I have some stuff to say.
First, the guy who said the input stage resistors are tweaked is VERY wrong... They are high only for high gain purposes. Even for a 1k resistor the transistors are biased and working prefectly on active mode.
Second, I improved frequency response by replacing some capacitors with larger ones.
Third, I made an emitter follower as input stage, and replaced with a coupled pair the gain stage transistors.
The new schematic will be shown later, but keep in mind taht one was only a scketch, not the real one I'm about to build. Of course I won't build anythink before I'm sure it'll work and be safe.
First, the guy who said the input stage resistors are tweaked is VERY wrong... They are high only for high gain purposes. Even for a 1k resistor the transistors are biased and working prefectly on active mode.
Second, I improved frequency response by replacing some capacitors with larger ones.
Third, I made an emitter follower as input stage, and replaced with a coupled pair the gain stage transistors.
The new schematic will be shown later, but keep in mind taht one was only a scketch, not the real one I'm about to build. Of course I won't build anythink before I'm sure it'll work and be safe.
ashade said:
Hi Ashade,
I think I was that person
. I was referring to the fact that one bias resistor was 200k and the other I think was 35k. Why are they what they are, and why not equal? What Q2, Q3 collector DC voltage does that give? Probably when they were equal, that collector voltage would either be close to the +15 or close to the -15? (I would vote for the -15, since NPN's generally have higher beta's).Jan Didden
ashade said:
Ashade,
Something I forgot to mention. You probably increased the output cap to get lower freq response. This type of cap is normally used to block DC in amps that have only a single supply voltage. The output stage is biased to midpoint of the supply to get max output level, and the cap is needed.
In your case where you want to use dual supplies, the cap can go, saving you money and improving the lf response (flatter and less phase shift). The output stage then should be biased close to ground for DC. There are several ways to do it, very often you see an opamp used as a DC negative feedback amp to null the output DC.
Jan Didden
ashade said:
Well, I'm not sure how to do it without a total redesign. IF you can get the first stage biased to about gnd, you can DC-couple to the output followers getting rid of C7 and C8 as well and including a bias network. Anyway, you have to bias the stages to about gnd to get symmetrical output levels, so even if you wanted to keep C2, which you wouldn't, the need for accurate, stable bias is still there.
To get stable bias to prevent thermal runaway you really need small (< 0.5 ohm) emitter resistors in the output followers and a bias setup at their bases.
Jan Didden
doing an ac analyses on the circuit with an 1mF output capacitor, I saw that the amplify is constant for about 200Hz+. So, I'm thinking about plugging a high-pass filter at the input of the amp, with ressonace at 100Hz, with a gain of about 2 times in the ressonance frequency. I think this should work. Maybe the filter won't be stable because of parameter instability, so this could not be a good solution, but it's still a solution.
ashade said:
Sorry, but this is elementary. You have a 1000uF output C with a 4 ohms load. That gives the lf roll off at 40Hz. There is also the 1000uF across R2 and R1, but that has already less than 1Hz 3-dB point. So where does the 200Hz come from? Can you look at the freq respoonse at the first stage? It is pretty useless to start talking about filters and whatnot if you don't know what the problem is.
Jan Didden
I think it would be pointless to fix something that is not going to work in the real world.
Jan is right in the sense that the DC output from the 1st stage is high unstable: the current going through the two transistors should be the same, but are determined (independently tho.) by the two resistors. So what you end of having is to find the collector current at which the two transistors' Vce and voltage drop off the RE resistors will be 24v. When the two transistors are different (hft, Vbe, etc.), you will have one of them less turned on than the other, and large swings in the DC output voltage.
It would be a much better learning experience for the original poster to understand how the 'box' works so that he can do a better job thinking out of the 'box'.
Jan is right in the sense that the DC output from the 1st stage is high unstable: the current going through the two transistors should be the same, but are determined (independently tho.) by the two resistors. So what you end of having is to find the collector current at which the two transistors' Vce and voltage drop off the RE resistors will be 24v. When the two transistors are different (hft, Vbe, etc.), you will have one of them less turned on than the other, and large swings in the DC output voltage.
It would be a much better learning experience for the original poster to understand how the 'box' works so that he can do a better job thinking out of the 'box'.
this is the new schematic with some changes proposed here and some changes by myself. I've put an emitter follower at the input with about 65k input impedance (MPSA18 is biased at Ic = 1mA). All the voltage souces are +-15V now. I changed the gain stage transistors to the coupled pair BC459C/BC559C. An temperature sweep told me the amp is working with not significant chages from -40ºC to 100ºC. Resistors R7/R8 can vary a lot without modifiying too much the output signal (a parameter sweep test told me about +-30% changes do not affect tha output that much).
Now you can go on and post your comments talking about what is or is not wrong in the design. Even you might think it's not the best way to learn, I can tell you that I'm learning a lot with your comments here and I would apreciate a lot if you to be keep posting on this topic.
Now you can go on and post your comments talking about what is or is not wrong in the design. Even you might think it's not the best way to learn, I can tell you that I'm learning a lot with your comments here and I would apreciate a lot if you to be keep posting on this topic.
Attachments
Ashade,
OK, you asked for it:
What are those units on the magnitude scale of the freq response? Volts? Can you do it with a dB scale? I'll bet you go down to 30Hz or so with less than 3dB loss. Not good, but not as bad as you think. C1 is still your limiting factor, throw it out!
What is the DC level at the U3, U4 collectors? Fix that and you can throw out at least 3 electrolytics. I still think you are trying to fix without knowing what the problem is. Whatever you do in input buffering has absolutely no impact on the freq response and the bias problems. Why not first look what's broken and fix THAT? Then you can get fancy.
Temp plots are totally irrelevant now because the amp will never work the way it is now. Waste of time.
Jan Didden
OK, you asked for it:
What are those units on the magnitude scale of the freq response? Volts? Can you do it with a dB scale? I'll bet you go down to 30Hz or so with less than 3dB loss. Not good, but not as bad as you think. C1 is still your limiting factor, throw it out!
What is the DC level at the U3, U4 collectors? Fix that and you can throw out at least 3 electrolytics. I still think you are trying to fix without knowing what the problem is. Whatever you do in input buffering has absolutely no impact on the freq response and the bias problems. Why not first look what's broken and fix THAT? Then you can get fancy.
Temp plots are totally irrelevant now because the amp will never work the way it is now. Waste of time.
Jan Didden
ashade said:
now this is looking more like something that may work, primarily due to the feedback introduced by R5/R6: you will have lowered distortion, more bias stability, and a manageable gain (those R5/R6 are still too high - I would use 22k there).
the problem with the circuit now is that as R5/R6 control the gain AND DC offset AND LF response of your amplifier, it is going to be very hard to optimize for all of them, some of them conflicting, at the same time.
As you have veyy small input signal, I was wondering where you want to use this. Maybe if you tell us where and how you want to use the amp, we can give you better advice.
But you will definitely benefit from a thorough reading of electronics 101 type books.
C7 and C8 are used to bias both U1 and U2 in active mode. Without them, I don't know how to do this. Maybe they can get slower without lowering frequency response. C5 and C6 are there for the same reason. The circuit without the output capacitor has about -200mV DC output and much worse temperature response. If I lower the capacitor too much, the AC response starts to have a very high ressonance at low frequencies, wich is not good.
PS: the plot vertical scale is Vout/Vin ratio.
PS2: 5mV is the signal of a normal computer microphone at about 100dB sound intensity.
PS3: 22K resistors would produce half the gain I have now. However, you are right, I'm changing the 800K to a 100K resistor.
PS: the plot vertical scale is Vout/Vin ratio.
PS2: 5mV is the signal of a normal computer microphone at about 100dB sound intensity.
PS3: 22K resistors would produce half the gain I have now. However, you are right, I'm changing the 800K to a 100K resistor.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.