• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

EF86 cathode bypass cap size... 500uF?

Not a typo, they are talking about getting extremely low hum when using it at very high gain, any supply hum will affect the cathode current and this will modulate the grid-cathode voltage directly. Note they are talking about µV levels of quietness at the grid earlier in that section of the datasheet. The whole AN is about getting the ultimate performance.
 
I don't think it is a typo — everything else on the datasheet (whew! what a long read!) seems factually correct — so the question is “why”?

Snell is correct — the use of an extra-large cathode-resistor-bypass-capacitor serves very little to suppress induced hum, in general. IF one is using a 2.2 kΩ RK, then using
Z ≈ 1,000,000 / ( 2πFC ) (C in µF, Z in Ω, F in Hz)​
with little surprise, it can be inverted to
C = 159155 / FZ … with F = 60, Z = 2,200 Ω
C = 159,155 ÷ 132,000 µF
C = 1.21 µF​
This would be the –3 dBV 'corner' of response. But consider what the RK + CK combo does. It stabilizes the quiescent bias voltage, set by RK and smoothed by CK. The bigger the CK, the lower will be any induced (k)cathode voltage fluctuation, either relative to input signal, or to A/C induced onto the kathode by the A/C filament windings. I think that's what they're trying to suppress.

Moreover, the specification for 500+µF was written in the era where “good, cheap, reliable, compact, readily available” electrolytics in the 16 V and 25 V blocking class were attaining values of 220 µF and up, for only pennies-a-pop. So, the advice to use a larger electrolytic than strictly necessary, could defensively not be a typo.

Just Saying,
GoatGuy ✓
 

PRR

Member
Paid Member
2003-06-12 7:04 pm
Maine USA
www.diyaudio.com
...Say what? 500uF? Typo? Joke?..

The local hum may be 3V. The "inaudible" hum level desired may be 3uV. This seems to suggest 1,000,000:1 reduction.

500uFd at 50Hz is nearly 6 Ohms. If hum comes in at a cathode of about 1k impedance this will reduce it 166:1. Clearly the hum does not come in at full 6V but something far less. Maybe low enough for most uses. The 1k:6r divider gives further reduction for hum-critical stages.
 

disco

Member
2006-04-17 6:27 pm
Holland
But consider what the RK + CK combo does. It stabilizes the quiescent bias voltage, set by RK and smoothed by CK. The bigger the CK, the lower will be any induced (k)cathode voltage fluctuation, either relative to input signal, or to A/C induced onto the kathode by the A/C filament windings. I think that's what they're trying to suppress.
I've once red a strophe on the effect of bypassing a power tube (so, not small signal) kathode with a large size capacitor, ending in a bias shift with loud program because the cap gets more loaded one way. It was referred to as 'diode effect'. Can't remember where, probably Morgan Jones or on this forum. Can anyone concur?
 
Wow, so it could be that big! Interesting! Thanks everyone. I usually eye anything so big outside of a robust power supply as suspicious. I'm working on an mc phono pre design, and I was modelling with 20-30 uF bypass caps. I may need to rethink this.

I think the large capacitor causing a bias shift may refer to blocking distortion, but as I understand it, that occurs with the decoupling cap.

(Btw, Hello Jaap!)
 
For what it's worth, I built an MM phono preamplifier with EF86 input valves some seven years ago. I decoupled the cathodes with 220 uF and used an AC heater supply. The centre of the heater supply (artificial centre made with two 100 ohm resistors) was connected to a bias voltage that was positive with respect to the cathodes, even though I didn't notice any difference when I connected it to ground instead. The EF86's grid was connected straight to the cartridge, so no DC blocking cap at the input. The heater wiring was done with balanced microphone cable: twisted and shielded.

As long as the power supply transformer was at a large distance from the amplifier, the MM amplifier had no audible hum at all. When I turned up the gain very much, I heard some noise, but no hum. Unfortunately, since I mounted the transformer on the same chassis as the amplifier, one channel does have a very slight hum - to weak to hear during normal use, but audible when you turn the volume up way beyond its normal level.

By the way, I found the RIAA- and A-weighted noise of a triode-connected EF86 to be optimal somewhere around 1.75 mA cathode current. Much lower currents increase white noise and much higher currents increase 1/f-noise.
 

PRR

Member
Paid Member
2003-06-12 7:04 pm
Maine USA
www.diyaudio.com
> working on an mc phono pre design, and I was modelling with 20-30 uF bypass caps. I may need to rethink

Many MM preamps resorted to hard-grounding the cathodes. Signal levels are so very small that it won't overload (even less in the first stage of an MC preamp). A large plate resistor stabilizes current, select to taste. And there is "NO!" hum on the cathode this way.
 
For what it's worth, I built an MM phono preamplifier with EF86 input valves some seven years ago. I decoupled the cathodes with 220 uF and used an AC heater supply. The centre of the heater supply (artificial centre made with two 100 ohm resistors) was connected to a bias voltage that was positive with respect to the cathodes, even though I didn't notice any difference when I connected it to ground instead. The EF86's grid was connected straight to the cartridge, so no DC blocking cap at the input. The heater wiring was done with balanced microphone cable: twisted and shielded.

As long as the power supply transformer was at a large distance from the amplifier, the MM amplifier had no audible hum at all.

It’s like you really mind! I’ve been mulling over all of these issues, the largest of which is whether to use the pentode at all. I keep coming back to the idea that it’s best to get as much first stage gain as possible and keep the noise as low as possible. I’m also going to rethink my bias point too based on what you recommend. Thank you! :up:
 
We tend naturally to think of any capacitor larger than 50uF as extreme, based on looking at schematics from 70 years ago. "I don't think Hank done it thisaway." Makes no sense today, when capacitors are cheap and pretty darn good, even electrolytics if they're biased with some DC voltage.


Performance of circuits with cathode resistor bypass caps increases with larger capacitance in all respects (assuming perfect capacitors, yada, yada), including stability with long loop feedback and noise. Best is as PRR says, bolt the cathodes to signal ground and add bias with a battery. No, they're not noisy.


DIY valve audio in general has an over-appreciation of the distant past. I'm sometimes guilty of using the term "Golden Age" too loosely and not making the irony part obvious enough. Folks back in that misty bygone era would have loved to have had the inexpensive excellent capacitors and resistors we have today. Modern valves? Maybe not so much.


All good fortune,
Chris
 
Last edited:
It’s like you really mind! I’ve been mulling over all of these issues, the largest of which is whether to use the pentode at all. I keep coming back to the idea that it’s best to get as much first stage gain as possible and keep the noise as low as possible. I’m also going to rethink my bias point too based on what you recommend. Thank you! :up:

I chose the EF86 because it is designed for low noise in the audio band, low hum with AC heater supply and low microphony. There are valves with lower audio noise, but they usually have far more capacitive coupling between the heater and the control grid.

I had several reasons to go for a triode connection, but one of them was to get rid of the partition noise. I don't know if it makes much of a difference though, as low-noise pentodes were also designed for low screen grid current and, hence, low partition noise. With a reasonable amount of gain in the first stage, second and further stage noise should be negligible.

By the way, Frank Blöhbaum has come up with a very creative way to get rid of partition noise, which he calls the BestPentode circuit. It boils down to using a current follower to reunite the screen grid current with the anode current to cancel the partition noise. Mind you, it is patented, so depending on the patent law in your country you can either not sell it or not use it at all.
 
As far as I recall, that was mainly criticism of the noise measurements in Frank Blöhbaum's Linear Audio article. A plain old triode connection should be essentially equally effective against partition noise as the BestPentode technique (in fact even slightly better), but for some reason he usually found lower noise values for his BestPentode circuit than for a triode-connected stage with the same valve. In any case, that doesn't make his circuit any less creative.
 

PRR

Member
Paid Member
2003-06-12 7:04 pm
Maine USA
www.diyaudio.com
.... Best is as PRR says, bolt the cathodes to signal ground and add bias with a battery. No, they're not noisy.....

I did not say "battery" in this thread.

I'm not opposed to battery bias. But there is no need at phono levels.

I agree that anything over 8 or 32 mFd is "huge" in old-school. Among other things electros went bad in a few years. But today what is the cost and risk of 500uFd 6V electro? Mouser has KEMET 6.3V 820uF 105C 3k Hour Radial $0.32, 820 uF 6.3 VDC 105 C Life: 8000 Hour Rubycon $0.33....

Hmmmm..... 500uFd through a 100k tube and resistor is like a Minute time-constant. Ah, but it is always best to warm-up the system many minutes.
 
I don't think the charging time is an issue. The differential resistance seen at the cathode is 1/gm, no way that will be anywhere near 100 kohm on any low-noise tube. Of course you might argue that all charging current for the cathode has to come from the anode and screen grid bias resistors, but when the circuit is designed for a bias current around 1.75 mA, the initial voltage rise will be greater than 3.5 V/s.

A problem with biasing at 0 V VGK is that the grid will actually start conducting a bit, causing an increase in input noise current. I haven't a clue how much, though. It is sometimes solved by using AC coupling and an excessively large grid leak resistor, but then the bias point becomes very dependent on the grid leakage current.
 
Last edited:
I mean, imagine the EF86's cathode would abruptly get hot. Initially the EF86 would then charge the 500 uF cathode capacitor as quickly as the anode and screen grid resistors allow, which means with several volts per second. When the cathode voltage approaches the desired voltage, it would settle with a time constant of 500 uF/(gm + gm_screen), which is in the hundreds of milliseconds range. The whole process would be done in just a few seconds. In real life, the charging waveforms will be smoother because the cathode heats up gradually, but it still won't take very long.
 
I did not say "battery" in this thread.

I'm not opposed to battery bias. But there is no need at phono levels.


Sorry for running things together and putting words in your mouth. We could argue for days about how much is a minimum bias for a 12AX7. Our parents would have, maybe even grandparents (6SL7, 6SC7, yada, yada.).


All good fortune,
Chris