Improving the Fisher SA-100 with EFB II

[notben]

Hi Dave,
I finally got around to pulling the SA100 apart and installing the ESBII circuit. Although I could use some diagnostic help.

The screen supply is only giving me 150vdc.
For the MOSFET: I have 400+vdc on the Drain (2 red), but I am only getting 150vdc on the Gate(1 green) and the Source (3 black).

I have solid B+ of 400-445vdc (I have it plugged into my variac to check voltages) feeding the circuit, then the 91k resistor to the Gate with a 12v Zener then connecting that to the Source/Screen.

Also, I sanded a few areas and installed ground lugs and then connected all the grounds to new buss wire. Signal on one and the PSU on another meeting up at the ground on the input jack.

Do you see anything obvious that I am missing?
Excuse the messy lead dress. It was such a bear taking everything out. It is such a lovable rats nest.

Thanks for any help you can offer.

-Ben

 

[dcgillespie]

Hi Ben -- Barring a defective MOSFET, I would concentrate on the wiring/components of the Gate lead. Basically, the Gate resistive divider network applies about 75% of the main B+ source to the gate. Therefore, the voltage at the Source should be about 75% of the available main B+ voltage. If your variac adjust main B+ is 400 vdc, then you should see about 300 vdc at the Source terminal.

The fact that you are seeing much less implies something is dragging the gate voltage down. If that is the case, and both Gate resistors are on spec, then the Gate cap you are using becomes the next suspect by way of excessive leakage. Also, don't rule out a Zener that has been marked backwards -- I have such an animal in my junk box!

I can't see all of your wiring due to the Gate cap blocking some of your shot, but what I can see looks correct. Of course, until this issue gets resolved, the output of the EFB bias supply regulator will also be (should be) very low as well.

I hope this helps -- keeps us posted!

Dave

 

[notben]

One thing I have learned when using silicone FETS or diodes is to buy extra; and to wear a static discharge wrist strap. I will take that one out, and the Zener to test them. I'll go easier on the leads of the MOSFET also. I do heat sync the leads when soldering, but I can lower the temp of the iron.
Thank you.

 

[notben]

Dave, I have a question about this quote of yours from a related thread:

Regarding the grounding straps, I'm not sure if my pics showed it or not, but I did reinstall them in the finished amplifier, but not in the manner you might think...Therefore, if the pics didn't show it, the original straps were reattached on each output strip (where I made the test points available) -- not to ground the test points, but to connect the test points back together again in each channel after the adjustments are made. Therefore, during the adjustment session, each tube stands on its own resistor, but during operation, the cathodes in each channel are connected together, grounded through a common 5 ohm cathode resistor. This approach allows for maximum performance from the output stage, allowing proper setting of the quiescent current in each tube

I don't quite follow how you did this. Did you run four wires from the X and COM posts to the A&B test points? (The original 2 and then 2 more to tie all 4 together at the test point during operation.

Turns out the MOSFET was bad, probably static.

 

[dcgillespie]

Not -- Glad you found your problem with the EFB screen grid regulator. I was comfortable that it was either going to be a defective mosfet, or possibly a leaky gate cap. In any event, I'm glad you've got it resolved.

In the modified SA-100, the original cathode wiring and cathode resistors were completely removed, with the original three terminal test point strip then left unused. Also, all of the wiring associated with the Com and X terminals on the output terminal strip of both channels was also removed.

Then, each output tube cathode terminal (pin #3) had a close tolerance 10 ohm resistor connected between it and ground. Also, the two output tubes for channel A had a wire run from each cathode terminal, with one lead going to the Com terminal, and the other going to the X terminal (it doesn't matter which goes to which) on the output strip for the channel served by those tubes. The same thing was done in the other channel. Finally, the two shorting straps originally used on the original 3 terminal test point strip were removed from that strip, and installed on the output terminal strips, so that when each one was strapped, the Com and X terminals of each channel were shorted together.

With this configuration, when a strap is uncoupled, then the Com and X terminals act as two independent test points in that channel to allow for individual measurement of the cathode current in each tube of that channel. When the strap is reconnected after adjustments are made, then it shorts the two cathode connections in each channel together, producing maximum performance. The other channel works in exactly the same fashion.

I hope this helps!

Dave

 

[Wharfcreek]

Nice to know........... WC

 

[notben]

That is pretty much what I thought. I put new wiring from each cathode to the respective X and COM terminals and then dropped the 10ohm resistors from there to the single ground point.
What confused me was tying them back to the original test terminal; because both X and COM for channel A would run to the A test point and the same thing for B. I was confused thinking that would throw off the impedance since X and COM would be strapped together via the leads to the A test point.

I knew from your explanation that A and B would benefit being tied together, thus lowering the IM distortion, but I kept second guessing that all four X and COM points needed to be run to the original A and B test points.

I have some new 1/4w .1% resistors on the way also for the output rK instead of the 1w wirewounds I have there now: I didn't have any other 10ohm. I bought a bunch to keep in the bin.

Thanks for everything. All I have left is to balance the bias as R52 and R54 are dropping too much voltage for me to get -11v on the control grids. Right now I can only get -9v and the bias is way to hot.

-Ben

 

[dcgillespie]

Not -- The voltage supplied to R52 and R52 is now provided by the EFB control grid regulator -- which itself takes its que from the screen grid voltage supplied by the EFB screen grid regulator. This is the very heart of EFB operation, where in the control grid voltage tracks relatively with the screen grid voltage, which itself tracks relatively with the plate voltage. Keeping all the voltages locked together relatively is what delivers such an increase in performance.

Design of the EFB modification is such that the original quiescent conditions are largely left in tact -- except of course for the reduced current draw now possible because of the installation. But this current draw should be available within the range of the original DC bias and balance control design -- with or without the EFB modification installed. Since it is installed however, check that under quiescent conditions that the voltage supplying these two resistors is about -28 vdc -- which is also about the same voltage as supplied at this point by the original design. If it is low, then check that your screen voltage is on target. If all is well there, then check the tolerance of the resistors associated with the EFB control grid regulator. I had no problem achieving the proper current flow with the original values of R52 and R54 installed. Let us know.

I hope this helps!

Dave

 

[notben]

I definitely have something else going on. The plate voltage is stable at 400vdc and the screen voltage is stable at 300vdc.
The bias supply puts out -29vdc to R52&54. I replaced those with some 1% 4.7k resistors and get the same problem with current draw in the 50ma range.
The 10k resistors off the bias control to ground measure fine and I have 180uf of filtering on the sweeper (couldn't be too much? could it?)
I checked all of my grounds and they seem good.
-Ben

Edit: Now I am a little worried. I took out all of the tubes and checked my resistances again. Everything was fine on the 7247's and almost all the way around on the power tubes until I got to the plates. Pin 7 on the 7189's and pin 8 on the 5ar4 measure in the Mohms. Did I fry my OT's? With the tubes out I should get around 250ohm.

 

[dcgillespie]

I wouldn't worry about the OPTs just yet. If you're getting -29 vdc, then check the 180 uF caps for excessive leakage. That magnitude of capacitance is hardly necessary, as the EFB regulator provides substantial filtering itself. 10 to 25 uF is just fine at the wipers.

Dave

 

[notben]

Again, thanks for you help. Hopefully I will get this working because I know the amp well enough to hear any changes. It has been my main amp for a good 6 years.

I changed the bias caps to 47uf and still get the same problem.

I am going to go through the entire grounding scheme again with a fine tooth comb. The fact that I am getting proper resistances on everything makes me think the loads on the screen are fine. I replaced all out of spec resistors in the AC and DC balance circuits, screen resistors and plate resistors.

I think I'll go check all the coupling caps for leakage and the coupling caps in the balance circuit. I keep forgetting about those.

 

[dcgillespie]

I was going to suggest the same thing. Where are you only able to get -9 volts at? If it's on the DC balance control side of the grid resistor, that implies a problem with the bias circuit. But if it's on the coupling cap side of that resistor, while a much greater negative voltage exists on the balance control side, then the coupling caps are leaking.

Let us know!

Dave

 

[notben]

Well, I get -29vdc at r52 and r54;
then that is when things go haywire. I can get -9 on the other side of r54 with the A bias pot all the way down and then also -9 on the center lug on the wiper that feeds the DC bias for Channel A.
However, on channel B I just realized that the bias was all the way up with -18v on the first lug and -9vdc on the center. When I turn the bias down on B, the current draw goes shoots up into the 60's on both channels and the tubes start to red plate.

I have gone over every connection in the AC and DC balance circuits, the negative feedback tests fine, but here is one odd thing I found: the ground wire that connects the DC balance circuit at the ground of the 47k resistors R18,22, and R35, 37 connects to the ground lug on the output speaker terminal. I didn't connect this wire. Should this connect to the output cathodes?

I think I need to go through all my grounds again. The amp ran fine before adding the buss and maybe there is a cold solder joint somewhere.

It's frustrating, but also fun because I can't wait to have it running cooler and with better screen regulation considering how hard this circuit is on 7189s or the most robust EL84's. I would love to run my nice quad of Dutch rx1 D getter el84s that I got from Andy Bowman years ago. Plus, I finally have a slight grasp of the balance circuits and think I may be able to fine tune the AC balance better when it's stable.

 

[notben]

To answer simply, I'm getting 170vdc on pins 1 and around 110vdc on pins 3 and then getting -7vdc after the coupling caps.

 

[dcgillespie]

Not -- your EFB control grid regulator wiring looks good, and is supported by the -29 vdc you are getting at the point supplying R52 and R54. If you can, post some pics of the remainder of the bias distribution circuit -- i.e., the bias and balance controls themselves, and the grid wiring of the output tube control grid circuits.

Dave

 

[notben]

Here are some closeup pics. Some of the solder joints that look shiny to the naked eye, look pretty ugly when I look at the pics. Especially where I de-soldered some huge blobs to test some resistors.
I still get good resistance readings, so something is out of whack for sure.
I always have the last resort of stripping the chassis and doing a total rebuild :sigh:

 

[dcgillespie]

Not -- It is hard to tell, but it appears that the caps you have installed at the bias control wipers are installed backwards. Most such caps contain a white strip to indicate the negative terminal, while blank side is positive. It appears that the blank side is connected to the wipers -- except that the voltage here is negative, meaning that if so, they are installed backwards. On the other hand, some of these caps use gold printing, making it hard to verify polarity from a pic. Could you please verify this?

Dave

 

[notben]

Oh yeah. That's it. When I was taking out the ground buss and switched out the 180uf and put those Cerafins in I ran negative to ground forgetting bias has to be reversed.

 

[larryderouin]

Bass Ackwards caps can do it everytime. B.T.D.T., got the T-shirt. Plenty of holes in it too from slinging old solder.

 

[dcgillespie]

Assuming no damage to the caps or other components then, reverse the caps, and you should be good to bias the tubes to the new lower EFB quiescent level.

Let us know how it turns out!

Dave