John Curl's Blowtorch preamplifier

Status
Not open for further replies.
Jump to Latest
#9,882
chascode said:


Hi, I am about 15 days behind on this thread, but I wanted to mention another way DC gets into the mains. We have several 3-phase 250 hp motors. When they start the three phase currents slew asymetrically due to the transient reactance of the motors. Two of the three phase currents slew for several seconds. This is not due to asymetrical waveforms (such as from half-wave rectification). If the line currents are equal, the fundamental waveform in the neutral cancels, leaving the zero-sequence current (ie. DC) as well as the third harmonic (from approaching the motor core saturation), which sums rather than canceling in a three phase 4-wire system.

Another interesting visual effect is that the high voltage pole-top power lines make a big jump as the motor is started. The "low" 480 volt side is underground, but I suspect those feeders also take a jump in the ducts and conduits.

This mechanical jump scales down to lower levels as well. I wonder if their is any miniscule motion in the discrete wiring in audio equipment that may affect the sound, as opposed to the constrained tracks of a pc board.
Click to expand...

For me, there's two or more transformers between me and DC offset on the mains.
It helps 🙂
 
#9,883
Re: speaker protection

Edmond Stuart said:


Hi Bob,

That wasn't Jan's point. He was concerned about a situation where a fault condition also affects the servo. Of course, this wouldn't happen if the servo has is own PSU, but normally the +/- 15V for the servo are derived from the main supply lines. If the latter are protected by fuses, and if one of them (that is, the 'wrong' one) is blown, the servo might not turn on Q1 as in Alex's circuit.

I think that a dedicated circuit, one that also monitors false supply lines, is much more desirable. As pointed pointed out by Arthur, the TA7317P is a good candidate. It also has provisions for disabling the speaker relay during power switch on/off transients.

Cheers,
Edmond.
Click to expand...


Hi Edmond,

These are good points. The +/- 15V that I usually provide for powering protection and other auxilliarly circuits always comes before any rail fuses, so those circuits would remain in operation if a main rail fuse blew.

Bob
 
#9,886
Bonsai said:
Some good ideas here - especially using th e integrator output as a trigger signal.

However, we need also to address the point raised by Forr:- using a DC offset correction circuit unbalances the front end and this leads to increased distortion.

How do we get around this problem?
Click to expand...

Hi Bonsai,

Regarding offset correction and input stage imbalance, see my post #9842 and let me know if you agree.

Cheers,
Bob
 
#9,887
Here are a couple more thoughts on servo operation.

Consider an amplifier with a closed loop gain of 20, having a 19k feedback resistor and a 1k feedback shunt. Assume an inverting servo integrator with a 1 meg series resistor and a 1 uF capacitor. The integrator will have a gain of 0.16 at 1 Hz.

Assume a unity gain servo inverter at the output of the integrator. Assume a 29k resistor from the servo inverter to the inverting input of the power amplifier, for a servo attenuation factor of about 30:1. The servo loop gain at 1 Hz will be approximately 20 * 0.16/30 = 0.106. This also means that the LF cutoff enforced by the servo is at about 0.106 Hz (the servo-based LF cutoff is simply the frequency where the servo loop gain falls to unity).

We have an LF cutoff frequency of only 0.1 Hz, not simply because we could and we wanted to really go low, but because the numbers will work out nice, especially in terms of servo signal-handling ability.

Assume that this is a 100 w amplifier with a 40V peak output. Assume that we trigger dc protection when the servo output reaches a threshold of 6V.

If the servo needs to correct a fairly large input offset of 100 mV, the servo output only goes to about 3V, well within its safe range. In reality it can handle nearly 200 mV offset before triggering dc protection and is still not even close to the output limits of the integrator op amp.

If a full-power 10 Hz sinusoidal signal of 40V peak is being produced by the amplifier, the sinusoidal output of the servo integrator will only be about 0.6V. It will be 10 times this if the input signal is at 1 Hz and will just trigger the dc protection.

If the amplifier fails shorted to a rail and produces 40V of dc, the servo integrator output will rise to the 6V protection trigger point within 0.25 second. A risk tradeoff is whether or not we think this is fast enough to protect a woofer.

Note that if we want to reduce servo influence at the expense of offset that can be handled, we can simply increase the servo injection resistor from 29k to 100k. This will reduce by 10 dB the opportunity of servo noise and distortion to get into the signal path. This will also reduce the LF cutoff to 0.03 Hz. As long as the amplifier design and architecture does not cause a turn-on settling issue that must be corrected, there is probably no problem with this lower LF cutoff frequency.

If for some reason we want to go to a higher LF cutoff frequency, we must increase servo loop gain. We can do this by reducing the 29k servo injection resistance at the expense of greater opportunity for servo garbage to get into the signal path. Alternatively, we can do this by increasing the integrator gain. Just going to a 0.1 uF integrator capacitor would increase the integrator gain by a factor of ten and bring the LF cutoff up to 1 Hz. Unfortunately, the servo is now more easily overloaded by large LF signals at the amplifier output. A full-amplitude 10 Hz signal at the output of the amplifier would cause triggering of the dc protection. This is just a design decision and it may be OK.

There are many possible variations to this kind of servo, but this seems to be a decent illustration of typical operation and tradeoffs.

Cheers,
Bob
 
#9,888
Hi Bob,

Good illustration of all those interdependencies. It is one reason why I like to have a DC protection at the output and not use the servo for that; I'd like to decouple the interdependencies. The servo is to servo 😉 .

The other reason is I want the chain not to include too many links that may fail. Suppose the servo fails in some way (say, its +15V zener shorts out), the output protection will either do nothing if it doesn't lead to an unacceptbly high offset, or go into protection if it does.

As you say, design decisions. Either way you can build a fine amplifier!

Jan Didden
 
#9,889
chascode said:


Hi, I am about 15 days behind on this thread, but I wanted to mention another way DC gets into the mains. We have several 3-phase 250 hp motors. When they start the three phase currents slew asymetrically due to the transient reactance of the motors. Two of the three phase currents slew for several seconds. This is not due to asymetrical waveforms (such as from half-wave rectification). If the line currents are equal, the fundamental waveform in the neutral cancels, leaving the zero-sequence current (ie. DC) as well as the third harmonic (from approaching the motor core saturation), which sums rather than canceling in a three phase 4-wire system.

Another interesting visual effect is that the high voltage pole-top power lines make a big jump as the motor is started. The "low" 480 volt side is underground, but I suspect those feeders also take a jump in the ducts and conduits.

This mechanical jump scales down to lower levels as well. I wonder if their is any miniscule motion in the discrete wiring in audio equipment that may affect the sound, as opposed to the constrained tracks of a pc board.
Click to expand...
Yea. Thanks. So much for the 'infinite buss' model of the mains that I was taught. (though it does work for most applications)
My (limited) experience would indicate that point to point wiring sounds better (more 'open'?) than PCB layouts. Perhaps "constrained" is the operative?
 
#9,891
h_a said:


D. Self tried that once:



from Audio Power Amplifier Design Handbook.

Have fun, Hannes
Click to expand...



You can easilly make a DC detector to shut down the amplifier that will operate within a fraction of a second whilst still being able to reject a full voltage 20Hz sinewave. Make a multiple pole low pass filter with a quad opamp. You can also power the relay coils from the aux/control circuitry supply rails as a fail safe incase these fail.

Cheers,
Glen
 
#9,892
To point out a few important things:
First, it is possible to build a servo that has virtually any time constant that you want. The trade off is the settling time at turn-on. However, IF you make a basically DC amp or preamp that is low in offset in the first place, there is no problem, and it is difficult how to show how the servo could effect the signal in significant time or distortion performance.
Second, it is usually more practical to make a servo a little faster, just to keep turn-on settling to a minimum length of time, without too much potential effect in sound quality.
Third, the control capacitor in the servo can be decoupled from its effect to the audio signal, because it is usually bypassed by the feedback resistor, and virtually any amount of decoupling can be designed in, by making the servo have to act significantly, before ANY real offset is adjusted for. This just means that you have to decouple (large ratio) the output resistor from the servo that ratios with the feedback resistor to ground. Typically 10-100 to one is possible, which that the servo output might have to be 1 Volt just to control an offset of 10mV, but you will reduce servo range before clipping, if you are not careful and go too far.
This also means that you have to respect the limitations of the servo IC and its cap, and not accidently include it significantly into the audio path, by expecting the servo to control the low frequency bandwidth or to fix a large initial offset, because you cannot bother to match the input devices. Then the servo may become a disappointment, and be one good reason why many designers avoid it.
 
#9,893
I haven't really researched it, but would you agree that the requirements for a servo opamp are similar as for any audio-signal-path opamp?
I'd think that to work well at higher frequencies (no signal leakage leading to unwanted feedback) the integrator opamp should present a low, non-inductive output Z to the integration cap. Which means it (the opamp) must have good hf performance.

Jan Didden
 
#9,894
Jan, the servo amp sees a less than 1Hz low pass filter at its input. Do you really think that it can find much high frequency at all? That is one of the reasons that servos are practical. If we had do make the servos with exactly the same quality and speed as the forward through-path, then they would not be very practical at all. Of course, some might insist that I am wrong in this. That's OK, some just love to do it the hard way.
 
#9,895
G.Kleinschmidt said:




You can easilly make a DC detector to shut down the amplifier that will operate within a fraction of a second whilst still being able to reject a full voltage 20Hz sinewave. Make a multiple pole low pass filter with a quad opamp. You can also power the relay coils from the aux/control circuitry supply rails as a fail safe incase these fail.

Cheers,
Glen
Click to expand...


Attached below is a 20Hz full amplitude tone burst response sim of a 10Hz 8-pole butterworth low-pass filter that can be built up around a TL074 opamp or similar.

If this is followed with a comparator with a +/-5V DC trigger threashold it will respond to output swinging to one of the rails (in a 100W/8ohm amp) in ~60ms.

Cheers,
Glen
 

Attachments

  • dc.gif
    dc.gif
    64.1 KB · Views: 571
#9,898
john curl said:
To point out a few important things:
First, it is possible to build a servo that has virtually any time constant that you want. The trade off is the settling time at turn-on. However, IF you make a basically DC amp or preamp that is low in offset in the first place, there is no problem, and it is difficult how to show how the servo could effect the signal in significant time or distortion performance.
Second, it is usually more practical to make a servo a little faster, just to keep turn-on settling to a minimum length of time, without too much potential effect in sound quality.
Third, the control capacitor in the servo can be decoupled from its effect to the audio signal, because it is usually bypassed by the feedback resistor, and virtually any amount of decoupling can be designed in, by making the servo have to act significantly, before ANY real offset is adjusted for. This just means that you have to decouple (large ratio) the output resistor from the servo that ratios with the feedback resistor to ground. Typically 10-100 to one is possible, which that the servo output might have to be 1 Volt just to control an offset of 10mV, but you will reduce servo range before clipping, if you are not careful and go too far.
This also means that you have to respect the limitations of the servo IC and its cap, and not accidently include it significantly into the audio path, by expecting the servo to control the low frequency bandwidth or to fix a large initial offset, because you cannot bother to match the input devices. Then the servo may become a disappointment, and be one good reason why many designers avoid it.
Click to expand...


These are all good points, John.

With respect to Jan's question about HF sneaking through the integrator, I have at times experimented with putting a passive LPF between the output of the servo and the servo injection point (by splitting the servo injection resistor and taking a capacitor from there to ground). The idea was to further keep noise and distortion from the servo op amp out of the signal path, but it would also address Jan's HF concern as well, if that is a problem. However, one needs to watch out for subtle effects this technique can have on the frequency response at low frequencies, since the frequency-dependent impedance seen looking back into the servo injection network acts like it is in parallel with the main NFB shunt resistor to ground. This added capacitor is also slightly in the effective signal path, so its quality matters.

A compromize approach that would at least mitigate Jan's concern would be, in the case of an inverting integrator followed by an inverter, to put some capacitance across the inverter's feedback resistor, giving the inverter an LPF response. In any case, the new pole added should be well above the servo bandwidth point of, say, 0.1 Hz. This constraint is easy to satisfy.

Cheers,
Bob
 
#9,899
Bob Cordell said:
[snip]A compromize approach that would at least mitigate Jan's concern would be, in the case of an inverting integrator followed by an inverter, to put some capacitance across the inverter's feedback resistor, giving the inverter an LPF response. In any case, the new pole added should be well above the servo bandwidth point of, say, 0.1 Hz. This constraint is easy to satisfy.

Cheers,
Bob
Click to expand...


... or a low-impedance, passive lowpass *before* the integrator....?

Jan Didden
 
Status
Not open for further replies.
Top Bottom