in another forum...

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Ahh thanks guys.

So effectively it is just moving the point at which the opamp output stage moves from Class-A, prolonging the onset of x-over distortion but eventually it will leave Class-A if the load is low enough?

Sorry for taking this off-topic, but what is the way to calculate the current needed for a particular load, given in datasheets, simple ohms law?

I.E 10v pk-pk into 10k load, I=10/10000, about a 1mA??

So you pick a value higher than that and calculate the resistor.....

Say 2mA on 15V rails.
R=15/0.002, 7k5 resistor from out to neg rail??

Again sorry to go off topic but this seems more worthwhile to me than whinging about PSW.....

Cheers Tom
 
So effectively it is just moving the point at which the opamp output stage moves from Class-A, prolonging the onset of x-over distortion but eventually it will leave Class-A if the load is low enough?

In a sense, yes. The goal is to supply the NPN output stage with more current than the load will draw.

So you pick a value higher than that and calculate the resistor.....

Right. The total load on the op-amp will be the input impedance of the next stage in parallel with the effective current load resistance. In your case the opamp with the 10k component load in parallel with the 7.5k load will end up with a 4.28 kohm total load, so make sure that you can handle seeing this small of a load in terms of the output stage.

Now to complicate things a bit, in theory at least. A resistor is not really the best current source--there are two problems that we face with this approach.

First, it will only give a constant current level when the opamp's output signal is constant. Since audio is a varying signal, the bias level will only be constant during the silent parts--which doesn't do much for us the rest of the time. This isn't to say that a resistor bias is stupid or useless, just that the bias level will vary with the output signal level, which isn't really optimal.

The second problem with this method is that the higher the op-amp's load impedance, the better the dynamic performance of the op-amp. That's one reason buffers help op-amps to perform better--they present an impedance in the megohms range to the op-amp. Using a resistor down in the kilohms range to bias the opamp's output really kind of negates this, huh?

So probably a JFET or CRD is the better method.

*edit--clarity*
 
> I.E 10v pk-pk into 10k load, I=10/10000, about a 1mA??

Except use Peak, not peak-peak. So 0.5mA. And round-up: you don't want to go all the way to the edge of starvation. So 0.7mA to 1.0mA.

I think in general a resistor isn't a great answer. If you need large voltage swing, the resistor must be so small that it loads-down the amplifier. In HiFi where we have 3V peaks and 10K loads, a 3K pull-down is no big deal for most chips. But in recording we may be pushing 10V peaks into low-Z or long cables. Sucking-down with 220 ohms is probably going to upset the chip worse than the crossover nonlinearity.
 
"So probably a JFET or CRD is the better method. "

It's a few more parts but I would use a bipolar, couple of diodes in series to bias the base, resistor from the base to either common or the + rail, and a resistor in the emitter.

OTOH with one more transistor you could make a complementary output buffer. Run it reasonably rich and you have a nice performing circuit, and you offload the self-heating in the op amp to the buffer.
 
I am careful with how I ground. I often use single point ground wired point to point or perf. I sometimes wire in 3D and return each ground and power wire for each section to a single point for power and one for ground(might be overkill) I do this in effects, guitar amps, guitars, HiFi, microphones.

If signals that can cause problems cross, try to cross at 90 degrees for lowest coupling. Short wires often helps. twisted pair can sometimes help.

Good caps in the power supply (Low ESR)

Try to balence the input current of the diff pair, Sometimes this is overlooked.

I did post in another thread that the PRR 5534 tip 50?headphone amp might be fun as a gain stage

And it has been posted in other threads and this one don't have the opamp drive a load it is not rated for.

There are good posts above this.
 
I think this speaks volumes for the folks here at the Lab, too. You just can't get away with posting bogus stuff here because too many folks know better. Another reason I love this place!

BTW, I'm posting this as I sit on the beach and watch the sun set in the gulf - isn't the internet wonderful!!
 
Except use Peak, not peak-peak. So 0.5mA. And round-up: you don't want to go all the way to the edge of starvation. So 0.7mA to 1.0mA.
Shame on me for not pointing this out in the first place.

In HiFi where we have 3V peaks and 10K loads, a 3K pull-down is no big deal for most chips. But in recording we may be pushing 10V peaks into low-Z or long cables. Sucking-down with 220 ohms is probably going to upset the chip worse than the crossover nonlinearity.

PRR--this holds true with smaller values, but we're still looking at a pretty large value (3.64k) for Tom to bias for about 1ma. Obviously, we face the typical have-our-cake-and-eat-it-too predicament, but
if we're looking at biassing everything except for the buffer stage, we should be okay in terms of loading, am I correct?

What about a JFET cascode as a current source? I'd love to your hear opinions about that...it seems to be a circuit that Horowitz and Hill like.
 
[quote author="bcarso"]"So probably a JFET or CRD is the better method. "

It's a few more parts but I would use a bipolar, couple of diodes in series to bias the base, resistor from the base to either common or the + rail, and a resistor in the emitter.[/quote]

I've done that, and it works quite nicely. I usually try to make sure the output never gets below 0.5-0.75mA. So if a circuit is putting out 10Vpk into a 5k load then it gets 2.5-2.75 mA bias from the CCS. When calculating the load, don't forget to include the feedback network!

OTOH with one more transistor you could make a complementary output buffer. Run it reasonably rich and you have a nice performing circuit, and you offload the self-heating in the op amp to the buffer.

Yes. I worry, though, about stability with fast ICs and another stage, espeically when they're facing who-knows-what-loading on the outside. So for the most part I've stuck with current-source biasing.

Peace,
Paul
 
[quote author="soundguy"]IC's killed music, please. All music? Thats just a flat out moronic thing to say. If it wasnt for the IC, there wouldnt be any electronica, period.[/quote]
[quote author="soundguy also"]
____________________
running%20cat.gif

chips are good with dip[/quote]

:green:

Keef
 
It always amazes me how many people have and express opinions about the tubes vs transistors issue without ever having done any research. It's very easy to get wrapped up in a "group lynch mob" mentality. I wonder how many tube advocates realize how many of their commercial golden recordings have passed through solid state circuitry during the recording, mastering and duplication stages. I have yet to see a disclaimer on a recording stating " No transistors were used in the making of this recording."

Tubes have and probably will continue to have a presence in the recording industry, but we cannot shun transistors just because somebody tells us they aren't good. Maybe their eveluation wasn't implemented properly. Overload a tube stage down with too low an impedance load and see how beautiful it sounds!
 
There are certain things that tubes do well, and that can be understood based on what we know about psychoacoustics and physics. There are certain things that have come to be associated with tubes and tube sound owing to the constraints that, as PRR says, "hollow state", have placed on the overall circuit design.

Tubes employ rather large structures compared to transistors. Consequently the thermal effects tend to be rather long-time-constant in nature. There are a lot of charge carriers involved, although they are a lot less densely distributed. There is a lot of thermal jostling of those carriers and probably less of an interaction of the associated noise with the signal---some have said that is is easier to ignore the noise in tube electronics, that it seems to float there independent of the signal somehow.

As well, the input capacitances tend to be more constant with operating point, reducing a distortion mechanism compared to solid-state devices used in similarly simple configurations.

To take advantage of the very high input Z of tubes and the highish output Z, we use transformers. Were it not for tubes the art of lower-frequency multi-octave transformer design would probably not have advanced nearly so much as it has.

Transformers are wonderful devices for isolating systems from the vagaries of interference like hum and much EMI. They also tend to be inherent bandpass filters. And like any other component they have limitations at high signal levels, and very distinct ways of responding to such overload that is usually measureably and subjectively benign.

Given transformers and their limited bandwidth, or even without transformers and just with the constraints of R-C coupling, it becomes difficult to apply large amounts of negative feedback around a multistage amplifier. But thus is sidestepped some of the problems involved with poorly-designed high-feedback systems.

In addition, triodes at least with high-Z plate loads can be remarkably linear voltage amplifiers over a useful signal range. But it is hard to extract this signal and drive a load without spoiling the performance.

Another characteristic of tubes that worked to constrain designs is their existence in only one polarity. This favored balanced designs from the outset when one was trying to achieve high performance, which conferred other advantages in many ways.

Bipolar solid-state devices were the first practical ones available. It was natural to insert them into tube-like circuits as those were all we knew, but they didn't work very well. They do have high transconductance though, and thus local feedback can often linearize their performance. Since they come in two different polarities a number of topologies become available that are difficult or impossible with tubes. But there was much to learn.

The transistor is a microscopic structure compared to a tube, especially when the two handle similar currents. It's instructive to crack open a plastic small-signal transistor and realize just how tiny the actual chip is. And then realize that what you see is a chip where most of the action is occurring very close to the surface, typically.

The transistor also has a variety of parameters that are sensitive functions of temperature. Couple this with the tiny thermal mass and you have a recipe for signal-induced shifts in those parameters---with which, by the way, affordable simulators don't even attempt to deal.

Bipolars do have an amazingly accurate conformance to an exponential dependence of collector current on base-emitter voltage, at a constant temperature. This can be exploited.

But for the simpler topologies the easiest thing to do is use lots of local and global feedback. With it comes all of the problems of transient response and stability against oscillations under varying conditions. And we know that for smaller amounts of global feedback there is even the effect of creating additional energy at higher harmonics in the distortion spectrum.

The availability of monolithically constructed, inherently matched and thermally coupled devices allows for a number of circuit enhancements. At the same time, putting whole amplifers on a chip requires close attention to the thermal effects, and usually entails a lot of parasitic capacitances. In this regard a mix of monolithic matched multiples and discrete devices can yield superior performance.
 
As well, the input capacitances tend to be more constant with operating point, reducing a distortion mechanism compared to solid-state devices used in similarly simple configurations.

So, you're proposing that physical size has something to do with the device's internal capacitance, am I understanding this correctly?

In this regard a mix of monolithic matched multiples and discrete devices can yield superior performance.

Superior to...solid state designs of the past? Or superior to tubes?
I know that new advances in semiconductor manufacturing and laser trimming have improved on-chip performance in the past few years...
 
"So, you're proposing that physical size has something to do with the device's internal capacitance, am I understanding this correctly?"

Really more the capacitance and its variation with voltage and current having to do with geometry and the material. It's complex, but for example to a first approximation the electron flow in a tube doesn't influence the plate-grid capacitance all that much (see PRR's answer to my query about how much of an effect there is, referencing the Radiotron Designer's Handbook 4th ed., around page 55).

"Superior to...solid state designs of the past? Or superior to tubes?"

Superior to the best that can be done with ICs, usually. Actually there are some possible hybrid tube/sand-state designs that may offer the best of both worlds compared to any designs past or present. And I should qualify this further: best performance, independent of cost, weight, parts count, physical ruggedness and reliability, and power consumption. This is an area of active investigation for me at the moment, time and money permitting.

"I know that new advances in semiconductor manufacturing and laser trimming have improved on-chip performance in the past few years..."

Absolutely---no argument there. But there will always be substrates and their capacitance and coupling---if it is a reverse-biased diode, then voltage-variable capacitances. Less of an issue for dielectrically isolated IC's. And the thermal stuff will always present issues, despite the many clever tricks of compensation and layout to ameliorate them.

At sufficiently high frequencies the benefits of low interconnect inductance put discrete circuits at a disadvantage, although with surface-mount devices this has pushed the transitional region up in frequency.

Unfortunately some of the really good new high frequency processes are not being made available in discrete form.

Don't get me wrong---I think properly used ICs are great and I couldn't design cost-effective equipment without them. Having said that, one of the last projects I did for a client used about 100 discrete transistors for various functions, based on the need to maximize performance and power consumption while constraining cost. Low power, decent audio performance op amps for example were just too expensive.
 
Well, the cool thing about DIY is you can build the classics for cheap.

So if someone comes in and says "Oh, I hate tubes!"
You show them your API's and Neves etc.

If they say "Oh, I can't stand transistors!"
You lead them to the tube rack.

You can't beat having a smogasboard of cool equipment, eh?
That's my DIY philosophy and I'm stickin to it. :wink:
 
threads like that one are why i love this forum. thank you everyone for a truly civil and an extremely informative place for me to grow.
 
[quote author="amorris"]threads like that one are why i love this forum. thank you everyone for a truly civil and an extremely informative place for me to grow.[/quote]

You got that right.! Amazingly, everybody seems to maintain a professional attitude on this forum. This is highly unusual on a public forum. Of course we aren't above a little humor; Tech stuff would be pretty dull otherwise.

It's interesting to note that even though most of us have never personally met, the personality / character of each member is revealed through the various posts.
 
Back
Top