Ras]

Hi Chris,

nice amp! very interesting choice of low power transistors, what made you decide on these?