The smallest amp I ever heard was a breadboarded spud with triode-wired E280Fs running at 180V 25mA or so, producing around 1/4 WPC. The second smallest was an SE OTL that used 4x 27GB5 per channel and burned almost 300W to produce 2x 0,75W. Both sounded surprisingly good through my old ScanSpeak 8454-based speakers
Thinking loudly…I was quite convinced that the speaker's impedance could actually go negative at the moment that the transient arrives.
Due to the inertia of the speaker, during that transient the speaker is a (momentary and weak) amp connected to our amp. An high-Zout no-feeeback amp will have less capability to control the movement of the speaker, but will then keep immediately the control back after the transient will be gone.
A low-Zout amp with lot of global feedback will control it better but then will “ring” due to “reverberation back and forth” of the feeeback.
A negative Zout amp… I don’t know…
This article discusses the effects of driving speakers from a negative impedance: https://sound-au.com/z-effects.htm#zn3A negative Zout amp… I don’t know…
To reconnect to the original question for a moment: A decade or so when "Schade" feedback was the latest fad in the tube world, one of the things people kept mentioning was how the resulting low Rp worked wonders to squeeze every single hertz of bandwidth out of cheap, small output transformers.
Yes, a transformer driven by a lower impedance will match the primary impedance (lows) at a lower frequency, and shunt capacotance (highs) at a higher frequency, therefore extending the bandwidth.
Indeed many transformers show optimistic data with real low driving impedances.
Indeed many transformers show optimistic data with real low driving impedances.
Consider that an OPT has an inductance in its primary winding. This inductance must be sufficient at the lowest desired operating frequency that its inductive reactance is not imposing a low enough apparent load to consume some of the needed audio power. Let's just say that the inductive reactance of the OPT is 1000 ohms at 20 Hz and the output impedance of the tube driving it is 1000 ohms, then about 1/2 of your precious audio power is burned up in the tube. Reduce that output impedance and more of that power gets into the magnetic field of the OPT.
An OPT also has several unwanted stray inductances and capacitances associated with it. These form a resonant circuit that often leads to a high frequency notch. Lower the resistance across that circuit and you reduce the "Q" of that resonance dropping the depth of the notch.
An OPT also has several unwanted stray inductances and capacitances associated with it. These form a resonant circuit that often leads to a high frequency notch. Lower the resistance across that circuit and you reduce the "Q" of that resonance dropping the depth of the notch.
The same mechanisms became apparent in a couple of cathode follower amps I made years ago. For better or worse they sounded more like (good) transistor amps than SE tube amps. The biggest difference was probably the increased damping factor. Those were low budget builds with cheap transformers and small(ish) tubes, would be interesting to try the same thing again with fatter bottles and bigger iron some day. I think I have a pair of Hammond 1640 OPTs and a box full of PL519 sweepers somewhere...
I clearly remember @Wavebourn saying some of his speakers play considerably better with the amp set at negative output impedance by using positive current feedback. I also remember other forumers saying that adding positive current feedback helped to stabilize too much negative feedback.This article discusses the effects of driving speakers from a negative impedance: https://sound-au.com/z-effects.htm#zn3
In each SE the frequency response with real load follows , less or more, the curve of the module of the loud speaker mainly in the bass reflex stuff where are two peak in the region where the energy is great.
Then the use of poor OT where the L at low frequency force the tube to have great distortion ( in addition to loss efficency) as for high frequency range for parasitic
My opinion ( and what I normally do) is the use of non standard value of impedance.
P.e. for 300B I use 5 ohm secondary ( only one tap) and 3-3k5 ohm for primary with always the greatest wire possible for the iron used to reduce one more parasitic value
In this way I lost some percentage in rms power but dinamically the results are very fine.
I post sometime these graph whre is displayed one test of primary coil (secondary open) for different OT
Is possible to see the resonance at mid frequency where there is the switch of the impedance module from inductive to capacitive ( also the phase)
Of course with the complete circuit the things are different but this test tell us something also how them works at low and high frequency
Then the use of poor OT where the L at low frequency force the tube to have great distortion ( in addition to loss efficency) as for high frequency range for parasitic
My opinion ( and what I normally do) is the use of non standard value of impedance.
P.e. for 300B I use 5 ohm secondary ( only one tap) and 3-3k5 ohm for primary with always the greatest wire possible for the iron used to reduce one more parasitic value
In this way I lost some percentage in rms power but dinamically the results are very fine.
I post sometime these graph whre is displayed one test of primary coil (secondary open) for different OT
Is possible to see the resonance at mid frequency where there is the switch of the impedance module from inductive to capacitive ( also the phase)
Of course with the complete circuit the things are different but this test tell us something also how them works at low and high frequency
All of the output transformers I tested on the Rohde & Schwarz vector network analyzer . . .
The phase reversal was between 500Hz and 2kHz.
If I remember, the phase reversal was mostly caused by the primary inductance and the primary distributed capacitance.
Tubelab_com talked about the inductive effect at low frequencies, and capacitance effects at high frequencies.
And the resonances from those.
In addition, there is often a resonance at very high frequencies, due to the leakage inductance between the primary and secondary, as it interacts with all of the capacitances.
Capacitances?
Primary distributed capacitance
Primary capacitance to the laminations and end bells
Secondary capacitance to the laminations and end bells.
All are according to the design details of the construction of the transformer.
Oh, I did not mention the distributed capacitance of the secondary (but if/when it was a factor, the Rohde & Schwarz with its extreme high frequency capability could find it.
I used to leave the primary un-connected, and then make lots of measuremts of the secondary.
Boy, I miss the convenience, powerful analysis capability, and great graphics, of the Rohde & Schwarz VNA.
The phase reversal was between 500Hz and 2kHz.
If I remember, the phase reversal was mostly caused by the primary inductance and the primary distributed capacitance.
Tubelab_com talked about the inductive effect at low frequencies, and capacitance effects at high frequencies.
And the resonances from those.
In addition, there is often a resonance at very high frequencies, due to the leakage inductance between the primary and secondary, as it interacts with all of the capacitances.
Capacitances?
Primary distributed capacitance
Primary capacitance to the laminations and end bells
Secondary capacitance to the laminations and end bells.
All are according to the design details of the construction of the transformer.
Oh, I did not mention the distributed capacitance of the secondary (but if/when it was a factor, the Rohde & Schwarz with its extreme high frequency capability could find it.
I used to leave the primary un-connected, and then make lots of measuremts of the secondary.
Boy, I miss the convenience, powerful analysis capability, and great graphics, of the Rohde & Schwarz VNA.
Last edited:
That is the reason for the resonant notch seen in most OPT's. In some old Hammonds with two secondaries that had to be interconnected that resonance was near or IN the audio frequency range. The 1628SE was one of the worst offenders. Wire it for 8 ohms nnd drive it with a wimpy tube and you might have a 30 db notch centered at 16 to 22 KHz, but the roll off it causes starts at about 10 KHz. The improved 1628SEA with a single secondary with two taps was considerably better. The frequency response is not bad when driven with a 300B squeezed to the edge of melting. They actually sound quite nice with a pair of fat TV sweep tubes in parallel UNSET, and the loads purposefully miswired to reflect 2500 ohms on the tubes. Well over 30 watts result.In addition, there is often a resonance at very high frequencies, due to the leakage inductance between the primary and secondary, as it interacts with all of the capacitances.
In a well made smaller OPT that notch can be in the 40 to 80 KHz range.
Boy, I miss the convenience, powerful analysis capability, and great graphics, of the Rohde & Schwarz VNA.
10 years ago you could get an HP3575A Gain - Phase Meter for $50, more like $175 now for a working one.
You'll need an XY capable scope for a display and a sweep-able signal generator. A Heathkit IG-1275 Lin/Log Sweep Generator works well. All things a designer/builder should have anyway.
With the Tek 576, you can just connect the stepper output directly to the cathode. the step amplifier can handle a couple of Amps for doing bipolar parts. My stepper V boost supply however, for tube tracing, can only handle 1/2 an Amp. For more than that you either connect a bigger external boost supply or put the P-Channel Mosfet under the cathode with a minus drain supply and step it's gate. (There will be a MosFET gate threshold offset voltage apparent in the position of the 1st curve then. There is a step offset adjust 10 turn pot on the tracer you could then remove that with.) The stepper output is from some big totem-pole bi-polar transistors with N-Fdbk V control to the step generator, So the voltage you get on the cathode with the direct connection is accurate. With the P-MOS driver you are contolling the steps on the gate instead. Not much difference, other than the offset V, if the P-MOS is rated for a few Amps.
Unset 6HJ5:
standard triode configured, 6HJ5:
Unset 6HJ5:
standard triode configured, 6HJ5:
Last edited:
smoking-amp,
Thanks to you for helping me recall some memorable times!
When I worked in Tektronix spectrum analyzer manufacturing group, then in the Tektronix spectrum analyser engineering group, I used lots of HP test equipment.
Then when I worked in Tektronix spectrum analyzer marketing, and later in Tektronix Support group (including sales support to our account managers, field application engineers, and marketing; as well as our customer's engineering groups) I had to compete when possible with all that HP test equipment.
But in all my measurement days, I never saw any stand alone-test and measurement equipment that had the convience, measurement dynamic range, accuracy, and powerful automatic calculation routines, and graphics of the Rohde & Schwars VNA that I drove for years while I was in the Tektronix support group.
Using that R&S VNA forced me to think of new and/or better measurement methods, parts and circuits to measure, and increased my engineering knowledge.
Another excellent test equipment . . . Was the Sony Tektronix 370B curve tracer that I used in the Tektronix support groups, including coming up with new ways to use it (I had to, for example I had a major customer who every 6 months would come up with another question of how to do a new measurement . . . he challenged me).
Thanks to you for helping me recall some memorable times!
When I worked in Tektronix spectrum analyzer manufacturing group, then in the Tektronix spectrum analyser engineering group, I used lots of HP test equipment.
Then when I worked in Tektronix spectrum analyzer marketing, and later in Tektronix Support group (including sales support to our account managers, field application engineers, and marketing; as well as our customer's engineering groups) I had to compete when possible with all that HP test equipment.
But in all my measurement days, I never saw any stand alone-test and measurement equipment that had the convience, measurement dynamic range, accuracy, and powerful automatic calculation routines, and graphics of the Rohde & Schwars VNA that I drove for years while I was in the Tektronix support group.
Using that R&S VNA forced me to think of new and/or better measurement methods, parts and circuits to measure, and increased my engineering knowledge.
Another excellent test equipment . . . Was the Sony Tektronix 370B curve tracer that I used in the Tektronix support groups, including coming up with new ways to use it (I had to, for example I had a major customer who every 6 months would come up with another question of how to do a new measurement . . . he challenged me).
Last edited:
Have you ever tested the utracer? It will for sure be needed to be modded in order to work with positive voltages.
I only have the Tek 576 curve tracer here, no uTracer or eTracer. The positive step voltages would be useful for grid2 or Crazy drive modes. You will need some current drive for them, maybe 100 mA. You only need the usual negative voltage stepping for doing UnSet curves (and usual grid1 curves). (but would need some current capability if driving the cathode directly )
I conversed with the maker of this 10X (non inverting) grid step amplifier below:
https://www.ebay.com/itm/266781451716
and he said he would be doing a new version with + an - V capability with significant current capability too. Maybe a while though. And no doubt more expensive.
The present model uses an Analog Devices HV Op Amp chip, good for 0V to -200V and 20 mA with pretty good bandwidth. (Might work fast enough with pulsed curve tracers.)
I conversed with the maker of this 10X (non inverting) grid step amplifier below:
https://www.ebay.com/itm/266781451716
and he said he would be doing a new version with + an - V capability with significant current capability too. Maybe a while though. And no doubt more expensive.
The present model uses an Analog Devices HV Op Amp chip, good for 0V to -200V and 20 mA with pretty good bandwidth. (Might work fast enough with pulsed curve tracers.)
Last edited:
Thanks @smoking-amp , being the p-mosfet voltage always positive (Vk=0 is the maximum current point of the loadline), how the negative voltage stepping can be used? By adding a positive reference and then step down?You only need the usual negative voltage stepping for doing UnSet curves (and usual grid1 curves). (but would need some current capability if driving the cathode directly )
Depends on where the grid1 voltage is nominally biased. (the Plate Fdbk resistor is varying that around positively too) The cathode only needs to come up to the max pos. grid1 voltage. You can run the grid1 bottom bias resistor down to a neg. bias V, adjusted so the max excursion is up to 0V when tracing. Then the cathode works between some neg. voltage and 0V. The grid1 is not drawing any current.
One -could- also arrange grid1 max +V to be positive and step the cathode from below that (maybe 0V ) up to the max +grid1 V. The stepping would then be operating in the positive range. But the cathode current is always into (electrons here) the cathode. (either way)
On the 576 you get a V offset control for the stepping range, as well as a stepping polarity (direction) push-button. And maybe an external adjustable power supply for the Vbias resistor source. I just push buttons and turn knobs until I get a good display! Can usually get it to "home-in" pretty quick.
(the external Vbias and the Offset control can be mutually adjusted to get a good curve trace set at different grid1 Vbias settings, if you want the effective avg. B+ at some specific setting -- too much bother most of the time, not much effect on the curve set )
The stepper output amplifier is totem-pole with N Fdbk. If it wants some voltage +/- X, that's where it will be, regardless of current direction. This makes for some "slop" allowance in setting things up for UnSet on the 576, but it works correctly wherever you finally end up Vbias-wise.
One -could- also arrange grid1 max +V to be positive and step the cathode from below that (maybe 0V ) up to the max +grid1 V. The stepping would then be operating in the positive range. But the cathode current is always into (electrons here) the cathode. (either way)
On the 576 you get a V offset control for the stepping range, as well as a stepping polarity (direction) push-button. And maybe an external adjustable power supply for the Vbias resistor source. I just push buttons and turn knobs until I get a good display! Can usually get it to "home-in" pretty quick.
(the external Vbias and the Offset control can be mutually adjusted to get a good curve trace set at different grid1 Vbias settings, if you want the effective avg. B+ at some specific setting -- too much bother most of the time, not much effect on the curve set )
The stepper output amplifier is totem-pole with N Fdbk. If it wants some voltage +/- X, that's where it will be, regardless of current direction. This makes for some "slop" allowance in setting things up for UnSet on the 576, but it works correctly wherever you finally end up Vbias-wise.
Last edited:
Here is some good practice before approaching UnSet set-up on the Tek 576 (not really this hard! ) :
4th picture down is video, click for full screen size, is that George with the blond hair on that yellow board?
https://www.thisiswhyimbroke.com/hydrofoil-stand-up-paddleboard/
4th picture down is video, click for full screen size, is that George with the blond hair on that yellow board?
https://www.thisiswhyimbroke.com/hydrofoil-stand-up-paddleboard/
Last edited:
- Home
- Amplifiers
- Tubes / Valves
- Strong local feedback + light global or the opposite?