oscillating...always oscillating

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Alright, so I designed this simple symmetrical small signal amplifier and I can't for the life of me get it to stop oscillating. Image of circuit is attached; any ideas?
 

Attachments

  • hpa1_1.png
    hpa1_1.png
    5.3 KB · Views: 404
This circuit has some problem. If You use this VAS stage, You can't put current mirror over the input stages. Replace them with simple resistors!
Is oscillation still exist, use compensation (c-b capacitors for Q10, and Q11), or use emitter degeneration resistors for the input stage, or resistors between the base of the outputs and the ground. The last two solution reduce the open loop gain.

sajti
 
You need to add feedback compensation in one form or another. The bulletproof way is to put a 100pf or so cap from base to emitter of both VAS transistors. Another way of doing it is to use feedforward, a small 20pf or so cap from the VAS collector (either) to the negative feedback node, bases of both transistors.

In passing, you have way too much output bias. Only two diodes are needed to compensate for 2 Vbe drops of the output devices. And that's perhaps too high, depending on the actual part variations. You really need an adjustable Vbe multiplier to set bias. Most amps have them.
 
I think the biggest problem is the current mirrors. With current mirrors, there is nothing to set the bias for Q10, and Q11. Solve this problem first! You have too high bias currant for the input stages (I count it about 8mA!!!) Reduce it about 2mA, or lower. Set resistors instead of current mirrors, with about 1kohm. This will results 7mA for the Q10 and Q11.
Use 330-470ohm emitter degeneration resistors for the input stages. This will increase the input impedance, and lower the distortion in the first stage, and also reduce the OL gain.
I think that 30mA bias current is OK for the output stage. They works in class A in most of the application. If the oscillation still exist, use some 22-47pF between the c-b of Q10 and Q11.

I don't know if You interest but I think non feedback buffer would be better.

sajti
 
Oscillating amp Design

Hi WoulYaShassu,
That's funny as you designed an amp just like the one I was going to try. I don't see a problem with the current mirrors. I would use a Darlington connected pair for the VAS transistors.
I agree with slowhands that you should add some compensation from the collector of the VAS transistors to the inverting input. Is this feedforward? I would could this feedback. Also a few 10's of pF over the feedback resistor may help.
It is absolutely important to decouple the VAS and inputstage from the outputstage by means of a small resistor of 10-30 Ohm and a decoupling cap or by a capmultiplier as in John Curls Parasound amplifiers.
Your circuit has a gain of one but could easily modified to a circuit with gain by adding a resistor Rg from the inverting input to ground. This will also improve the stability situation! Gain = 1+ R15/Rg.
I calculated some 6mA for the current sources for the inputstage which is a bit on the high side. Something like 2mA will do. See this thread for an idea making current sources with a LED and a transistor but it could also be done with FETs as Sonnya does.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=1794&highlight=
Two diodes will do for bias setting of the outputstage unless you are after class A. These diodes should be in close thermal contact with the output transistors or else you will get huge thermal runaway. Diode connected (base tied to the connector) transistors make better thermal contact. An amplified diode can be made adjustable but should be bypassed with a good cap.
:)
 

Attachments

  • hpa1_1[1]-gif-sm.gif
    hpa1_1[1]-gif-sm.gif
    29.9 KB · Views: 557
Hi WoulYaShassu,
I like the current mirrors also, Do try the phase compensation from one VAS collector to the bases of the feedback transistors. You will probably also need a global phase compensation capacitor across the global feedback resistor, as Elso mentioned. Its value may be about half VAS stage one is. They may be about 10pF and 4.7pF, respectively, give or take several pF. Also, place a 0.1uF capacitor across your bias diodes. If you are not using diode references for your constant current sources, you may as well just use a single resistor to each power rail. Also, don't forget to voltage divide the global feedback so that your circuit has gain. You can use a capacitor from the lower resistor to ground to preserve a unity DC gain.

Best regards
 
My problem, that the bias of the Q10, and Q11, can be anything between 0, and 90mA depends the collector voltage of the input stage... (I think that it's not so stable, if the temperature is change)
I don't use current mirrors, but i use double emitter follower between the input, and the VAS stage, if I need high speed. One of my last design produced 2MHz (-3dB) with this tecnology. The most important pole between the VAS, and the output stage, and that is not easy to pull up. I use high current VAS, with low resistors at the collector of the VAS trannies.

sajti
 
Yes but the part count is low enough that after the transistors are selected for appropriate gain, the current mirrors can be bonded with the VAS with silicon or something.

I played with the emitter follower buffers on the simulator. Now, I know that is not as good as real circuits, but the circuit without the buffers as shown by the simple configuration above, was way more stable. The key is to have as few active silicon junctions in the circuit path as possible. Sure, distortion suffers, but the total number of harmonics generated may be less with less stages too. I have almost always found that the lower one tries to make the distortion, the more instbility increases. Hmm, dunno.

But I have no doubt that your circuits work very well. I am just a minimalist and am willing to have trade-offs for small circuit size and simplicity. So, do not feel that I am knocking your ideas at all.:smash:

So also feel free to explain your ideas. I for one, do like to examine and who knows, may turn out using them.
 
I've built this kind of amp, and the result is IT WONT WORK. This is exactly the problem with Randy Slone's controversial design. I've built it and it produces big DC, not music.

Sajti is right. Anyone can tell what is the voltage in the base of Q11 and Q12? The VAS just dont have any reference, so it "dont know" its operating point, that is why it will always oscilate.
Replace the current mirrors with resistor, (or maybe put the resistor in the emitors of current mirror and take the base of VAS from this resistor drop, if you still want current mirror in it?)

Can anyone tell what is the voltage between 2 collectors? In this case the voltage of base Q11 and Q12?
 
AX tech editor
Joined 2002
Paid Member
lumanauw said:
I've built this kind of amp, and the result is IT WONT WORK. This is exactly the problem with Randy Slone's controversial design. I've built it and it produces big DC, not music.

Sajti is right. Anyone can tell what is the voltage in the base of Q11 and Q12? The VAS just dont have any reference, so it "dont know" its operating point, that is why it will always oscilate.
Replace the current mirrors with resistor, (or maybe put the resistor in the emitors of current mirror and take the base of VAS from this resistor drop, if you still want current mirror in it?)

Can anyone tell what is the voltage between 2 collectors? In this case the voltage of base Q11 and Q12?

Yes, the ref for R11, R12 seems undetermined. But it DOES work, otherwise there would be no oscillation. An oscillator is just an amp where the neg feedback has turned to pos fb, because of the phase shift, at a freq where there is still gain.

It is not sure to me whether this oscillates in the sim or whether it has been build. If it works in the sim, that may be because the transistors are perfectly matched and have infinite thermal capacity. In real world, it may just disappear in a puff of smoke.

Jan Didden
 
I agree. It can work, but maybe the bias of the VAS can be 0.1mA or 50mA depends by the parameters of the input, and cm transistors. Due the very high closed loop gain, the output DC level will stable and will be close to 0.
And if You start to use it, all the internal parameters will change time by time, maybe with 100% or more....

sajti
 
Hello WoulYaShassu

I am assuming you are trying to build a pre-amp and not a power amp.

With that in mind, I have some suggestions which you may like to consider. You can:

1) Do away with the driver stage Q13,14
2) Remove the 3 diodes at Q11,12 and take output at this stage instead. NFB also taken at this stage.
3) Set close loop gain with resistor to ground at inverting input.
4) To make your life easier, do away with the mirrors and use resistors. Get it to work first.
5) You may also need to incorporate some forms of DC offset adjustments, servo or otherwise.

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