Finally attacking TSL100 channel switching latency

[jamesmafyew]

I've seen posts on several different boards with people complaining about the "volume ducking" or "switching latency" or "reverb fade in/out" on the DSL and TSL series of amps. No one seems to have found a definitive solution for this though, that I know of. If they have, please disregard the rest of this post and tell me where to go read!

I have heard people say that it's a problem with the logic circuits in the footswitch. Well... I've looked at the schematics of the footswitch and can't seem to find a reason why that would be. Furthermore, when I use the buttons on the front of the amp, it still has the latency. I understand that Marshall designed this in as a "feature", and from what I can tell it's not built into the footswitch.

One person modded his DSL 401 to get rid of the ducking by clipping two transistors. He had a website up here which doesn't seem to be working anymore. I do remember though that he clipped T15 and T16, which are clearly labeled as "volume ducking" in the schematic (see attached images).

Essentially, I'm trying to figure out which transistors are analogous to these two in my TSL 100. I have figured out that TR3 on the TSL's main board is analogous to T15 on the DSL schematic, but I can't figure out what would correspond to T16. Could it be TR1? Is there something else I'm missing? (see attached image)

Would there be harm in taking TR1 and TR3 out to see what happens?

Full TSL 100 schematic is here, DSL 401 is here.

I'd appreciate any help in getting this figured out.

[Gregg]

Most of the amps especially higher gain ones these days have a dedicated mute circuit that will ground the signal for a short time while the realys are switching to avoid the usual clicks and pops they are producing.

Full TSL 100 schematic is here, DSL 401 is here.

This is not the ful TSL100 schematic. Get all the files from DrTube and on one of the PDFs - tl10-65-02.pdf you'll see two J174s - FET1 and FET2. They are part of the mute circuit.
Usually ou should look for P-channel FETs as J174 or J175 in parallel with the signal somewhere around the master volumes, before the FX loop, or after that.
Also the NSL32 is not very fast switching LDR and there's plenty of them in the schematic.
Be prepared however to "get" some pops after you "cut" the transistors in question.

[jamesmafyew]

I'm not looking to disable the mute circuit entirely, more just get rid of the annoying fade/swell when changing from the clean to either of the OD channels and vice versa... and so trying to replicate the aforementioned DSL 401 mod on my TSL. From what I can tell with my ears and looking at these two schematics, doing that would only require changes around the reverb return, or near CON7 on the TSL mainboard schematic.

It looks to me now like TR3 may be doing the job of both T15 and T16, since the TSL's FX loop and reverb structure is different from the DSL 401. I think one of the transistors in the DSL is for the reverb, the other for the effects loop. The TSL has them arranged differently, and appears to send the reverb signal out the FX send (I know this from using a slave amp hooked up to one of the FX sends, not from the schematic). So one transistor would mute/fade them both at their return point...?

However, given that my issue only happens going to or from the clean channel, these two transistors you're pointing out look like they're worth some attention. I may just have to try several different options and see what works.

Thanks for the pointer!

[jamesmafyew]

To follow up on this:

I took out all three transistors in question --- TR3 (main board) and FET1 and FET2 (clean board). As far as I can tell, the change has had almost no effect at all. I took them out one at a time and listened to the relative change... but there's not much difference. No new pops or noise were added. If anything, the switching is more stable/predictable, but still essentially the same problem.

I've always had a small pop when changing from clean to OD1 or OD2, but that change is always instantaneous. The only change that is NOT instantaneous is the change from either OD1 or OD2 BACK to the clean channel. It has an annoying gap in the sound and then the volume fades back up over about 200ms to the full clean volume. The problem gets less severe the longer the amp has been on.

All three of the removed transistors had 22uF caps on their gates, which led me to believe that they were collectively somehow involved in the "fading in" effect -- but that's apparently not the case. Could this "fade-in" effect just be a limitation of the NSL32 opto-isolators? If so, why would all switches except OD1/OD2 -> Clean be instantaneous?

Then again, there are two more P-channel FETs that I haven't touched -- TR1 and TR2 on the main board (see 3rd pic in original post). Could either/both of these be responsible?

[Mullard Man]

Giving this thread a friendly bump.

Did you ever find a solution to the delay in channel switching on the TSL?

The latency on my TSL-100 drives me nuts in a live performance setting. I realize that Marshall did this 'ducking' on purpose, perhaps to help in effects spillover and/or to reduce any pops. I still want faster channel switching.

[jamesmafyew]

After sitting with it a while, I realized that clipping FET1 and FET2 on the clean channel PCB and TR3 on the main PCB does help. It doesn't get rid of the latency entirely, but is a definite improvement. Removing TR1 and TR2 did nothing to the latency and changed the tone for the worse, so I put them back in.

At this point it's livable, but for what it's worth I know that the opto-isolators used in the TSLs are sloowwww. There may not be much you can do by altering the muting circuit. Who knows?

[Mullard Man]

Thanks for the update.

I'll try your recommendations when I get a chance.

[jazzyseb]

After sitting with it a while, I realized that clipping FET1 and FET2 on the clean channel PCB and TR3 on the main PCB does help. It doesn't get rid of the latency entirely, but is a definite improvement. Removing TR1 and TR2 did nothing to the latency and changed the tone for the worse, so I put them back in.

At this point it's livable, but for what it's worth I know that the opto-isolators used in the TSLs are sloowwww. There may not be much you can do by altering the muting circuit. Who knows?

Hello guys out there,

In the last hours I've been reading lots of web pages about TSL100 mods, and tube amp mods in general, as I am seriously thinking about giving a try to lower noise level on my TSL head, especially that white noise coming out on the Lead Channel - but that's another story.

I remember having read this thread a few months ago, I also took a look at the schematics and tried to understand this switching latency issue, I also came to the conclusion that these NSL-32 optocoupler were mostly the cause of the problem with their inherent decay time (500ms max in the datasheet) when removing current in the LED branch.
Well, I might have some good news about that...
I found a thread on another forum where people were talking about something completely different, but the NSL-32 was concerned too.
There is actually another reference from the same manufacturer, with exactly the same package, which is NSL-32SR3.

http://www.alliedelec.com/Images/Pro...es_6993012.pdf

The interesting part is that its decay time falls to 10ms, with a 5mA current condition instead of 16mA in NSL-32 datasheet - not exactly the same condition but it surely will decay faster at the same current rating.
On the other hand, Vf voltage is slighlty higher (2,5V instead of 2V) and max forward current is lower (25mA instead of 40mA) - but it should work eventually.

I find the schematics a bit hard to read, but if I understand well, let's take an example with OPT3 OPT5 and OPT7 in series with R10 (2K2) and a LED1 (red LED) to ground - when RL1b is closed. (TL10-61-02 sheet)
We have +24V on R10 if my connector chart is correct :-)
There is 2V Vf on each Opto and let's say 2V Vf for LED1 -> this leads to (3*2 + 2) = 8V drop on these guys.
So R10 sees 16V, that means 7mA through the branch.
It's about the same for OPT2, OPT4, OPT6, just with a yellow LED instead.

Another example with LDR3 on TL10-60-02 sheet.
Series resistor R45 is 4k7, there is 2V drop on the optocoupler's LED, and let's ignore Vce on TR5 - we are in the worst case scenario.
We have then about 22V on R45 -> about 5mA.

With a similar method, OPTO1 and OPTO2 see about 8mA

So after this quick check showing we are far below 20mA, I think it could be worth a try replacing NSL-32 with NSL-32SR3, and see if the switching latency is actually reduced.
Just hope this will not reveal any other unwanted stuff, like pop or noise of any sort...

When I find some time to try it, I'll post my feedback here.
In the meantime, if you have some remarks or impressions, or if you see any mistakes, or if you already found a solution, I'll be happy to hear from you..

Bye
Seb

[jamesmafyew]

It's funny you should bump this, I was thinking of resurrecting it last week as I was working on the amp. I was actually doing exactly this -- trying to replace NSL-32s with the SR2 (not SR3) type to see what happened.

Here's the thing: the standard NSL-32 goes to full conductance (low resistance) in 3.5ms - fast. All three types short in 10ms or less. However, the standard type takes 500ms to get to 100k resistance when current is removed (80ms for SR2, 10ms for SR3). So the decay is not the same turning on and turning off. My logic was that since the delay is when switching from EITHER of the drive channels back to the clean channel, the problem must be in one of the optos that is turning OFF (going high resistance) when the clean channel is turning on.

This means that it would NOT be either of the two optos on the clean channel PCB (they short when clean comes on, connecting volume and tone stack). The one opto on the mainboard (LDR3) is also NOT involved, since it shorts while clean is on -- dumping the output of V1B to ground, since it is not used by the clean channel. The six optos on the lead channel PCB are possible culprits. OPT2/4/6 affect the crunch channel, being shorted when it's on and open when it's off. OPT3/5/7 do likewise for the lead channel.

My thought was that the slow rise time was causing either OPT6 or 7 to load the output of the clean channel for 500ms, which is about the delay time I hear in the amp. There's a 470k series resistor after the clean channel's volume pot and opto, which seemed like it could form a voltage divider with the momentarily heavy load from the optos in the lead/crunch channels slowly rising in resistance.

The output from the clean channel's volume pot goes through its own opto, OPT01 on the clean PCB, then a 470k resistor R12, and comes onto the lead board at pin 3 of CON7 (labeled "MUTE") and connects to the top side of R27 ("MUTE" again... it's confusing). Weird part: I replaced OPT6 and OPT7 with NSL-32 SR2s, and the problem got worse. As in, the fade-in of the clean was deeper and took longer. That left me scratching my head. So I put the old NSL-32s back in OPT6 and 7, and tried the SR2s in OPT4 and 5. No effect at all this time, delay was exactly as stock.

Having the SR2s in OPT6 and 7 was the only time I've ever managed to get a significant change in the switching delay, but it made it worse. You're right to point out that the TSL is not giving the optos their rated current - it's actually less than half what the stock NSL-32 wants, and it's worse with the SR2 and SR3 because they require more. I thought about changing out R9 and R10 on the lead PCB to address this (1k2 would get SR3s to about what they're rated for), but have not tried this. I also now have some SR3s but have not tried them in the amp. It's possible that you'd need to change ALL the optos so that they're all the same type to get good results, as well as the current-limiting resistors.

I'd appreciate any thoughts or experiences. It's not a pleasing prospect to think about changing all the optos in the amp for SR3s, only to find out that it either did nothing or made the problem worse.

NSL-32 datasheet: NSL-32 pdf, NSL-32 description, NSL-32 datasheets, NSL-32 view ::: ALLDATASHEET :::
SR2 datasheet: http://www.alliedelec.com/Images/Pro...es_6993010.pdf
SR3 datasheet: http://www.alliedelec.com/Images/Pro...es_6993012.pdf

TSL lead PCB: http://www.drtube.com/schematics/mar...tl10-61-02.pdf
TSL mainboard: http://www.drtube.com/schematics/mar...tl10-61-02.pdf
TSL clean PCB: http://www.drtube.com/schematics/mar...tl10-65-02.pdf

[Gregg]

Maybe you should try replacing those NSLs wit VTLs. AFAIK they are faster.

[jazzyseb]

Hi jamesmafyew,

It's funny to see how the timing thing goes sometimes... Seems the time is up for that switching problem here... :-)

Your approch with the logic I think was very good, the only thing is that I would have recommand to set new Optos on a full channel to be more coherent with the circuit, for example exchanging OPT3/5/7, and try to switch from Crunch to Clean, and the other way round, see how it goes.. but I'm not sure this would have really change results... Plus, the way it is actually working with these NSL-32SR2 sounds really strange to me too. It might be caused by the way resistance goes up in the decay time, it goes up faster, but starts lower, and goes to higher values than the stock ones, hard to know how they are evolving comparingly, and the dissimetry between the 2 references inside a single channel may be a factor too...
And definately at the "MUTE" point, there must be a strange fight between the 3 signals from each channel as one LDR is coming down very quickly, another is coming up slowly, and the third is staying quiet. I'm not sure to understand how that's supposed to work well and smooth when switching channels...

Anyway, this made me think alot and study the schematics again especially what's happening at the moment you switch back from crunch (or lead , cause it is the same really) to clean.
Maybe we can assume the optos work right (as they eventually do when switching between Crunch and Lead, and from Clean to Crunch or Lead) - they might not have anything to do with the fade-in thing.

There are several things that embarassed me. First, RL1c contact. It switches signals from OPT2 and OPT3 (crunch/lead) right back to ground instantly - with the speed of the mechanical connection of the relay, when the realy coil loose its command current, that means when you realease OD/Clean switch (SW1A p. 65-02).
The switch actually causes TR1 base to be tied to ground -instantly - (via R2).
You also have RL1b (another contact from the same relay) coming back to its normal position, instantly tying to ground the clean channel opto led branch that was floating before.
Nothing here could explain this big latency, even the crunch/channel optos I think, as optos from clean channel should go ON very fast and the signal would superpose to Crunch/Clean fading out slowly - however with the little nuance that CLean has a 470k series resistor between his volume pot and MUTE signal, whereas the other don't.
This is instictively how I understand it, may be wrong though...

What I finally wanted to coming up with, is that the FET1 / FET2 temporary mute has better chance in my opinion, to be the culprit. They are the ones that are supposed to hold the signal to ground while switching occurs.

In fact, there is a very interesting clue for that.
TR1 transistor base is tied to ground to go back to clean channel.
Relay coils on its collector are not powered anymore to 24V because the path to ground is open, the flyback diode comes in, and the collector is connected to C40 a 22uF cap - not willing to let flow any DC.
It is the only transistor that has a cap on its drain like this.
I don't know exactly then what would be the voltage on the collector pin when the transistor gets blocked, how the coils and flyback diode exactly react, but it would be much believable that Vce comes up a few volts.
Now, what is connected on C40 :
a 100k resistor in series with a diode (about 1V drop) and then directly to the 2 JFET gates. (reverse biased by the way).
It is much believable too, that there is still a few volts at that moment on the JFET gates, keeping them ON and tying clean channel sound to ground. That's pretty much how they must have been designed for.
Now the great part of it is that C40.R11 is a very remarquable value :
22u.100k = 2.2s
Considering the fact that D3 series diode drops about 1V, the JFET have a Vgs around 0.7V, a possible scenario would be :
"There is about 4,5V on C40 when TR1 goes OFF. This voltage is immediately discharging through R11, loosing 1V across D3. So at the switch time, JFET gate see 3,5V. After a certain amount of time, the gate voltage reaches 0,7V, the JFET is OFF, and the clean channel sound appears, fadin up as the gate voltage fades out...
How many time ? About -RC*ln(1-0.7/3.5)=0.5s."
Sounds familiar ?...

Could be one explanation, with different voltages, different amount of hundreds of milliseconds of latency...
This could be easy to verify : check the voltage on JFET gates and see what happens.
If it is a kind of low continuous voltage drop, like a capacitor discharge, in a time quite similar to what we hear, could be our man here.

I don't have my TSL with me at the house these days so I won't be able to check that before next week probably.... If someone else takes a chance with it, please let me know :-)

(Hope this was not too heavy to read from top to bottom, but as I can't make the real tests in real life in real time, I have to tell the entire story, just for peace of mind..)

Hope to read from you soon :-)

[jamesmafyew]

Jazzyseb,

You are right about FET1 and FET2 on the clean board being part of a muting circuit. Same is true of TR3 on the main board, which mutes the reverb output on a channel change (TR1 and TR2 on that board are gain/voicing changes). Thing is, I've removed them all... so my TSL effectively has no muting circuit. FET1 and FET2 have no effect because they aren't in my amp (see post #4 above).

That's where my train of thought re: optos began.

RL1c grounds the grid of V1B, which is only used for the crunch/lead channels. LDR3 on the mainboard performs the same function on the output of that triode. Clean signal goes from V1A to OPT02 on the clean board, through the tone stack and gain pot, then into V2A via CON6 on the mainboard.

Do you think LDR3 on the mainboard could be loading the clean signal from CON6 before V2A? Again, forming a voltage divider where R80 is the series component, and when LDR3 is low resistance, there would be 471k to ground through R25 and LDR3, meaning roughly a 6dB suckout. That value would rise to well over 1M as the LDR goes high-resistance, making the signal increase by 4-5dB?

EDIT: did the calcs, LDR3 would cause a 1.8dB rise over 500ms (going from 140R to 500k) at CON6 when switching to clean channel. Could be part of the story, but probably not all of it.

EDIT 2: Changed out LDR3 for an SR3. Little to no change.

EDIT 3: Changed out all six optos on lead PCB for SR3s. Same problem as before --- the fade-in is actually worse now.

[Enzo]

I did not read all this as closely as I might, so forgive me if I missed something.

No idea how much difference it might make, but there was talk of reducing current through these things to improve the recovery time. The dimmer the LED glow, the less the cell has to recover from. My concern is that possibly we reduce current enough that the ON condition starts to suffer - not enough LED light to fully On.


But whereas it is fun to think through a problem and try to fathom it as a mind experiment, why not just apply a steady tone to the amp, and scope at each point the signal behavior. That should show where the signal dynamics are slowed.

[jazzyseb]

Jamesmafyew,

Okay... I should have read your previous post again before thinking too much about the FETs, I remembered reading about that but did not actually realize it was the same thread and you tested it already...
Anyway...
On the schematics, TR1 looked interesting also, because its gate was powered by a voltage coming from C45, R82 and R46, going from ground to 24V switching back to Clean. With a RC~220ms.
But you tried to remove it did not work...

I find it quite strange that you got worse a behaviour when changing to SR3 optos... It find it quite hard to understand how everything is going on with these circuits, as the OPTO resistance value is changing with the time, and when you take a quicker OPTO it also goes much higher in OFF resistance...
But the facts are there.
Changing them does not improve things.

I really looked hard at these schematics, thinking about all that stuff.
The only other idea I'm seeing right now is to test the MUTE signal.
The difference between Clean and Overdrive here is that Clean Volume goes through LDR, then 470k, then to MUTE, whereas Overdrive channels' volumes go straight through LDR then MUTE.
The thing is, this MUTE signal is charged with R27 (2M2), C20 (220n) and R24 (600k) to ground. FX send signal is taken on the top of R24, let's forget about that.
These values are quite high.

If we consider that the OPTOs are not really the cause of that latency (conclusion from real life experiments), and that their resistance change fast enough to be considered as switches (less than 50ms for instance - even for stock ones, 500ms is a max value in the datasheet, they might switch faster actually), there is something interesting here.

When you switch back to Clean channel, there is RL1c (relay coil) that switches Overdrive signals to ground, these signals are on V1B grid.
If you follow the signal path, the output of V1B goes to the point where Clean and Overdrive signals are merged through R79/R80. Then it goes through V2, V3, and back to LEAD PCB in CON11.
At this point they are split, each one goes through its own path, Volume Pot for Clean, LDR + EQ + Volume Pot for each Overdrive. After the volumes pots, all three channels melt again through their own LDR, the only difference being Clean signal the only one to have 470k in series.
I don't know exactly how fast the signals go through each path, but it could be possible that the signal pulled to ground instantly by RL1c on V1b input would go right down the path until the MUTE point, which would then really be "muted" because 0V would be "propagated" very fast.
At this time, Clean signal begins to show up and sees 470k in series (R12) then (2M2-220nF-600k) to ground. A lowpass filter with RC~800ms.
When you switch to Overdrive, the Clean signal might disappear quickly because its LDR go OFF quickly, and let's say the Overdrive LDR go ON quickly. Even if there is a point where there is a very low or no signal on MUTE, when Overdrive signal shows up it does not have that 470k series resistor, it goes straight to V4A input through C28 without any lowpass effect.

This might be interesting to test, as it there is a very simple thing to do to verify the theory, because MUTE signal is pulled to ground when the guitar jack's unplugged...
So, with the guitar unplugged, make the open string vibrate, or any other noise, with the amp already in Clean channel.
Plug the guitar. You should hear the fade-in.
Then, unplug the guitar, switch to Overdrive (any of them).
Make sure the strings still vibrate (or make any noise you want with it), then plug the guitar. The sound should be there straight up, without fade-in.
This would demonstrate the OPTOs are not involved, because they would have enough time to reach their final state between the moment you switch and the moment you plug in.
If the fade-in still appears on Clean channel, then the problem would be on MUTE signal circuit, or somewhere after MUTE in the circuit - that means no OPTOs in the way.
Actually after MUTE there is V4 and then the power stage, there is no more difference between Clean and Overdrive I think, so it would mean that MUTE circuit is really the culprit.

I'm going the rehearse wit my band tonight, I'll give it a try...
If someone else makes the test by the time, please let us know... :-)

[jazzyseb]

But whereas it is fun to think through a problem and try to fathom it as a mind experiment, why not just apply a steady tone to the amp, and scope at each point the signal behavior. That should show where the signal dynamics are slowed.

Enzo, you're right. If I'd have the amp with me I'd have directly scope every part of the circuit instead of just reading the schematics...
Within the same amount of time I'd have already tested different possibilities and would be closer from the truth... :-)