This is an humble attempt to sketch a general guideline for designing a DIY feedback amplifier for home audio purposes.
I'm a newbie to hi-fi electronics design, so please forgive for any ingenuity.
Version 3.0 updated with your suggestions.
0) Evaluate the kind of load and the level of power to drive. Set rail voltages.
1) Define how much and what kind of distortion is allowed.
2) Set all other amplifier's specs (closed-loop gain, Zin, Zout, input sentitivity, THD, noise, etc.).
3) Design the feedback network to achieve the specified closed-loop gain and distortion level.
4) Design the basic open-loop amplifier A - from the differential input stage to the output network (Zobel/Boucherot) .
5) Estimate the open-loop frequency response of A; locate poles and zeroes through analysis and/or SPICE simulation.
6) Check stability of the closed-loop amplifier Af using Bode plots.
7) If (almost certainly) Af turns out to be unstable, then proceed to frequency compensation of A.
8) Check again to confirm stability of closed-loop amplifier Af (enough phase margin).
9) Design the input signal conditioning circuit.
10) Design ancillary circuitry for loudspeaker protection, soft-start, ecc.
11) Build prototype.
12) Check overload behavior, rail sticking etc.
Any further suggestions are welcome!
Angelo
I'm a newbie to hi-fi electronics design, so please forgive for any ingenuity.
Version 3.0 updated with your suggestions.
0) Evaluate the kind of load and the level of power to drive. Set rail voltages.
1) Define how much and what kind of distortion is allowed.
2) Set all other amplifier's specs (closed-loop gain, Zin, Zout, input sentitivity, THD, noise, etc.).
3) Design the feedback network to achieve the specified closed-loop gain and distortion level.
4) Design the basic open-loop amplifier A - from the differential input stage to the output network (Zobel/Boucherot) .
5) Estimate the open-loop frequency response of A; locate poles and zeroes through analysis and/or SPICE simulation.
6) Check stability of the closed-loop amplifier Af using Bode plots.
7) If (almost certainly) Af turns out to be unstable, then proceed to frequency compensation of A.
8) Check again to confirm stability of closed-loop amplifier Af (enough phase margin).
9) Design the input signal conditioning circuit.
10) Design ancillary circuitry for loudspeaker protection, soft-start, ecc.
11) Build prototype.
12) Check overload behavior, rail sticking etc.
Any further suggestions are welcome!
Angelo
Last edited:
The output Zobel should be considered to be part of the open loop amplifier
in the first step. It will affect the later steps.
Last edited:
The approach that I've been taught starts with Jan Didden's step 0 (although in a professional context that's often someone else's problem), then your step 3, then the rest.
The choice of the input stage is determined by the noise requirements, the choice of the output stage by the signal voltages and currents that are needed to drive the load. Insert stages between input and output stage as needed to get an adequate loop gain, but don't overdo it as that will complicate frequency compensation; a many-stages amplifier that requires brute-force frequency compensation can end up having less loop gain at the upper end of the band of interest than a somewhat-fewer-stages amplifier that can be frequency compensated in a more subtle manner.
It's all based on an oversimplified theory, so deviate from the procedure whenever you have a good reason to do so.
The choice of the input stage is determined by the noise requirements, the choice of the output stage by the signal voltages and currents that are needed to drive the load. Insert stages between input and output stage as needed to get an adequate loop gain, but don't overdo it as that will complicate frequency compensation; a many-stages amplifier that requires brute-force frequency compensation can end up having less loop gain at the upper end of the band of interest than a somewhat-fewer-stages amplifier that can be frequency compensated in a more subtle manner.
It's all based on an oversimplified theory, so deviate from the procedure whenever you have a good reason to do so.
Last edited:
What are the iron clad constraints, which you absolutely must obey for this amplifier?
Some hobbyist projects begin with the constraint: I MUST use such-and-such transformer. Because I already own it and I refuse to buy a different transformer.
Other hobby projects begin: I MUST use a vacuum tube output stage. Because this is a tube amplifier and I seek to avoid solid state components.
Others begin: I MUST use so-and-so chassis + heatsinks (perhaps re-purposing from a commercial product by Yamaha or Kenwood etc).
Others begin: I MUST achieve a continuous RMS output power into 4 ohms, of so-and-so number.
etc.
Some hobbyist projects begin with the constraint: I MUST use such-and-such transformer. Because I already own it and I refuse to buy a different transformer.
Other hobby projects begin: I MUST use a vacuum tube output stage. Because this is a tube amplifier and I seek to avoid solid state components.
Others begin: I MUST use so-and-so chassis + heatsinks (perhaps re-purposing from a commercial product by Yamaha or Kenwood etc).
Others begin: I MUST achieve a continuous RMS output power into 4 ohms, of so-and-so number.
etc.
Thanks Mark. I modified the list to include project specifications and requirements in step 0:
0) Specify the amplifier (closed-loop gain, Zin, Zout, output power, THD, noise, etc.)
1) Design the feedback network to achieve the desired closed loop gain
2) Design basic open-loop amplifier A - from the differential input stage to the output (Zobel) network.
3) Estimate the frequency response of A; locate poles and zeroes through analysis and/or SPICE simulation.
4) Assess stability of the closed-loop amplifier Af using Bode plots
5) If (almost certainly) Af is unstable, then proceed with compensation of A
6) Check again to confirm stability of closed-loop amplifier Af
7) Design the input signal conditioning circuit.
0) Specify the amplifier (closed-loop gain, Zin, Zout, output power, THD, noise, etc.)
1) Design the feedback network to achieve the desired closed loop gain
2) Design basic open-loop amplifier A - from the differential input stage to the output (Zobel) network.
3) Estimate the frequency response of A; locate poles and zeroes through analysis and/or SPICE simulation.
4) Assess stability of the closed-loop amplifier Af using Bode plots
5) If (almost certainly) Af is unstable, then proceed with compensation of A
6) Check again to confirm stability of closed-loop amplifier Af
7) Design the input signal conditioning circuit.
Specify the amp.
For commercial , or here at DIYA ?
A - economics vs. performance.
B - sourcing vs. performance.
C - sophistication vs. performance.
These would apply to either OEM/hobby use.
For a project - ,
Single sided toner transfer capable (sophistication).
Through - hole/ SMD , this is a current sourcing issue.
No "unobtainium" (Jfets , etc.) This could affect A-C.
"C" can affect the PITA of current/future support for the design.
Or , OEM warranty costs.
"B" can affect support and the re -writing of BOM's.
"A" can determine how widely adopted the design is.
Absolutely.
OEM or hobby - warranty / or even credibility ?
Design should have sufficient de-ratings in both sourcing and the
native design for a degree of "resiliency".
OS
Last edited:
Hi, Angelo!
Not so good approach...
First of - you need to start from task, what kind of load and for what levels of power you will really want to drive from any giver source.
This will let you to determine gain and OPS/supply properties.
Next you need to clear the task - how much and what kind of distortion allowed.
This will let you to determine needed feedback depth and freq correction complexity.
Next - work on topology for achieving known gain + FB depth + correction.
This is a workflow.
1, Looking for a philosophy.
2, Build your amplifier around it.
Same here. Lately I'm thinking about the importance or non-importance of wide bandwidth amplifier. With VFA I have always tried to lengthen the bandwidth. But with CFA I started to think that there must be a maximum where beyond that there is nothing to gain but to lose.
1. Check bank balance.
2. Introduce idea to wife/partner.
3. Agree that some of (1) is theirs to spend as they wish.
4. Success!
If you extend a CFA to >2.5Mhz , not only does it have the 20Khz gain margin
for PPM operation , but it is also fast enough to negate both AB and class H
crossovers.
At @3Mhz , if still stable .... will negate non-optimal setups.
Most VFA's only manage 6db margin 20K , use fancy compensation.
I'm sold on a "all out" 3Mhz CFA = 20db margin @ 20k.
You are right , the 2.5-3Mhz region is the "point of diminished returns".
OS
Seems like what i think too. But to simplify things i like to look at how far the OLG response can still be flat (2k-3k). By setting this as benchmark, loop gain will follow or be adjusted accordingly. And distortion is not an issue as it is already below threshold.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.