John Curl's Blowtorch preamplifier

Status
Not open for further replies.
Jump to Latest
#9,841
I have slowed down my research, as there is no place to conveniently publish it.
In the early days, we published in 'Audio', and maybe the AES Journal. Later, we published in 'The Audio Amateur', and a few other publications. Now, I don't have the inclination to do the extensive research necessary to do something for publication. However, since 1990, I have worked seriously with power line interference, Bybee devices, diode transient response, power transformer problems, electrolytic cap quality, wire, cable transmission theory and measurement, and use of Teflon caps for phono EQ.
I do not change what seems to work, but I have designed a number of amps and preamps for Parasound. I have 4 models in the 'Absolute Sound' recommended components as of today. My efforts tend to making those products better and that does not require much more research than I have done already.
 
#9,842
anatech said:
Hi Bob,
Most servos operate by applying a DC bias to one to the gates/bases. This upsets the DC balance of the diff. pair, no?

If you an design an amplifier without the feedback capacitor to ground I think you are further ahead. This causes gain at DC, but you should be able to buck that out with the servo. A trim pot also works fine as long as your circuit stays in balance with changing temperature.


A "proper design" is rare from what I have seen.

I don't think that is true. To correct for a DC offset, the diff. pair will shift the static currents so that one conducts more than the other. This would shift the operating point to one side. Most DC servos are designed to amplify the DC error (and filter out audio), and apply that to one side of the pair. This does the same thing as the dif. pair would do without the servo, but to a greater degree.

If I'm wrong here, can someone explain to me how this works? I'm serious about this.


I have to disagree with you on this Bob. When a product like an amp or preamp is manufactured, there is no guarantee that the other piece of equipment is safe from DC offsets. Krell did this. Explain to the customer why a DC offset in teh preamp output caused a big bang and force the amp into protection. The servo op amp in the preamp had high DC offset all by it's little self. The muting transistors can also fail DC, outputting on heck of an offset as well. Those 2SC2878 transistors often had about 15 VDC reverse E-B bias. Even though these were designed for that, a relatively high failure rate exists.

My view is that those coupling capacitors are really required. Some amplifiers do not even have DC offset detection / protection (think Bryston 4B for one). The chances of equipment being changed out is too high to argue that a specific situation is safe. It may not be safe in 5 years.

Now, if we had to make a choice as to which capacitor to keep, I vote for the cap on the input of an amplifier. That way you have a DC block for sure,

-Chris
Click to expand...



Hi Chris,

It looks like you are expressing two main concerns about what I have said about dc servos. First, you are concerned that the correction applied by the servo causes an imbalance in the input differential stage. Second, you are concerned about my mention of the possibility of eliminating the amplifier’s input coupling capacitor and instead allowing the dc servo to swallow any dc offset from the preamp. These are both very legitimate concerns. Let me try and put them in perspective.

In the case of the input coupling capacitor, this is strictly a choice in regard to managing risk. The risks you mention are real, but true dc-coupled power amplifiers out there will have this same risk. Some of those have the option of switching in a blocking capacitor at the input. This is a good idea that can and should be applied in the general case if one uses a servo and chooses to optionally skip the input coupling capacitor.

My philosophy in using a dc servo is that it should do as little as possible to the circuit, and its effect should be as subtle as possible. The dc servo is there to eliminate the “natural” offset of the amplifier that would be there if we did not employ a capacitor in the feedback return leg (I always assume that if we employ a servo, we do not employ a capacitor in the return leg). The servo should not be more powerful than is necessary to do this job with some margin.

Consider an amplifier with a gain of 20 and a natural output offset of 200 mV. It will have an input-referred offset of about 10 mV. This 10 mV could come from numerous different sources. It could merely be the input offset of a JFET input differential pair. Or it could be offset caused in a perfectly matched BJT input pair where the input bias currents on the positive and negative inputs flow through different dc resistances, resulting in an input current induced input offset voltage. Or it could be input offset that results from an imperfectly biased drive of the VAS (as could happen if the input pair tail current and the current required to flow through the input pair load resistor are not in the right 2:1 relationship when the output is zero).


The generally-accepted definition of an input pair being in balance is when the two collector or drain currents are equal. This is not necessarily the same as saying that the two bases or gates are at the same voltage. Indeed, they will differ by the intrinsic offset voltage of the differential pair.

If the amplifier is properly designed, the output will be at zero when the currents of the collectors or drains of the input pair are equal. If the servo is required to correct for an input pair offset voltage, its correction under these conditions is merely canceling out the intrinsic offset of the input pair, and certainly does not drive the input pair to a state of imbalance.

If, on the other hand, the servo must correct an offset whose origin was input base current flowing through different dc resistances on the two inputs, then the servo is merely supplying a voltage to cancel out this externally-created offset at the input of the pair. Again, the servo does not act to cause an imbalance in the input pair. It merely allows the input pair to be in its balanced state when the output voltage is zero.

Finally, if the natural offset of the amplifier originated because the load resistor on the input pair was not right for the conditions, the input stage will be out of balance when the output is at zero, but this would have been the case with or without the servo. Here the servo makes things no worse.

When a current mirror is used instead of a load resistor for the input pair (watch out in applying this to complementary input pairs), the matched current mirror will tend to force the input pair into a state of balance as long as the load presented to the input stage by the VAS is quite light (think Darlington VAS or MOSFET VAS).


One final observation. Often, when we go from an amplifier with a capacitor in the NFB return leg to one where we do not employ that capacitor, we may often incur worse input base current induced offsets if using a BJT input pair. This is because the inverting input to the amplifier will typically have a low dc impedance to ground, while the non-inverting input to the amplifier will have a comparatively high dc return path resistance, since we don’t want to have an unduly low amplifier input impedance (which would also drive up the cost of the input coupling capacitor). This needs to be taken into account in determining how much correction range the servo will need to be able to apply.

The use of a complementary input stage will reduce the effect of input base current because the NPN and PNP input base currents will oppose each other. However, if betas are not matched between the NPN and PNP pairs, we may incur perhaps a 2:1 difference in beta, reducing the advantage gained to only a factor of two compared to a unipolar input stage.

Cheers,
Bob
 
#9,847
I've used this simple combination of a non-inverting servo (so it can be used into an inverting input of the amp) and protection. As soon as the output voltage of the servo is over approx. 7V it resets the protection of the amp.

Alex
 

Attachments

  • servo_protection.gif
    servo_protection.gif
    3.8 KB · Views: 804
#9,852
john curl said:
Well said Bob. I do believe that some sort of amp DC protection is useful and necessary IF the servo goes outside its limit, but that also means that something is wrong, either with the amp itself or the DC supplied by a poorly designed preamp.
Click to expand...


Thanks, John. I agree with your point.

I usually monitor the output of the servo integrator and if the voltage gets beyond a certain point, I open the speaker relay or otherwise engage protection. Depending on the design parameters, however, this can sometimes interfere with full-power testing at frequencies below 20 Hz.

Cheers,
Bob
 
#9,853
Bob Cordell said:



Thanks, John. I agree with your point.

I usually monitor the output of the servo integrator and if the voltage gets beyond a certain point, I open the speaker relay or otherwise engage protection. [snip]Cheers,
Bob
Click to expand...


Bob,

Wouldn't it be wise to have an offset protection mechanism looking at the amplifier output to the speaker? One could possibly have a scenario where the servo still looks as if it is in it's design range while the amp has too high offset from another fault condition. In other words, detect excess offset where it matters, at the speaker jack.
Then this sort of protection would automagically also protect against servo failures and amp problems that drive the servo out of it's design range (including input offsets), so no separate servo error detection scheme would be necessary.

Jan Didden
 
#9,854
janneman said:
detect excess offset where it matters, at the speaker jack.
Click to expand...

Jan,

That is exactly what the servo does, and if this offset present for long enough the servo output goes out of range and trips the protection. That what my circuit does and Bob is apparently using a similar approach. Obviously, the servo may fail but so may the protection circuit itself and after all there is such thing as a direct lightning strike 🙂 .

Alex
 
#9,857
x-pro said:


Jan,

That is exactly what the servo does, and if this offset present for long enough the servo output goes out of range and trips the protection. That what my circuit does and Bob is apparently using a similar approach. Obviously, the servo may fail but so may the protection circuit itself and after all there is such thing as a direct lightning strike 🙂 .

Alex
Click to expand...

Well, yes, a chain is only as strong as the weakest link. That's why I like as few links as possible in my chain. Like one 😉

Jan Didden
 
Status
Not open for further replies.
Top Bottom