A Zobel across the output terminals of any loop feedback amplifier is valuable, and for pentode outputs, often needed for no-load stability. Current DIY fashion is to copy pre-War theater amplifiers or the French-Japanese style of the 1990's, or whatever the current fashion, so actual engineering is hard to find lately.
YOS,
Chris
YOS,
Chris
Snubber network, or zobel network: In this context they are the same thing. My first post on page 1 of this thread covers your question. And the answer is essentially yes. Not as common as with solid state amplifiers, but quite often done in tube amps. See my later posts for more design detail.
My simple answer is NO, a zobel if used should be adopted to the speakers impedance.
But a resistive load , 100 - 200 ohm permanently connected will protect the tubeamp from a no-load situation without other negative sideeffects. A 5w reistor will do fine.
But a resistive load , 100 - 200 ohm permanently connected will protect the tubeamp from a no-load situation without other negative sideeffects. A 5w reistor will do fine.
Not difficult to find. Since the purpose is the same as with a solid state amplifier, the same engineering applies.
The impedance of the zobel network should be the conjugate of the speaker impedance. That is, model the speaker as R + L in series, of impedance R + JX. Then the zobel network should be R + C in series, R - JX. In other words, the zobel network should have the same impedance as the speaker (ignoring bass resonance) but capacitive instead of inductance.
This makes the impedance seen by the amplifier as a pure resistance, which maximises stability and minimises pentode or tetrode distortion.
Any instability that remains after adding such a zobel network is due to a different cause, and the appropriate solution applied. Eg instability due to phase rotation within the amp, then apply lead compensation or adjust coupling caps.
It will help protect, but should not be relied upon. 100 to 200 ohms is 10 to 20 times the speaker impedance and will still allow a kilovolt or 2 to occur on the output tube anode upon tube cut-off, and may be sufficient to cause tube breakdown.
Many tube guitar amps (eg Fender) use 1 or 2 db of negative feedback - this is not intended to lower distortion but to make the amp tolerant of open circuit speakers, which it does quite effectively, because any dramatic increase in output voltage is substantially reduced.
thanks, will consider this, also the voltage ratings of the cap should be at least 3x B+.....
Tony, The resistor also should have a voltage rating 3x B+. Modern carbon film resistors are not good, unless you put 2 or more in series.
I've tried 2W 100R resistors across the output and blown the crap out of them...
The nice thing about my trioded sweep amps is the lowish B+ of 320V and the 7kV+ rating, and the fact I use power toroids as OPTs makes an amplifier that fool proof enough to function properly even if the load is suddenly removed (tripped over the wire?) while playing at full blast. It's also nice when you get a dud of a tube that arcs over on power up that nothing else gets destroyed excpet maybe the 10R 5W WW cathode resistors. They tend to open like a fuse on an arc over event.
More on topic, I've never used a Zobel on an OPT even when doing tetrode or UL amps, but I use them on my TY-250P audio isolator. Gets rid of ringing on 10kHz square waves in my case.
There is two issues here : A zobel will compensate reactive load ( non resistive load from a speaker). To be effective it has to involve the speaker used.
The other issue is to protect the amp from a no-load . There 100 ohm 5w across the speaker output will do well.
The other issue is to protect the amp from a no-load . There 100 ohm 5w across the speaker output will do well.
i used a 10 ohm 2 watt resistor and 0.01 ufd caps across speaker terminals in my tube power amps, and unlike in my ss amps, never had the 10 ohm burn out in use...
To be MOST effective, the zobel network must be selected to match the speaker impedance (the true R+L impedance, not just the nominal eg 8 ohm resistive impedance)
However, to be QUITE effective, it need not exactly match. A dodge became common in solid state amps 30-40 years ago to make it independent of the speaker used - put a small inductor in series with the speaker "hot" terminal. This works equally well with tube amps, of course, because the function of the zobel network is the same in both cases.
A 100 ohm, or 200 ohm resistor has some protective effect against flashover on no load, but unless there is some other feature helping (eg negative feedback, or a capacitor on the transformer primary), it will not generally be sufficient. This is seen most obviously by imagining the tube cut -off. If the standing (idle) tube current is sufficient to pull the primary fully down (say 80% of HT) with the load connected, then with 100 ohm connected, cut off must produce a peak of 100/8 x 80% of HT, ie 10 times HT. (in practice interwinding capacitance in the transformer will lower this quite a bit, but you can see that kilovolts is possible.
An EI power transformer is not so good as the relatively high leakage inductance can cause flash-over even if the speaker circuit doesn't go open. The low leakage inductance of a toroid transformer alleviates or eliminates this problem.
But, the 7Kv+ rating of a sweep tube only applies to the anode, not to the screen, so if you triode connect it, the screen rating applies. This is typically only 200 or 250V. Apart from that, sweep tubes make very good, robust, audio power stages. Some designers get around the low screen voltage rating by feeding it from a separate output transformer winding and HT rail, usually half the main HT voltage, or by series resistance and shunt gas discharge tube operated below strike voltage (not a method I would recommend - gas tubes are less reliable and other methods are cheaper).
Some very nice sounding commercial guitar amps used sweep tubes in the output stage, with the screens fed direct from a HT rail at half the main HT voltage. But I believe these were not a commercial success because of the anode top cap connection. Non-technical guitarists like to change a tube right away if one fails (you don't get paid if there's no playing) without worrying about electric shock hazards. Or going to a music shop and finding they don't stock sweep tubes.
For really severe use conditions folks might consider RC snubbers (look just like Zobels) across the primary halves of the output transformer. Solo, they don't give the same loop feedback stability advantage (because xfrmr leakage inductance doesn't appear in series) but may be enough alone.
Primary snubbers have the advantage of being directly across primary inductance where large di/dt voltages appear, and not separated by leakage inductance and other coupling imperfections.
YOS,
Chris
Primary snubbers have the advantage of being directly across primary inductance where large di/dt voltages appear, and not separated by leakage inductance and other coupling imperfections.
YOS,
Chris
There are some misconceptions in Chris Hornbeck's post #54.
1. Since the di/dt voltage across the primary is NORMAL at up up to twice the HT voltage, this is not an advantage of putting a zobel network across the primary. As has been pointed out multiple times in this thread, you'll need a capacitor AND a resistor having a voltage rating in excess of twice the HT voltage. These aren't the usual parts, they are more expensive parts.
2. Snubber or Zobel? The correct name is "zobel". as the function is teh same regardless of which side of the transformer it is on, i.e., neutralise the reactive impedance of the speaker. The term "snubber" is correctly applied if it is to suppress a switching transient, eg transistor switching off in an inverter, mains on/off switch, rectifier diode, etc. A snubber can be an R-C combination, just a capacitor, a zener diode, and many other circuits/devices, depending on circumstances.
3. A zobel network (or snubber if you insist on calling it that) at the output should never be used to fix "really severe" cases. It is more of the final icing on the cake. It shouldn't be expected to do much, either for stability or open load tolerance. That's why, although a very old idea, it often was not done done in commercial tube amps, which were designed far more with an eye on saving cost, that is modern practice.
If you have an amplifier that has a severe stability problem (ie not just a little bit of ringing on square waves WHEN a speaker is connected and not a test resistor load) use the proper fixes for this - ensuring internal load resistors are not too high, feedback phase compensation etc. Use a quality output transformer of adequate core size.
The one advantage of an RC network on the transformer primary is that it can prevent flashover due to the excessive leakage inductance of a very poor quality output transformer. This is a bit like using a heavy oil to fix a noisy car engine (a trick long used by dodgy used car sellers). It's better to fit a quality transformer.
1. Since the di/dt voltage across the primary is NORMAL at up up to twice the HT voltage, this is not an advantage of putting a zobel network across the primary. As has been pointed out multiple times in this thread, you'll need a capacitor AND a resistor having a voltage rating in excess of twice the HT voltage. These aren't the usual parts, they are more expensive parts.
2. Snubber or Zobel? The correct name is "zobel". as the function is teh same regardless of which side of the transformer it is on, i.e., neutralise the reactive impedance of the speaker. The term "snubber" is correctly applied if it is to suppress a switching transient, eg transistor switching off in an inverter, mains on/off switch, rectifier diode, etc. A snubber can be an R-C combination, just a capacitor, a zener diode, and many other circuits/devices, depending on circumstances.
3. A zobel network (or snubber if you insist on calling it that) at the output should never be used to fix "really severe" cases. It is more of the final icing on the cake. It shouldn't be expected to do much, either for stability or open load tolerance. That's why, although a very old idea, it often was not done done in commercial tube amps, which were designed far more with an eye on saving cost, that is modern practice.
If you have an amplifier that has a severe stability problem (ie not just a little bit of ringing on square waves WHEN a speaker is connected and not a test resistor load) use the proper fixes for this - ensuring internal load resistors are not too high, feedback phase compensation etc. Use a quality output transformer of adequate core size.
The one advantage of an RC network on the transformer primary is that it can prevent flashover due to the excessive leakage inductance of a very poor quality output transformer. This is a bit like using a heavy oil to fix a noisy car engine (a trick long used by dodgy used car sellers). It's better to fit a quality transformer.
Last edited:
tube output traffos have what Patrick Turner called "funny business" above the audio band, around 70khz as a result of resonance due to leakage and shunt capacitance. my thoughts are to damp this resonance using the RC network across the opt primary...
Tony T: Re your post #56:
It's very true that transformers exhibit a resonance above the audio band. But as it is a resonance (R-L-C, L due to leakage inductance, C inter-turn & interlayer capacitance), attempting to cure it with a C-R network is not the right way. You might stop any tendency to instability, but you will be left with sub-optimal performance in other respects. The only way to make a resonant circuit appear a pure resistance at all frequencies is to add an anti-resonant circuit across it - ie a series R-L-C combination, which is impractical - the added L would be another big wound component.
Note that when a speaker is connected, designing a transformer such that the leakage inductance has negligible effects in an amp is not difficult. Of course, if you use an EI-core power transformer as an output transformer, all bets are off. Power transformers are not designed to minimise leakage inductance and so don't have sectionalised windings as quality output transformers do. As reported by another poster, using a toroid power transformer is usually ok, as toroid construction minimises leakage inductance.
It seems to be forgotten or not known in tube amp circles a fundameantal design rule known to professional engineers: For stability with negative feedback, or just a good flat response free from funnies without neg feedback, the high and low frequency roll off should each be set at ONE POINT ONLY. For the low freq end, this should not be the output transformer, as otherwise you will get harmonic distortion on bass frequencies. It should be a coupling capacitor - the one feeding the output tube(s) grids(s) is best. For the high end rolloff, you can let the transformer do it, but a shunt capacitor within the amp circuit is better.
Follow this rule and you won't have any trouble with instability or funny effects like neg feedback seemingly making distortion worse instead of better.
If you know what you are doing, and have the test instruments to let you see what's hapenning, you can use coinicident or near coidicdent coupling roll-offs by using zobel networks within the amp circuit or other special techniques. But ONLY if you know what you are doing. Electronic engineering students at universities do a special subject, usually over a whole year on that.
It's very true that transformers exhibit a resonance above the audio band. But as it is a resonance (R-L-C, L due to leakage inductance, C inter-turn & interlayer capacitance), attempting to cure it with a C-R network is not the right way. You might stop any tendency to instability, but you will be left with sub-optimal performance in other respects. The only way to make a resonant circuit appear a pure resistance at all frequencies is to add an anti-resonant circuit across it - ie a series R-L-C combination, which is impractical - the added L would be another big wound component.
Note that when a speaker is connected, designing a transformer such that the leakage inductance has negligible effects in an amp is not difficult. Of course, if you use an EI-core power transformer as an output transformer, all bets are off. Power transformers are not designed to minimise leakage inductance and so don't have sectionalised windings as quality output transformers do. As reported by another poster, using a toroid power transformer is usually ok, as toroid construction minimises leakage inductance.
It seems to be forgotten or not known in tube amp circles a fundameantal design rule known to professional engineers: For stability with negative feedback, or just a good flat response free from funnies without neg feedback, the high and low frequency roll off should each be set at ONE POINT ONLY. For the low freq end, this should not be the output transformer, as otherwise you will get harmonic distortion on bass frequencies. It should be a coupling capacitor - the one feeding the output tube(s) grids(s) is best. For the high end rolloff, you can let the transformer do it, but a shunt capacitor within the amp circuit is better.
Follow this rule and you won't have any trouble with instability or funny effects like neg feedback seemingly making distortion worse instead of better.
If you know what you are doing, and have the test instruments to let you see what's hapenning, you can use coinicident or near coidicdent coupling roll-offs by using zobel networks within the amp circuit or other special techniques. But ONLY if you know what you are doing. Electronic engineering students at universities do a special subject, usually over a whole year on that.
Last edited:
I'd recommend connecting a suitably sized MOV across each half-primary winding. The MOV is effectively inert until the half-winding voltage flies off well in excess of the normal B+ swing, and then is a soft clip part.
At least one Fender guitar amp showed up parasitic oscillation (snivet like oscillation on part of waveform) on some amp examples - that amp (AB568) included speaker side feedback to PI stage. It caught many service techs out, as they tried a lot of fixes and often thought it was a bad speaker. The OT model used was the issue - it appears to have had a particularly peaky resonance circa 30-40kHz, and although the oscillation burst has been scoped the oscillation frequency hasn't been measured afaik. Swapping out the OT was the typical fix, although suppressing loop gain in the 10's of kHz by various typical primary side cap additions was also used. I'd expect a zobel on the speaker side could also be used.
At least one Fender guitar amp showed up parasitic oscillation (snivet like oscillation on part of waveform) on some amp examples - that amp (AB568) included speaker side feedback to PI stage. It caught many service techs out, as they tried a lot of fixes and often thought it was a bad speaker. The OT model used was the issue - it appears to have had a particularly peaky resonance circa 30-40kHz, and although the oscillation burst has been scoped the oscillation frequency hasn't been measured afaik. Swapping out the OT was the typical fix, although suppressing loop gain in the 10's of kHz by various typical primary side cap additions was also used. I'd expect a zobel on the speaker side could also be used.
Last edited:
ok, i have to give this up, i have searched high and low, never seen one done before....thanks guys...