[a1rocketpilot]

Since the Steinchen DB's are not finalized yet by the look of things, I have looked into other options for output devices. Many people have suggested placing BUF 634's or the equivalent within the feedback loop of the opamp. Apparently this is to control DC offset. I was considering using 2 LH0033's as the output, without any opamp. In other words, the coupling cap would connect to these directly and the output would feed into the output resistors. My questions are, will this work? How could this cause a DC offset problem? The LH0033 has provision to connect a trimpot to null the offset, so if there is an issue, could it be fixed using that?

Here's the datasheet for this chip. It is no longer in production, but Futurlec still has them in stock. Apparently they don't measure as well as modern buffers, but apparently they sound great and run in Class A.

Datasheet

Futurlec

Aditya

 

[tomb]

Nope. The DC is there because of the tubes, and the small film coupling caps are not sufficient to filter it. However, one of the ways to use an opamp is to zero offset. That's why they are Op-erational Amp-lifiers: they are able to control voltage gain. When placed in a circuit intended as zero, or "unity gain", then extraneous voltage that may be interpreted as additional gain is filtered out.

A buffer does not have this circuit "inelligence" and has no effect on voltage - good or bad. Since the tube passes DC regardless, you will get offset - no doubt.

 

[a1rocketpilot]

Wow, somehow I never realized that. It makes perfect sense though, given how opamps work. However, since this particular buffer has a provision to enable DC offset adjustment, wouldn't you be able to null that offset caused by the tubes using the trimpot?

Aditya

 

[tomb]

That's a good point - I missed that feature. However, if I am interpreting the datasheet correctly, the offset adjustment ranges from 5 to 25mV, depending on which variant. That sounds like something that would only be useful for balancing small output differences among multiple-paralleled buffers. The offset coming from a tube could be in volts, not a couple dozen milli-volts.

 

[a1rocketpilot]

I think what you are referring to is the amount of allowable offset error present in the actual IC. For example, if you look at the datasheet for the OPA2107, the same spec measures .1 to 1mv and that opamp has no built in provision for nulling offset. I think this refers to the manufacturing tolerances regarding how much DC offset is permitted in the chip.

Aditya

 

[tomb]

Well, it still looks like the offset adjustment is intended to zero out this the published variance - and it's on the order of tens of mV's. The application pages spell this out and imply that the offset adjusting pins simply automatically zero this variance, not some huge amount of DC volts.

That's my guess and I'm sticking it to it.

 

[a1rocketpilot]

Heh, fair enough!

I'm going to try and stick a pair of these in with an OPA2107 and see what's what when I get this baby up and running.

Aditya

 

[eVITAERC]

Quote:

Originally Posted by tomb /img/forum/go_quote.gif Nope. The DC is there because of the tubes, and the small film coupling caps are not sufficient to filter it. However, one of the ways to use an opamp is to zero offset. That's why they are Op-erational Amp-lifiers: they are able to control voltage gain. When placed in a circuit intended as zero, or "unity gain", then extraneous voltage that may be interpreted as additional gain is filtered out. A buffer does not have this circuit "inelligence" and has no effect on voltage - good or bad. Since the tube passes DC regardless, you will get offset - no doubt.

No offense, but this statement is actually quite ridiculous. The capacitor can block ANY amount of DC provided it doesn't blow up or meltdown. That's a characteristic of all capacitors, no matter how big or small. If you have a coupling cap, you can bet money (up to any satisfaction guarantee reimbursements by the cap manufacturer in case of aa defect ) that the DC coming out of the tube is dead zero.

The only reason you would want an OpAmp there to control DC offset (this is called a DC servo) is when you have some wildly behaving devices on the buffer input circuit. Sometimes these devices (fets or bjts both) have parameters that are very sensitive to temperature. As a consequence, depending on how long it's been on or what the season is, the output offset from mismatched devices can vary wildly between 2-50mV. The function of a DC servo is that they monitor offset at all times, instead of a static offset control which you implement with a VR and a multimeter. That way when the offset drifts over time, the opamp will see it and make adjustments as necessary. However, if you have an adequately competent buffer design and good ventilation, you very likely won't need it.

 

[tomb]

Ridiculous? This flies in the face of everything about the SOHA and its design. Perhaps I don't have the specifics quite exact, but every buffer tried that isn't contained within a feedback loop with the opamp is uncontrollable in offset. Witness the several highly documented threads by Steinchen and the buffers he's developing, all of which have their own specific DC servo.

The opamp in the SOHA behaves as a unity gain current buffer. Nevertheless, without local feedback the offset generated is on the order of several volts. You are perhaps correct that my interpretation of the cause being the small size of the coupling caps is wrong. Whatever the exact reason however, the fact that a simple buffer results in intolerable offset is a fact and has been tried.

One of these days I'll learn not to conjecture too far and give you guys an excuse for labeling a response intactfully ridiculous in spite of your being totally ignorant of the design.

 

[tomb]

Nah - I'll stand by my "quite ridiculous" assertion. The small film coupling caps are not enough to block all of the DC from the tube, period.

 

[eVITAERC]

Quote:

Originally Posted by tomb /img/forum/go_quote.gif Nah - I'll stand by my "quite ridiculous" assertion. The small film coupling caps are not enough to block all of the DC from the tube, period.

I don't think so man. If "smaller caps" cannot block DC, then the electronics industry is quite doomed from all the caps that have several hundred volts through them placed in mission critical positions. All the vacuum tube circuit PSUs in existence will have to be redesigned as well.

In fact, "small caps" are even better at stopping DC than large caps. This is simply because the high-pass filter formed by coupling caps have a lower cut-off frequency when you increase the capacitance.

Quote:

One of these days I'll learn not to conjecture too far and give you guys an excuse for labeling a response intactfully ridiculous in spite of your being totally ignorant of the design.

Seriously mate, I have built a SOHA as well and studied its designs carefully. Helping people on the forums is great and all but if someone else has a different opinion on the matter it helps if you attack the opinion instead of accusing the person making them as ignorant. Can't we talk about these physical laws objectively, and leave the subjective stuff to sound quality evaluations? The laws of physics on this level isn't really open to interpretation you know.

Quote:

Ridiculous? This flies in the face of everything about the SOHA and its design. Perhaps I don't have the specifics quite exact, but every buffer tried that isn't contained within a feedback loop with the opamp is uncontrollable in offset. Witness the several highly documented threads by Steinchen and the buffers he's developing, all of which have their own specific DC servo.

The JISBOS buffer doesn't have a DC servo. In fact he put one in and took it out on the nature of it being too complicated for the task. The Diamond buffer on the other hand does need one, because the inherent designs of a diamond buffer means the offset will drift like crazy. The JISBOS is quite stable, as you can see from Steinchen's own words here:
http://headwize.com/ubb/showpage.php...d=6711&fpage=1

Quote:

I haven't encountered offset drift with this buffer, the offset stayed well below 1mV over hours.

Quote:

I already changed the layout to corporate your servo suggestion and I really like it's sophisticated design. What makes me hesitate is the fact that this is a significant change that has to be prototyped before proceeding and may imply intricacies. I'm totally swamped with projects and work to do, hence I'd like to take the easy way, keep the circuit simple and go back to the trimpot version which does it's job in my SOHA just fine. PRR, thank you very much for your input and suggestions.

Quote:

proto finished, working fine and sounds excellent. I thought about bridging the 51 Ohms resistors and changing the offset trimpot from 100 to 200 Ohms but the present circuit easily adjusts more than 3mA IDSS difference at the input JFETs.

and here you can hear it from the horse's mouth that the tube does not affect the buffer offset at all

Quote:

there is a coupling cap between tube and buffer with the SOHA, tube rolling / aging does not affect dc offset, offset depends on the jfets only.

Hope this helps. The disadvantage of DC servos is that in order to sound good they must only work on the DC component on your signal, or else they are in your signal path. But if you filter out all the AC from the output of the opamp, then the opamp cannot keep itself stable. It's a tricky situation that needs careful tuning of the circuit, and if you can get by with a static offset trimmer then by all means it is almost always the better way to go.

 

[regal]

A capacitor blocks DC after power-up. The coupling caps on the SOHA are small because the input impedance of the buffer is high. The capacitance of these coupling caps has nothing to do with their ability to block DC. You could replace them with 470uF or 1F caps and you would still need a buffer in a feedback loop.

 

[SnoopyRocks]

Quote:

Originally Posted by tomb /img/forum/go_quote.gif Nah - I'll stand by my "quite ridiculous" assertion. The small film coupling caps are not enough to block all of the DC from the tube, period.

Sigh...the attitude is uncalled for, especially when you are incorrect in fundamental understanding and not subtlety as you seem to think. Perhaps eVITAERC could phrased things in a friendlier fashion. The DC offset issues that you are alluding to are not from the tube, since the coupling capacitor does block the DC from the tube, but from how the buffer is implemented/biased. Instead of stringing together more jargon in a fashion that doesn't make much sense, you could respond constructively by simply asking for a technical explanation so that everyone benefits.

 

[tomb]

Quote:

Originally Posted by SnoopyRocks /img/forum/go_quote.gif Sigh...the attitude is uncalled for, especially when you are incorrect in fundamental understanding and not subtlety as you seem to think. Perhaps eVITAERC could phrased things in a friendlier fashion. The DC offset issues that you are alluding to are not from the tube, since the coupling capacitor does block the DC from the tube, but from how the buffer is implemented/biased. Instead of stringing together more jargon in a fashion that doesn't make much sense, you could respond constructively by simply asking for a technical explanation so that everyone benefits.

Very well, please explain then, where the offset originates in a power up condition when the charge rate for a 0.1uF coupling cap is next to nothing. Also, where is the biasing in the basic opamp output stage? There is sufficient "jargon" as you call it to imply that differential spikes may not be controlled by the caps in question. So, yes, please, explain this subtlety.

Thank you for your attitude assessment.

 

[tomb]

OK, I will leave it at this - noting that I wasn't the one that put this conversation into the mode of "quite ridiculous" or "jargon." It may seem absurd to argue that a coupling cap doesn't block all DC, but again, it appears that we are in the state of ridiculous and jargon. So, I would just like to know - is the coupling cap in question perfect, and does it perfectly block all effective DC at all conditions?

Here's an interesting take on it:
Quote:

== Do capacitors block DC? == It is often said that capacitors block DC or equivalently, that a capacitor is an open circuit at DC. Is this true? What is true is that the DC steady state current through a capacitor is identically zero. In DC steady state, all circuit voltages and currents are constant. By the fundamental capacitor equation: the capacitor current must be zero if the capacitor voltage is not changing with time. Does this imply that a capacitor blocks DC current? Clearly, the answer is no. According to the equation above, a constant (DC) current through a capacitor can exist if the voltage across the capacitor changes at a constant rate. For example, a DC current of 1mA through a capacitor with capacitance of 1000uF causes the voltage across the capacitor to change at the rate of 1V per second. Theoretically then, a capacitor does not block DC current. Obviously, for any real current source connected to a real capacitor, the capacitor voltage cannot continue to change forever. Eventually, the capacitor dielectric would break down or the DC current source would reach its maximum working voltage. These real world limitations do not take away from the fact that the capacitor does not block DC current. '''But isn't the impedance of a capacitor infinite for DC?''' The formula for the impedance of a capacitor is given by: where f is the frequency of the AC voltage and current associated with the capacitor. It is often said that "DC is just AC with zero frequency". In a sense, this is true. Thus, it seems reasonable to believe that setting the frequency to zero in the equation above would give the impedance of the capacitor 'at DC'. Mathematically, this is problematic because division by zero is undefined, however it is clear that as the frequency approaches zero, the impedance increases without bound so one can properly say that the impedance of a capacitor 'goes to infinity' as the frequency goes to zero. What ''exactly'' does this mean? To answer this, we need to understand how the formula for the impedance of a capacitor is derived. The first step in any deriviation of the impedance is to assume that the voltage across the capacitor is a sinusoidal function of time with ''constant'' amplitude, frequency, and phase. This step is ''crucial'' to the derivation of impedance and is equivalent to requiring that the circuit has settled into '''AC steady state'''. Now consider the following: if the frequency of this sinusoid is set to zero, ''the voltage across the capacitor becomes constant''. In other words, setting the frequency to zero is equivalent to requiring that the circuit has settled into '''DC steady state'''. We already know that the DC steady state current through a capacitor is identically zero so it is reassuring to find that the impedance formula gives the same result. The question then becomes: do all circuits have a DC steady state solution? Consider the case of a sinusoidal current source connected to a capacitor. Using the impedance form of Ohm's law: we find that the voltage across the capacitor increases without bound as the frequency decreases to zero. How is this to be interpreted? The voltage given by the formula above is a phasor voltage. This phasor gives the peak amplitude and phase of the sinusoidal voltage across the capacitor. However, the phasor representation assumes that the circuit is in AC steady state or, if the frequency is zero, in DC steady state. For a circuit to be in DC steady state, the circuit must have a DC steady state solution in which to exist. The circuit composed of a non-zero constant current source and a capacitor ''has no DC steady state solution''. This was made clear earlier when we found that if the current through a capacitor is constant and non-zero, the voltage must be changing at a constant rate. Thus, for a circuit that does not have a DC steady state solution, we should expect a meaningless answer from a method that requires that the circuit is in DC steady state. A infinite voltage amplitude is such a meaningless answer.

By the way, in principle, I accept eVERITEC's latest post - I had forgotten that Steinchen took out the DC servo in the JISBOS. I generally agree with Regal's post, too. Neither really explain the startup spikes, though. In a classic output cap coupling scenario, when the caps are sitting right on the output and are much larger, you can attribute this to some discrete time for charging. I don't see how that's the case with 0.1uF.

The point is, you guys don't come back and accuse us of not having a civil conversation about someone's mistatement when the first thing you start out with is "quite ridiculous," etc.