Ad Widget

Collapse

Announcement

Collapse
No announcement yet.

Those Inter-stage Coupling Caps

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Those Inter-stage Coupling Caps

    An inter-stage coupling cap re-biases the signal for the following stage. It forms a low-cut filter. It works well for signals possessing little nearly DC content, and removes content at very low frequencies. It needs to pass 20Hz signals (or somewhat higher for guitar).

    What happens when the driving stage is clipping one edge of it's input waveform and flipping it over? We see (in our mind in this case) that the DC level of the waveform shifts away from the clipped edge, somewhat proportional to the amount of clipping, and this shifting is provided at a low-frequency rate.

    The shift is good for a distortion generator. By shifting the bias, the inter-stage coupling cap's circuit creates duty-cycle asymmetry, providing "nice" even harmonics.

    It all works great for constant amplitude audio frequency sine waves and their sums, but consider signal dynamics. If the driving stage clips one edge, and the receiving stage clips the other, you might think that you would get a constant amplitude output when both triodes are clipping. You don't. Once the driving stage starts clipping, increasing it's input level increases the change in DC level of the driving stage's output, shifting the level of the previously clipped edge that the receiving triode sees. This edge passes through the receiving stage unmolested, maintaining around half the dynamics of the original signal. The receiving stage's haircut clipping and following filter similarly reduces but does not eliminate the signal dynamics. You get around 1/4 of the dynamics, not none. And the even harmonics increase, even beyond the point where both stages are clipping.

    I've not seen discussion of this. RDH4 and the original designers would simply recommend that you turn down the amp. I recently read a comment in an article quoting John L. Murphy, former Carvin wizard, and he mentioned the even harmonic effect of bias shift, which got me thinking. It raises lots of questions. Does any amp modelling software take this into account? Can and does a Kemper-type amp capture it? Surely a few people here have played with different values of the cap in question, the associated grid leak, etc. What does the effect on pick-attack sound like?

    I discovered the concept of prosody when I looked it up in Wikipedia, and I want to increase the prosody of guitar amps - make them better communicators. Heavy clipping kills dynamics as it increases sustain and harmonic content. Making the dynamics live on as changes in harmonic content sounds like a good thing (or maybe it sounds like an auto-wah). It lets how hard you pick control a variable of musical prosody.

    It should be possible to remove the shift in signal offset, at least up to a certain frequency, for clean distortion, whatever that is. It could even have a knob.

    Of couse, if one string is played hard enough to clip the amp, the effect on the signals from other strings playing softly is dramatic. The weaker signals aren't just modified - They're unrecoverable. It's the sound of rawk, but I think the metal ERG guys are doing it just to annoy me. I'm already annoyed with the way chords and arpeggios sound when I adjust my amp for singing over-driven sustain on lead lines. Polyphonic distortion would help a great deal, but digital modelling the amp chains by brute force needs maybe four filters and three triode emulations time 8 strings, just for the preamp. Add more for the power amp, which shares similar problems. That's 32 filters and 24 look-up tables. I can't see any easy way to combine the filters or triodes due to the non-linearity of the triodes. I don't think a program can keep up in real-time. I can imagine an FPGA running fast enough to process eight channels through one filter structure and one look-up table structure, but it's a big job. Twelve 12AX7s is pretty unappealing, from a reliability standpoint alone. 24-40 JFETs starts to look good.

    Understand that I'm not interested in modelling exactly. If you want something that sounds just like a Plexi, you should probably buy or build a Plexi. Models should be judged by how they sound, not what they sound like.

  • #2
    Very interesting stuff. Here are a few thoughts that might be relevant (but I expect you know them already).

    If a triode stage is cathode-biased with a bypass cap on the cathode resistor, the voltage from cathode to ground tracks the 'average' cathode current over a cycle. So changes in duty cycle of this current will change this voltage, which forms the voltage bias for the tube (but only if the average voltage at the grid is zero).

    If the input voltage swing to the grid drives it positive relative to the cathode, then we get grid conduction. This adds two further bias-shifting effects:

    The grid current comes out via the cathode, adding to the anode current, increasing the average current through the cathode bias resistor and shifting the bias colder.

    When grid current is flowing the time constant for the incoming decoupling cap is much faster than when grid current is not flowing. This tends to shift the average level of the incoming waveform (at the grid) more negative, as the top of the waveform is clamped and cannot go much above 0V (relative to the cathode). Again, effectively making the bias colder.

    When grid current is not flowing at all during the cycle; as a result of the decoupling cap the incoming wave settles so that its average value (at the grid) is zero. So if we have a duty cycle where the voltage is up for 90% of the cycle and down for 10% we could get something like the following at the grid: +0.5V for 90% of the cycle and -4.5V for 10%.

    The above comments are really about steady state conditions. Under dynamic conditions, where the amplitude of the signal is changing and where duty cycles are varying, we get dynamically changing bias-shift and recovery, according to the relevant time-constants. In extreme cases we get blocking distortion, and for less extreme cases we get some nice dynamically changing harmonics.

    A long time ago I think someone experimented with an independent pick-up for each string, with 6 output channels each with their own 'fuzz-box'. I have also heard rumours that some bands have gone to the trouble of recording 'crunch rhythm chords' one note at a time. This avoids intermodulation distortion between the various notes of the chord.

    I also wonder how much of this stuff is really represented in digital amp modelling.
    Last edited by Malcolm Irving; 01-29-2015, 08:50 PM.

    Comment


    • #3
      TLDR

      any schematic/visual to help ADHDers?

      Comment


      • #4
        Does any amp modelling software take this into account?
        To be blunt, all competent amp designers have been aware of this phenomenon since what... 1970's? ...at least early 1980's...? That's stuff discovered several decades ago. There are several patents discussing this phenomenon and at least few of them were applied in early 1980's. There's really nothing new to this.

        So think about it? Why would NOT the modeling algorithms/software take it into account? I don't see any good reason. In fact, not so surprisingly even the very first Line 6 and Yamaha patents discuss methods to turn "static" waveshaping into more realistic "dynamic" process. So yes, they accounted it from pretty much day one.

        You might want to search for a whitepaper titled "A Review of Digital Techniques for Modeling Vacuum-Tube Guitar Amplifiers" for further reference because I'm sure as hell not going to summarize it here. The authors are Jyri Pakarinen and David T. Yeh.

        Comment


        • #5
          Originally posted by Malcolm Irving View Post
          ...Under dynamic conditions, where the amplitude of the signal is changing and where duty cycles are varying, we get dynamically changing bias-shift and recovery, according to the relevant time-constants. In extreme cases we get blocking distortion, and for less extreme cases we get some nice dynamically changing harmonics.
          IMO - this topic * time-varying parameter shifts * is hugely important and almost completely neglected in texts and technical discussions. It's coupled with non-linear circuit behavior, which classic circuit design and analysis methods mostly punt on. This is a unique problem with music (mostly guitar) amps, so it gets little attention beyond that circle. These shifts happen almost everywhere there are caps.

          There is a book waiting to be written about this. I've started notes on one myself, but have not made a serious start yet. Frankly, I'd be open to a collaboration on this. I think the topic is bigger than any one person's views and it will take some creative insights and methods to grapple with it. Glad to see this issue getting some print.
          “If you have integrity, nothing else matters. If you don't have integrity, nothing else matters.”
          -Alan K. Simpson, U.S. Senator, Wyoming, 1979-97

          Hofstadter's Law: It always takes longer than you expect, even when you take into account Hofstadter's Law.

          https://sites.google.com/site/stringsandfrets/

          Comment


          • #6
            Originally posted by teemuk View Post
            ...
            There's really nothing new to this.
            ...
            You might want to search for a whitepaper titled "A Review of Digital Techniques for Modeling Vacuum-Tube Guitar Amplifiers" for further reference because I'm sure as hell not going to summarize it here. The authors are Jyri Pakarinen and David T. Yeh.
            I don't think I was claiming anything new. In fact I stated that I thought OP would already know it!

            Thanks very much for the interesting reference paper, though. I'll read it thoroughly.

            Comment


            • #7
              Originally posted by teemuk View Post
              ... whitepaper titled "A Review of Digital Techniques for Modeling Vacuum-Tube Guitar Amplifiers"
              ... Jyri Pakarinen and David T. Yeh.
              Very interesting paper. But note that they state:

              "In conclusion, the complicated interdependencies and dynamic nonlinearities in vacuum-tube amplifiers make their accurate physical modelling extremely demanding. As a result, approximate models simulating only some of the most notable phenomena have been developed by the amplifier-modelling community."

              But that was in 2009, so progress continues no doubt.

              Comment


              • #8
                Interesting subject. I've been experimenting in this area for quite a while. One thing not mentioned so far is the influence of the power supply. What you see at the first preamp stage is a slow ramping of the B+. Down when you play a note and up when you release. Some of this makes it's way to the grid after the coupling cap, shifting the bias. It's hard to see the actual effect on the waveforms and you certainly don't see it with continous signals. What is really needed is a gizmo that will record and playback the guitar signal at the input, and a waveform recorder that will capture the output of a stage or complete amp and allow analysis. Some amps minimize the effect with large caps on the B+ for the preamp.

                The output transformer of a tube amp won't put out DC. But you can observe a baseline shift when an asymmetrical waveform is being produced. There is a time constant for this shift, I theorize it has to do with the low frequency performance of the OT. Is this what produces speaker cone bobbing? I started a thread on the subject, but no definitive answer was produced.

                Edit: Link to old thread: http://music-electronics-forum.com/t37023/
                Last edited by loudthud; 01-29-2015, 08:14 PM.
                WARNING! Musical Instrument amplifiers contain lethal voltages and can retain them even when unplugged. Refer service to qualified personnel.
                REMEMBER: Everybody knows that smokin' ain't allowed in school !

                Comment


                • #9
                  One of the main issues of DSP is that when it's computationally taxing it also tends to introduce quite a bit of latency. The need to upsample everything that creates additional harmonics (read: distortion) definitely won't help in that.... Highly realistic models of tube amplifier performance can be produced, no doubt, but it makes little sense if the result has too much latency to be any useful in so-called "real time" or if the hardware costs so much no one can practically afford it. So a lot of DSP and amp modeling is not about inventing how to make stuff perform "realistically" (they probably know it already), but about inventing how to do that as efficiently as possible. So that the result is actually usable. If you look at stuff patented about "amp modeling" it's not often so much about algorithms for waveshaping and such but about means of efficient computation, like when to access lookup tables instead of trying to calculate everything on the fly.

                  Also, it is interesting that many of these so-called "digital modeling amps" also utilise plenty of downright analog modeling. Analog "Valve Reactor" of Vox amps is probably no surprise to anyone, but how many were aware that Peavey's Vypyr series amps actually generate the distortion tones with an analog circuit similar to that of analog TransTube amps? Yes, one channel of the CODEC is solely dedicated to loop through an analog distortion circuit, not to mention the whole "T-Dynamics" power amp is a very clever 100% analog circuit. Roland? You'd be surprised how much analog modeling / soft clipping goes on in some of the "COSM" Cube series amps. Many of them, with certain settings, switch on an analog complementary class-AB circuit - just before the power amp - which soft clips nicely, and when overdriven enough also starts to bias shift and introduce crossover distortion like many class-AB tube amps. Line 6? Familiar with certain Duoverb, Flextone or HD147 models? Their analog distortion circuit mimicking power amp clipping vs. B+ sag and bias shift induced crossover distortion is surprisingly similar to circuit Quilter Labs managed to patent several years later, and of which's realistic "tubeyness" everyone is now currently hyping about while reaction to Line 6 stuff is..... well... I guess it's about hearing with eyes.

                  So beware: It's not always even 100% digital you're evaluating.

                  The appearance of analog solutions in amps widely regarded as "DSP-based" could be a sign of something, then again, it could be just a coincidence. I don't think same people develop the DSP that develop the analog stuff and if one method works why reinvent the wheel?
                  Last edited by teemuk; 01-29-2015, 07:38 PM.

                  Comment


                  • #10
                    I think that Aiken may be referring to the effect described in this topic as 'swirl', see Technical Q&A

                    Q:What is "swirl"?
                    A: "Swirl"is a dynamically-changing, slightly "phasey" sound as a note or chord decays, which is common to some tube amps. Typically, "swirl" is caused by a midrange "dip" or varying duty-cycle change in a clipped square wave that changes position as the note decays, giving a sort of mild phase shifter effect.

                    What happens is that first the phase inverter or output stage clips and produces a flat square wave. As the note decays, the signal level decreases, and the midrange frequencies start getting "unclipped" (either by the fact that their frequency band level is lower, or by phase cancellations due to the unequal phase shift with respect to frequency caused by tone controls and other RC phase shifts that occur in a gain stage) and show up as a "dip" in the top of the square wave, which will move back and forth along the top as the fundamental and other harmonics shift the operating point. Even if the clipping ratio isn't extreme enough to show the "dip" on the scope, the duty-cycle of the square wave will usually be dynamically changing as well.

                    Since the preamp stages are all AC-coupled to each other, the operating point shifts as the signal gets smaller, due to slight "blocking" distortion, where the gain stage clamps the top peak to a point slightly above it's cathode voltage, while allowing the wave to still increase in the negative direction. As the signal decays, it shifts upward and changes the duty-cycle of the clipping. It is this ever-changing shifting of the operating point that causes the "swirl" effect. The trick to good "swirl" is in the correct staging of the gain and frequency breakpoints of each gain stage in the amp, particularly in the phase inverter and output stage.

                    A similar effect can be caused by too much drive from the phase inverter to the output tubes. As the note decays, a riding "buzz" can be heard coming in and out. This is crossover distortion aggravated by too much signal swing to the output tube grids. Reducing the signal levels at the output of the phase inverter will cure this.

                    Another cause of a "swirly" sound is a parasitic oscillation that is riding on the output signal, causing intermodulation distortion.
                    My band:- http://www.youtube.com/user/RedwingBand

                    Comment


                    • #11
                      Originally posted by Malcolm Irving View Post
                      I don't think I was claiming anything new. In fact I stated that I thought OP would already know it!

                      Thanks very much for the interesting reference paper, though. I'll read it thoroughly.
                      I think he was responding to me, and I caught him before his coffee.

                      Comment


                      • #12
                        Originally posted by uneumann View Post
                        IMO - this topic * time-varying parameter shifts * is hugely important and almost completely neglected in texts and technical discussions.
                        Fortunately, it is discussed at some length in my book
                        Click image for larger version

Name:	BlockingDist2_zpsa25ae7bb.png
Views:	1
Size:	2.6 KB
ID:	836715

                        Comment


                        • #13
                          If "all competent amp designers have been aware of this phenomenon since what... 1970's? ...at least early 1980's...?", I can only conclude that most amp designers are not competent. This forum is unusual.

                          Comment


                          • #14
                            What makes you conclude that amp designers aren't competent? From what I see, in many, many amps there has been a lot of thought put on "gain staging", particularly on the aspects of overall signal levels and gains, grid conduction characteristics, stage bias affecting overall symmetry/asymmetry, coupling capacitances, etc. A lot of thought. This couldn't have happened without competence.

                            If people at Internet forums know this why wouldn't people who design amps for living, and have done so for practically most of their lifetime, wouldn't? I'm just a plain construction worker myself and I'm aware of this phenomenon. You can't really put me in the same category as the guys who earn their living working for Peavey, FMIC, Marshall, Korg, etc.

                            Comment


                            • #15
                              Originally posted by Tooboob View Post
                              If "all competent amp designers have been aware of this phenomenon since what... 1970's? ...at least early 1980's...?", I can only conclude that most amp designers are not competent. This forum is unusual.
                              Don't read what wasn't written

                              If you want to expand it,
                              "all competent amp designers have been aware of this phenomenon since what... 1970's? ...at least early 1980's...?"
                              clearly refers to analog amp designers, many Tube but also a few SS , such as, just to name a couple: Mike Soldano , Jim Kelley, Randall Smith, whoever designed the Randalls, the brilliant guy (Murphy?) who designed the variable duty cycle distortion for Carvin and which is the basis of excellent sounding Crates, Ken Fischer , Pearce, .... such brilliant guys, and not even delving in the earlier Classic names, all included because the clause was by/since 1970

                              And that is being compared to modern algorithm and software designers.

                              To be more precise, it's being conceded that they probably know how to do things, just not fast enough.

                              Don't understand how you can read despise in that realistic comment.

                              By the way, I suggest you download and read Teemuk's book, a Bible of analog design applied to Musical Instrument amplification.
                              Juan Manuel Fahey

                              Comment

                              Working...
                              X