If low noise, precision, etc. is indeed required then why not just use the circuit architecture of countless mic preamps, which is discrete differential stage amplified by any generic opamp. Naturally matched, low noise, etc. transistors need to be handpicked for the duty, but that is really no different than what one would do when building an entire discrete opamp for the purpose.

I'm quite sure that the simple discrete design (which they can house on those tiny boards) is no match for a much more complex integrated circuit.

I fail to comprehend what particularly makes it superior to other opamps. Just being discrete isn't exactly a merit by itseli.

Thank you!

Most of them you actually do find from internet. Gibson master service manual, for example, covers a lot.

Those Estey (Magnatone) M-series schematic I have never encountered...

QuoteSo if it limits the low freq to some degree then it obviously alters the frequency response curve as that FB is increased/decreased.

Well, the point is pretty much altering the frequency response. Similarly to high output Z tube power amps, voltage gain to load increases when load increases. With loudspeaker, a reactive load, the amp boosts gain at (moderately low) resonant frequency of the speaker and towards upper high frequencies where impedance increases due to voice coil's inductance.
Yes, RC networks can also be implemented to the loop itself. They commonly serve at least one of these functions: Phase correction, "damping" highest impedance peaks.

QuoteI notice the Jordan 440 schematic sweeps both types of FB (via that pot in red circle) allowing for some adjustment either way. I think I've seen Peavey circuits with similar ideas.

Yes, it basically adjusts between low output Z (where voltage gain is not interacting with load Z) and high output Z (where voltage gain is affected by load Z and amplifier's frequency response is altered by the load). So with generic loudspeakers is goes between "level response" and response with boosted highs and lows.
IMO, the effect is quite often rather subtle so continuous control (e.g. potentiometer) usually adds no more benefit than a simple three-way switch that adjust damping between "high", "medium" and "low".

One can tweak the feedback circuit to produce lower voltage gain to lower impedance loads. This will simultaneously limit current delivered to those lower impedance loads, and thus power dissipation of the output devices. The "current sensing" resistor additionally prohibits a complete short circuiting of the load.

It's not maybe best of protections, most likely far from it. But at least it limits lowest load impedance to resistance of the current sensing resistor and does not try to deliver almost full rail voltage to low impedance loads. I think one could expect the circuit to handle momentarily short-circuited loads but it probably fails rather quickly if the condition is sustained for long periods of time. Anyway, it's certainly better than nothing.

---

I'm not too sure if Fahey's earlier comments about no bias provision are correct. This does not look like a class-B design to me. Correct me if I'm wrong, but I believe that "diamond buffer" is providing the class-AB bias: In reference to circuit's input (the opamp output) voltage at base of driver transistors is one emitter voltage higher (or lower), pretty much like it would be when dropping over a a generic diode junction. Diamond buffer simply exploits the base-emitter junction of the cathode follower. The output pairs are Sziklai so they introduce just a single junction where voltage drops, thus two diode voltage drops per whole complementary pair should be enough. At least in theory.

Adequate temperature tracking could be an issue with such a simple setup though. Especially because that output stage is configured to introduce voltage gain, which means it likely amplifies its thermal coefficiency.

For example:


As you see, it's pretty much the generic current feedback topology. In just few years this basic circuit would convert from being a short circuit protection to being a tube amp emulation feature instead. Simultanously gaining worldwide popularity as that.

Yes, information was much harder to acquire back then. Therefore I beleive there must have happened something very groundbreaking that explains why this scheme spread to numerous solid-state guitar amps like a wildfire in the early 1970's - and particularly NOT as a protection feature but as a tube amp emulation feature.

When I started to research all this, what I discovered was that many of these kinds of features (current feedback) were actually introduced much earlier than people commonly think.

There's nothing new in mixing in different feedback topologies. A lot of stuff about that was discovered already in the early days of tube amps and some of the inventions are still exploited. Certain Bogen tube amps from few decades past were famous for employing -positive- current feedback because it - like negative voltage feedback - decreases effective output impedance of the amplifier. If interested, you likely find a few related magazine articles and patent documents with a Google search. Anyway, what the solid-state stuff is doing is just inversion of the idea: Apply negative current feedback, because it increases effective output impedance of the amp.

Overall this scheme must have not been -that- uncommon for amplifier designers. Especially for those who earn a living with it.

Anyway, I wonder what groundbreaking happened in the late 1960's, because after that we see a boom of this basic "current feeback" -topology making an appearance in many solid-state amps, and yeah, particularly as a deliberate attempt to emulate tube amps.

Ca. 1969 Triumph introduced amps that employed negative current feedback and had a moderately high output impedance. However, the circuit documentation still describes the associated circuit as a crude short circuit protection scheme.
Then all suddenly... The self-powered cabs of the Fender Super Showman system (ca. 1969) employed current feedback and associated tube/SS -tone switch. Jordan amplifiers (early 1970's) employed current feedback, with or without adjustment. Randall, Risson, Polytone, etc. In early 1970's the scheme is employed all over the place, and I even heard that there is a German book about solid-state guitar amplifiers from the time that describes the scheme. Several early 1980's patents (at least european ones) refer to the scheme as well-known "prior art" so basically every designer for every bigger brand out there must have at least been aware of this stuff, whether they chose to implement it or not. I wonder what happened?

Let's put this to proper context in time line: All this stuff happens about 15 years -before- the famous "Carver challenge".

Agreed. It's most likely not glue but thermal paste. Yes, sometimes it hardens with age and sort of sticks like glue, sort of, but all in all you should be able to pull the transistors off the heatsink by carefully applying some pressure to them. ...Unless some idiot really glued them, but I'm quite sure it didn't leave the factory that way.

One can read between the lines and make few deductions:

They call the "Restricta (TM)" circuit a "power amp driver", so it's not neccessarily a power amp at all, it could be just a stage that drives it. "Restrictor" sounds a lot like limiter.

They call it a "logarithmic driver", so I guess the stage has a logarithmic gain function instead of linear one. Hmm... "The early stages of its PA distortion really just compresses the sound a little, giving it a fatter tone as the volume is turned up. This also progressively reduces the dynamic range, the louder you play until, finally, the brain interprets it as distortion."

Sounds suspiciously like ordinary "soft clipping". Hmmm... Even generic diode clipping actually introduces a said logarithmic gain function: Small signal levels do not forward bias the diode but at higher signal levels dynamic impedance of gradually forward biasing the diode attenuates overall gain in logarithmic manner. Instead of linear gain function that abruptly turns to hard clipping the diode introduces a round "knee" to the transition area from linear operation to full clipping. Basically, the transisition to distinct clipping distortion becomes less abrupt at the cost of increased overall signal distortion.

Basically some power amp headroom is sacrificed to achieve softer transition to clipping distortion. You can find many examples of such guitar amp designs. Without soft clipping the power amp would generate higher output power within the quoted THD spec but transistion to clipping would be equally more abrupt.

I'm not sure if the power amp stage features the "Dynamic Feedback" of their "RetroTone" mods, but that is basically just generic current feedback. It's been somewhat a "standard" feature of SS guitar amps for a few decades.

Generally I would agree with Fahey...

But in this case lack of Zobel might have something to do with the amp section's purpose of driving a tweeter only. Generally the basic RC zobel introduces some capacitance to compensate increasing inductance of a generic dynamic loudspeaker at higher frequencies. ...but what happens when the load is something like a piezo tweeter and behaves very much like a capacitive load.

This is somekind of Behringer's active powered speaker, bi- or tri amped, right? I'm sure they did not produce several thousand units that were prone to fail from the start. I'm also suspecting there's a generic Zobel in the "LF amp" section, which probably parallels a couple of LM3886 chips for increased power.

In practice, there aren't too many "digital" amplifiers in 100% meaning of the term. Most, inherently, require quite a lot of analog circuitry for overall assistance. You need analog circuitry for IN and OUT interfacing with CODEC chips, for multiplexing analog control information and converting it to digital form. SMPS supplies are practically analog and technically there aren't "digital" power amps either, there are even very few class-D designs that would feature digital data inputs (for internal DA conversion).

Also, since beginning many of these "digital modeling" units have exploited analog circuitry even in signal processing duties. "Hybrid" amps like Vox Valvereactor series and Peavey Vypyr series implement huge parts of their modeling with analog signal processing circuitry. "Valvereactor" is basically an analog effect processor circuit plus power amp in same package that emulates operation of a push-pull tube power amplifier. Peavey's "T-Dynamics" power amp is in many ways very similar circuit. Their Vypyr series amps even feature DA/AD loop in the signal path to produce harmonic distortion in an entirely analog circuit instead of generating it with a DSP algorithm. In last NAMM Vox announced to introduce modeling amplifiers with similar operating principle. The new Quilter power amps include an analog tube power amp emulation circuit very similar to one featured years ago in some "digital" Line 6 amplifiers. Even Roland's "Cube" series amplifiers feature analog "soft clipping" and "tube emulation" circuitry. You can't solely judge what is "digital" just by cosmetic appearance of things. Many early preamp processors - accused to sound too "digital" - in fact employed digital circuitry only for user interface and generic DSP effect chips could only produce time-based effects, like echos and reverbs. Overdrive, filters, gain-modulation, etc. was all handled by 100% analog circuits that had (digitally assisted) reconfiguring signal path / circuitry. Everything just -looked- digital.

In practice, by little bit of research you can easily find some general guidelines for what typical CODEC chips require as analog interface. Input voltage range and current should preferably be limited, there are single-ended and push-pull inputs and outputs, many CODECs recommend external analog circuitry for filtering both input and output signals to ensure optimal fidelity. etc.

Control information is, in many "digital" units, generated and read in very similar manner. (e.g. potentiometers supplied by low DC voltage potential, wiper voltages multiplexed and finally analog information converted to digital).

The actual DSP processing, and how its implemented, can vary a lot. There is a trend to belittle possibilities of digital modeling, saying it "simplifies" things too much to be realistic, but DSP algorithms of even the early mid 1990's modeling units were - in fact - quite dynamic by nature and a lot of research had been put to emulate tube amplifier performance in very realistic degree.

In practice algorithms are largely limited by memory and processing time involved. Today's digital signal processor get faster and faster, though, and optimising algorithms, data, etc. to decrease processing time, for example, forms a major part of content of all patents relating to "digital modeling" and alike.

For good readout I suggest you Google search for Jyri Pakarinen and David T. Yeh's whitepaper titled "A Revied of Digital Techniques for Modeling Vacuum-Tube Guitar Amplifiers". Matti Karjalainen has also published many interesting whitepapers like "Wave Digital Simulation of a Vacuum Tube Amplifier". This certainly IS NOT a topic one could cover in single post in an Internet forum. You can find hundreds and hundreds of pages of reading if you truly want to research the topic.

I agree with Enzo, "DSP" is as descriptive term as "analog", so - IMO - not descriptive at all. If we assume quality ADA conversion then "DSP" is practically entirely dependent on algorithms of signal processing introduced. And then we come to realize that there is equivalent variation in them as in "analog" signal processing circuits in general.

Additionally, many of these "DSP" amplifiers exploit analog signal processing as well: Vox Valvetronix series has DSP-based preamplifier but the tube power amplifier emulation implemented within the output section is entirely analog. Peavey Vypyr series has largely DSP-based preamp but the overdrive sections are actually analog circuits. Tube emulation implemented within the power amp section is again entirely analog. Roland's "Cube" series amplifiers also exploit analog circuits that mimic overdrive characteristics of push-pull amps, and I recall a few Line 6 amplifiers that also had an analog push-pull tube power amplifier simulation implemented to the power amp stage.

If it looks "digital" don't be so sure it is 100%. You can't judge a sound by eyes, but do note that when your brains process sensory information you are troubled by this thing called "prejudice bias", which will alter your perception if you think you know what you are listening to. I remember several people mocking some Digitech preamps for their "digital" tones eventhough associated signal processing circuits were 100% analog and digital circuits implemented to the units were practically used for controlling signal path switching and for the user interface. Do note that there are plenty of "DSP" -thingies where DSP is used only for applying crude time-based effects, like echos, delays, reverbs or choruses. Everything else could be 100% analog. Few years ago people ridiculed Axe-FX because it had this "digital fizz"... only to discover that what they were hearing was actually crossover distortion, similar to that introduced by many overdriven push-pull tube amps. So "digital fizzyness" was actually just more realism in the tube emulation, and as such an entirely user-controllable parameter. Unlike in tube amps. Which is nice thing in many modelers: They can do things tube amps couldn't.

Then there are "DSP" amps where pretty much everything is done digitally and only analog stages are those required by ADA conversions. AXE-FX perhaps? Even if there is a power amp it could be a class-D design that receives digital input instead of analog signal and does not implement an analog modulation like most class-D design. We can never know that by hearing.

My point is, one should be absolutely sure of what he is hearing before making too many accusations or drawing too many conclusions. In many cases we only -think- we know what we are hearing. ...Like when listening "DSP" tones that are actually partially produced by analog signal processing circuits. That happens a lot.

Quote from: Enzo on January 26, 2016, 06:09:20 PM
And for that matter there are a few designs that bring the cathode of some stages down to a negative supply, I assume to provide increase headroom.

Or to allow DC coupling between stages.

QuoteIt refers to any power supply source distributed through a circuit. 

This.

And in some systems the distribution scheme involved using an actual metal rail for the task. In typical electronic schemes wires, circuit board traces and such tend to replace a genuine "rail" but nevertheless the function, distribution, remains.

For example, a generic breadboard. It should be pretty obvious to spot what "rails" in this case are. In practice various subcircuits that draw current need physical connections to power supply. Conductors distributing those currents can be considred "power rails".