Amp Project - About to Begin Build

[steeledriver]

Hi guys,

After months of pouring over ideas from TAG, EL34World, and here, and reading Merlin's excellent book many times, I've put together a design for a 40W 2x6L6GC amp that has an effect loop and mixer (a la Merlin), tube driven reverb (a la Channel Road), and an EQ from Dumble amplifiers, except with the master volume moved out of the EQ and placed between V1A and V1B. I'd like this amp to be mostly clean, but break up when you dig into a note.

I've attached the schematic. This is my first build, so I'd like to know if anyone can spot any glaring problems with the design.

I, for one, am concerned that I might just have too many gain stages (for a clean amp). V1A and V1B are series gain stages with a volume knob in the middle, and I'm not sure how lossy the EQ will be, so I might be sending quite a large signal to the FX send and dry channel of the mixer - enough to overdrive it even at low volume. If that's the case, I might be overdriving the input of my PI and the input of my tube reverb driver (which has about 4Vpp of clean headroom). My options would be to change the pre volume to a dual taper pot and put the second pot after the EQ to attenuate even further, or drop the gain of V1A and V1B by either switching them out for a 12AU7 or using local NFB like a Dumble.

Food for thought. I appreciate all of the fantastic work that goes on on this site, and I want to get the input of some of the experts here before I move forward with the build.

Thanks in advance for your time and advice.

Dave

[Chuck H]

I don't get the tone stack. It looks like a variation on the Dumble stack, but the differences might make it less effective as a tone stack and more like some alternative voicing control.

The treble mid boost bypass value seems large and will probably pass too much LF.

The parallel mid/bass arrangement with the large mid pot will make the mid and bass pots behave pretty much the same.

The bright switch probably won't be that effective without a smaller load on the downstream side. That series resistor depends on that load for attenuation. If the resistor doesn't attenuate much, the bridged cap can't make much difference.

I think you're right that the effects loop will be too hot.

Boy that's a lot of power supply nodes! You surely don't need that many for filtering, decoupling. If you pair up out of phase stages on nodes you're pretty safe WRT interactions.

Very ambitious project considering the fine tuning and troubleshooting you'll encounter by designing that far out of the box.

[pdf64]

Agree; also do you really need presence, resonance and feedback level?
It depends if this is intended as a flexible prototyping blank canvas for trying out all the variables, or to be a usable amp that won't be too confusing when you need to tweak something whilst being distracted by making music.
The PI may be missing an input dc blocking cap.

[Chuck H]

The PI may be missing an input dc blocking cap.

Yes it is.

[Malcolm Irving]

Won't C8 and C16 block DC from the input to the PI?

Edit: Ah, sorry guys - you are talking about blocking the DC from coming back from the PI to the post volume pot - which would make it sound scratchy and upset the bias of the PI. My mistake.

[steeledriver]

I don't get the tone stack. It looks like a variation on the Dumble stack, but the differences might make it less effective as a tone stack and more like some alternative voicing control.

The treble mid boost bypass value seems large and will probably pass too much LF.

The parallel mid/bass arrangement with the large mid pot will make the mid and bass pots behave pretty much the same.

The bright switch probably won't be that effective without a smaller load on the downstream side. That series resistor depends on that load for attenuation. If the resistor doesn't attenuate much, the bridged cap can't make much difference.

I think you're right that the effects loop will be too hot.

Boy that's a lot of power supply nodes! You surely don't need that many for filtering, decoupling. If you pair up out of phase stages on nodes you're pretty safe WRT interactions.

Very ambitious project considering the fine tuning and troubleshooting you'll encounter by designing that far out of the box.

Sorry about the late reply - I was at work all day. I had some time to think about your response.

You're absolutely right, the treble/mid-boost bypass cap is too large. I'll replace it with a 500pF cap and see how that sounds in the build. I swapped out the mid for a 100k pot and the bass for a 250k pot.

To solve the high gain problem, I've replaced the 1M pre-level pot with a 500k dual wiper pot, and placed the second pot in the tonestack. This was to a) help load V1A more heavily and reduce its AC gain, and b) have an extra step of attenuation, and a place where I can move the bright cap instead of on the 220k bridge resistor. The second step was to replace the 100k loaded 12AX7 (AC gain of 50ish by my load lines) with a 20k loaded 12AU7 (AC gain of 12 ish by my load lines). It'll be harder to overdrive the 12AU7 since it'll be biased at -6V, and produce less gain for the phase inverter, meaning I can use less feedback. It also lets me test out a NOS Rogers 12AU7 that I have. All around good. The last step was to take the input for the 6GH8 reverb driver pentode from the cathode of the effects mixer instead of the plate. Since it has such a high input impedance, it won't load the cathode and should provide a nice, smaller signal I can use.

PDF and Malcolm - thank you for noticing the post-level coupling cap! I had it in my first iteration, and later couldn't remember why I had put it in since C8 and C16 were blocking the HV, so I removed it. I forgot about the tail voltage.

I'll attach a revised schematic shortly.

Great catches!

[MarkusBass]

I think that there are still (steel) mistakes on the schematic.
There are two VR13 trimmers (first valve). I don't think that trimmer will work correctly there (especially after several years). And value 1.5k is not typical. Maybe 1k will be better? Dumble used resistors there - why do you need trimmers?
The resistor R45 (1M) is too low. It should be 3.3M, or even 4.7M.
The Deep switch doesn't look correctly. The R42 resistor should be rather parallel to C21 - please double check your source schematics.
Missing capacitor on the input of the power amp -> scratchy Volume pot.
DC on the Scale pot (1.5k ???) -> scratchy pot.
The Presence pot - 30k - quite unusual value.
The Mid boost switch was usually wired differently - the capacitor was permanently in the circuit and the switch was shorting it.
In general, the schematic is overloaded with neat ideas (Resonance, Ground Lift, Feedback Adjust) but at the same time it contains basic errors. You should at least simulate it to make sure that gain and frequency response is as expected.

Mark

[steeledriver]

I read a general rule of thumb once of "no more than two inverted gain stages on a single capacitor". I took that to mean one tube - one capacitor, unless there were cathode followers, and decided to give each tube a node to play it safe. You think I could get away with two pairs of out-of-phase stages on each node?

And PDF, you're correct - this is more or less a flexible amp for trying out all the variables. It's a personal project, and it doesn't need to be user friendly as long as it's friendly to me.

[steeledriver]

Thanks for your post, Mark. I'll take another look at the Dumble EQ. My source schematic for this one used a 1M - but I've found disagreeing schematics. I've heard some sound clips of the Rock/Jazz switch, and I'll probably throw that in too.

I put the trimmers in to give me the flexibility to change my headroom without resoldering - that way I can swing the bias resistance from 820 ohms to 2.3 kohms (beyond what I'll need). I had considered that they may be noisy - probably best to play it safe, especially if time will kill them.

The switches in Dumble EQs are DPDT, but don't necessarily have to be - I have a bunch of SPDT switches lying around, so I attempted to shuffle the connections to make those work. I'll double check.

I don't think I can keep DC off of the Scale pot, because it's part of the DC biasing resistance. I know it'll be scratchy, but I haven't come up with an alternative yet.

[Chuck H]

Won't C8 and C16 block DC from the input to the PI?

Edit: Ah, sorry guys - you are talking about blocking the DC from coming back from the PI to the post volume pot - which would make it sound scratchy and upset the bias of the PI. My mistake.

Actually, in this case the DC blocking cap needs to be there to prevent DC from getting OFF the grid. The LTP circuit has the grids biased on the top of the tail resistor. Having another resistive path to 0V upsets the bias condition.

[Chuck H]

If you change C24 to 500p and C26 to 1n you'll get a net treble circuit capacitance of (BING BING BING BING!!!) 333p. That's damn close the 330p value that the Dumble uses. And your boosted circuit will be 1n. Plenty big enough for a fat switch and not too fat so the treble pot becomes a volume control.

For the bright circuit you need to figure the voltage division. Right now you have a 220k series (that the cap bypasses) and a 1M load. If you have, say, 10VAC going into this circuit (which you won't but it makes for easier calculation ) you'll get 8.2VAC out. That means the bright cap is only operating on less than 2/10 of the signal. Sort of like the effectiveness of a bright cap on a volume circuit, but the knob is stuck on 8! So, not very effective. You need to change the series resistance or the load in order to make this more effective. This would, or course, also mean a subsequent loss in net signal level. Which should be ok since you're headed for an effects loop anyway and you have a little leeway. In fact, a 50/50 voltage divider might get the signal into your CF to the right level for an effects send. You want about 1VAC effects send at max volume. So you may need to simulate this. If you can increase voltage division further it will only add effectiveness to the bright circuit.

[steeledriver]

I'm having trouble attaching a new PDF, and I don't think I can edit my original post, so I've put a picture of the EQ up here. Chuck - I followed your advice for C24 and C26. I had originally made C24 1nF and C26 500pF after the suggestion in your first post, but by switching the two, the mid boost will be more prominent. I've added the Dumble Rock/Jazz switch to the EQ, where the Rock setting is a short to the original circuit, and the Jazz is a 0.047uF shunt capacitor to ground and a 100k series resistance leading to the bass pot.

I also changed the pre volume pot to a 500k dual taper - to load V1A more heavily and reduce its AC gain, and so that I could put the second taper in the EQ. This creates more division for the bright cap and adds a second step of attenuation before the CF - helping to solve the gain problem. I changed V1B to be biased around -0.9V with an 820R resistor so that V1A can overdrive the valve a bit when the pre volume is turned up.

Mark - as for the value of the presence potentiometer, you're right, it's a little unusual. I calculated the amount of feedback I wanted using 47k/3k feedback resistors, which are also non-standard. Merlin says in his book that the presence pot should be much larger than the feedback resistor it is in parallel with so it doesn't affect the feedback calculation, so I chose a value 10x higher. With the 0.15u cap, I get a -3dB cutoff frequency of about 370Hz. It's unusual, but it could work.

[MarkusBass]

There is much simpler circuit that you can use for the Mid boost. Just connect the C24 cap to A and B lugs and just short these two points with a switch. The idea behind this (and the R45 resistor) is to avoid pop sound when you use the switch. I've seen versions where the C24 cap is 1n but I would use no more than 390 pF (at least for the start).
I don't like the idea of using both C20 and C42 capacitors (both 500 pF). I think that Dumble used 270 pF as the bright cap but I don't recall using both circuits together (especially with both caps 500 pF). Or, to be precise, he used 500 pF as a bright cap but it was never combined with the other cap.
I don't understand the Presence pot. It is true that Merlin says that the pot should be higher than the cathode resistor so it does not influence the calculated amount of feedback but he doesn't say: "multiply it by 10". Instead, he says "multiply it by 10 and choose the first available value (e.g. 47k)". Anyway, the pot has variable value. How are you going to solve the problem? Using parallel resistors, or you order unusual value of the pot from factory?
I see that you corrected the Deep switch circuit. Now it looks OK.

PS: what software did you use for the drawing? I could try to draw at least the Mid boost circuit correctly.

Mark

[steeledriver]

Hi Mark,

Looking at a few different Dumble circuits, you're correct - he either used the 500pF bright cap across the volume pot, or a 390pF cap across the 220k resistor, but never both together. I removed that cap from my schematic. As for the mid boost cap, I'll start with 330pF because I have one lying around, and move up if I find there's too much cut/boost.

I used the student edition of PSpice Capture for the drawing. I've modified my schematic to have a DPDT switch in it for the mid boost, with one direction shorting the mid boost cap (attached). Is this what you were suggesting? Also, the 270k resistors in series with the deep switch - their purpose is to suppress switch pop too?