QuoteNow if you want pure studio super clean compression DON'T bother,, but if you are wanting something that comes remarkably close to soft clipped Valve power stage compression then this would be very hard to beat.

There's already the A110 chip and involved circuitry, for "tube like" clipping. It employs "soft clipping" characteristics of the CA3094 OTA and drives Iabc terminal with input signal to achieve dynamically varying degree of asymmetric clipping (and related effects to harmonic distortion).

Compressor part of L5 is actually incredibly simpe. There's a JFET source follower working as a "unity gain" amplifier and a CA3080E OTA gain stage essentially working as its variable "feedback loop". In order to properly feed one of OTA inputs the signal must be heavily attenuated (input range is few millivolts). Varying gain of this stage also varies the gain of the entire follower circuit. The stage can be considered as plain "open loop" amp (with lowish gain), but gain can also be controlled by applying signal to Iabc terminal.

Signal from JFET source advances to three places: aforementioned OTA input, power amp input and circuit that generates Iabc control signal for the OTA.
 
So other circuitry is there pretty much just to generate that control signal: First there's a plain resistive attenuator for the "input" signal, then signal is amplified. Two diodes and an inverting amp full wave rectify the signal, C139 and R194 make up the RC circuit for "filtering" the rectified signal. This circuit primarily establishes attack, sustain and decay characteristics of the compression. Very "crude" circuit for anything remarkable on that regard, BTW. The result signal drives transistor Q102 and it's collector circuit in turn drives both the OTA Iabc input and another transistor, which simply drives a LED indicator lamp.

If an amp has an OTA-based compressor then the circuit most likely follows pretty much this architecture. Follower, OTA, rectifier + RC filter, BJT drive for OTA. What you have in L5, IMO, is about the most minimalistic example of an OTA based compressor circuit. Really.

QuoteYeah, that's it! This is a fixed-bias stage that's driven from the cathode instead of the grid.
Would then the transistor be simply a common-emitter stage, with the tube working as it's "collector resistor" and at the same time, the transistor working as the tube's "cathode resistor"?

It's a "cascode" so it's simply direct coupling two amplifiers. "Up", we have common grid amplifier with speaker load reflected to its plate via transformer (as usual). Since grid of a common grid amp is held in "common" potential the amplifier is effectively driven by its cathode current/voltage. Instead of modulating grid voltage with AC signal (and shunting AC at cathode to common), cathode voltage is modulated (while AC at grid is shunted to common). The tube naturally just cares about grid-cathode voltage difference, which really makes the whole arrangement sensible (insofar how upper section of the cascode is both driven and biased). At the "bottom" we have a plain common emitter amp, which effectively controls the cathode current/voltage. Voltage drop across the BJT also establishes bias point for the upper device (along with DC voltage potential provided at the grids).

What's "load" in this type of circuit anyway. It's something to ensure proper current flow and proper voltage potentials. One can use plain resistors, or substitute them with current sources, or ampliers (gain stages) acting very much like such. What is a transistor anyway? It's shortened from "Trans-Resistor" so think of it as nothing but a variable resistor controlled by base-emitter current.

QuoteThe tube works on the relationship between grid and cathode, not necessarily those elements and ground.  Hold cathode at ground and wiggle the grid, you get signal at the plate.  But hold the grid to ground and wiggle the cathode, the same thing happens.

Pretty much, but there are some charactistic differences between these two circuit types. One difference is, fo example, that "grounded grid" amplifier does not invert its output signal. Basically you can consider grid of the tube as inverting input and cathode as non-inverting input.

QuoteIn practice, the amp examples I cited are running closer to class B than something like a fender twin Reverb would be.

This. They are also heavily influenced by anything driving the cathode (e.g. bipolar transistor). In so much that Peavey went on to patent a circuit (VTX series amps) which provides "softer" crossovering characteristics to their circuit (because in plain format, like that of Music Man amps, this kind of cascode arrangement has "hard" and quite bias-sensitive crossovering characteristics of a bipolar transistor circuit).

...Speaking of Music Man amps. They are known for their high plate voltage vs. reliability.... but few seldom mention that they also operated at exceptionally low (on generic guitar amp standards) screen voltage. The same little detail was also the key difference of prototype Ampeg SVTs (that failed catastrophically even with extremely reliable and rugged output tubes) and later SVTs that didn't (which no longer had tubes of equal ruggedness but simply much lower screen voltage). Something to ponder at. 

This.

In practice, the first tube stage will not clip at all, but it has a lot of voltage gain, and more importanly output voltage swing. Given that second tube starts to grid clip already at few volts of input signal, and has enough gain to start clipping at plates about simultaneously you do get plenty of ovedrive from the second tube.

What the diodes do is limiting the input voltage range to about 600 mV max (peak). At this point the final tube is already driven to "square wave" clipping and there's very little to gain from that point on. The diodes prevent excessive ovedrive of tubes (e.g. excessive grid current draw, which could be hazardous to tube or introduce "nasty" high order harmonic distortion) and also enforce fairly symmetric signal input to the tube waveshaper stages. At "in between" regions you get a little bit of harmonic distortion from the diodes softly compressing the signal by clipping, and more harmonic distortion from subsequent tube stage, which already at point of diode compression is going to clip rather much.

Datasheet example circuit:


It's principally the same idea described by Enzo, but with a little bit of extra twist.

Supply is unipolar, so chip's "V+" terminal is connected to the positive terminal of the DC power supply (V+) and "V-" terminal is connected to the negative terminal of the power supply (GND/COM). The chip - like many other opamps - works in bipolar-supply fashion and therefore needs its DC offset voltage adjusted to about half of the V+-to-V- voltage. This - at least in simplified opamp theory - provides equal "headroom" for the circuit to operate at both negative and positive voltage excursions. (In practice it will not always apply).

If supply is bipolar (e.g. +15VDC/-15VDC) then "ideally correct" DC offset voltage is 0V.
If supply is unipolar (e.g. +30VDC/GND) then "ideally correct" DC offset voltage is +15VDC.

Note that in both cases V+ terminal's potential referenced to (no pun intented) reference voltage is +15VDC, and V- terminal's potential (with same reference) is -15VDC.

Basically it's only about shifting DC references, while potentials remain the same. The chip only cares about valid supply voltage potential differences provided to its terminals so as long these are not violated you can practically shift the DC references anywhere. Signal path naturally needs to be AC coupled so you need coupling caps in input and output.

In the datasheet application circuit chip is DC biased to these conditions by supplying the input circuit with DC reference. Input is "DC referenced" via resistor RA 75k - not to ground (0V) - but to the "DC offset bias" voltage (which is approximately half of V+). The reference voltage is generated with resistive divider RA91k / RA 100k. CA provides additional filteration to keep offset steady.

The internal ground of LM3886 (pin 7) is similarly hooked - not to power supply ground - but to DC offset bias reference. The twist is that the reference voltage to this terminal is provided via "buffer", here a simple emitter follower. It's basically a voltage regulator, which keeps emitter voltage steady despite of magnitude of current draw by emitter circuit. Base circuit also inherently needs only very low currents to operate.

As long as we can retain ground (reference) currents low, we can keep impedance of the reference voltage source high. This has some benefits, like lesser current draw by the reference circuit itself, and much lower filtering capacitance requirements for the voltage reference node. If ground current draw/sink is higher the reference circuit must be lower impedance. "Buffering" the reference circuit with a (low output Z) current amplifier is common practice to achieve this without sacrificing high impedance of the reference circuit itself.

Yes, it would probably work also in simplified form (sans regulator) referred by Enzo, but I have a hunch that pin 7 draws or sinks considerable currents and the 91K/100K divider circuit would be too high impedance to be compatible. You can probably estimate magnitude of currents within pin 7, and calculate compatible impedance requirements of the resistive divider. I have a hunch that ideal values would prove to be an excessive waste power load, which can be largely reduced by "buffering". The benefits are probably vast enough to support one cheap transistor more in the BOM.

---

QuoteI went with a single supply because the tube preamp. I couldn't figure out how to do a dual supply that wouldn't leave the tube voltage unreferenced to anything (which intuitively I assume is unstable/dangerous). Is there a way to mix in a tube on a dual supply?

Well... You don't say much about details of that power supply. If it's single supply of tube circuit + dual supply for rest then both simply have the same ground reference and that's it. If you need to operate the tube circuit from the same dual supply used by rest of the circuit then you simply substitute ground reference of the tube circuit with V- (its technically as low impedance as ground so why not) and use capacitor coupling (which you'd probably do anyway). Not rocket science.

Final part of the full-wave rectifier made out of two forward biased half wave recitifiers (the diodes) and a differential amp (Q3 & Q4).

I think the point of that BJT differential is that it operates less ideally than a traditional FWR and somewhat "smoothens" the transitional region to cutoff state. If a sinusoidal input is fed to traditional FWR you get a "pulsating" output (IOW rectified sine wave) at 2x input frequency (IOW an "octave up" effect). When you feed the sinusoidal input to this circuit it tends distorts so much near the cutoff state that the output resembles more a sinusoidal wave (at 2x input frequency) than a traditional rectified sinusoidal wave.

QuoteI notice ESP mounts the bias sense transistor right on the predrive, hum?

The output is configured as "Sziklai pairs": Collector of the driver drives base.

Unlike in "Darlington pair" (emitter of the driver drives base), where we need thermal tracking of the output transistors, in Sziklai pair we must thermally track the driver transistors.

Similarly, it is beneficial to thermally isolate driver and output transistors in Sziklai pair while in Darlington pair you need to thermally couple them through common heatsink.

QuoteHow high can you push these rather simple power amp designs like my Laney KD circuit when it comes to higher voltage rails?

Depends.

Your power transistors will have some maximum collector-emitter voltage limit, and they they will have a certain "safe operating area" or SOA, which is dependent on many things, such as rail voltage and load.

Keeping the output devices within their "SOA", for instance, becomes much harder with reactive loads because of different phase in voltage and current, which might lead to much higher dissipation requirements than those estimated to purely resistive loads.

Since these "simple" design employ either a very simple scheme to force output devices to stay within their "SOA" (simply schemes simply can't do this reliably) - or no such scheme at all (even worse) - I wouldn't have high hopes that they fair decently in reliability department as soon as you change from using purely resistive dummy loads to reactive speaker loads. And those chances to survive would pretty much decrease along with increases in rail voltage.

Oh, I also like how the article is written largely from very objective theoretical point of view. It addresses certain phenomenon as is, and does not even try to pull the discussion about them into discussion about digital vs. analog signal processing, solid-state vs. tube amps, or anything of that manner (at least any more than neccessary).

Yeah, it seems to be written with largely final studio mastering stages etc. in mind, so it's not really addressing "warming" of individual instruments, like electric guitars. Nevertheless, I think the overall concept - which is trying to address that there is no "good" or "bad" clipping, or "good" or "bad" harmonics, when all that is largely dependent on application and related preferences - is on point.

Just bumbed to this little whitepaper.

T. Serafini S. Barbati, "A Perceptual Approach on Clipping and Saturation", 2002

http://www.simulanalog.org/clip.pdf

Short (4 pages) and written in layman's terms. It basically tries to address the issue that "musical" special effects can't be evaluated with same principles as high fidelity sound reproduction and therefore some pre-established concepts about "good" or "bad" harmonics/clipping need to be reconsidered depending on application. Totally worth reading.

Oh, and....

Refer to this diagram concerning the correct diode orientation.

QuoteOh I added the PDF for LaneyKB80, yes that 100k resistor (R2) does go back to common at the header plug, clearly drawn top right header plug.

Those Laney printed circuit boards are quite "modular": Same power amp boards are used in a wide range of amps, and if you take a more careful look even the "pin order" of that connector is retained throughout a wide range of Laney amps.

Probably makes experimenting in R&D process much easier when you can simply in "plug&play -style" combine almost all preamp boards to almost all power amp boards, within some rational limits of course.

Some Laney amps have a limiter circuit in the preamp that is in some form driven by an attenuated speaker output signal. If the amp has a limiter the power amp board connections support the feature (you know, that "modularity" thing), if it doesn't the reduntant wire is simply used to "double" common connection for little additional bit of reliability.

According to my schematic archive JX50, JX40, and JX30 have practically identical preamp sections. Maybe JX25 does as well.

Pretty much a plain shunt diode clipping arrangement for O/D tones: Switch provides more voltage gain pre clipping diodes (thus more overdrive) and more low-pass filtering post clipping diodes (thus "smoother" O/D tone with less annoying high frequency "fizz").

Few points worth noting: 1S1555 diodes has few hundred millivolts higher forward voltage than a generic silicon diode. ...If there even is such a thing as "generic" silicon diode. A failry high series resistance is employed post clipping to limit diode current. These features provide a moderately "soft" clipping performance.

Quote from: Enzo on May 07, 2015, 11:02:39 PM
There is nothing special about using a 3886 for a guitar power amp.  It will just be a power amp.  What makes a guitar amp special for the guitar is the preamp.  And of course an appropriate speaker.

...Then again it's just a matter of adding some external circuitry to turn the power amp "special" as well:



As much as I respect Enzo, I do believe engineers of likes of Pritchard, Quilter, Peavey, Vox, etc. would entirely disagree with him. I basically falls down to whether you want to design just another "HiFi" chip power amp with LM3886, or a power amp that behaves more like a tube power amp... also using an LM3886 chip. LM3886 is just one high power "opamp" and like with everything else regarding opamps it's the surrounding circuitry that defines the function.