I have experimented that an RC shunt (built according to the Zobel loudspeaker network formula Speaker Zobel / Impedance Equalization Network Circuit Calculator , but reducing C at one third, one fourth) the tube amp sound much better. I'm using full-range speaker, so easy to get parameters
R = Re * 1.25
C= Le/R^2 * 0.3 (or Le/R^2 * 0.25)
Do you usually put an RC shunt just after the output transformer and how do you calculate it?
tHANKS
R = Re * 1.25
C= Le/R^2 * 0.3 (or Le/R^2 * 0.25)
Do you usually put an RC shunt just after the output transformer and how do you calculate it?
tHANKS
Last edited:
Many commerical tube amplifiers did not use a zobel RC network across the output terminals - for several reasons. Many amps had such low amounts of feedback it wasn't necesary; lower levels of stability was considered acceptable then; the lead compensation in the feedback did the job.
Clearly, if there is no negative feedback, there is no need for a zobel network for stability. However, if the amp has tetrode or pentode output, the network is still necessary so that the rise in speaker impedance doe not cause trebble boost and harmmic distortion.
In those that did, the RC network, or often just C, was placed across the transformer primary. It works just the same but the capacitor value is much smaller.
In the early transformer coupled transistor amplifiers, it was often on the transformer primary for the same reason.
With a network across teh speaker terminals, since the idea is to present the amplifier with a frequency independent pure ressistance, how the rigfht values are calculated is the same as for a solid state amp. That is, independent of the amp circuit, and entireley dependent on the loudspeaker. The ressitor should equal the woofer/loudspeaker dynamic resistance (eg 8 ohm). With today's loudspeakers, their efficiency is so low, you can take this as 110% of the dc resistance. The capacitor should be sized such that its reactance is the same as the loudspeaker inductive reactance at a convenient high audio frequency.
Often it is necessary to assume the amp can be used with any speaker for which the characteristics are unknown. A good compromise is 8.2 ohm in series with 470 nF.
If you put it across the trasnformer primary, scale the values up in the same inpedance ratio as the trasformer.
Clearly, if there is no negative feedback, there is no need for a zobel network for stability. However, if the amp has tetrode or pentode output, the network is still necessary so that the rise in speaker impedance doe not cause trebble boost and harmmic distortion.
In those that did, the RC network, or often just C, was placed across the transformer primary. It works just the same but the capacitor value is much smaller.
In the early transformer coupled transistor amplifiers, it was often on the transformer primary for the same reason.
With a network across teh speaker terminals, since the idea is to present the amplifier with a frequency independent pure ressistance, how the rigfht values are calculated is the same as for a solid state amp. That is, independent of the amp circuit, and entireley dependent on the loudspeaker. The ressitor should equal the woofer/loudspeaker dynamic resistance (eg 8 ohm). With today's loudspeakers, their efficiency is so low, you can take this as 110% of the dc resistance. The capacitor should be sized such that its reactance is the same as the loudspeaker inductive reactance at a convenient high audio frequency.
Often it is necessary to assume the amp can be used with any speaker for which the characteristics are unknown. A good compromise is 8.2 ohm in series with 470 nF.
If you put it across the trasnformer primary, scale the values up in the same inpedance ratio as the trasformer.
Last edited:
Zobel network for an amplifier that uses a transformer coupling to the loudspeaker? Whatever for??
It will have no consequence. Valve amplifiers have never used them and have never needed them. To add one seems a bit silly. If they were needed, I would have thought, they would have been fitted before, in over 80years of valve design and use.
It will have no consequence. Valve amplifiers have never used them and have never needed them. To add one seems a bit silly. If they were needed, I would have thought, they would have been fitted before, in over 80years of valve design and use.
Interesting! I thought to do so but I didn's know how to size.
So if primary/secondary turns ratio is 20, and my best found RC on seconday is:
R=15 ohm
C=0,22nF
is correct to move on primary as:
R=15*20^2=6000 ohm
C=0,22nF
??
If you do that, the impedance for the anode load will become non linear and screw the loading up! All that can be added on the HT side of the transformer, is a couple of reverse biased diodes to stop the negative spikes when being over driven.
You have no idea what you are talking about.
Go read any textbook on pentode output stages, right back to the dawn of pentodes, the 1930's. Any textbook on pentode and tetrode output will tell you that the rise in speaker impedance causes a rise in harmonic distortion.
And with neg feedback, the inductive impedance of a loudspeaker reduces the phase margin, leading to instability and distortion, just as it does in solid state amps. However, as I said before, there is more than one way to address this.
R-C networks, or just capacitors, across the transformer primary, were quite common.
If you move a CR network to the primary you have to scale C as well as R.
I can't see the point. Given a good amp it is not necessary. Given a bad amp the most it can do is roll off the treble a little, and increase treble distortion - better to use a tone control.
In a wireless set the anode capacitor had an extra useful role: it got rid of any IF which had sneaked through the audio stages.
I can't see the point. Given a good amp it is not necessary. Given a bad amp the most it can do is roll off the treble a little, and increase treble distortion - better to use a tone control.
In a wireless set the anode capacitor had an extra useful role: it got rid of any IF which had sneaked through the audio stages.
You need to scale the capacitor as well. So that 0.22 nF (=220 pF) becomes 0.55 pF. Clearly 0.55 pF is a nonsense value - stray wiring capacitance will be at least an order of magitude bigger. But so is 220 Pf on the secondary side. Did you mean 0.22 uF?
If you found that 0.22 nF improved sound quality, it measn your amp is oscillating at some high (probably many megahertz) frequency. The cause should be identified and corrected, as a zobel network on teh ouput is not likely to provide a permanent cure.
Typing mistake, again
original values were 0.220uF=220nF which will result as 550pF reflected.
The issue is that I think is better to have on secondary due to the high voltage on primary.
I should find 550pF at 1000 V at least, which i don't have easily, and it would come ceramic too!
I have a 220nF at 100V and also poliester type!
I'm wandering the sound different to have it on primay or on secondary...
original values were 0.220uF=220nF which will result as 550pF reflected.
The issue is that I think is better to have on secondary due to the high voltage on primary.
I should find 550pF at 1000 V at least, which i don't have easily, and it would come ceramic too!
I have a 220nF at 100V and also poliester type!
I'm wandering the sound different to have it on primay or on secondary...
DO NOT use a ceramic capacitor in this application. Ceramic caps vary their capacitance with voltage; they also display considerable soakage. The result is added harmonic distortion.
Some amplifiers that have an ultra-linear or other split load topology put the RC network between screen and anode. This still lets you use more convenient resistor ohm and cap capacity values, but does not need parts with high voltage ratings.
Note that the resistor in an RC network across the full primary will also need a voltage rating greater than twice the anode supply. Or use 2 or 3 resistors in series.
The carbon composition resistors, used when tubes were king, by virtue of their construction had enormous tolerance of voltage, as the voltage was evenly distributed all along teh resistor length. But the modern carbon film sort does not like high voltage due to the high electric stress across the scribe lines. A typical "0.5 W" carbon film has a voltage rating of only 250V peak.
Some amplifiers that have an ultra-linear or other split load topology put the RC network between screen and anode. This still lets you use more convenient resistor ohm and cap capacity values, but does not need parts with high voltage ratings.
Note that the resistor in an RC network across the full primary will also need a voltage rating greater than twice the anode supply. Or use 2 or 3 resistors in series.
The carbon composition resistors, used when tubes were king, by virtue of their construction had enormous tolerance of voltage, as the voltage was evenly distributed all along teh resistor length. But the modern carbon film sort does not like high voltage due to the high electric stress across the scribe lines. A typical "0.5 W" carbon film has a voltage rating of only 250V peak.
As far as compensating for loudspeaker impedance rise, which is the purpose of a zobel network on the output, no it won't give a different sound.
However, from what you have said, it isn't merely lopudspeaker impedance rise that is the problem. Your problem is that your amp is oscillating at a very high frequency. That is a whole diffrent ball game, and where you put any capacitor may make a lot of difference. It may even be different depending on how close you install it to the tube pins, even when scematically you have it in the same place.
A zobel network, regardless of whether you put it on primary or secondary, is the wrong way to fix HF oscillation. It's the wrong way because some small change, tube characteristic drift, hot weather, different speaker leads etc etc may cause the problem to re-appear, unless you slug the amp down so much the trebble response is affected.
If your amp is oscillating at HF, make sure you have "grid stopper" resistors in series with the grids (usually a control grid stopper does the job, but its good to use a screen stopper as well). It is essential to install the resistors right at the tube socket pins.
If you have no CRO, you can detect HF oscillation if you have a scanner radio. Alternatively, if you have a multimeter, switch it to ACV and with series capacitor, see if there is any volts on the transformer windings with no signal. But BE CAREFULL - many multimeters are useless at HF. Some read very low, some read very high, some don't detect any HF at all. Also, some amps only burst into oscillation when signal is present.
Last edited:
I don't have a grid stopper (I have a screen stopper) and I cannot put since everything use PCB, unless cutting a path.
I will try with a radio, because the multimeter could show the hum value.
I will try with a radio, because the multimeter could show the hum value.
OK, show me an example. I have worked and do work on many different manufacturers valve amplifiers. None, even the FU29 use Zobel networks as they are of no use.
That's weird. The control grid stopper is the more important one, as the control grid gm is the greatest. Who designed the PCB? You may have to cut a track.
Don't make silly claims or make stuff up - you merely cripple your reputation.
The fact that loudspeakers show an impedance with frequency is a widely known fact. That pentodes (& tetrodes) show an increase in harmonic distortion with increased load impedance is a widely known fact. Vast numbers of amplifiers and radios were churned out in the tube days with a capacitor or capacitor-resistor combination across the output transformer primary for just this reason. Not by any means all, because as I said there are alternative methods (eg neg feedback), but vast numbers all the same.
Zobel networks are as old as amplifiers, and they are still used. The data sheets for common audio amplifier IC's show a zobel network across the output - eg LM3886 datasheet page 6 which I have to hand right now. Same with TDA2030A application notes.
Zobel networks are common. They are common because they do indeed do something very useful - as I have explained. By making the loudspeaker appear as a flat resistance they increase the phase margin and reduce distortion. While mild neg feedback usually makes a zobel network not essential, it is the nature of negative feedback that it gives the greatest benefit when there is less for it to do. Hence a zobel network also assists negative feedback to get distortion down.
See Radiotron Designer's Handbook 4th Ed, F Langford-Smith, 1953 (In England, published by Illiffe, in other countries, RCA). Section 13.1.iii on page 546 discusses how pentodes and tetrodes show much higher distortion on loudspeaker loads compared to resistive loads due to the rising impedance as I stated above. Section 21.1.iii on page 881 in discussing the network between the tube and the loudspeaker lists the ways in which this increased distortion can be mitigated - item (3) in the list is "shunting the load with a capacitance or a capacitance and a resistance in series" - this is of course a zobel network.
Since you are English, you may care to look in issues of the magazine Wireless World. In the 1940's and early 1950's, they used to publish circuits of commercial equipment. No doubt there were trade publications in England, serviceman's guides and the like, that are compilations of circuits of commercial products. Check your library and make a survey.
Last edited:
Just realise that I cannot cut beacause is multi-layer PCB.
I could put 2K..60K stopper resistor between the DC-block capacitor (220n) and the grid capacitor (470K), so soldering to one floating leg of the capacitor to the PCB. Which value do you reccomend for EL34/KT66 ?
HH Scott 299D 0.027u in series with 68R across 16R to gnd terminals
Heathkit W3-AM -.-5u in series with 47R across 16R to Gnd terminals
Heathkit W4-AM 0.1u @200 in series with 47R across 16R to gnd terminals
Heathkit W5-AM 0.1u in series with 47R across 16R to gnd terminals
Harmon Kardon HCII, .039u in series with 39R from common to 16R terminals
Bogen CHB-35A .0012 1KV in series with 3.3K from anode to CT of transformer
Bogen DB212 1000pf in series with 1.2K from anode to Ct of transformer
Use Silver Mica, and not ceramic for caps used for Zobels if connected to anodes.
Heathkit W3-AM -.-5u in series with 47R across 16R to Gnd terminals
Heathkit W4-AM 0.1u @200 in series with 47R across 16R to gnd terminals
Heathkit W5-AM 0.1u in series with 47R across 16R to gnd terminals
Harmon Kardon HCII, .039u in series with 39R from common to 16R terminals
Bogen CHB-35A .0012 1KV in series with 3.3K from anode to CT of transformer
Bogen DB212 1000pf in series with 1.2K from anode to Ct of transformer
Use Silver Mica, and not ceramic for caps used for Zobels if connected to anodes.