Debug that drove me to the brink of madness

[pullshocks]

I have been working on a test bed for the 12DW7 transformerless reverb circuit provided by 2Deaf in another thread


A schematic is attached.  It is based on the familiar Hoffman Princeton Reverb no tremolo amp circuit available here on el34world.  I cut and pasted 2Deaf's driver in place of the Fender transformer type driver.  The recovery circuit is the standard Fender.  2Deaf's driver needs a 350 volt B+ so I modified the power supply to provide a 350 volt node, and separate nodes for the other preamp tubes. I also added a middle control, changed the power amp to cathode bias, and added the LarMar PPIMV.


I had an idea of how to do a bread board in an existing gutted amp chassis I had.  It is the bastard love child of a turret board and a terminal strip--plastic with 6-32 tapped holes for the connection points.  It turned out to be pretty fussy and cumbersome to work with, and I have to be vey careful to assure there are no unintended connections of component leads.  And it is just plain hideous as you can see in the pictures.


On the good side, it made it easy to disconnect the reverb circuit and test the basic amp circuit first.


No fuses blew, and the light bulb limiter test went OK.  But being in the category of "knows just enough to be dangerous" I went down the rabbit hole of checking voltages while the amp was still plugged into the limiter.  Several hours spent chasing low heater and B+ voltages before the, uh, light bulb went on (rim shot please) that there is a reason they call it a "limiter."


When I turned it on without the limiter, no fuses blew, but it didn't make any sound either.


Suspect # 1 was the kludged up nature of the construction.  Every component was checked and rechecked, traced against the schematic, tubes verified to work in other amps, speaker verified, connections, cleaned, cathode resistor changed.  Hours and hours.


I started consulted the Fender AA1164 schematic and found some of my voltages were high.  Power supply was revised accordingly


I finally started to make some progress when for some reason I tried to take a voltage reading on the grid of V3a.  Not a check you would normally do.  But the amp began making sound.   For a few seconds of playing.  Then it fritzed out again.
I thought it must be a tube socket / pin 2 issue......another rabbit hole.


Eventually I realized it was not a mechanical contact issue, but something else related to the grid of V3a.  I found I could repeatably get the amp to play for a few seconds by touching the V3a grid with the voltmeter probe (other probe grounded).  Then I found if I turned the volume waaaay down, I could play for extended periods, but it would fritz out as soon as I turned up the volume more.


With no guitar plugged in, turning the volume up caused squealing feedback type noises.  Which made me wonder if the feedback circuit / output transformer primaries were hooked up wrong.  Fortunately, those connections were kind of hard to get to, so I did not go down that rabbit hole....


What I finally figured out after reading Merlin's preamp book was that when I disconnected the reverb, the grid of V3a was no longer referenced to ground.  The breadboard made it easy to patch in a 1 meg resistor, and the world suddenly became a better place.  This thing sounds GOOD.  How close to a real Princeton?  I don't know, the closest I ever had was a silver face Deluxe Reverb, and it has been 6 or 7 years since I sold it.  But this build does have a certain richness and fullness to it.


Stay tuned, I will report soon on the transformerless reverb circuit.

[PRR]

> for some reason I tried to take a voltage reading on the grid of V3a.  Not a check you would normally do.

Actually, from bitter experience, I like to know ALL grid voltages. Sure, they are usually zero. And if the amp plays happy I would not poke grids. But an unhappy amp, IMHO, begs for grid-checks. It is the most sensitive part of the tube.

Some grids can't be checked. Self-biased cathode followers and long-tails will un-bias due to meter loading. Then go by the other electrode voltages making sense.

[shooter]

Self-biased cathode followers and long-tails will un-bias due to meter loading.

could you go a little deeper on this, scopes also?, guessing "invalid" measurement, or actual tube loading that causes the "device" to change tube static operation?

sorry for the momentary hijack.

[sluckey]

Both circuits have a very high input impedance due to bootstrap biasing. The impedance is so high that even a 10MΩ DVM will load the circuit down and shift the bias voltage. Look at this schematic, in particular, V2, a typical LTP PI. Now look at the voltage chart. The grids measure 48V and the cathodes measure 67V. What? How can that be? That tube should be in hard cutoff with the grid sitting at 29V less than the cathode.

     http://sluckeyamps.com/18w/18w.pdf

Well it can be because the LTP PI uses bootstrapped bias and the grid impedance is so high because of the bootstrap that when you connect a meter to the grid, the meter significantly loads the voltage down. The meter shows only 48v but that's impossible. The real voltage is probably only a couple volts less than the measured cathode voltage.

The same thing happens with a self biased (bootstrapped) cathode follower. Look up bootstrap bias. I think Aiken has a good read on it. Merlin probably does also.

[PRR]

> grids measure 48V and the cathodes measure 67V. What?

Right. A difference like that is typical, though with wide variation. At first look "impossible"; you explained the Why.

So a grid-voltage check is useful if you know the meter-loaded voltages; and semi-useful if you can predict what would be "bad". Here, zero voltage suggests a mis-wire, and simple trace-out is quicker than theory.

BTW, OT-- mis-wiring is everywhere. I have a backhoe. Many old hoses. I burst my 3rd ($100) hose, so I replaced 5 in that area. Now the dipper and bucket interact (seem cross-wired). Indeed it looks like I put the hoses back in the right order, but in the (long drawn-out) process, managed to swap two pipes inside the arm. I can't see up that 1967 arm, but my 2016 cellphone camera peered up there and shows pipes not going where they should go. Argh, another 5 gallons of fluid to spill.

[HotBluePlates]

... V2, a typical LTP PI. Now look at the voltage chart. The grids measure 48V and the cathodes measure 67V. What? How can that be? ... The meter shows only 48v but that's impossible. The real voltage is probably only a couple volts less than the measured cathode voltage. ...

What I usually do here is measure the tube-end of R13 and write it down.  Then measure the non-tube-end of R13, write it down.  I assume the LTP grids are sitting at the same voltage as the non-tube-end of R13 (which will usually be higher then the loaded-down measurement of the actual grids).

Either way, R13 is the self-bias resistor, and the total voltage across it is typically the bias for that tube.  The only drawback to this is it assumes neither C1 nor C18 leak d.c.  That might be best checked by quick replacement, or observation of very-wrong voltage for the non-tube end of R13.

[sluckey]

I assume the LTP grids are sitting at the same voltage as the non-tube-end of R13

That's exactly what I started doing with my later stuff...

     http://sluckeyamps.com/dual_lite/dual_lite.pdf

[shooter]

R13

Thanks guys, R13 was my ""missing link", I was just lookin at R9 n 17 in || with a 10M meter, just like V1's R3.
I'll go back n re-read Merlin.