AB763 Stunflower

[romberg]

Having never built or played a classic black face Fender, I decided to build one for my next project. I like amps in the 50 watt range so I decided to build a vibroverb in a head cabinet. The schematic I'm using is below.



Like many of us, I could not resist the chance to tinker with the circuit a bit. The idea is to have a mostly stock tremolo channel and a more hotrod normal channel. I also wanted to more or less use a stockish circuit board. Which I have mostly done. I did add a few more eyelets to support the circuit modifications.

To start things off, I moved the mixing resistors for each channel to just before the reverb send as described by Rob Robinette here: https://robrobinette.com/AB763_Modifica ... Reverb_Mod. I like the idea of giving effects to both channels and since I'm going to make the normal channel sonically different than the trem channel, it would be nice to have them both in phase so that they can be joined with a jumper.

I was initially going to make the normal channel something along the lines of a marshall 2204. But then I saw the great bandmaster ultraphonix schematic on this site here: https://ampgarage.com/forum/viewtopic.php?f=4&t=35308. To my eye, it looks like Dumble was doing something similar so, I've made my normal channel use the ultraphonix values. But I've moved the tone stack to after the cold biased second stage. On my layout I'm just swapping the triodes used for stage 1 and 2 to make this simpler.

I also, used the ultraphonix phase inverter. It uses smaller coupling caps to the power amp, which should tighten things up. After the PI I added a PPIMV which should give some more control over the volume as well as the ability to overdrive the PI at low volumes.

The tremolo oscillator was also tweaked to match the ultraphonix schematic. Mostly it adds a couple of caps (which reduce ticking I think). I have plans to modify this to Steve Luckey's tremulator circuit. But I think I'll do this after the amp is up and running (this has now been done an included in the schemati).

Inspired by Martin's wonderful delux reverb build https://ampgarage.com/forum/viewtopic.php?f=8&t=34541, I chose to clean up the power supply a bit as well. The bias circuit has been changed to the one described by Merlin Blencowe. And I have added two additional caps to the preamp. So now the first two stages of the normal and tremolo channels have their own power supply nodes. The value of the dropping resistors were bumped up in order to maintain similar voltages to the case where all six triodes shared a single node. I have a new node for the reverb recovery tube as well. So, I now have three premp nodes branching off of the PI node C. I'm also going to run individual grounds for nodes C, D, E and F (this kinda is a mess with the stock circuit board). Nodes A and B are going to continue to share a ground as node B is used for the trem and reverb. For all I know this may be part of the blackface sound. So, I decided to leave it alone.

stunflower.pdf

Here are a couple of shots of my build so far:

guts1.jpg

boards.jpg

Here I have the heaters wired up and tested. The bias supply and rectifier also seem to function. I'm generally trying to use cloth wire for most of it. But there are some spots I chose not to. I'm running shielded cable on the first stages of each input channel as well as the runs from the normall volume and PPIVMV. I think my next step will be to install the boards and start to wire them in.

Mike

[xtian]

Bitchin'. You're going to be pleased.

Your photos remind me of Under The Skin.

[romberg]

Your photos remind me of Under The Skin.

LOL. I had some lights and a backdrop already set up for another project. So, I just took the pictures with that setup. Followups will probably be on my bench .

Mike

[Reeltarded]

Great choices! We are all going to love your amp!

[pdf64]

5k and 100R feedback network?
Even off the 16R tap, that’s a negligible degree of feedback, basically open loop.

Bear in mind that the treble peaking cap on the normal channel mixing resistor will tend to have an inverse, treble dulling effect, on the other channel.

[romberg]

5k and 100R feedback network?
Even off the 16R tap, that’s a negligible degree of feedback, basically open loop.

Leo seems to have used 820R and 100R on a 2 ohm tap. Somewhere I remember seeing a rule of thumb that advised doubling the 820R if the impedance was doubled (this may not have been good advise). So, that would mean around 6k5 100R on a 16 ohm tap. Or the advise was not good to begin with. I'm not sure if 2x the impedance is a 2x change in the voltage. It may not be.

In my case, There is a few percent more feedback as I'm using the Ultraphonix bandmaster values of 100k/4k7 (the values in parenthesis like this) on a 16 ohm tap. And of course, I'm more or less killing the whole feedback thing anyway with the PPIVMV . I may experiment a bit with the values of these after it is up and running.

Bear in mind that the treble peaking cap on the normal channel mixing resistor will tend to have an inverse, treble dulling effect, on the other channel.

I think I saw that mod on Robinett's site. I slapped that on without thinking much about it as it sounded like an easy way to make the normal channel sound different from the trem channel. I'm sure you are right about this also having an effect on the highs on the trem channel. It has made me think more about the whole circuit surrounding the mixing resistors. I guess one needs to factor in not just the mixing resistors but also the coupling caps and plate resistors of each state feeding the mixing resistors? So, the AC signal coming out of a channel sees a voltage divider consisting not just of the mixing resistors themselves but of the coupling caps and plate resistors. And the B+ feeding each plate resistor is a sort of ground for the whole thing? I think I may leave this peaker out and try clipping it in after the build is up and running. That way I can decide if the tradeoff is worth it.

Mike

[romberg]

After some real world distractions, I've found some time to get a bit more work done on it. The boards are now mounted and the front face pots, switches and jacks have been partially wired up. I've also connected all the positive connections from the doghouse cap board and the main board.

top1.jpg

pwr-test.jpg

All the power supply caps have their ground side connected under the cap board and I'm running an individual ground wire for the C, D, E and F nodes. So, for now I've just connected the A/B ground connection to the main board. The next task will be to connect all these ground wires and make sure to connect all the front and back component grounds to the proper node.

But for now, I can test the power supply rail and see if it holds voltage or has any shorts to ground. There are too many extra wires flying around right now for me to feel safe taking it up to full voltage. But I did hook the rail up to an external 24v power supply and things look good.

pwr-test.jpg

I tested the rectifier earlier so I have some confidence that the power supply looks to be in good shape. So, next I'll wire in the grunds and signal connections and it should be ready to fire up for real.

Mike

[romberg]

Got a little bit more wired up. Both the tremolo and hotrod channels are now wired up along with the reverb driver.

guts2.jpg

I suppose that at this point one could probably plug a small 8 ohm speaker into the reverb transformer secondary and have a low power amp to test out. But there are still too many dangling wires for me to be that brave.

Instead I set up a low cost signal generator to log sweep a sin wave from 50Hz - 8kHz and display it on my oscilloscope to make some crude bode plots. I learned this trick on the EEVBlog youtube channel and it works kinda nice. I'm injecting signal before the tone stack and can see the effect on the scope as the tone controls are adjusted.

bode-setup.jpg

Here are the bode plots of the marshall and fender tone stacks. You can see the fender being more scooped in the mids (as well as having a much lower output (I had to adjust it up quite a bit to fit on the same screen as the marshall one). Here channel two is just displaying the sync pulse comming out of the signal generator. So, 80Hz is at the start of one cycle and 8khz is at the end. All controls set on 5.

bode-marshall.png

bode-fender.png

Mike

[romberg]

Got it all wired up. It works! Only trouble I had during start up was a bad solder joint in my filament string. These EH 12ax7s just don't glow enough and I missed this when testing earlier in the build. The fact that I was getting voltage on the plates but no current through v1 and v2 quickly lead me to the culprit.

Here are a couple of shots of the top. I left some notes for myself under the doghouse as all the grounds are tied together there under the board.

top-1.jpg

top-2.jpg

Here is a picture of the front and back. The PPIVMV is the extra knob on the front. On the rear I added bias jacks. Also the ground switch is not connected to anything and I ran each tap of the 4-8-16 ohm transformer to an individual speaker jack. I did not think I had enough room to wiggle in an impedance selector switch.

front-1.jpg

back-1.jpg

Finally here are the guts.

guts-3.jpg

At this point I'm close to done. Everything works and there are no extra noises to track down. The tremolo channel sounds pretty dang good and can get nice and crunchy starting about 5 on it's volume pot. However my modified normal channel is a bit of a let down.

It also works. And does noticeably sound different than the blackface channel (more midrange punch for sure). However it is not putting out as much crunch/gain as I would have hoped. In fact at five on the dial it puts out about half the peak voltage as the tremolo channel on 5. I'm not expecting to get trainwreck express levels of gain/distortion out of just two stages. It does start to crunch a little bit when cranked to 10. But I was shooting for a little more.

I think the culprit may be the marshal speced tone stack that I have driven with the cold biased (10k) and unbypassed second stage. This stack is probably too low impedance for that to work without a cathode follower. My guess is that I'm not getting it to clip as much because it is heavily loaded down by this stack. So, I'm thinking of experimenting with some of these ideas:

• Bypass the second stage. Maybe with 0.68u (marshally) to 25u (ultraphonixish)
• Monkey around with the bias of this second stage. Either go colder (ala SLO) or hotter. See if hot or cold clips easier than the other.
• Take out the 470k divider before the volume pot. It is certainly knocking down some gain. But only about 30% or so. Might still need this if bypassing increases the gain too much.
• Increase the 33k slope resistor to 47k like the ultraphonix stack. This may be easier to drive but also sound more fendery.

I would welcome any ideas other may have on

Mike

[sluckey]

Maybe consider this idea...

Convert your AB763 tremolo circuit to the Trem-O-Nator circuit. Better sounding trem plus you free up V5B to possibly use in the Normal channel.

[pdf64]

…
Bypass the second stage. Maybe with 0.68u (marshally) to 25u (ultraphonixish)
Monkey around with the bias of this second stage. Either go colder (ala SLO) or hotter. See if hot or cold clips easier than the other.

I don’t think you’ve got sufficient gain stages to make one of them unbypassed cold biased.
Hotter bias and at least partial bypass of the second stage should help to get some preamp overdrive.
Bear in mind that the clipping (3rd) stage of a 2203/4 uses a 820R cathode resistor; stock it’s unbypassed but a very common mod is to bypass it.

Increase the 33k slope resistor to 47k like the ultraphonix stack. This may be easier to drive but also sound more fendery.

The slope resistor + the mid resistor is load on the preceding stage, 55k is a heavy load for that unbypassed cold biased common cathode stage, its gain will be reduced.

[romberg]

Maybe consider this idea...

Convert your AB763 tremolo circuit to the Trem-O-Nator circuit. Better sounding trem plus you free up V5B to possibly use in the Normal channel.

I'm very seriously working over the details of retrofitting a Trem-O-Nator. Thanks for putting togeather all the wonderful docs on it!

Besides being a better sounding tremolo I really really like the triode it frees up. It is problematic that the available triode is located so far away from v1 and v2. So, would probably require a bunch of shielded cable runs to use early in the preamp. But...

This whole thing seems similar to the modification Tim Caswell did on the marshall 1959T (S.I.R. #39). There the tremelo pedal was lost so he stole the fourth tube to add an extra gain stage before the tone stack. In my case the 3rd gain stage on the reverb recovery tube is sitting right there. It has me thinking. But I think I'd need to make a fouth gain stage switchable. So, then a relay? It is tempting me alot...

Mike

[romberg]

I don’t think you’ve got sufficient gain stages to make one of them unbypassed cold biased.
Hotter bias and at least partial bypass of the second stage should help to get some preamp overdrive.
Bear in mind that the clipping (3rd) stage of a 2203/4 uses a 820R cathode resistor; stock it’s unbypassed but a very common mod is to bypass it.

You are totally right! A quick experiment yielded a (good) result that surprised me. I'm studying Merlin's preamp book again. It seems bypass caps dpn't just roll off the bottom end (which I had mistakenly thought). Unbypassed cathode resistors also add to the overall amplification and output impedance. Which makes my little stunt of driving a marshall tone stack with an unbypassed cold biased stage even more silly. I'll make another post once I better understand the results I got by dropping a simple 0.68uf cap across the 10k cathode resistor.

Mike

[pdf64]

… I'll make another post once I better understand the results I got by dropping a simple 0.68uf cap across the 10k cathode resistor.

That alone should have made quite a difference ?!

[romberg]

… I'll make another post once I better understand the results I got by dropping a simple 0.68uf cap across the 10k cathode resistor.

That alone should have made quite a difference ?!

Oh it did! You see the whole subject of input/output impedance used to make my eyes glaze over. So, I mostly thought bypassing the cathode resistor just rolled off the bass response. Reading Merlin's chapter on this subject again with my recent experience in mind really opened my eyes. The big thing is that the resistance of the cathode resistor is added to the internal plate resistance of the tube when you consider both the amplification factor and the output resistance of the tube. But it is not simply added. First it is multiplied by the mu of the tube. So the cathode resistor of an unbypassed triode really has a large effect Especially when the load on it is relativly small.

I wrote some of Merlin's formulas down into a simple python script so I could play around with the numbers:

Code: Select all

u = 100             # 12ax7 amplification factor.
ra = 81000          # 12ax7 plate resistance.
Ra = 130000         # plate resistor
Rk = 10000          # cathode resistor
Rl = 55000          # load
Rku = Rk * (u + 1)  # extra unbypassed plate resistance.

# Ampification factor (bp = bypassed, un = unbypassed)
Abp = (-u * Ra) / (Ra + ra)
Aun = (-u * Ra) / (Ra + ra + Rku)

# Output impedance.
Zbp = (Ra * ra) / (Ra + ra)
Zun = (Ra * (ra + Rku)) / (Ra + ra + Rku)

# Load factor (amplitude lost from output/load divider).
Lbp = Rl / (Zbp + Rl)
Lun = Rl / (Zun + Rl)

# Total Amplification.
Tbp = Abp * Lbp
Tun = Aun * Lun

print("( Ra:", Ra, "Rk:", Rk, "Rl:", Rl, ")")
print("bp A: %.1f" % Abp, "Zo: %.0f" % Zbp, "Lf: %.2f" % Lbp, "T: %.1f" % Tbp)
print("un A: %.1f" % Aun, "Zo: %.0f" % Zun, "Lf: %.2f" % Lun, "T: %.1f" % Tun)

The output for the values I'm using (I kinda guessed at the impedance of the tone stack):

( Ra: 130000 Rk: 10000 Rl: 55000 )
bp A: -61.6 Zo: 49905 Lf: 0.52 T: -32.3
un A: -10.6 Zo: 116159 Lf: 0.32 T: -3.4

And this is what I saw before and after clipping in a 0.68uf bypass cap:

unbypassed.png

bypassed.png

Not quite 10x as the above formulas would predict. But the cold clipper is now kicking in so a good chunk of the waveform has been lopped off. The tremolo channel is putting out about 1.2Vpp at the same volume setting. So, the peak to peek is nearly the same. I may mess around with some different bypass cap values and cathode resistors to tweak to taste. But my normal channel is now starting to sound mean around 4 on the dial.

Mike

P.S. For fun here are some of my scripts outputs for some more typical values:

( Ra: 100000 Rk: 1500 Rl: 1000000 )
bp A: -55.2 Zo: 44751 Lf: 0.96 T: -52.9
un A: -30.1 Zo: 69925 Lf: 0.93 T: -28.1

( Ra: 100000 Rk: 1500 Rl: 150000 )
bp A: -55.2 Zo: 44751 Lf: 0.77 T: -42.6
un A: -30.1 Zo: 69925 Lf: 0.68 T: -20.5

( Ra: 100000 Rk: 39000 Rl: 1000000 )
bp A: -55.2 Zo: 44751 Lf: 0.96 T: -52.9
un A: -2.4 Zo: 97573 Lf: 0.91 T: -2.2

Super cold bias (ala the SLO) would be a really poor choice for a plate driven tone stack