• The move to the new server is done. There are some software and database maintenance updates in process. This has us passing the hat around to help out. We appreciate any donations. Seriously, even a dollar helps. The payment page may be found here - https://www.audiokarma.org/support.html

Emory Cook and Gus Jose's "The Ultimate Amplifier" (Long Post Warning)

I agree, the writing style is annoying and unhelpful. I wonder if the magazine's editors (or Gus Jose, whose role in the article/design is not clear) added this stuff to make a fairly technical article more appealing to a general audience. Emory Cook wrote an editorial response to various questions about the amplifier in a later issue and that's devoid of all this weird language.

"Wreck it purposefully" seems to be hyperbole, since elsewhere the article claims that the amplifier is only down 3db at 100k. I think "The remaining stages have to be so good that just a matter of a 1.0-mfd condenser can upset them" is referring to the need to minimize stray capacitances related to the 1uf coupling caps. IIRC, the article says not to use metal capacitors with their cases attached to the chassis for this reason.

In any case, I'll replace the parallel resistors with a single properly-sized high precision resistor and retain the mica capacitors. Ordering a proper sig gen now so I can do frequency response and square wave testing on the working unit. This should provide a decent reference. I think the guy who I bought this from told me it had issues with "rumbling noise" that occurred when the variac was brought to full line, but he claimed he sent the amp to his tech for fixing. So who knows if the assembled example I have is even fully stable.
 
Last edited:
It is weirdly phrased but purposely limiting the bandwidth of an amplifier to a useful range in order to keep it stable is a common thing. A whole lot of amps will have a cap from the first amp plate to ground in order to make this work.

The resistor thing was a common trick back before cheap and common access to odd value, low tolerance parts was a thing. Hand-matching resistors is another thing that is mostly not needed but back when 10% parts were common and you needed a pair to match within 1%, that was just how it was done. Now you can order 1% or less parts without any particular trouble.
 
which also has a cap to "wreck" the response a bit :) Runs to B+ so it grounds via the filter caps but same idea.

Some of all this might just be the state of output transformers at the time this was written. Possible they decided all of these shenanigans were required in order to get the desired performance with a transformer that really didn't want to do it. Current feedback and variable damping was also a 'thing' for a while but it faded away. Numbers and data and sciencing the **** out of everything was just how it was for a long time. This "does it sound good" nonsense probably came from some long-haired hippie.

I can appreciate both a simple and well executed design as well as a very Rube Goldberg sort of approach. Probably not overly inclined to build a Goldberg amp but they are cool to study from a technical perspective.
 
I'm not even going to claim that I knew what I was getting into. I could tell from the listing's images that the amplifier had a lot of parts and stages, but I didn't know the design would be so eccentric. I thought it was a slightly modified Langevin 101-3101 with additional stages and screen regulation. In reality, there's little resemblance beyond output iron and tube complement.
 
Ok, so I'll have a good update post in a bit. My current plan is to essentially build two examples of the amplifier from scratch using the parts I have -- the true build-from-schematic experience Cook intended. The wiring in the unit I purchased on ebay is very messy and there is PCB oil all over the interior of the chassis. I am not trying to solder that and inhale dioxins (a byproduct of PCB combustion).

I'm mostly done speccing the chassis and preparing the BOM, but I do have one question. Can anyone explain to me why this adjustable 3k resistor that drops the B+ for the VRs is specified at 50 watts??? My understanding is that the regulated supply current should not exceed the maximum VR current, or 40 ma. If it's 40 ma, a 10 watt resistor seems more appropriate than a 50 watt. Am I missing something, or is this part just seriously over-specced?

VRshuntressitor.png
 
overkill de-rating? 10w would be 2x, 50w would be 10x.

Not really sure why that needs to be adjustable though. The voltage should be more or less constant within tube tolerances so it should be possible to pick a fixed resistor to get the job done. The instructions say "Adjust the variable 50w 3k in the power supply until the voltage regulator tubes light a pleasant blue *but not too brightly)" which tells me the current flow isn't that critical. Figure a resistor that will flow 20-30 ma and call it good enough. Unless the gas regs or the output tubes are wildly variable it should behave itself just fine.
 
AlexHempel said:
The mica caps connect the plates of the 12AX7 triodes to ground before the coupling cap
Click to expand...

When a signal passes through an amplifier, each stage contributes phase shift. This shift adds to the phase shift in the output transformer, and at some high frequency, the total becomes 180°. At this frequency, the negative feedback loop becomes positive, and if the amplifier as a whole exhibits more than unity gain at that frequency, it will oscillate. Even if gain is below unity, the amplifier will ring at that frequency in response to a signal with a fast rise/fall time. The purpose behind the step network in the driver (and, when present, the capacitor in the feedback loop) is to reduce the gain at the frequency where the 180° shift occurs, so that the amp is stable and ringing is minimized.

Here are a couple links on this topic at Max Robinson's excellent site that you might find useful:

https://www.angelfire.com/electronic/funwithtubes/Amp-Compensation.html
https://www.angelfire.com/electronic/funwithtubes/newcomb_d-10_followup.html

It's also possible to find a satisfactory relationship between gain and frequency empirically, using cut-and-try. This is the approach recommended by
Bruce Rozenblit. The 10kHz square wave is your friend through this work, but it's important to remember the results don't have to be perfect in order to achieve an excellent amplifier.

Jack
 
Last edited:
Thanks for that concise and straightforward explanation of the step network's function. I've read about phase shift and oscillations etc., but still trying to fully understand everything.

Regarding the adjustable resistor, that is definitely true. But I measured the resistor in the assembled amp at 2.6kish, so I assume the majority will be used and this probably doesn't justify more de-rating than like 30% or so. I think I'm going to get a 25 watt part instead of the 75 watt I have currently (it's taking up a lot of space in the chassis). I might also take Gadget's advice and try to use a fixed resistor.
 
Why is there a big cap (80uf) across the OD3 regulators? I thought the rule was you should never have a cap greater than .05uf or so across a VR tube, in order to avoid oscillation.
 
I've also heard that, and I'm not sure why/if it works here. But it's certainly connected that way on the example I have.
 
probably works by virtue of high current load. Where you have trouble with relaxation oscillators is when the cap charges up to strike voltage, causing the bulb to switch on and draw it down. Once the voltage drops enough, the lamp goes out and the cap can recharge. If you keep the current flow high enough, the cap can never discharge and allow it to turn on and off.

The only trick is that under full output the current flow through the regulators would drop since more would be pulled by the screens. The idle current through the tubes has to be high enough to keep them lit all the time. If the screen voltage starts bouncing because the supply is oscillating the output of the amp will also bounce right along with it.
 
Salectric said:
Why is there a big cap (80uf) across the OD3 regulators? I thought the rule was you should never have a cap greater than .05uf or so across a VR tube, in order to avoid oscillation.
Click to expand...
I'm not clear on the purpose of the 47K resistors, either. Series glow tubes don't need equalizing resistors, and the power supply already has bleeders.

Jack
 
gadget73 said:
probably works by virtue of high current load. Where you have trouble with relaxation oscillators is when the cap charges up to strike voltage, causing the bulb to switch on and draw it down. Once the voltage drops enough, the lamp goes out and the cap can recharge. If you keep the current flow high enough, the cap can never discharge and allow it to turn on and off.

The only trick is that under full output the current flow through the regulators would drop since more would be pulled by the screens. The idle current through the tubes has to be high enough to keep them lit all the time. If the screen voltage starts bouncing because the supply is oscillating the output of the amp will also bounce right along with it.
Click to expand...

Interesting and makes sense in this context. Whereas amps like the RCA MI-9377a use gas regs for the screens only, this design also uses the regulated supply for 12AU7 and 12AX7 B+. Definitely more current draw than screens alone.

But perhaps this also explains why an adjustable resistor is called for? There might be a relatively specific resistance where this whole equation balances out.
 
gadget73 said:
which also has a cap to "wreck" the response a bit :) Runs to B+ so it grounds via the filter caps but same idea.

Some of all this might just be the state of output transformers at the time this was written. Possible they decided all of these shenanigans were required in order to get the desired performance with a transformer that really didn't want to do it. Current feedback and variable damping was also a 'thing' for a while but it faded away. Numbers and data and sciencing the **** out of everything was just how it was for a long time. This "does it sound good" nonsense probably came from some long-haired hippie.

I can appreciate both a simple and well executed design as well as a very Rube Goldberg sort of approach. Probably not overly inclined to build a Goldberg amp but they are cool to study from a technical perspective.
Click to expand...

Fisher threw "Z-matic" variable damping circuitry at everything, even when they really didn't need to. I stripped it off of my 70AZ and rebuilt it to match my older 70A.
 
gadget73 said:
probably works by virtue of high current load.
Click to expand...

But maybe not. I've always taken it on faith that any shunt capacitance greater than the specified maximum of 0.1uF across a glow tube or string of glow tubes is likely to create a relaxation oscillator. Have never seen seen the effect though, except with NE-2 neon lamps many years ago. Reviewing a few of my older books this evening regarding the use of glow tubes, I find now that the articles don't all carry this warning. In fact, the first one I opened has a schematic with a 5651 shunted by 40uF, and current draw is extremely small.

Just to settle this for myself, I connected a pair of 0A2s in series a few minutes ago and powered them through a 4K 10W resistor from my bench supply. I shunted them with a 1uF film cap, then a 22uF electrolytic, varied the power supply through the full range of tube current (< 5mA to 30mA), and I did this with and without a 100K load (3mA) across the tubes. There was nothing but DC under any and all conditions.

If there's another way to make these oscillate, I'm all ears.

Jack
 
More results... The tubes will oscillate if the feed resistor is made significantly larger in value. Adding 10K to the 4K wasn't enough, but adding 39K produced a sawtooth at roughly 40 Hz (1uF parallel cap). This only occurred at very low currents through the tubes, well under the 5mA recommended minimum. When tube current was increased to 5mA or so, the oscillation ceased. Also, adding the 100K test load resistor required a higher supply voltage and more tube current to encourage oscillation. This might indicate that the combination of a very high voltage supply, perhaps 1KV or greater, and its attendant large value feed resistor, could give rise to these same oscillations. However, no condition under which this admittedly limited testing caused oscillation was analogous to any configuration likely to be used for regulating screens in most amplifiers. I have to conclude from all this that the risk of oscillation in most applications is extremely small.

Jack
 
^^^ this is what I love about Audiokarma right here.
People with part tools and knowledge to go empirically check an idea out, and the willingness and curiosity to do so.

Regulator circuits are a rabbit hole that I'm going to have to go down some day if Im going to be building larger tube amps.
Bravo.
Somewhere I've got a chart with all of the data for the various gas regulator tubes compiled in one spot.
Regulated voltage, start voltage, current range, base type, pin configuration and I took note of the color of the glow.
I'll dig it up and shoot you a picture.
 
About 40 years ago -- the last time I played with these things -- I had very different results -- using series connected OA2s passing about 15 mA as a shunt regulator to supply 300 vdc to a preamp (itself drawing 10 mA), I noted the tubes would oscillate -- and rather violently so (flashing) -- if any significant value of capacitance (over say 1 uF) was shunted across the tubes. The tubes were brand new, and fed from a 6K 10W resistor from a 450 vdc raw supply. Maybe because of the higher load current?

Dave
 
You must log in or register to reply here.
Back
Top Bottom