New build with dual-mode tremolo, pentode overdrive stage, DC coupled output section, common plate mixer

[cdemike]

I've been meaning to re-work an amp I built for my wife a while ago since it came out having a significant amount of 60hz hum that I wasn't able to fix over several iterations of the amp. In the process of redesign the amp, I developed an interest in directly-coupled output sections since I was curious about reports of the compression the driver lends in amps like the Steel String Singer. My wife's amp is a small, light amp similar to a 6G2 Princeton that was made to be easily be carried to coffee shop gigs and other lower-volume venues, so the relative improvement in output section efficiency subsequently got me interested in trying out a small signal pentode as a source of overdrive ahead of the phase inverter (similar to tubenit's Tweed ODS, Carolina ODS, and D'Mars). Something else I was working on in the abstract was a dual-function tremolo circuit that works both as standard amplitude tremolo and as harmonic tremolo and thought this might be a good test case since the original version of the amp omitted the tremolo circuit. I initially was playing around with designs influenced by the common Brown Fender harmonic tremolo circuits but eventually came across descriptions of using 18w Marshall/Watkins Dominator circuits' two channels jumpered to essentially get harmonic tremolo by modulating the brighter tremolo channel in parallel with the darker "normal" channel. I decided to take that as the basis my tremolo design concept since it potentially offered a significantly simpler solution than the Brown Fender designs and could work with more commonly available parts (e.g., no need for a 10M intensity pot).

The design evolved from that set of factors and produced a working amp that sounds great. I wanted to share the design since it incorporates a variety of unusual design elements, hence the title. Two things in particular I thought might be helpful were the tremolo and output section designs, since I was unable to find anything similar to either. Although I've seen videos and read descriptions of the jumpered 18w harmonic tremolo, I wasn't able to find any designs using that as the basis for harmonic tremolo rather than the more common Fender topology. I initially set the amp up to use the same LFO as the 18w Marshall but found the amp had terrible oscillations. I was able to mitigate the oscillations increasing use of shielded wire, but with the amp stable and actually amplifying guitar, I found I wasn't able to get the tremolo to work. I made several changes to the amp to try to get the 18w tremolo to work but eventually decided to re-design the tremolo set up. Where initially I designed the first gain stages to be parallel stages similar to the 18w Marshall with a bright always-on tremolo channel and a dark parallel channel which would get mixed in at the common plate mixer for harmonic mode, I decided to work in a Vibro Champ style bias wiggle circuit to wiggle the bright channel's cathode in the common plate mixer. That required splitting the mixer's cathode, which actually ended up being a good thing since once I got the tremolo working I found the channels were imbalanced in harmonic mode with the dark channel overpowering the bright one, resulting in weak-sounding harmonic tremolo. Splitting the cathodes allowed me to better fine-tune the relative volumes of the two channels by biasing the dark side significantly colder. Abandoning the Watkins/Marshall style LFO also opened up options in terms of LFO; I wanted to have a wide range of speeds available and consequently borrowed Sluckey's modified Supro LFO design which I am unsure would have worked in the Dominator/1974 style circuit given the differences in AC loading between the circuits. But the LFO works great with the source follower (credit to RG Keen). The only downside I can find in this set up is that the common plate mixer sounds very different with the parallel channels running into it in harmonic mode and the "bright" side running into it alone in amplitude mode. To my surprise, the amp is significantly darker sounding when plugged in the "amplitude" input, which I suspect may be due to loading effects viz the 33k plate resistor. That value was chosen, though, because of the very limited range on the volume control before the 5879 starts to overdrive (it starts to overdrive around 9:00). I'm not particularly bothered by the early overdrive having played mostly Marshalls with large bright caps but wanted to have some granularity in the volume control which the 33k provided.

Drivers in DC coupled output sections obviously aren't new, but I haven't seen any designs using them in smaller "micro" style builds. I wasn't sure the juice would be worth the squeeze in a "micro" design but I've noticed three advantages that I really enjoy in this amp that I attribute to the output section topology:
-As many have noted in their SSS builds, using 12AX7 as a driver introduces a fair amount of compression, which in this micro build helps offset the relatively over-powered power transformer. In other words, when working hard, the amp feels more like a larger amp, in my opinion, due to the compression.
-Because the output section topology itself is the source of compression rather than the power supply, I used heavier filtering than I would have otherwise and the resulting amp doesn't feel stiff. As I mentioned earlier, the impetus for rebuilding the amp was an intractable 60hz hum, so one certainty coming into the rebuild was that this iteration would have significantly heavier filtering. I was initially concerned the amp would be unforgiving and stiff, but I have been pleasantly surprised by how squishy the amp can feel.
-I haven't been able to elicit blocking distortion despite the significant signal amplification happening in the preamp. The 47nF coupling capacitors feeding the driver were carry-overs that found their way into the new amp based on availability of capacitors I have on-hand. I anticipated needing to replace them based on the significant increase in preamp gain and resulting increase in bass amplitude reaching the output stage. I doubt the amp is operating deep in class AB2 (if at all) but the amp has significantly higher headroom than in its previous form. The increase in headroom allowed for the 5879 stage to really shine as the primary factor shaping the overdriven sound, which helps the amp sound more a bigger pentode-powered amp despite having the "micro" style power amp.

[lonote]

Really cool design, thanks for sharing your work!

I will have to dig deeper into both the text & drawing to really get a grasp of how the two signal paths work. Totally ignorant about DC-coupled circuits, is it just the driver tube to power tube connections that makes it DC-coupled?

I have a small collection of alternative trem circuits (one of your drawings included) & started fooling with a hybrid SS LFO/tube trem circuit before getting sidetracked. I would like to get back to it soon.

[cdemike]

Thanks! The DC coupling refers to the output of one tube passing directly to the input of the next stage without a coupling capacitor. In addition to the effects of the driver itself (AFAIK, use of a high plate resistance tube like 12AX7 is the source of the compression in the SSS), direct coupling avoids charge time associated with capacitor coupling resulting in a "quicker" feeling bass response. Capacitively coupled output sections are also unable to sustain positive grid voltages for meaningful amounts of time since that requires DC flow into the grid, so the two ways I've seen AB2 output sections coupled to drivers or using interstage transformers. I would be interested to know if my amp is actually operating near AB2 come to think of it -- I'm guessing the loading effect of the multimeter would interfere with signal readings at the ECC99 grid. Is there a way to measure that without a scope or signal generator?

Flattered to hear about the trem circuits -- thanks again for your help with that Marshall's trem circuit. Revisiting the trem on that amp is second on my project docket currently, and I planned on installing a source follower in that amp. Will report back in the related thread!

[cdemike]

Recorded a sound demo showing the tremolo modes. I'm a geek about NASA so I name my amps after different NASA programs; Gemini seemed like a natural fit given the dual mode tremolo. I realized I showed a lot more of the harmonic mode in the demo and I think that might be related to how dark and dirty the amp is when in amplitude mode. Seems like I might need to mess with some values around V2 to balance that out, but things improved with a 100nF cap bypassing the bright triode's cathode in V2. This was recorded with the new cathode bypass cap installed.



This gut shot, however, was not.

IMG_1910.jpg

I'll play with some values and report back. Right now the node supplying V1 and V2 has ~230V, and there's 1V on the bright channel's cathode, so I'm thinking I'll start with fine tuning the bias of that stage to see if that helps without impacting the tremolo too much.

[dorrisant]

That thing sounds mean, and I like it!!

[maxkracht]

Cool project, nicely done!

[cdemike]

Thanks so much! As promised, here's an update after I tried to get the channels to sound more alike and to get more headroom. I determined that most of the distortion was coming from the common plate mixer which also was making the amplitude tremolo side sound significantly darker than expected with its 500pF coupling cap when running independently. I experimented with getting the stage better loaded to get better performance including asymmetrically split plate load resistors (i.e., a series resistor from the always-on triode's plate to the dark channel's plate) and increasing the load overall to 100k and 220k with corresponding adjustments in cathode resistors to get somewhere near center-bias. My takeaway was, heartbreakingly, that I'll likely need to split the plates and go with a plain resistive mixer. Bummed since I came to believe that a common plate mixer should perform at least reasonably closer to a plain triode gain stage, but I just wasn't able to make it work. I suspect a larger plate load and a high-value cathode bypass capacitor would have made the difference by reducing the loading effect of the unused triode, but under such conditions I wasn't able to get the tremolo to work with adequate depth on the one hand while on the other hand maintaining a reasonable amount of headroom in the second stage.

With the second stage where the tremolo gets injected split, I decided to experiment with AC tremolo coupling since DC coupling was a major contributor to my troubles with the second stage's headroom at higher tremolo intensity settings. Due to space constraints on the Hiwatt style turret board, I couldn't fit the EQ network that splits the signal into bright and dark components in the regular 6G4A/6G5A circuit and the 6G4/6G5 EQ network has an even higher parts count. So for anyone still interested in this alternate method of getting harmonic tremolo based on the Watkins/Marshall quirk, I stuck with the same concept of dissimilar coupling and cathode bypass capacitors to split the signal rather than parallel high and low pass filters. I tweaked the plate bypass capacitor on the "dark" side, but I think I'll likely go back to the 4n7 that was there before since the harmonic mode lost the swirly/phasey sound that it had before.

The trouble I'm having now is that I'm getting very weak signal to the cathodes with this setup, so I'll need to to think through ways to get better modulation at the cathodes. I settled at 5K6 at the second gain stage cathodes since the 7k2's I started with were way too cold which led to a nasty always-distorted quality that reminded me of crossover distortion. So I could go three different routes I think:
- My preference would be to find a way to increase the LFO signal reaching the cathodes. I've read that some people have had luck with lower plate resistor values, but it seems like that would reduce the amplitude of the signal by reducing the gain of the stage?
- Find a way to DC couple both cathodes to the LFO signal. Maybe the regular Vibro Champ setup but the LFO's anode feeding signal both to the cathode follower and to a gain stage with local NFB to invert the signal. The plate of the gain stage could then be DC coupled via a 2M resistor from the plate, which should reduce the DC level to be roughly equivalent that from the cathode follower when it reaches the intensity control. A dual ganged 25KL pot would then feed the signal to the two cathodes.
- DC couple one of the cathodes: use the same cathodyne setup as-is but with the cathode resistor running in series with the bright cathode in the second gain stage. The dark cathode could remain AC coupled but the lion's share of modulation would be affecting the bright side.

Option 3 would likely be the most effective, but an additional concern that I'm trying to figure out is that the tremolo ticks in amplitude mode and thumps in harmonic mode at higher depth settings. With DC coupling I'd think that there would be even greater signal transmission, making the ticking worse? Would love any input. I may try option 3 later since it's the simplest, but any thoughts on getting better signal? The schematic for the first two gain stages and tremolo set up is attached. I didn't label the intensity control, which is the pot between the LFO's anode and the cathodyne's grid. It's 1MA currently.

[cdemike]

New round of changes done! I stuck with the resistive mixer since that was providing much more predictable performance than the common plate mixer. To get the tremolo operating better, I first tried messing with the LFO to see if I could increase its output but found that 470k was indeed providing the best signal strength. With option 1 seeming to be a dead end, option 3 would be the most conservative rework prior to needing to comprehensively rework the driver to be an anode follower while also working the source follower back into the circuit as option 2 requires. DC coupling the cathodyne's cathode dramatically increased modulation amplitude and to my surprise was significantly quieter than the AC coupled version. I thought the frequency filtering from the coupling capacitors would reduce noise, as would the limitations on the LFO signal amplitude, but with the cathodyne's cathode resistor moved to the junction of the bright gain stage's cathode and cathode resistor the thumping nearly disappeared. The dark side at this point remained AC coupled to the cathodyne's plate, but with the output of the bright side grounded, I found I wasn't getting any audible LFO signal on the dark channel. Nevertheless, the harmonic tremolo function still worked well, so it seems to reinforce that the Watkins-inspired harmonic tremolo may be easier to set up given that only one side need to be modulated. With the general direction forward seeming clearer, I removed the connection between the plate and the dark stage's cathode leaving essentially the rump cathodyne functioning as a cathode follower which was current-restricted by the remaining plate resistor.

My next step was to hone in the voicing the harmonic mode and to fine-tune the range of the depth control. Reinstalling the 4n7 plate bypass cap returned the swirly quality that I missed when aiming for similar corner frequencies as the original 6G4A/6G5A signal filters. Regarding the depth control's range I figured increasing the ratio of signal preserved at the gain stage's cathode would be important to setting the maximum depth, while I could attenuate the signal reaching the cathode follower in the AC coupling stage. Following that line of thinking, I did find improvement as I lowered the value of the cathode follower's cathode resistor with corresponding increases in the plate resistor to keep the overall current levels somewhat steady. Once I reached a point where the cathode follower's plate resistor was 220k and the cathode resistor 27k, I began thinking this was starting to look basically like a Supro-style driver where the cathode follower outright shares its cathode with the gain stage it modulates, rather than having a resistor in series with the CF driver and gain stage's cathode. My next step was then to rework the driver to match the 6424.

Since revising the board to accommodate the new driver topology would require removing the board to access under-board jumpers, I took the opportunity to make some voicing changes since the amp sounded much more mid-scooped than expected. I was surprised since I've read the 5879 tends to be a mid-forward tube, and the 6G2 style tone stack does not scoop mids. While I expected the harmonic channel to sound mid-scooped compared to the amplitude channel since the filtering required to separate the signal into bright and dark components, I expected there to be a mid scoop in the way that the Fender approach similarly leaves midrange filtered out. However, with the scoop remaining on the amplitude channel, I thought the solution for the mid scoop would likely mean re-voicing the signal outside of the parallel second gain stage. While reviewing Supro schematics to guide the tremolo driver revisions, I noticed a note in the 1690t schematic on the Hoffman site describing "squawking mids" resulting from a split plate load resistor arrangement where a 5nF capacitor connects to ground (New round of changes done! I stuck with the resistive mixer since that was providing much more predictable performance than the common plate mixer. To get the tremolo operating better, I first tried messing with the LFO to see if I could increase its output but found that 470k was indeed providing the best signal strength. With option 1 seeming to be a dead end, option 3 would be the most conservative rework prior to needing to comprehensively rework the driver to be an anode follower while also working the source follower back into the circuit as option 2 requires. DC coupling the cathodyne's cathode dramatically increased modulation amplitude and to my surprise was significantly quieter than the AC coupled version. I thought the frequency filtering from the coupling capacitors would reduce noise, as would the limitations on the LFO signal amplitude, but with the cathodyne's cathode resistor moved to the junction of the bright gain stage's cathode and cathode resistor the thumping nearly disappeared. The dark side at this point remained AC coupled to the cathodyne's plate, but with the output of the bright side grounded, I found I wasn't getting any audible LFO signal on the dark channel. Nevertheless, the harmonic tremolo function still worked well, so it seems to reinforce that the Watkins-inspired harmonic tremolo may be easier to set up given that only one side need to be modulated. With the general direction forward seeming clearer, I removed the connection between the plate and the dark stage's cathode leaving essentially the rump cathodyne functioning as a cathode follower which was current-restricted by the remaining plate resistor.

My next step was to hone in the voicing the harmonic mode and to fine-tune the range of the depth control. Reinstalling the 4n7 plate bypass cap returned the swirly quality that I missed when aiming for similar corner frequencies as the original 6G4A/6G5A signal filters. Regarding the depth control's range I figured increasing the ratio of signal preserved at the gain stage's cathode would be important to setting the maximum depth, while I could attenuate the signal reaching the cathode follower in the AC coupling stage. Following that line of thinking, I did find improvement as I lowered the value of the cathode follower's cathode resistor with corresponding increases in the plate resistor to keep the overall current levels somewhat steady. Once I reached a point where the cathode follower's plate resistor was 220k and the cathode resistor 27k, I began thinking this was starting to look basically like a Supro-style driver where the cathode follower outright shares its cathode with the gain stage it modulates, rather than having a resistor in series with the CF driver and gain stage's cathode. My next step was then to rework the driver to match the 6424.

Since revising the board to accommodate the new driver topology would require removing the board to access under-board jumpers, I took the opportunity to make some voicing changes since the amp sounded much more mid-scooped than expected. I was surprised since I've read the 5879 tends to be a mid-forward tube, and the 6G2 style tone stack does not scoop mids. While I expected the harmonic channel to sound mid-scooped compared to the amplitude channel since the filtering required to separate the signal into bright and dark components, I expected there to be a mid scoop in the way that the Fender approach similarly leaves midrange filtered out. However, with the scoop remaining on the amplitude channel, I thought the solution for the mid scoop would likely mean re-voicing the signal outside of the parallel second gain stage. While reviewing Supro schematics to guide the tremolo driver revisions, I noticed a note in the 1690t schematic on the Hoffman site describing "squawking mids" resulting from a split plate load resistor arrangement where a 5nF capacitor connects to ground (https://el34world.com/charts/Schematics ... _1690t.pdf). I wasn't sure how this would change the midrange since the signal path "forward" out of the gain stage had significantly lower impedance than the parallel path to ground through the plate load resistor, but a Spice simulation was helpful. With a plain 220k plate load resistor, a non-bypassed gain stage's frequency response looked like this:

220k plain.PNG

With the 5nF capacitor between series 110k plate resistors, there was a slight high end roll-off:

220k supro.PNG

Trying different values, it appeared that a value around 1-2nF would provide the greatest emphasis on midrange. This is the response for a 2nF cap:

220k 2n.PNG

I also worked in the low pass filter common to the 1624t, 1690t, and 6424 to further reduce treble response. I placed this filter after the first stage while simultaneously reducing the coupling cap to 4n7 to simultaneously reduce bass response and therefore increase midrange content rather than just making the amp darker. Though reducing the 100k plate resistor to 68k at the first stage helped increase headroom given the cold-clipper adjacent operating point of the second stage, splitting the plate resistor ended up providing a better solution with greater control. I opted to put the "squawking midrange" split plate load/cap arrangement on the bright side of the second stage since the schematic noted the behavior "when pushed hard."

Overall, these changes got the amp sounded much closer to how I would like. I'm very happy with the tremolo, and the revisions to the voicing were steps in the right direction voicing-wise. I'm especially happy with how touch response the amp remains. Here are some sound clips:
Light overdrive (amplitude channel): https://soundcloud.com/mike-913267560/g ... -overdrive
Amplitude tremolo: https://soundcloud.com/mike-913267560/g ... itude-trem
Harmonic tremolo: https://soundcloud.com/mike-913267560/g ... ic-tremolo

I was hoping to get input in terms of how to better voice the amp to reduce the fuzzy quality of the overdrive. Over 1/2 way up the volume control the overdrive sounds more like fuzz to me. Here's the schematic as it stands now:

Gemini Mk IIb schematic cropped.PNG

Here are some sound clips of the overdrive. The clip starts in the harmonic tremolo channel which still sounds pretty scooped to me. Once I switched to the amplitude channel, it still sounds surprisingly scooped to me, but in both cases the overdrive sounds too fuzzy in my opinion: https://soundcloud.com/mike-913267560/overdrive-tuning

I reworked the values at the 5879 stage and upon review of the 5879 datasheet it looks like there should be a decent amount of headroom, so I'm not sure if the culprit is the 5879 getting completely overwhelmed with input signal. I have noticed that the Tweed ODS and related amps have significant differences in plate to screen current ratios, and, maybe more importantly, there is significantly more current flowing through the tube in that setup. For example this schematic shows ~3ma at the cathode vs 2.4ma in mine:


I similarly tuned the phase inverter's bias and added a grid stopper. I'm somewhat confused by the phase inverter remaining at such a high bias voltage despite the voltages being similar to those described in the Paul C mod, but perhaps the PI's operating point is contributing to the fuzzy distortion? I'm skeptical that my the phase inverter measurements are accurate based on their dissimilarity to others' measurements and the known issues measuring grid voltage with a regular meter. FWIW, the measurements are taken from the divider's junction with the grid stopper, not at the grid itself, which as I understand it should improve the accuracy of the measurement.

Is there anything else that might be causing the fuzzy sounding overdrive and mid scoop? I'm coming up dry searching here and the Hoffman forum, and I'm especially confused by the EQ response being so dissimilar to others' descriptions of 5879s' sound.

[cdemike]

I had a chance to do a lot of fine tuning on the amp in the past few days to figure out what was going on with that harsh fizzy sound I mentioned in my last post. As I suspected, a major part of the issue was the cathodyne's operating point, which was biased way too cold. Raising the bias about another 30v (exact new voltages in updated schematic) cleaned the amp up, eliminating the hash in the overdrive and improving the note decay. The overdriven sound improved also with removal of the cathode bypass cap on the output section. I didn't expect the amp to dip into AB2 after doing more reading on what's required to operate tubes with a significant amount of grid current. I was curious if I could see any sign that the amp was getting into grid current, though, so the bypass cap was meant to give the amp the best chance of doing so. It turns out that I like the way it sounds better without it. While trying to dial out the fizz, I did some digging on plate bypass capacitors operating as smoothing caps, as well as tubenit's "enhance cap" which he describes as also helping reduce fizz. I wasn't able to find any documented examples of people who have tried it on a cathodyne, so I wasn't sure how it would differ versus the documented applications in LTPIs. I found I liked the sound better over the cathode vs the anode. I'm not exactly sure why, but it didn't really make that much of a difference in the sound across the anode, whereas the PI seems to be much more stable with the cap across the cathode resistor. I'm thinking that may be related to the same loading phenomenon that occurs when cathodynes are AC coupled to power tube grids when the tubes start conducting grid current, i.e., the conditions where the "nipple" appears in the signal trace. Seems like the cap may have the opposite effect by stabilizing the cathode when it's across the cathode resistor?

Interestingly, the amp sounds a lot less scooped to my ears with the PI operating closer to center bias. I further boosted the mids by bypassing the top leg of the voltage divider with a 4n7 cap between the 5879's plate and the PI's grid.

Overall, I think the amp is sounding really good. The harmonic tremolo still works well despite the changes to the EQ brought on by reducing the grid leak resistor for the 5879, and I'm pretty pleased with the amp's transient response which is neither too soft or too firm for my preferences, and often has a vowel-like quality when the amp is driven hard. I'm also pretty happy with the design working as-intended: this was my first build that I'd consider an "original" design (i.e., not based on someone else's amp) and I'm liking the sound despite numerous unusual design features like the unorthodox harmonic tremolo circuit, DC coupled cathodyne, and DC coupled low-wattage output section.

Below are some clips. There's a fair amount of hum, but my guitar's pickups really prone to amplifying noise my router and other nearby noise sources, which I am not really able to get that far away from in my apartment. I also recognize these clips have WAY more overdrive than any of the original recordings, but I wanted to demonstrate how well the amp cleans up using the guitar's volume control.
Overdrive on the amplitude tremolo channel: https://soundcloud.com/mike-913267560/h ... d-dynamics
Overdrive on the harmonic tremolo channel: https://soundcloud.com/mike-913267560/h ... ic-channel

[B Ingram]

... a major part of the issue was the cathodyne's operating point, which was biased way too cold. Raising the bias about another 30v (exact new voltages in updated schematic) cleaned the amp up, eliminating the hash in the overdrive and improving the note decay. ...

Your final schematic shows 94v across the cathode-load; we will assume the plate-load is an equal resistance and therefore has an equal voltage-drop. It turns out you just re-discovered a rule of thumb for split-load inverters, which is to leave about 1/3 of the supply voltage across the tube, and 1/3 of the supply voltage across each load resistor.

... I was curious if I could see any sign that the amp was getting into grid current, though, so the bypass cap was meant to give the amp the best chance of doing so. ...

You have about 15v indicated as the cathode-voltage of the output tube sections. Measure peak-volts at the ECC99 grids using a peak-hold function of a meter, or your scope. When the peak input voltage approaches or exceeds 15v, you're certain to have grid-current (unless the cathode voltage rises due to unequal current draw by each ECC99 section when driven).