Improving a Fisher 500B Power Amp Section – Advice?

[nj pheonix]

Rich,
I'm asking to learn, not to suggest a method.
I'd guess stock value would be in the ballpark (obviously I could be entirely wrong)
What I was thinking was maybe test original cap (to see if its spec or at least have an idea where your starting. Leave it in circuit and vary resistance in 10% increments (both ways) moving towards best waveform. Once there leave resistance value and vary cap 10% in each direction until best waveform.
Like I said I could be entirely wrong in my method. It makes some sense to me.
You could try that backwards also.
Assuming that worked ( I tend to think there is a flaw in my approach). I don't know how that would effect stability.
IIRC Kevin made a KT120 amp (last year?) And was playing with feedback networks. It shouldn't be to hard to find and probably helpful. Nice work so far!

 

[kward]

I assume you're talking about R95 and C59? When you change the value of R95 you are changing the amount of feedback applied. It is usually wise to set the amount of feedback first and then adjust the c59 to give the best response you can obtain. You may need a step compensation network also to help flatten out the square wave tops. This is a series combo of resistor and cap from the plate of the first tube to ground. I'm pretty sure Dave added one on one of his Fisher projects that you could look at to get an idea for reasonable sizes.

Also unless you think there is too much feedback on the stock unit it seems you would want to leave R95 at the stock value and tapped from the 16 ohm secondary.

 

[njcanuck]

I think you should be doing this only on the 16 ohm tap to exercise the whole winding and to keep the variables to only one - the feedback network. Now to get the latter part of the waveform quieting down faster.

Sorry, I guess I wasn't completely clear. The scope traces are all taken from the 16 ohm tap and the load is always across the 16 ohm winding. The 4, 8, and 16 ohm taps were all tried just for feeding the NFB loop, which I believe is fair game?

 

[njcanuck]

Andy: You could be right! I'm just pretending to know what I'm doing. It's much easier to vary the resistor because a pot makes a great substitute and you can see the effects immediately on the scope - you can dial from a crappy waveform to a semi-decent one in the blink of an eye. On the other hand, I only have a few cap values. If you have the whole range in 10% increments, I may have to come visit. :yes:

 

[dcgillespie]

That is absolutely how you should do it -- except that as you move from tap to tap, you don't want to just try different FB resistors willy nilly, but scale them to always maintain the same FB level regardless of which tap you are using.

Therefore, if the original FB resistor was 3.9K from the 16 ohm tap, then when using the 8 ohm tap, the FB resistor should be 2693 ohms, while from the 4 ohm tap, it should be 1840 ohms.

If the feedback levels were not leveled out in your tests between the taps, then the results cannot be conclusive. Therefore, they should be repeated using the correct resistor as indicated above to get a true indication of which tap is most appropriate for the NFB signal to be taken from. Fisher OPTs often favor the 4 ohm tap for the supplying the NFB signal.

Dave

 

[njcanuck]

Kward: Yes, R95 and C59. Thanks for that - this approach makes sense to me based on my limited experiments so far. Lowering R95 from 3.9k to 1.1k - 2k gives me the squarest waveform and sharpest transients with the 8 ohm tap delivering the smoothest wave top. I need to go back and check for consistency with the capacitance values that I have.

Off now to research step compensation. I was pondering how to get that flatter.

 

[njcanuck]

Thanks Dave: I knew you would be along soon! :yes:

Ok, I see what you are saying. But why not shoot for the squarest waveform on the 16 ohm tap first and then scale that as you describe? Wouldn't that establish the optimum feedback level?

 

[nj pheonix]

Rich,
I'm sorry what range of caps are you working with. I'll check my stock. I also have a resistor and cap decade box if you wan to play. I'm not sure if cap box goes low enough though. Lowest cap decade is in 100pfs

 

[kward]

Therefore, if the original FB resistor was 3.9K from the 16 ohm tap, then when using the 8 ohm tap, the FB resistor should be 2693 ohms, while from the 4 ohm tap, it should be 1840 ohms.

I was curious exactly how this was calculated. So after a little bit of math, I found it is calculated in this manner:

R1’ = Vr*(R1+R2) – R2

Where:

• R1’ is the new feedback resistor size to achieve the same level of feedback but taken from a different tap
• R1 is the original feedback resistor size
• R2 is the lower cathode resistor size
• Vr is the voltage ratio of the secondary taps. If the feedback resistor is currently connected to the 16 ohm tap and you want to move it to the 8 ohm tap, the voltage ratio is 0.707; if moving it from the 16 ohm tap to the 4 ohm tap, Vr is 0.5. It also works in the inverse—if the feedback resistor is currently on the 4 ohm tap and you want to move it to the 8 ohm tap Vr is 1.414, and to the 16 ohm tap, Vr is 2.

So in the case of njcanuck’s 500B as an example, moving the feedback resistor from the 16 ohm tap to the 8 ohm tap and keeping the feedback level the same as original, then:

R1 = 3.9K
R2 = 220
Vr = 0.707

thus R1’ = 0.707*(3900+220) – 220 = 2693 ohms.

So, I learned something new today that I didn't know yesterday, and thought I would share. .

(Note: I derived the above formula knowing that in order to keep the feedback at the same level as in the original design, it requires that the voltage at the junction of the voltage divider created from the feedback resistor and the lower resistor in the cathode circuit be unchanged from the original circuit. Given that, running a voltage loop equation creates a formula that can be simplified down to what is given above.)

 

[njcanuck]

Christmas duties are impacting my ability to get back to this. Good times, though.

Andy: Thanks for the offer. Values appear to range from about 300pf to 700pf. If I get stuck, I'll reach out to you. :yes:

Kward: Thanks for posting that. I'll add it to my knowledge base. :thmbsp:

Happy Holidays Guys! :tresbon:

 

[nj pheonix]

I have some loose 10s 20s 5s. Maybe a loose 47
Probably enough to piece together 5-100

 

[njcanuck]

Hope everyone had a great holiday!

My wife joined me in retirement on Dec. 23rd - sure seems harder to find bench time now. ara:

Had some time to try Dave's scaled approach using a constant 330pf cap across: a 3.9k on 16Ω tap (original config), a 2.69k on 8Ω tap and a 1.84k on 4Ω tap thus maintaining the same level of feedback on each tap. To my untrained eye, the 8Ω tap appears to marginally give the best square wave.

Using my own willy-nilly approach of putting a 330pf cap across a 10k pot and then twiddling the knob on each tap to get the cleanest waveform, I come up with 1.29k on the 8Ω tap. I have no idea how much feedback this represents?

Pics:

1. 330pf 3.9k 16Ω tap

2. 330pf 2.69k 8Ω tap

3. 330pf 1.84k 4Ω tap

4. 330pf 1.29k 8Ω tap

 

[dcgillespie]

Nearly double (192.7%), since the FB resistor has basically been halved. In terms of dbs, Fisher typically used about 18 db of NFB in the global NFB loop, so doubling that would add another 6 db, for a total of about 24 db of NFB. With this much increase in NFB, I would again stress that you do some stability testing before connecting this amp to your favorite speakers, so no damage can be done to tweeter elements.

Also, it will cause the input sensitivity to increase from requiring a rather typical 1.8 vac rms to drive the amplifier to full power output, to over 3.5 vac rms. This will still work with many preamps, but is well outside the typical norms, and will cause the volume control to operate in a less than optimum position with regards to loudness functions, and rate of change.

And the experimentation continues on!

Dave

 

[kward]

I'm surprised at 24 dB feedback you didn't run into instability already. Must be some pretty good output iron on these Fishers to achieve that level of negative feedback.

 

[dcgillespie]

K -- IIRC, you have been used to using some physically bigger transformers, which are often harder to deal with, as the physically larger core and winding structure can make for some interesting HF response characteristics.

Smaller transformers, like those in the these Fishers, can be kinder in that regard, but testing should still be done.

Dave

 

[njcanuck]

Next experiment. Tried a few parallel networks.

This was the best I came up with using both 16 ohm and 8 ohm taps. Could be worse, I think. Sensitivity is still compromised.

Pics:

1. NFB schematic

2. 10khz square wave into 16 ohm resistive load

3. 10khz square wave no load

4. 10khz square wave into .23uf capacitor

 

[nj pheonix]

Wouldn't the OPT see the other network through the winding? Am I missing something?

 

[dcgillespie]

Rich -- The proper configuration for a dual tap network is to determine which will be the highest impedance tap to use, and then let the total value of the FB resistance from that tap be of the appropriate value as previously discussed (3.9K from the 16 ohm or 2693 ohms from the 8 ohm tap). However, that resistance is actually made up of two resistors in series, with values such that where they connect together then represents the appropriate place for the lower tap to connect into. Therefore, for example, if using the 16 and 4 ohm taps, then the feedback "resistor" would actually consist of an 1840 ohm resistor, and an 2060 ohm resistor connected in series (making up the original 3.9K value), with the 2060 ohm end being connected to the 16 ohm tap, and the 1840 ohm end connected to the cathode circuit of the AF amplifier stage. The 4 ohm tap is then connected into the point where these two resistors connect together, by way of a cap only.

Again, you might review the network I came up with for the X-101C, which uses this very configuration. The cap from the lower tap is one more constant to adjust in achieving the best HF transient performance. This approach allows the network to maintain the original FB factor, while still taking the FB signal from multiple taps on the secondary winding.

Dave

 

[njcanuck]

Dave: Thanks for the explanation and direction. I did actually try your X-101C example earlier without much benefit but that was also using the X-101C values.

Is my last dual tap experiment inherently wrong or just unorthodox??

Here's two more using the proper dual tap methodology. While I see some slight smoothing of the wave tops, there appears to be no improvement in the transient response. The only way I seem to get noticeable improvement is by lowering the sensitivity. :scratch2:

Pics:

1. Stock NFB reference

2. Dual 16 & 4 tap schematic

3. Dual 16 & 4 tap trace

4. Dual 8 & 4 tap schematic

5. Dual 8 & 4 tap trace

 

[dcgillespie]

Your dual tap approach was incorrect on two counts:

1. It increased the amount of NFB because of using the wrong value resistance for each tap, and

2. It has basically doubled the amount of NFB (on top of the wrong value resistors) because you have now caused the the point where the NFB is inserted into the AF amp stage (top of the 220 ohm resistor) to be a summation point for the two networks: Since both networks are dumping feedback voltage across this resistor, the NFB voltage developed across it has basically doubled. Using the approach I gave you prevents this from happening, yet still lets the HF characteristics of both taps be reflected in the NFB signal.

You're only going to get so far juggling the NFB network alone. At some point, the HF tailoring network (made up of the 47K input resistance and 7 pf cap between the plate and grid of the AF amp stage) will require optimization as well for best results.

Dave