off board DC servo - feasible?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I posted a thread about this board in the Analog Line Level forum, but I thought I would also throw out the idea here, focused on using the DC servo in paralleled chip amps like in National's BPA200, etc. Please see the main thread for more details on the board.

Initially I intended the DC servo to be used at line level to kill DC offset. I was modeling a "Linkwitz Transform" circuit that I am designing and I realized that, with high gain at low frequency (e.g. boosting a subwoofer driver's low freq response) also causes DC offset to be boosted along with the AC. The op amps in the circuit have inherent output offset, and if there is any input DC offset the resulting DC offset at the output could be almost a volt! I considered putting DC blocking caps in a couple of places in the circuit, but realized that one could just live with this but follow the circuit with a DC servo that effectively rolls off the response at very low frequency (e.g. less than 1 Hz) and completely removes DC.

Once I started playing around with some sims of the circuit, I recalled seeing the application of the DC servo in the National tech app (e.g. Application Note 1192) on the BP200 amp. This uses paralleled LM3886 chip amps, each with a DC servo so that there would not be DC current flowing chipamp-to-chipamp. I started wondering if I could just add on this kind of servo to an existing LM3886 chip amp (or any other amp) as an outboard add-on circuit.

Is this possible? Are there any problems related to locating a feedback loop off the board? RF interference? Phase issues? Does anyone know of any other implementations that use an off-board DC servo?

-Charlie
 
The feedback loop itself should be close to the amp and have a short path. One reason is that a long path increases inductance, creates a low-pass filter and slows the feedback loop down. Another reason is that a long feedback loop is likely to pick up noise and inject it into the amp.

There are solutions with feedback loops that are not short, e. g. active speakers sometimes get feedback from the speaker's movement. That is a question of the effort you want to put into the design and the possibilities you have to measure your circuit.

The DC servo loop itself is not as sensitive, because it operates at very low frequencies down to DC. It can be put off the board.
 
The DC servo ties in across the upper feedback resistor.
The DC servo injects signal right into -IN pin and will come out of the speaker in amplified quantities.

Keep the DC servo daughter pcb very small and hook it in with very short leads

Find Gootee's answers to the DC servo question. A very informative thread.
 
What can the servo inject into the amp's inverting input that is bad?

1) noise picked up between the amp's output and the servo's input will mostly be filtered out by the servo's own action. What is not filtered out is so far below the audible range in frequency, it won't do any harm.

2) noise picked up between the servo's output and the amp's inverting input. With a little care when choosing the path, the effect should be negligible.

3) distortion due to parasitics of the path between servo output and amp input. The servo output signal is far enough below the audible range that those parasitics won't have a noticable effect on the signal. What size can the parasitics have? And what size would it take to distort a signal that is usually below 1 Hz?
 
RE: noise injection by DC servo

Thanks for all the comments. I would try to keep the leads to/from any off-board DC servo as short as possible. Makes good sense.

The comments on additional noise from a DC servo are valid, but I found that:
(a) the amplitude of the signal that is fed back in to the amplifier falls off at higher frequencies. If you make the servo "slow" by using a very low corner frequency this also causes the servo's AC output to be down by, for instance, 55dB at 100Hz and it keeps falling, down by almost 100dB at 10k Hz. So noise injection would be negligible.
(b) the AC component can be reduced even further by adding to the servo's output an RC LP stage with a corner frequency equal to 1/tc of the servo. Some design issues on that are described here:
Website of Wayne Stegall - Simple DC Servos

-Charlie
 
I think that many of the inaudible and should be negligible are precisely why some do not advocate DC servos.
Do they particularly not advocate servos that are mounted outside of the amp PCB with long leads in unfavourable locations. Or do they not advocate servos in general?

I think most agree that any deleterious effect they can have on the final sound must be designed out.
If there is a deleterious effect on the final sound, yes. The question remains, is there any? Or is it only a deleterious effect that does not break through to the sound?

(b) the AC component can be reduced even further by adding to the servo's output an RC LP stage
In the case of a chip amp that is not feasible, because the servo output must attach to the inverting input. The capacitor would come to rest in parallel to Ri and influence the NFB loop. You would have to add a buffer in between servo output and inverting input which creates a new source of interference.
 
In the case of a chip amp that is not feasible, because the servo output must attach to the inverting input. The capacitor would come to rest in parallel to Ri and influence the NFB loop. You would have to add a buffer in between servo output and inverting input which creates a new source of interference.

Not exactly. That's only true (servo connected to inverting input) if the chip amp is configured as the non-inverting configuration. For the inverting configuration, this would work fine.

Take a look at National's white paper (AN-1192) on the BP200 to see an RC filter on the servo for a non-inverting amp. They take the servo input from a LP RC filter on the amp's output (Figure 14). This cuts the HF input to the servo, which should reduce feedthrough to the feedback loop of signal, noise, etc. They also show a servo for an inverting configuration but without the RC filter on the servo output (Figure 15).

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