Tweedle Dee - 5E3-D

[Stephen1966]

So, I think the answer to the problem is two pronged: increase the value of the cathode resistor and replace the power tube with one that has a greater voltage drop. I am doing the numbers now.

[Stephen1966]

I've crunched a few of the numbers and made an order for new resistors and tubes.

I checked the specs of my build against the Hammond PT. The heater voltages and current are all in line with what the amp is delivering and I don't have a problem there. The rated 355-0-355 B+ is kicking out around 371V to the rectifier plates though and so I've ordered a Ruby 5Y3GTC (not a Sovtek ). If I get a 60V drop with this tube, I would not be surprised, or alarmed because if it brings my voltage down to around 355V on my power tube plates that sounds like a good start stage for increasing the resistance of the cathode resistor.

Using the cathode resistor bias method I measured 245 Ohm across the resistor and a voltage drop of 23 V giving me a plate current of 0.0938A. Multiplying that by the plate voltage (COM to GND) which today, measured 425V gave me a plate dissipation of around 20W... per tube! If I don't alter the figure for the voltage drop (I don't know how to calculate it) it takes a 350 Ohm resistor to cool the tubes down to around 12W. The output transformer method of calculating bias appears to provide a more accurate method of calculating plate dissipation but the headroom of the cathode resistor method seems as good a place to start as any.

This next part is a bit of wait and see because increasing the resistance of the cathode resistor is likely to decrease the current but drive the plate voltages higher again.

A little extra reading took me to the Fender Edge Deluxe which has a similar tube compliment (with a 5Y3) but they are not giving up any information on the power transformer they used alongside the 250 Ohm bias resistor. I contacted Mercury Magnetics and received a polite but firm reply that the Edge transformer was a proprietary model and that as such, they couldn't give any information. (Ah well, worth a try, I suppose.) Reading an independent review of the amp though, one reviewer measured the plate voltage at 440V - my best guess is that it is around 320-0-320 peak. If I were to build this amp again, that is the B+ I would try.

I've ordered a pair of JJ6V6S' as well, along with a variety of resistors so that I have some different values to play with.

I am guessing the first job is to replace the cathode bypass cap but from there, I am not sure of the best way to proceed. Perhaps the least invasive path would be to swap the 6V6's first, and then the rectifier, and from there, work on the resistor?

[Stephen1966]

So I've been working on the amp and I have the bias down to a level I think I can live with.

The first thing was to replace the 5AR4 with the 5Y3. As expected that brought the voltages down but the current was still hot so then I replaced the cathode resistor with one rated at 330 Ohms. As I suspected this sent the voltages higher, though not so high as the 5AR4 but it brought the current down to a more acceptable level.

I am posting the test data here for anyone who is interested or who goes down this path in the future and I measured the bias first using the cathode resistor method and then the output transformer method. These gave me slightly different numbers but both see the bias at around 100% max dissipation using the JJ6V6S's which seem to be handling it quite well - no signs of redplating.

With thanks again to Charlie Wilson, and Marcus who came forward with more revealing information about the original Tweedle Dee and the clone Marcus made. I think as a note to my future self that if I have any further problems with this now, I would look to replace the power transformer with one that doesn't kick out so much juice. The rig might take a 355-0-355 power transformer but mine appears overspec'd and the current output at 115 mA is higher than I have seen in other reproductions of the fender standards. As I understand it, Dumble used stock transformers though - no doubt - they were carefully selected.

I haven't recorded my B+ voltages in this attachment, but I'll document those before I button the cabinet up and get playing with this thing. So far, it sounds great and very dynamic.

Voltage data.pdf

[Stephen1966]

Here is the comparison of voltages of my amp after all the mods against Charlie's original readings:

Voltages compared 28 Feb 2020.pdf

Also, here are a few more pics of the interior - before I close it up forever

SAM_7186.JPG

SAM_7182.JPG

SAM_7190.JPG

[Stephen1966]

I felt it was time to revisit this, my first build. I've had a lot of fun with the amp in the meantime but with what I learned it became obvious I could do better. I knew next to nothing about lead dress back then and there was a little noise which went unnoticed when playing but which was always present and got louder as the volume increased. The back of the new board I made is much cleaner and having ground points in the chassis, like the Tweedle Dee and the Deluxes before them, runs a very low sound floor now - no shielded cable anymore, either. So, good grounding and carefully managed lead dress will give you a dead quiet amp without any special measures. As a nod to modern Fenders I also added a thermistor for inrush current limiting and an X-class cap for a bit of mains filtering.

In this shot, you can just make out some of the solder prep for the grounds. It takes a really hot iron (150W) to get the solder flowing on such a big heat sink. I had some concerns it might damage the silkscreen on the front but it was fine - just very quick in, and quick out.

IMG_20240125_184144.jpg

The circuit you see already installed here is the VVR.

Here's the new board, front and back:

IMG_20240121_114210.jpg

IMG_20240121_114415.jpg

[Stephen1966]

As a space saving measure as well, I used the switched output jacks for single 8 Ohm or 4 Ohm output or dual 8 Ohm. The 16 Ohm tap goes redundant in this scheme but it could be utilised for impedance matching with tubes.

IMG_20240128_164832.jpg

[Stephen1966]

So, one of the defining characteristics of the TD compared to the Tweed Deluxe, 5E3* are the TD's higher voltages.

Lots of builders around here have met this conundrum. It may be possible to get a modern 380-0-380 PT like the old Triad Charlie mentioned but loads of builders have followed the path of using a "typical Fender" 355-0-355 PT. Trouble is, it invariably runs too hot and with a 250 Ohm cathode bias resistor we get a very hot max plate dissipation - anything up to about 20 Watts. Voltages are usually higher than documented as well and making it all worse is the GZ34 rectifier. The voltages are brought into a more reasonable range with a 5Y3 rectifier but the cathode bias resistor needs to be increased to bring the operating point into a lower, safe operating area (under 14 Watts for JJ 6V6S tubes). Two steps forward, one step back!

In the rebuild, I thought I would try a lower voltage PT, the Hammond 290BEX compared to the old Hammond 290BBEX I had in previously. Of course, it didn't go according to plan. Biasing the power tubes to 14 Watts, my voltages overall were on average, 30 Volts higher and the PI was hard clipping with all the current it was seeing. You would think that using a PT rated for 330-0-330 would have tamed things a bit but now I had a 470 Ohm cathode bias resistor, and quite a helping of crossover and blocking distortion at the very highest volume levels. I took an experimental approach to the problem of distortion and tried a number of things involving grid stoppers and attenuating networks but nothing really worked until I put the old 355-0-355, 290BBEX transformer back in. By this stage, it had already dawned on me that it wasn't so much the voltage of the transformer that mattered but the amount it was going to sag. Both the transformers I used have around 7% regulation, which is good by modern standards I guess, but the main difference was the rated current available on the HT. The 290BEX is rated for 140mA the higher voltage 290BBEX is rated for just 115mA and this makes a lot of difference. Interestingly, people who used the Classic Tone 40-18021 reported better results and the reason appears to be the 100mA rating on 355-0-355. Unfortunately not an option for European users but certainly seeming to be heading in the right direction. A typical 5E3 consumes around 90mA and the higher voltage TD about a 100mA so I am thinking that to get something like the documented TD voltages with a GZ34 and a 250 cathode bias resistor, we need a PT that sags a lot and barely charges the filter caps before they are "discharged" with each cycle. Meeting the current demand at idle would drive the voltages lower. So the secret, if there is one here, seems to be one of throttling the current to the bare minimum. I have no idea what the original Triad was rated for but logic dictates it probably wasn't even capable of 100mA.

40-18021.pdf

290BBEX.pdf

290BEX.pdf

Edit: Looking around, it isn't so hard to find lower voltage/current transformers for 5E3. Maybe contact Mojotone if you go with their PT though, I've seen three sets of documentation for a single export transformer and they are all different!

*5E3 because even though the documented TD is in a 5C3 chassis, the TD uses the cathodyne PI like the 5E3, not the paraphrase inverter of the 5C3. The basis for the TD is really the 5E3 circuit despite appearing to be based on the 5C3!

[Stephen1966]

Another of the defining characteristics often mentioned is that the cathodes of the first stage are split in the TD and that this leads to less volume interaction. The prevailing wisdom today, is that the interactive volumes - providing a bass boost as they do - is a virtue to be exploited. I don't know if MrD saw it that way and splitting the cathodes was his attempt at fixing a flaw in the design, or whether it was a way of minimizing the effect for more predictable results. Splitting the cathodes doesn't remove the interaction altogether but what I found interesting isn't that he simply splits the cathodes, but that he alters the bias point of the first stage as well.

I don't know when it became a thing to replace the 12AY7 with the 12AX7 for more gain in the Tweed Deluxes but I'm assuming when it did, players didn't bother adjusting the bias. It's interesting to see what happens when you follow the evolution from the 12AY7, to a 12AX7 in the 5E3 and then how MrD treated it with higher B+ voltages as well.

Deluxe 5E3 - V1 - 12AY7 curves.jpg

Centre bias -3V actual -2V

Deluxe 5E3 - V1 - 12AX7 curves.jpg

Centre bias -1.35V actual -1.35V

TWEEDLE DEE - V1 - 12AX7 curves.jpg

Centre bias -1.6V actual -1.2V


[Edit: I had to make some corrections here, it was just compounding an error I made when compared the tube charts visually. From the design perspective though, I think the basic premise holds true: the TD was taking its cues from the 5E3 with a 12AY7 in the first stage.]
What I find curious is that the effectively smaller cathode bias resistor of the TD is almost perfectly centre biased for a minimum of distortion. [Correction! Not true... it biases warm...] More like the original 12AY7. [Deleted: incorrect statement] The effective 1.64k cathode bias resistor is pretty close to the ordinary 1.5k so when MrD dropped that to just 1k and you have a B+ around 40V higher, you can be sure he was considering the bias, the headroom, the output swing and the distortion.

[Stephen1966]

Just for completeness, I'm posting the charts for the second stage of the amp which in many ways is a conventional gain stage. MrD added the extra filtering stage after B+3 in the TD. The 10k dropping resistor he used to create B+4 only dropped the final node's voltage by about 23 volts but it allowed the extra level of filtering in the power supply. Together with the smaller coupling caps for a tighter low end in the TD the increased voltages allow this stage significantly more headroom - from about -3.2V to cutoff in the 5E3 to about -4.1V in the TD. The following impedance the triode sees is bootstrapped by the cathodyne's 1 Meg grid leak resistor in the 5E3 to a massive 18.081 Meg. This effectively puts the AC load line of the second stage here, parallel with the DC load line. The following impedance of the TD I calculated comes out at 13.52 Meg. Again, barely enough to rotate the AC load line around the idle point. The AC load lines shown here are derived before the bootstrapped input resistance of the PI was calculated providing the most conservative estimate of the AC load.

Deluxe 5E3 - V2a - 12AX7 charts.jpg

TWEEDLE DEE - V2a - 12AX7 charts.jpg

[Stephen1966]

I'm throwing the schematics for the 5E3 and TD in here. The 5E3 voltage data is straight from the 'Standard 5E3 DC Voltages' section published on Rob Robinette's site https://robrobinette.com/How_The_5E3_Deluxe_Works.htm and the TD voltage data are measurements I took using the VVR to set the voltage at the reservoir cap to the same, 398 V; see the interpolated voltages. You will see some disparity with the voltage data I collected from my build and that predicted in the charts which have been using the extra voltage data published in Charlie's layout sketch where possible; see the schematic with Charlie's voltages. That said, the predicted voltages are all in the ballpark of the measured voltages.

Charlie's layout is here if you want to check it out: https://ampgarage.com/forum/viewtopic.p ... 53#p305953

Deluxe 5E3 voltages.jpg

Tweedle-D Schematic with Charlie Wilson voltages.jpg

Tweedle-D Schematic with voltages.jpg

[Stephen1966]

Here are the charts for phase inverter of the 5E3 and the TD.

Deluxe 5E3 - V2b - 12AX7 curves.jpg

TWEEDLE DEE - V2b - 12AX7 charts (2).jpg

[Stephen1966]

Perhaps the most obvious place to start with the differences between the Fender and Dumble PIs is with the higher gain we see in the Dumble load line. This isn't an ordinary gain stage though and both the Fender and the Dumble have a gain just a tad less than unity: 0.97.

What the charts reveal though is that the TD has more swing between saturation and cut-off (213V) compared to the Fender (145V). Earlier, in the preamp, he was able to increase the headroom by increasing the HT and little else but here he practically doubles the value of the plate and tail resistors and the bias is slightly cooler as well. No doubt enhancing the non-linear harmonic distortion. Fender's design uses equally matched 56k resistors for the plate and tail, with a 1.5k bias resistor which barely tips the balance between the output signals at the plate and cathode. Dumble's design however, exploits imbalance with a 110k plate and (an equivalent) 99k tail resistors thus rendering it practically impossible to balance the output voltages of plate and cathode.

Another major difference between Dumble's PI and Fender's is the addition of LNFB between the plate and grid of the Tweedle's cathodyne. The most significant difference in my view, but perhaps not for the reasons you might think. The 5E3 does not use any additional LNFB circuitry. The cathodyne already functions with a theoretical 50% negative feedback. In the Fender cathodyne you have a very efficient and economic PI perfectly suited to the amp. Because very little voltage drops between the bottom of the grid leak resistor and the top and very little current is lost, the cathode voltage is very close to the grid voltage and so it appears the resistance of the grid leak is much higher than it actually is. The resistor is bootstrapped or amplified to a much greater resistance and this is the AC load seen by the preceding stage. Fender's design thereby increased both the headroom and swing of the previous stage. Dumble took it to the next level when he added the LNFB cap and resistor.

The cap in the LNFB is not there to really provide any frequency filtering, it's cut-off is quite below the range of the guitar at around 20Hz, but it does effectively block DC between the plate and the grid. The 3.3M frequency resistor performs a double duty by attenuating the closed loop signal passing from plate and reinserting it back to the grid, and it also acts as a voltage divider along with the grid leak to form a virtual input resistance in the absence of a physical grid stopper. The feedback is shunted back to the grid and combined with the open loop unattenuated signal to reduce the signal size. In a kind of Alice Through the Looking Glass move though, the signal does not realise it is any smaller, rather as far as the signal is concerned, the headroom around it has grown bigger. So now instead of having a Vg swing of 4V it has something closer to 6V and that converts to the increased headroom for driving the next stage, the power tubes. And, when the grid leak became part of the voltage divider it lost it's ability to be bootstrapped and so this now loads down the previous stage. But even if we consider it has lost its magical amplification factor it is a small price to pay, that particular gain stage had already had its headroom increased as a consequence of the higher HT.

The amount of negative feedback from the LNFB has a relatively modest feedback factor of approximately 1.5 or 3.3dB. Not enough to significantly alter the amount of headroom but enough to effectively filter a portion of the noise and distortion. Making for a lower sound floor and a cleaner sound with more bandwidth. The technical analysis and findings of the LNFB are dealt with more comprehensively here: https://ampgarage.com/forum/viewtopic.p ... 15#p463915.

There is one other element to the Dumble cathodyne that is not found in the 5E3, the 10k trimmer in the tail resistors.

[Stephen1966]

The main job of the trimmer below the bias resistor in TD is to tune the tone of the amp. The tail of the TD cathodyne phase splitter is comprised of the 10k trimmer, a 56k fixed resistor and a 33k fixed resistor in series. These three resistive elements have their counterpart in the single 56k resistor found in the 5E3. The cathodyne is an excellent phase splitter for low to medium gain amplifiers and when Ra = Rk, or the load at the plate matches the load at the cathode, it offers not only an economy of parts but also a low THD making it especially suitable for Hi-Fi applications. It's my understanding that Dumble's cathodyne was conceived with a quite different set of goals to Fender's. Fender, was basing his circuit on hi-fi principles, principles of balanced outputs that persist to this day. Attitudes to distortion were different and the 5E3 had it's performance issues with a tendency for blocking distortion - farting out at lower frequencies. It's not possible to say what the exact design goals were in Fender's workshop but the cathodyne possibly presents an easier phase splitter to work with and balance than the conceptually similar paraphase phase splitter of earlier models including the 5C3, and a balanced cathodyne may well have been an attempt to minimise the problems with the amp's low frequency power handling. It wasn't Fender's first foray into the cathodyne way, which I believe was with the short-lived Harvard but it's one that stuck with the 'Deluxe' model going strong to this day.

It's fair to say that Dumble's approach was informed by a different musical aesthetic. The changes he made to the cathodyne enhanced the headroom (through LNFB) and following the maths we can see he was shaping the tone in a much more purposeful way. Non-linear distortion is an inevitable and unavoidable feature of this amp, the beloved happy accident of tone in the 5E3 became a statement of intent in Dumble's revision of the circuit. It is telling that it was from this point in the signal chain, the phase splitter, that Dumble held off making any further changes to the circuit. A simple solution to the problems of blocking distortion would have been to reduce the coupling caps connecting the cathodyne outputs to the grids of the power tubes. Doing so would have reset the high-pass filter that contributes so significantly to the problem but it would also have stripped out a portion of the low end which makes this little amp sound so huge. There are degrees of acceptable though and caps of half the size would have improved the situation without a great or really very noticable impact on the low-end response.

I consider the whole. First, the common cathode bias resistor and cap are decoupled in the first stage. This doesn't remove all the interactivity of the two channels but it removes a lot of the bass frequencies that present themselves in the empty channel when its gain is turned up. Next he reduced the coupling caps after the first stage to the driver in the second stage (V2a) From .1uF to .022 - trimming the bass frequencies. The voltage is increased with the higher gain rectifier this improves the voltage swing and headroom of the preamp and in turn, reduces the propensity for distortion. He then places a LNFB network on the cathodyne which increases the headroom again and reduces the noise floor. Everything suggests he had enough faith in the power section to leave it alone but that the real improvement to the quality of the tone was in the changes he made towards cleaning up the signal in the early stages and trimming its harmonic distortion profile for the right balance of clarity and power.

[Stephen1966]

The TD's power section is remarkably similar to the 5E3 and if we figure in the interpolated voltages, this is what we get.

Deluxe-5E3---V3-and-V4---6V6GT-chart.jpg

Tweedle-Dee---V3-and-V4---6V6GT-chart.jpg

[cowboyblues]

Amazing job documenting this iconic yet mystic amp. I built a similar one last year and really struggled with voltages as you have. I started with a Modulus clone of the Triad OT and PT. The amp sounded good but had about 18 watts on the JJ6V6S. Also, noticed that regardless of how good my heater line dress was there was low hum at idle that drove me nuts. I decided to swap in a Hammond 290BEX PT and a 1750E OT. Unfortunately, the voltages went even higher with 19watts on the 6V6S. I swapped back to the Modulus transformers. The Modulus Triad clone is running 325-0-325 @ 120ma. I also changed the cathode resistor from 250 to 300 ohms. This put me spot on 14watts for the 6v6S. As for the hum, I finally gave in and installed a 250ohm humdinger pot to equalize the heater voltage while elevated. My experience was this is a very cool amp that has awesome gain and more headroom than the standard 5e3. But definitely a headscratcher to build to get voltages in spec and quiet at idle.