I have this Fisher amplifier I'm slowly getting in shape for actual use. It's one of those "console" amps that's a separate chassis, but when you remove the rest of it (tuner, preamp) the remaining circuitry within the amp chassis is all screwed up. One problem - after getting the voltages close to the schematic values - is the gain is way higher than I'd like.
Referring to the schematic, the 1st 12AX7 stage has a 470K plate resistor, the highest value I've ever seen (besides a 1M once). I'm thinking of reducing it, but, I'm also thinking doing that would change the feedback gain as well. Given the way that's all coupled to the phase splitter, I'm wary to touch it.
I could simply attenuate at the input by adding a level control - means drilling the as yet "unmollested" chassis for it. Or a resistor after the 47K from grid to ground. That would swamp whatever the 3pf cap does, perhaps changing whatever character the amp once had... Perhaps I shouldnt worry about it.
Any ideas? Thanks!
Referring to the schematic, the 1st 12AX7 stage has a 470K plate resistor, the highest value I've ever seen (besides a 1M once). I'm thinking of reducing it, but, I'm also thinking doing that would change the feedback gain as well. Given the way that's all coupled to the phase splitter, I'm wary to touch it.
I could simply attenuate at the input by adding a level control - means drilling the as yet "unmollested" chassis for it. Or a resistor after the 47K from grid to ground. That would swamp whatever the 3pf cap does, perhaps changing whatever character the amp once had... Perhaps I shouldnt worry about it.
Any ideas? Thanks!
A bigger issue, with some Fisher's and H.H.Scott's of the Golden Age (pronounced with air quotes) is that the output valves' summed cathode currents supply 150mA DC to some offsite preamp valves' heaters. 33VDC would mean three in series. Sounds like you've already made a stab at that. The other big issue is modern AC line voltages, and that's tougher.
Two things we can't do are to disturb DC conditions within the amplifier or to increase feedback percentage, as you've pointed out.
One approach would be to remove the inner positive feedback (R17 to R10) and remove the first stage cathode resistor bypass (C4), all helping to reduce open-loop gain. We could combine R9 and R10 into a single resistor, and bring loop feedback to the top of it - maybe about 1K8 or 2K0; selected to keep DC conditions as-is. Feedback parts R18 and C10 would then be scaled up in impedance to maybe 20K and 75pF or something in that neighborhood. R17 would ideally be made equal to R16 (both 56K or 62K?), so some adjustment of the value of the composite of R9 + R10 might be made to keep the idling DC voltage at pins 2 = 6 at about 100 VDC (measure at pin 3).
C18 should be replaced in any case with an actual Zobel network, maybe 10R in series with 0u1F, a useful modern conception not available to the original designers.
All good fortune,
Chris
Two things we can't do are to disturb DC conditions within the amplifier or to increase feedback percentage, as you've pointed out.
One approach would be to remove the inner positive feedback (R17 to R10) and remove the first stage cathode resistor bypass (C4), all helping to reduce open-loop gain. We could combine R9 and R10 into a single resistor, and bring loop feedback to the top of it - maybe about 1K8 or 2K0; selected to keep DC conditions as-is. Feedback parts R18 and C10 would then be scaled up in impedance to maybe 20K and 75pF or something in that neighborhood. R17 would ideally be made equal to R16 (both 56K or 62K?), so some adjustment of the value of the composite of R9 + R10 might be made to keep the idling DC voltage at pins 2 = 6 at about 100 VDC (measure at pin 3).
C18 should be replaced in any case with an actual Zobel network, maybe 10R in series with 0u1F, a useful modern conception not available to the original designers.
All good fortune,
Chris
Ah, I get it. I'll try that first, then onto Chris's suggestions on R17's value and path to ground / + feedback connection. Pretty much all the schematic's there; though I should have added in the voltages for the screens, the bias and the input tube's V+ rail. I'll follow up with that - thanks guys - hopefully can get this hunk of iron sounding a bit better! (Zobel idea registered too)
So here's what I got so far. It's for sure sounding better than it did. When I grounded R17, the 180K R12 had to go - sounded "ring-modulated". I dropped R8 from 470K to 390K, to get the voltages about right, per original, above.
The power supply is a SS diode voltage doubler. I figured the impedance of it is about 180 Ohms, by observing a 20V drop as the tubes come on and guessing the total current at 150 mA. So, adding a small resistor in series with the transformer secondary winding wont make it that much more sloppy, so I tried a 33 Ohm. That brought down the 427V B+ to 396, with ~7 VAC across it for about 1.5W; used a 5W part.
All the other voltages fell as well. I can keep it in circuit and adjust everything else back up to "spec" by replacing the powersupply resistors (which I've replaced the originals with all new values already) to get the above readings. What do you think?
Last edited:
Note R12 in the original; also the connection to R10 R18 NFB loop.
I don't see why the innards need to be twiddled. To shave gain, increase R10 to 330 or 470. Any other path requires understanding some of the 'cleverest' circuit design of the century, and my head is not that big.
Or using a simple passive attenuator in front, but I know this offends some religious authorities.
Another idea would be to replace the 12AX7 with a 12DW7 and the associated resistor value changes...
Closed loop gain can be reduced by simply increasing R10, but that increases loop feedback, which may or may not be OK for stability. Simply removing C4 decreases open loop gain, but doesn't much effect closed loop gain. 12DW7 would be a good choice but doesn't effect any gains because the phase splitter has a fixed gain of about 2.
The original circuit was optimized (very well) for a higher closed loop gain. We should look at ways to trade that extra gain for some benefits, as possible.
All good fortune,
Chris
The original circuit was optimized (very well) for a higher closed loop gain. We should look at ways to trade that extra gain for some benefits, as possible.
All good fortune,
Chris
Those EL84's are gonna cook like hell with that 400+ volts fed to them.
They like 300-325V much better, and -12V on their cathodes, and last longer.
They like 300-325V much better, and -12V on their cathodes, and last longer.
Just replaced the 33 Ohm with a 25 Ohm, 10W part. It was hot... Putting my Fluke 123 across the 25 Ohm showed a voltage waveform typical of a full wave bridge current and read ~10V RMS; perhaps a better reading than the Wavetek DMM was giving me. Still got me into the 390s for B+. Other than that, I dont know what to do reasonably to get the power supply down 75 - 100V lower.
I noted in another Fisher schematic the use of a potentiometer in series with the resistor on the cathode side of the phase splitter, along with a corresponding procedure to adjust it for minimal harmonics at 1kHz. So I put that in place and dialed it up to match the 47Ks exactly on the plate end for now.
Those adjustments can be done later to tune the thing up using REW.
Now for the really drive you nutz dept. I swear I'm sometimes measuring B+ at 394, then 385, then 394 again. I thought the 33 Ohm was failing, so I replaced it with the 25 at double the wattage. Same thing again. Without the resistor, 420 - I see 427 above. <scratch...> My wife "cooks" garden soil using a small countertop oven and it's plugged into the same AC elsewhere in the detached garage my workbench is in. I'm sure its cycling, once it reaches temperature and making the AC line go up and down...
Part II nutz driving; when I listen to "The Jazz Groove" using the web page interface, it's waaay louder than when I play anything using Foobar 2000. Windows volume control is like 20 of 100 max on the JG opened in Chrome. So I endeavored to put the JG URL into Foobar and back to sanity on the volume level. That's with both apps levels maxed, using the windows level to control the volume. Who designs these things? One would think at least an output level would be consistent across "apps".
I noted in another Fisher schematic the use of a potentiometer in series with the resistor on the cathode side of the phase splitter, along with a corresponding procedure to adjust it for minimal harmonics at 1kHz. So I put that in place and dialed it up to match the 47Ks exactly on the plate end for now.
Those adjustments can be done later to tune the thing up using REW.
Now for the really drive you nutz dept. I swear I'm sometimes measuring B+ at 394, then 385, then 394 again. I thought the 33 Ohm was failing, so I replaced it with the 25 at double the wattage. Same thing again. Without the resistor, 420 - I see 427 above. <scratch...> My wife "cooks" garden soil using a small countertop oven and it's plugged into the same AC elsewhere in the detached garage my workbench is in. I'm sure its cycling, once it reaches temperature and making the AC line go up and down...
Part II nutz driving; when I listen to "The Jazz Groove" using the web page interface, it's waaay louder than when I play anything using Foobar 2000. Windows volume control is like 20 of 100 max on the JG opened in Chrome. So I endeavored to put the JG URL into Foobar and back to sanity on the volume level. That's with both apps levels maxed, using the windows level to control the volume. Who designs these things? One would think at least an output level would be consistent across "apps".
Last edited:
Wire the (now unused) extra filament winding on the PT in series with the 120V primary - as a bucking winding.
That will reduce the B+, as well as compensate for the lack of loading from the preamp since it's not being used.
That will reduce the B+, as well as compensate for the lack of loading from the preamp since it's not being used.
I do appreciate that suggestion, but...unfortunately...that winding got used to generate my negative fixed bias, allowing me to ground all the OPT's cathodes through a 10 Ohm resistor, which I setup to measure ~110 mA.
I'll need to work on it some more, to provide measured values here.
I'll need to work on it some more, to provide measured values here.
Then it would not be a Fisher, eh?
It will be stable at gain of 3 or less. Another 14dB-20dB of NFB can really smooth-out these AX7 Fishers, though that is a matter of taste. (I've run the top of the line 7189 Fisher at unity-gain, tight like a DC-300.)
Probably a positive cathode voltage? Like the +33V cathode (+21V grid) in post #1?
Yeah, Fisher did run these amps hard.
PRR,
Speaking of clever design, that reminds me of the Fisher 500C and Fisher 800C.
Each channel power amp section consisted of just a 12AX7 and a pair of 7591 output tubes in Beam Power Mode.
Its good performance relied on global negative feedback, and . . .
. . . A very very good output transformer that had good rise and fall times and did not have any ringing.
Those output transformers work very well for 2A3s, 300Bs, 6CK4s, etc.
Just the experience of many club members in my area.
Speaking of clever design, that reminds me of the Fisher 500C and Fisher 800C.
Each channel power amp section consisted of just a 12AX7 and a pair of 7591 output tubes in Beam Power Mode.
Its good performance relied on global negative feedback, and . . .
. . . A very very good output transformer that had good rise and fall times and did not have any ringing.
Those output transformers work very well for 2A3s, 300Bs, 6CK4s, etc.
Just the experience of many club members in my area.
So this 6BQ5 based Fisher amp produces 14.4W into 8 Ohms resistive, with an input signal ~0.6V RMS. I recorded THD at just under 4%, using REW.
Backing down to 8W, I could adjust the above mentioned phase splitter balance pots to reduce the distortion a fraction of a percent, on one channel I got it from 1.28% to 1.16%. It seemed the control could trade off 2nd with 4th harmonic levels and of course does nothing to the third, as one "nulls" the 2nd level. I wonder if the phase of the 2nd harmonic is different, depending on which side of the null setting?
I took a look at the harmonic structure at the output and found a similar pattern looking on the screen supply. Adding 100u there did nothing to the output harmonics chain. Adding it across the main B+ reduced the "spurious" harmonics about the base of input 1kHz tone and higher harmonics - somewhat. I wonder what's generating those?
I tried adding back the cap I clipped out in the cathode circuit and it did reduce distortion some, albeit barely perceptible by REWs Spectrum Analyser.
The thing that made the most difference was the output tube bias. Assuming all 4 of the matched quad I bought share equally, at ~21mA per tube I read 4% THD, at ~25 mA - 2.64% and ~31 mA it dropped to 0.98%. This is adjusting to 8W output each time the bias was changed.
Here's the spectrum at ~8W, with the added 100u cap to B+, showing the reduction of the spurious stuff. At least it did something; not that I'd add it into the amp based on this.
Here's the frequency response; other channel is same. I wonder why the third harmonic nulls at 70 Hz, of all places?
REW is a great tool!
I bought a single ended tube amp from another member here and running a similar spectrum, I find that the 2nd and 3rd are at about the same level. Somehow I thought that amp would have a majority of even harmonics; not so. On a P-P amp, you can null the second, or make it bigger using that balance adjustment. Cant touch the third...
How does one design a tube amplifier with minimized odd harmonics?
Thanks!
Backing down to 8W, I could adjust the above mentioned phase splitter balance pots to reduce the distortion a fraction of a percent, on one channel I got it from 1.28% to 1.16%. It seemed the control could trade off 2nd with 4th harmonic levels and of course does nothing to the third, as one "nulls" the 2nd level. I wonder if the phase of the 2nd harmonic is different, depending on which side of the null setting?
I took a look at the harmonic structure at the output and found a similar pattern looking on the screen supply. Adding 100u there did nothing to the output harmonics chain. Adding it across the main B+ reduced the "spurious" harmonics about the base of input 1kHz tone and higher harmonics - somewhat. I wonder what's generating those?
I tried adding back the cap I clipped out in the cathode circuit and it did reduce distortion some, albeit barely perceptible by REWs Spectrum Analyser.
The thing that made the most difference was the output tube bias. Assuming all 4 of the matched quad I bought share equally, at ~21mA per tube I read 4% THD, at ~25 mA - 2.64% and ~31 mA it dropped to 0.98%. This is adjusting to 8W output each time the bias was changed.
Here's the spectrum at ~8W, with the added 100u cap to B+, showing the reduction of the spurious stuff. At least it did something; not that I'd add it into the amp based on this.
Here's the frequency response; other channel is same. I wonder why the third harmonic nulls at 70 Hz, of all places?
REW is a great tool!
I bought a single ended tube amp from another member here and running a similar spectrum, I find that the 2nd and 3rd are at about the same level. Somehow I thought that amp would have a majority of even harmonics; not so. On a P-P amp, you can null the second, or make it bigger using that balance adjustment. Cant touch the third...
How does one design a tube amplifier with minimized odd harmonics?
Thanks!
Last edited:
I guess that's something everyone knows; https://www.ebay.com/itm/124269823127. My amp cost 1/10. Just look at the iron in those OPTs. The ones in mine are like thumbnails in comparison...