If that is the case, you may be able to watch the 2nd harmonic slowly descend if you play a 20Hz tone strong enough and watch the FFT. I've seen this before.
It is also what I do, but typically at 20 - 50Hz depending on the transformer. Over some number of hours it generally works well.
I have a sig gen that goes up to 20 V RMS so I tried some 20Hz and ramped the voltage down slowly on the 1676, the FFT afterwards definitely showed reduced 2nd order harmonic (as much as 10dB lower in the 20-40Hz range). The 2nd order is also bumpy, rising and falling in odd patterns instead of smoothly rolling off. Still, the overall THD didn't change much because the 3rd order is so dominant.
Hard to imagine how it could be magnetized, I've never used the transformer for anything other than testing and neither my sig gen or sound card show any measurable DC component. I'm trying to remember if I ever measured ohms with my multimeter, but I'm not sure. I only did the 20hz for 2 minutes or so. I can setup a SPCI script on my PC to run the sig gen overnight and slowly ramp it from max voltage down to zero over any number of hours, not sure if it's better to stay at full power for a long time and then ramp down just over the last hour or slowly ramp the whole way. Also I'm not sure, they list 22V RMS as the max for the secondaries when in series, so in theory the max should be 5.5V RMS when they're in parallel, but I'm not sure what level is appropriate for 'degaussing' I'm assuming you want to be somewhat into saturation.
I have another LL1676 sitting in a box, I'll get that wired up too, perhaps I'll wire it as the exact data sheet 1:4 w/ 600 ohm input and 10k termination, and see if I can hit their 1%, 30Hz, 22Vrms level specified.
Hard to imagine how it could be magnetized, I've never used the transformer for anything other than testing and neither my sig gen or sound card show any measurable DC component. I'm trying to remember if I ever measured ohms with my multimeter, but I'm not sure. I only did the 20hz for 2 minutes or so. I can setup a SPCI script on my PC to run the sig gen overnight and slowly ramp it from max voltage down to zero over any number of hours, not sure if it's better to stay at full power for a long time and then ramp down just over the last hour or slowly ramp the whole way. Also I'm not sure, they list 22V RMS as the max for the secondaries when in series, so in theory the max should be 5.5V RMS when they're in parallel, but I'm not sure what level is appropriate for 'degaussing' I'm assuming you want to be somewhat into saturation.
I have another LL1676 sitting in a box, I'll get that wired up too, perhaps I'll wire it as the exact data sheet 1:4 w/ 600 ohm input and 10k termination, and see if I can hit their 1%, 30Hz, 22Vrms level specified.
Most amorphous core transformers (practically speaking all I use except for power transformers) need to be conditioned to remove residual magnetism. I did this habitually with SUTs until I decided I liked jfets better. 😈
The best transformer in many cases is no transformer at all, unless you really do need it - galvanic isolation being one very good reason and why I still use them in balanced applications. The other of course being impedance transformation in the output stage of a tube power amplifier.
I used to have a whole lot of audio transformers in my stereo system, now many of them are AES/EBU transformers. LOL I am now down to just 12 audio transformers, six of which are in the balanced inputs of my amps, 4 are output transformers and 2 are in my SACD player. Eight of them may go away in the next year or so. At one point the count was 22, and I don't miss the ones that have left to date.
The best transformer in many cases is no transformer at all, unless you really do need it - galvanic isolation being one very good reason and why I still use them in balanced applications. The other of course being impedance transformation in the output stage of a tube power amplifier.
I used to have a whole lot of audio transformers in my stereo system, now many of them are AES/EBU transformers. LOL I am now down to just 12 audio transformers, six of which are in the balanced inputs of my amps, 4 are output transformers and 2 are in my SACD player. Eight of them may go away in the next year or so. At one point the count was 22, and I don't miss the ones that have left to date.
You lowered the 2nd harmonic so maybe this solves the issue. But I wonder what happens if you use a capacitor before the transformer, maybe set to 5Hz corner? Maybe more strictly enforcing balance will make the 2nd harmonic less erratic?
I thought Lundahl trafos were shielded, but it also occurs to me that if there is something magnetic rotating nearby or turning on and off, it could be affecting the core while the measurement is taking place.
At some point the level of magnetization will go below the level that would be caused just by normal use. When magnetization distortions << core saturation distortions, there is diminishing returns, at least from a technical point of view.
I thought Lundahl trafos were shielded, but it also occurs to me that if there is something magnetic rotating nearby or turning on and off, it could be affecting the core while the measurement is taking place.
At some point the level of magnetization will go below the level that would be caused just by normal use. When magnetization distortions << core saturation distortions, there is diminishing returns, at least from a technical point of view.
I have had problems with capacitors in series with transformer primaries and that is that it creates a series resonant circuit and a huge peak in the LF response at the resonance. I try to set the LF corner somewhere below 1Hz to avoid this. (Unavoidable in parafeed applications)
Most of these transformers have rudimentary mu-metal shields that are moderately effective. I have seen all of the phenomena described in testing here. Reducing source impedance reduces distortion to a significant extent, while also causing significant HF peaking just before roll-off - usually Lundahl specifies loading and source impedances for the stated performance - changing anything alters the picture in ways that may/may not be desirable.
Real world example in my long dead 6N6P parafeed line stage thread with LL1930 transformers.
Most of these transformers have rudimentary mu-metal shields that are moderately effective. I have seen all of the phenomena described in testing here. Reducing source impedance reduces distortion to a significant extent, while also causing significant HF peaking just before roll-off - usually Lundahl specifies loading and source impedances for the stated performance - changing anything alters the picture in ways that may/may not be desirable.
Real world example in my long dead 6N6P parafeed line stage thread with LL1930 transformers.
Last edited:
If I do try to run them for many hours to break in I may do some back-to-back very large electrolytics just to be absolutely sure I'm not introducing any DC when running them that long.
Second harmonic is do to core magnetisation, core not reset or dc. For example, Beyer, the manufacturer of a tiny microphone transformer having an uncut Ni core, warns of measuring the resistance with a multimeter because even that tiny current could magnetice the core enough to impair performance.
Best to make absolutely sure theyr is not the sligthest amount of dc current possible, and that the serial cap is totally discharged and did not regain the sligthest trace of charge before connecting.
Best to make absolutely sure theyr is not the sligthest amount of dc current possible, and that the serial cap is totally discharged and did not regain the sligthest trace of charge before connecting.
Last edited:
That's the Hofer patent I meant. I once attended a demo by him where he disabled the compensation in an AP analyzer and we saw the xformer distortion shoot up.
Jan
At Calrec in the late 70s, we had an A4 sheet of 'standard' building blocks for circuitry. One of them was the output stage for M series which used bridge OPAs, a big Lundahl and output winding resistance cancellation. The date on that note is just before that of the Lundahl note and their patent. Ken Farrar has passed on so I can't find out which came first.
The Hofer patent looks like a single ended version of this. You need the temperature compensation cos as the effective resistance goes to zero, a large subsonic peak develops. You have to deal with that peak so need it accurate and stable. If the effective resistance becomes negative, yucky things happen.
Need to engage my single remaining brain cell and analyse Hofer's version. Anyone got a schematic with values?
This is from S1. The comp resistance is inside the winding and indicated with a circled 2.
Jan
Jan
Even harmonics due to DC bias - more magnetic flux swing one way than the other.
I have always held the opinion that so-called "tube sound" is in reality magnetic sound. Electrons in vacuum are tasteless, odourles and only react to their surroundings, magnetic devices for example.
The solution? get rid of the nasty core and its unpleasant quirks, like this for example:
https://www.diyaudio.com/community/...rs-lets-think-big-see-big.312699/post-5194856
A few tens of kg of copper should be no obstacle for the brave, no-compromise audiophile.
Mind you, I did an actual test (small scale obviously) and it went surprisingly well.
This is slightly different, but it also works splendidly:
https://www.diyaudio.com/community/...-evil-mad-scientist-style.224128/post-3252823
The solution? get rid of the nasty core and its unpleasant quirks, like this for example:
https://www.diyaudio.com/community/...rs-lets-think-big-see-big.312699/post-5194856
A few tens of kg of copper should be no obstacle for the brave, no-compromise audiophile.
Mind you, I did an actual test (small scale obviously) and it went surprisingly well.
This is slightly different, but it also works splendidly:
https://www.diyaudio.com/community/...-evil-mad-scientist-style.224128/post-3252823
For what it's worth, I made a python script to run my sig gen for hours with a gradually decreasing 30hz sine wave close to saturation on the transformer. I definitely saw a change in the even order harmonics. However, interestingly there was an increase in the higher-frequency second order harmonics (300-1khz), with a decrease in the lower frequencies (30-100Hz).
I discovered that REW (Room Eq Wizard) does the frequency sweep in a way that definitely produces a strong imbalanced low frequency signal. The initial low frequency sine wave starts at full power on the positive side, and the frequency increases significantly before the negative half of the wave, so there's a strong initial imbalance. Starting the sweep at higher frequencies (20Hz instead of 5-10Hz) shows much less ugliness from 20-100Hz. Subsequent runs do show weird changes in the 2nd order over a few runs, so it's still doing something to the core, but I'm not seeing a large increase in LF distortion at least. But these signal levels are low enough it's realistic regular recordings could do this anyway.
Beyond that, it does drive home two points for me.
I discovered that REW (Room Eq Wizard) does the frequency sweep in a way that definitely produces a strong imbalanced low frequency signal. The initial low frequency sine wave starts at full power on the positive side, and the frequency increases significantly before the negative half of the wave, so there's a strong initial imbalance. Starting the sweep at higher frequencies (20Hz instead of 5-10Hz) shows much less ugliness from 20-100Hz. Subsequent runs do show weird changes in the 2nd order over a few runs, so it's still doing something to the core, but I'm not seeing a large increase in LF distortion at least. But these signal levels are low enough it's realistic regular recordings could do this anyway.
Beyond that, it does drive home two points for me.
- I think I had unrealistic expectations of what unsaturated transformer distortion levels were like.
- At least for the transformers I have (lundahl inputs, some edcor interstages, some old line outputs, some generic pro audio isolators), adding typical tube plate drive impedances dramatically increases distortion throughout the low and mid frequency ranges. Like a 15dB increase across most harmonics, so 20-30 times more odd order harmonic distortion even with 'low Rp' type tubes.
Good catch. This problem shows up everywhere, even in simulation, because there is no balanced way to "start" or "stop" a sine wave. You can start mid-wave to try and balance the effect out, but where you have to start depends on the frequency response of the system you are driving. To complicate it even further, higher frequencies don't penetrate as far into the core due to eddy currents (depending on lamination/particle size), so if your frequency is not low enough you may never fully demagnetize the core.
So demagnetization is not really effective below the excitation levels of normal use.
The turns ratio does not change, so in terms of pure induction there is no distortion from one winding to another. What does change is the current required to magnetize the core at the desired level of induction, and due to the finite input winding resistance it is not normally possible to deliver all the required current. This is why we can virtually eliminate distortion by canceling the winding resistance as shown in the whitepapers shared previously in this thread. So any additional resistance as in a tube stage will dramatically worsen the distortion as you have seen.
So demagnetization is not really effective below the excitation levels of normal use.
The turns ratio does not change, so in terms of pure induction there is no distortion from one winding to another. What does change is the current required to magnetize the core at the desired level of induction, and due to the finite input winding resistance it is not normally possible to deliver all the required current. This is why we can virtually eliminate distortion by canceling the winding resistance as shown in the whitepapers shared previously in this thread. So any additional resistance as in a tube stage will dramatically worsen the distortion as you have seen.
The topic of frequency sweep stimulation signal, and how that may affect certain measurements like distortion, seems well worthwhile to explore further.
One aspect could be to automate a 'demag' sweep prior to starting a formal measurement sweep, or as part of a maintenance/service action.
Another aspect could be to automate a 'max to min' frequency sweep, like the FRA4Picoscope frequency sweep software uses. Or alternately if there is pre-cursor practical ramp up of signal amplitude that can be optioned, prior to the start of a measurement window/sweep.
Something that John Mulcahy may be able to comment on with REW.
It's not the same as room modes, but seems to have some measurement similarity, where it takes time for a mode to establish itself if wanting to make a steady state measurement at low frequency.
One aspect could be to automate a 'demag' sweep prior to starting a formal measurement sweep, or as part of a maintenance/service action.
Another aspect could be to automate a 'max to min' frequency sweep, like the FRA4Picoscope frequency sweep software uses. Or alternately if there is pre-cursor practical ramp up of signal amplitude that can be optioned, prior to the start of a measurement window/sweep.
Something that John Mulcahy may be able to comment on with REW.
It's not the same as room modes, but seems to have some measurement similarity, where it takes time for a mode to establish itself if wanting to make a steady state measurement at low frequency.