Announcement

Collapse
No announcement yet.

Finally attacking TSL100 channel switching latency

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

  • #16
    I don't have a scope (can't afford one yet), so the logic process is all I have to work with. I believe I've learned basically what I could learn with a scope though -- just with a lot more effort and soldering -- that the area around "MUTE"/OPT6/OPT7 is where this is happening. Specifically, the spot where the three channels join together after each master volume, before the FX send and PI. I still believe the voltage divider formed by the 470k series resistor (R12 + OPT01/clean PCB) and the load at the 3-way join is somehow responsible.

    After the seemingly backward result of putting faster optos in the lead PCB and having the fade-in worsen, I started thinking of the process as a kind of crossfade. Drive channel fades out, clean fades in. By putting the faster optos in only the lead PCB, I was cutting half of that crossfade and leaving the other mostly intact. So I decided to reverse my approach.

    So I put the old NSL-32s back in, ONLY at OPT6 and OPT7. The other SR3s in LDR3, OPT2/3/4/5 all stayed. Sound went back to stock. So it really is the stuff at the join. Next, I put an SR3 in the OPT01 spot on the clean PCB, and also halved the value of R1 on lead PCB (current limiting resistor for the clean channel optos), effectively doubling the current through OPT01 and OPT02. This FINALLY caused a desirable change -- though with some qualifiers.

    In the current state, with the lead or crunch master vols all the way down, the clean fade in is as stock. As I turn up the volume on either of the drive channels, the transition from that channel back to clean becomes more seamless, until at about 3, there is no audible fade-in. This is with the clean volume at about 2.5. This confirms to me that loading is part of the issue. However, interestingly enough: now for the first time, with the clean volume all the way down, switching to either of the drive channels produces a fade-in! Not nearly as pronounced and not something I'm going to worry too much about, but it's there.

    I'm thinking at this point that just replacing OPT01 on the clean board with an SR3, and increasing the current through it, will mitigate the problem to most people's satisfaction (since in a live performance situation, the volumes aren't going to be that close to zero, where the effect is most pronounced).

    It's possible that ONLY increasing the current through the stock clean optos would produce the same effect... something to try. This makes me wonder about the relationship between current and rise/decay time. I understand that final values of resistance/conductance will be affected by current according to the curve in the datasheet... but I'm wondering about the logic that less current makes the transition faster. Maybe it makes the decay faster, but not the rise, or vice versa. Maybe it makes both slower. The assumption is that since the resistance has farther to rise with more current, it will take more time. I'm not sure about that. At least OTOH with conductance, it seems that applying higher current causes the resistance to drop farther, but in the same amount of time (based on my experience here, not on any hard data), thus mitigating OPT01's addition to the 470k series component of the voltage divider issue.

    EDIT: I tried increasing current through the crunch channel but not the lead channel (doubled it). It actually slowed things down. I'd guess that increased current shortens rise time, but lengthens decay time. So it might be possible to address the issue without replacing optos, by increasing current to the clean channel optos to about 16mA, but decreasing it to the lead/crunch channel optos (maybe to around 5mA? less?). Though, my approach has not been very disciplined or scientific -- so current may have little to do with it and it may just be a case of changing OPT01 for an SR3.

    EDIT: Forgot to point out from before that there's something else that may or may not matter to some people (matters to me): independent of the fade-in issue, there is a split-second delay on the actual mechanical RELAY engagement when switching from either drive channel back to clean. That is, RL1 and RL4 are instantaneous when switching TO the drive channels, but delayed when switching away. This has no bearing on the fade in, it just means that the whole channel-switching reaction doesn't even start until some time after you press the button. You can demonstrate this by powering the amp up in standby and pushing the channel switching buttons on the front. The relay click is instant EXCEPT when going from Lead/Crunch back to clean.

    The reason is C40 on the lead PCB. It controls the muting time for FET1 and FET2 on the clean PCB. Since I had removed FET1 and FET2, I didn't need C40 anymore, so I took it out (some boards have a C39 instead -- same function). The cap has 0V across it while RL1 and RL4 are engaged, but when TR1 pulls up from ground, the cap starts to charge. Thus current continues flowing through RL1's and RL4's coils for a split second after the channel change. Removing C40 solved this issue. Again, does NOT affect the fade-in, but in a performance situation, when changing channels on a specific downbeat, it does matter. I can't say whether removing it would add noise if FET1 and FET2 were still in... but my amp has no muting circuit, so taking this out was helpful for me. Just FYI.
    Last edited by jamesmafyew; 12-03-2010, 07:17 PM.

    Comment


    • #17
      Hi James,

      Very nice, you finally managed to find a working combination with these OPTOs, good job !

      Just to let you know I've made the test I mentioned before while I was in rehearsal tonigth, to plug the guitar in with the amp in Clean channel, to see if the fade-in would appear.
      There's not any. Clean sound is heard straight on. Same with the Overdrive channels.
      This confirms that the fade-in issue has nothing to do with R27 and C220 on the MUTE signal.

      You've definitely found the crucial point being the meeting of OTP6, OPT7 and R12.
      I'll have my TSL at home next week, I should be able to run some tests, like changing only current through the optos, see what's going on.
      I also think that a higher current would probably lead to a longer decay time. But it would take a deeper look into these components characteristics to clarify that...

      I did not really look for other optos, which would have a similar resistance but with a shorter decay time. It should be worth giving a try to find that kind of components, as it would only reduce time and not resistance, changing only one variable at a time.
      But I don't know if such an opto exist. We'll see...

      Till next time :-)
      Seb

      Comment


      • #18
        Hey,

        I know this thread is quite old, but did someone investigate things further with the TSL switching latency?

        Thanks!

        Comment


        • #19
          You'll get the most bang for your buck by just clipping C39 on the lead panel PCB. That cap is responsible for the slow switch from either drive channel back to the clean channel. In some cases it might be called C40 (this is what it is on the schematic for that board). Its discharge time, through the control coils, delays the relays RL1 and RL4 from clicking off.

          Changing optos isn't really worth it, because the results are so unpredictable. If clipping C39 isn't enough, you can clip TR3 (reverb muting) on the mainboard, and FET1 / FET2 on the clean panel PCB (clean channel muting), but it might introduce popping -- though in my experience, popping is only audible immediately after the amp is switched on.

          Comment


          • #20
            Exactly, C40 would 've been my first mod.

            I do have another issue with this amp (TSL100): on the clean channel, with gain and volume on '0', there still is coming a very small amount of signal through, along with some 'hiss'. I made a new shielded wire to properly seperate the gain and volume 'hot' leads (con 3 on the clean PCB to con 7 on the crunch PCB) which did help some, but still some hiss is left. Parasitic oscillations?? Anybody got an idea what's best to solve this?

            Best Regards,
            Koen

            Comment


            • #21
              It's just crosstalk, mostly from the reverb PCB being so close by. If you have gain=0/vol=0/reverb=10, you will hear an un-reverbed quacky sound. Turn the gain up and you will start to hear reverb. At gain=5 it will get quieter, because the signal on the gain pot is inverted from the input signal, and is cancelling... then with gain past 5 it will get louder again as more signal bleeds in. If you have the gain all the way down, turning the volume up or down will not change the volume of the quacky sound.

              I think if you're going to shield wires, you would do better to shield con1 from the clean PCB and con1 and con3 from the reverb PCB.

              In any case, there are certainly multiple sources of the crosstalk, but none of them are really that significant when the amp is being used at band volume. The only circumstances I can think of where it would be an issue is if you're using the clean channel as some kind of overall amp mute, or if you're playing at bedroom volume.

              The hiss, well -- the amp is hissy, it's got a lot of gain. Mine doesn't hiss noticeably on the clean channel with the volume down. It's possible you have a noisy preamp tube, or a plate resistor acting up.

              Comment


              • #22
                Ok, thanks alot! I'll report back when I've shielded those other wires!

                Comment


                • #23
                  Most of those wires are allready shielded, I couldn't gain much there.
                  But you're right, at normal volumes it doesn't mather, it's just a fine amp!!

                  Best Regards,
                  Koen

                  Comment


                  • #24
                    One other thing to think about is the fact that -- since with gain=0, the volume pot has no effect on the quack sound -- a good bit of the crosstalk is being injected after the volume pot. So something to think about there is the wires that carry the signal from the clean volume pot to the phase inverter. So look at con3 on the clean PCB -- where does it go from there? Trace that signal all the way to the phase inverter, and see if there's anything left to shield.
                    Last edited by jamesmafyew; 08-20-2011, 04:39 PM. Reason: typo, misread post

                    Comment


                    • #25
                      9 years later this is still an ongoing theme.
                      So, if I check schematic TL10-61-02.DGM (Lead Channel PCB), I see the transistors turning the relays on and off with C40 across the collector and emitter of transistor TR1. Is this the capacitor that needs to be removed? On the board it's C39, which is not even soldered into that position, but rather bridge. It's connected between the collector and emitter of TR1, though.
                      Removed it, seems to have fixed it.
                      EDIT: this didn't really fix the problem.
                      I'm wondering if someone ever tackled this situation and what the ultimate solution is.
                      I have a bunch of new NSL-32SR2 and NSL-32SR3 opto-couples over here and the question is if they can fix this issue, as I have a feeling it has to do with the switching off time-out.
                      I particularly noticed that OPT2 and/or OPT3 are slow when switching off.
                      Is it possible that C4 and C5 on the Lead board are too large and therefore cause the leds in the optos to "fade out" slowly?
                      Last edited by beeltink; 11-27-2020, 09:11 AM.

                      Comment


                      • #26
                        I don't have a TSL on my bench, but have customized many amps, and built my own channel switching circuits... Here is my take based on a schematics review. Search for any p-channel JFET that is a channel mute where there is an RC circuit on the gate. The R*C value controls the switching time. Example, oscope the rise or fall time on TR1 and reduce the C value for C45, and measure the rise/fall time, get it down to a few 10's of milliseconds, check there is no oscillation or ringing on the gate... repeat for TR3 & C26...

                        Comment

                        Working...
                        X