Tech: Amp guts - complete Marshall TSL mods

How to make the TSL-100 good at almost everything
(but especially sludge/doom)

This post is Google bait, so that someone else won't have to do all the searching I did in researching these mods. All of them require some experience soldering components on PCBs. They are presented in the order I did them.

Schematics here, with links throughout. Before doing these mods, please be sure you know about the voltage hazard inside the amp from the supply caps. The TSL (at least my 2005-made model) has bleeder resistors, so you can leave the amp unplugged for an hour and the caps will discharge themselves --- but ALWAYS CHECK WITH A VOLTMETER before diving in. A quick web search should yield whatever info you need here.


1. Get rid of the fizz
Cost: ~$1
Perceived sonic change: large and immediately obvious

These amps have a reputation for being "fizzy" or "buzzy" in the lead channel. You can easily and cheaply get rid of this without making the amp sound dull. This will NOT affect the sound of the crunch or clean channels. This is from a thread on MEF.

Use two silver mica capacitors to create a 12dB/octave low pass filter at around 7.5kHz, in the preamp. This not only kills fizz in the preamp, it saves some headroom in the power amp and the power tubes break up more smoothly when they saturate. The treble you do keep is more pleasing.

Put a 390pf cap across R1 on the main circuit board (the one the tube sockets are on), and a 47pf cap across the lead channel volume pot -- VR2 on the lead channel panel PCB. Note that the pots used have 4 terminals; one of these is part of the pot case, for grouding/shielding. Counting from the side toward the gain pot, use terminals 1 and 3. There should be only one pin (pin 2) between the two pins connected to the cap, and one pin on the side toward the treble pot. You will have to remove the chassis, tubes, aluminum top plate, and panel PCBs to do this mod --- that includes all the knobs and pot washers. You will want to put pieces of tape on each of the wires you disconnect, labeling what they connect to.

If you want to do the same thing to the crunch channel, just add another 47pf cap to the outer terminals of the volume pot (VR2) on that channel's panel PCB. I tried this and didn't like it, so I removed it -- the crunch channel's voicing seems ok to me.

EDIT 3/30/09: I have since changed the VR2 cap to a 68pf, to remove even more fizz and increase smoothness. I left the 390pf cap alone, since I know now that it affects all channels ever so slightly. So now the filter on the OD channel starts at a lower frequency, and maintains a 6dB/octave slope up until 7.5 kHz, where it steepens to 12dB/octave.

EDIT 4/4/09: Upped the VR2 cap to 100pf, which brings the corner frequency down to about 3.5 kHz. I don't recommend this unless you are a) using the MMOT (see #7 below) and b) you love midrange crunch. I used to keep the treble at 1 even with the 47pf cap, now it's at 4 or 5. It sounds very British -- more like a JCM800 running close to meltdown.


2. Change the tubes
Cost: $100ish for complete re-tube
Perceived sonic change: medium

This is inevitably a taste issue. If you want more raunch and a warmer sound, JJ tubes do nicely. This is what I use --- ECC83s preamp tubes, and E34L power tubes. Eurotubes offers a full re-tube kit for the TSL with a variety of options. I used the high-gain preamp tubes (I love gain). I have tried JJ's KT77 power tubes and found them to be muddy/farty and fizzy compared to JJ's E34L tubes when playing downtuned or sludgier material. There are applications where they are appropriate for adding roundness to the sound, but the bass is too loose for full-on gutteral roar. JJ's straight EL34 is a good option for a more traditional Marshall sound, more like the stock Svetlanas but better. The E34L type seems to improve both low-end extension (without mud) and headroom -- great for more rumble without fartiness.

EDIT 1/12/09: For E34Ls to sound best, they need to be biased on the high side, at about 90mV per side, or higher if you don't mind decreased life. 90mV per side is Marshall's spec for the amp, but most agree that this is high, and anything in the 80-90mV range is ok. I found that below 90, the E34Ls sounded a little fizzy and cold at lower volumes. There is good info on how to bias the TSL/DSL here.

EDIT 4/4/09: The high-biasing advice above really only applies if you replace the OT (see #7 below!). If you don't, most settings between 80-90mV per side will sound similar.

EDIT 6/15/09: On the other hand, since the MMOT handles more power cleanly, you can bias lower and get actual usable headroom --- with the same nice warmth at a higher volume level. Biasing low, to 70-80mV per side or so, gets you noticeably more volume before things start to break up --- by "noticeably" I mean that it feels like the speaker cabinet is going to fall apart. For Rosetta's European tour, I converted the voltage and ran the amp at ~80mV per side, and turned it a notch louder to compensate for a less efficient speaker cab.


3. Clean up the rectifier
Cost: ~$40, or ~$1
Perceived sonic change: marginal/unpredictable

D3 through D10 (8 adjacent 1N4007 diodes) form the rectifier portion of the power supply. These are crappy, cheap diodes. Recently, using ultra-fast (low recovery time) diodes in guitar amp supplies has gotten really popular, with people saying they sound "less harsh" and "more tube-like." These claims may be spurious -- since ultra-fast diodes have nothing remotely like the occasionally desirable voltage sag in tube rectifiers -- but you can clean up hash and switching noise, and improve reliability, by replacing the diodes with faster ones.

There are two options: FREDs (Fast Recovery Epitaxial Diodes) or UF4007s. FREDs are the parts that are getting all the attention. They are huge, look like two-legged transistors, and have the lowest possible recovery time (40ns) at extremely high voltages and currents. They also cost $5-7 a piece, and you need 8 of them. The other option is just the ultra-fast version of the stock diodes -- UF4007s instead of 1N4007s. They cost about 25 cents a piece, with 75ns recovery time.

I went all-out with the FREDs, to see what it would do. They eliminated the supply switching noise, and as far as I could tell, they did change the tone slightly -- but not in the way I expected. Instead of being "smoother", I found the tone had more crunch and sounded more aggressive. Perceived headroom improved slightly. It was not objectionable, so I have kept this mod. The 1200V/16A FREDs I put in will probably be more reliable than the original diodes anyway.

To change the rectifier, replace D3 through D10 (8 diodes) on the main circuit board, but be absolutely sure to observe polarity -- connecting a diode backwards will pretty much blow up your amp. If using FREDs, cover any exposed metal with electrical tape so they don't accidentally short.


4. Add a choke
Cost: $35
Perceived sonic change: medium

A choke is an inductor that helps to filter ripple out of the power supply. Inductors in series act as a lowpass filter -- like in a crossover network -- so one that is large enough to filter out everything down to DC can smooth out the 60Hz ripple after the rectifier (it doesn't matter if you didn't understand that). They were often used in older amps back when large-value supply capacitors weren't as readily available. Now, large caps are easy to come by and cheaper than filter chokes, so they are used by themselves. However, with high inductance, chokes also seem to be able to store a certain amount of current, making that available to the power tubes in high-demand situations. I installed a Mercury Magnetics MC10H choke, and found that the amp became more responsive in high gain situations with a lot of low end in the sound --- palm-muting, octave pedal stoner riffs, etc. It had more attack in the bass and seemingly some more volume.

To install a choke, desolder R71 on the main circuit board. Mount the choke to the chassis next to the power transformer and run the wires through a small hole in the aluminum top plate. Solder them where R71 used to be.

EDIT 10/23/09: I notice that Mercury now has a 25H choke from the Axiom line. I'd be interested to hear people's experiences with that... might be even better for this application than the 10H.


5. Misc. cap value changes
Cost: less than $1
Perceived sonic change: medium

Comparing the TSL with its earlier cousin, the DSL, some find the DSL to have a slightly darker, bassier sound, though the amps are nearly identical. One reason for this may be the coupling cap C16 on the main circuit board. In the TSL, it's a 2.2nF ceramic, in the DSL, it's 4.7nF. I replaced C16 with a 4.7nF Orange Drop cap (nicer than ceramic anyway). The 4.7nF value rolls off frequencies below about 35Hz, which should be pretty insignificant unless you're playing an 8-string or hybrid guitar/bass. The TSL's normal 2.2nF rolls off frequencies below about 72Hz, which shouldn't be terribly noticeable with a 6-string tuned standard. But when you get below that, it certainly does make a difference. The preceding triode (V1B) is not run full-range anyway, so the coupling cap is adding to an existing lowcut.

I eventually found that the 4.7nF cap wasn't great with downtuning -- it didn't add anything I like, and made the amp muddier at high gain. 2.2nF sounds tighter to my ears, so I switched back.

EDIT 4/4/09: I recently noticed another (much more important) component value difference between the DSL and TSL -- C18 on the main circuit board is designed to bleed treble frequencies to ground coming off the V1AA triode. Since this triode is not used by the clean channel, this area is crucial to distortion voicing. In the DSL, the cap is a 470pf cap, but in the TSL, it's only 100pf. This means that more treble passes through to the next stage in the TSL than in the DSL. I have replaced the 100pf ceramic cap C18 with a silver mica 470pf cap, and this is a bigger and better change than the coupling cap swap. The tonality change between stages makes breakup in the later preamp triodes smoother and more mid-centric, but without altering the clean channel. I think this is probably a key source of the perceived difference between DSL and TSL.

Another cap to change is C9 (a 470pf ceramic cap) on the Lead channel circuit board -- this cap bleeds treble past the gain pot on the lead channel. It has no effect when the gain is maxed out (so I didn't notice it while playing with Rosetta), but makes the tone brighter when you turn down the gain. I happen to think it makes the channel sound "quacky". You can reduce the value to 100pf to move the corner frequency higher, or just clip it entirely. I clipped it and find the sound to be much more predictable at different gain points.


6. Fix undersized cap
Cost: less than $1
Perceived sonic change: none

C46, a 22pf/500V ceramic cap on the main circuit board, has been known to fail in the TSL. If it shorts, it can destroy all the power tubes and the output transformer. There is absolutely no reason a 16-cent part should be allowed to cause $300 worth of damage. I replaced this cap with another of the same value, but rated for 3150V.


7. Replace output transformer
Cost: $250
Perceived sonic change: substantial, but not necessarily better

The stock output transformer on the TSL is a Dagnall general purpose model, part #C3070, TXOP 00001. Mercury Magnetics makes a drop-in replacement -- the MAR100-OM -- that is bigger, more reliable, and is supposed to sound better. It's pricey. This mod requires no soldering and no drilling, just reconnecting the 7 wires to the right spade terminals on the new transformer. Mercury includes a diagram.

The new OT has wider bandwidth (read: deeper and higher), so it perceptibly improves clarity and attack. The lead channel seems fizzier at low volumes, but maybe a touch smoother at really high volumes. The "improvement" becomes more audible as you turn up the amp, which is to say, the breakup of the power section comes in more slowly, and there's more consistency from low to high volume. You may or may not consider this a good thing. There is quite a bit more bass available in the tone network, no doubt due to the extended low-end response of the MM. However, this can make the tone sound "tubby" at higher volumes with very low tunings.

I now realize that while this transformer is undoubtedly more reliable and cleaner than the stock model, it may be a little too polished (i.e. linear), lacking some of the "gravel" in the old one. The added bass also makes it initially sound less "tight" with high gain detuning. It's hard to say whether I became attached to the "imperfections" in the old OT, or had gotten so used to compensating for them that it's hard to make the MM work for me. The MM model may be suited to a more traditional style of playing, where the Dagnall seems to be (accidentally?) tighter and smoother for more extreme gain and tuning. I would say the Dagnall has a more (gasp!) "vintage" sound, and the MM is more modern. It may all be a matter of taste, and $250 is a lot to spend on something that subjective.

EDIT 4/2/09: I switched back and forth between the Dagnall and MMOT a couple of times, and found that even though the "tone" of the Dagnall was more pleasing, it was no substitute for the greater volume and bass response of the MMOT. The MM transformer opens up a lot of possibilities.

The real issue is that it was one of the later mods I did, so the earlier component and settings choices had been made unconsciously to accommodate or to flatter the Dagnall's limitations -- after I tweaked some of my existing mods (pulled even more treble out of the lead channel, most importantly) and started over from scratch with all the controls at 5, I found I could get what I was looking for. I can now hear a much bigger difference between different bias settings, and the tone network's behavior actually makes sense. I now have to pull the bass back a lot --- but doing so gets back a good deal of the "tightness" I missed from the old setup, and reveals some lower frequencies (below the center frequency of the bass control) that were never there at all before.

However, make no mistake: the MMOT will not get rid of the "fizz" in the amp, contrary to some other opinions I've seen, and will probably make it even more noticeable. There are advantages to it, but less fizz isn't one of them. To put it succinctly: if you think the amp is fizzy, and you hate that you have to turn it up too loud to get a good sound, don't get the MMOT. If you are modding out the fizz (or you like it as-is) and you wish the amp was louder and had more "oomph," then by all means go for it.

EDIT 4/6/09: On further reflection, I think that the undersizing of the Dagnall is mostly responsible for the "tonal qualities" I was hearing --- it is highly likely that its core was saturating when I pushed the amp hard. I say this because with the Dagnall in the amp, I would hit a wall at about 6 on the lead channel's volume knob, beyond which the amp would not get louder --- but would start to sound dramatically different, losing bass response and tightness. With the MMOT, that wall was removed. I had assumed before that I was hitting power tube distortion, but if that were the case, the limit onset would have changed with different bias settings (it didn't) and would not have changed with a new OT (it did).

Furthermore, the Dagnall OT would get very warm while playing loud, while the MMOT remains the same temperature as the surrounding chassis, even at extreme volume. I can only assume the extra heat from the Dagnall was output power that was lost due to inefficiency, core saturation, or both. This would also explain why there are so many stories of the Dagnall OT blowing up when people try to run the TSL on the clean channel with all the knobs at 10 (Plexi-style). The amp is in fact capable of much more volume and bass response than its stock transformer allows, at least from a perceptual (standing in front of the amp) standpoint.

This added power could be good or bad depending on your needs -- if the amp only sounds good at enormously high power levels which you could never use, then it's not an upgrade. For my purposes though, it's a great thing, although I never expected I would think an amp was "too loud" (yikes!). Eventually I may be able to eliminate my Marshall 3210 slave head and drive both of my 280W 4x12s into speaker breakup with only this amp. I couldn't do that before.

Digression: I have heard that Mesa apparently uses deliberately undersized OTs in their Rectifier heads (but NOT in their other models) to get an effect similar to what I describe above. Maybe that's more integral to the "Mesa tone" of "pleasing compression" than their much-vaunted tube rectifiers...? But then again, "cheap, small output transformers" doesn't make for good marketing copy, so who knows?

---

At this point, I wouldn't trade this amp for one 5 times as expensive. You can hear it on every track on Wake/Lift except (Temet Nosce), using all three channels. At the time of that recording, the amp had mods #1-4 above, with the 390pf/47pf version of #1. The other mods hadn't been done yet. I used some EQ on the Wake/Lift guitar tracks, mostly to compensate for the microphone's non-linearities, but also to get a sound closer to what I imagined in my head. The amp is closer to that sound now, by itself. It is a very good-sounding amp now, one of the best I've ever played through, as far as the qualities I prize most: smoothness, bass transient power that maintains tightness, bell-like cleans, enormous gain, and of course sheer volume. I doubt that --- short of a custom design --- I could find anything else that would be as satisfying, in stock form.

As far as comparisons, I suppose it has a Bogner-ish and/or hot-rodded JCM800 kind of character, but deeper and with a more Fender-y clean channel. It is very much a Marshall, though, maybe more than in its original form. The gain on the lead channel is outrageous but stays focused at high volumes (I have the gain at ten, volume at the edge of power tube & speaker breakup). Crunch is nicely versatile. The VPR circuit actually sounds decent now, too, and whether it's on or not, the amp sounds steadily better as you turn it up. I would characterize the sound as the "next logical step" if Marshall had continued producing amps with the philosophy and quality of the JCM800s, but with modern levels of gain.


Some thoughts:

+ The most effective mods are not necessarily the most expensive. I've tried to be as objective as possible here, balancing what I've heard (NOT properly/scientifically ABX tested, obviously) with what I know about electrical engineering and amp design.

+ No mod will improve an amp for everyone. Different tastes have different requirements. I don't see these changes as "fixing" the amp; instead it's more about keeping the things I liked and refining the things I didn't. Since so much of the tweaking of this amp involves controlling the treble: before you get out the soldering iron, put an EQ in the effects loop and see what you like and don't like in the voicing. Try an EQ in front of the amp, too.

+ Mercury Magnetics has quite a hype machine, but for the most part their products live up to it, and they are helpful to talk to. I would buy from them again.

+ As a taste issue, I also swapped the reverb tank on the amp. The stock one is very good, just not as dense as I'd like. EDIT 8/28/09: I had Accutronics make me a custom Type-9 reverb tank and it's even better than the one I had swapped in before. Nice and rich --- the part number is 9BB3C1D.

EDIT 8/17/09:
"Tightness" vs. "looseness" -- while a lot of this has to do with what kind of cab you're using, I've found that the modded TSL is WAY tighter sounding than any Mesa. Even A/B'ed against a VHT (Sig:X) or Engl (Fireball), it holds its own. You can boost the input, roll back the gain, and get amazingly percussive tones without fizz. It's crunchy but VERY controlled. It's the closest thing I've ever gotten to a Meshuggah tone without a POD. My 8-string sounds incredible (even on the low E-flat string!), with the tone controls all at 12 o'clock and the lead channel gain at 4, and a 6-8 dB boost in front of the amp (a little "frown curve" EQ helps too).

This is all quite opposite of what I originally intended the mods to do --- give me a thicker, smoother saturation for enormously high-gain sludge riffing. It really speaks to the true versatility of the basic design (as opposed to its initially compromised realization) as well as the retention of a lot of buried "Marshall tone" that only came out after modification.