Ta. So referring to post #16 again, why use a 5kohm source impedance (resistance) to drive the primaries unless the amp was going to operate in pentode mode?
Williamson used a 2.5kohm generator series resistance, and Partridge used a 3k3ohm series resistance in their datasheet for the WWFB frequency response measurement.
Edit: I had initially reasoned a quad would double the effective drive resistance, but it should be halved, so a triode quad would appropriately use 1.2kohm 'ish for frequency measurement purposes.
Williamson used a 2.5kohm generator series resistance, and Partridge used a 3k3ohm series resistance in their datasheet for the WWFB frequency response measurement.
Edit: I had initially reasoned a quad would double the effective drive resistance, but it should be halved, so a triode quad would appropriately use 1.2kohm 'ish for frequency measurement purposes.
Last edited:
I disagree on the abundance of parameters in Lundahl datasheets of output transformers. Except for declaring the leakage inductance as many other do, the high frequency region (which is the less straightforward) is not characterized at all. Not even a "typical" measurement shown. Zero.
I have used several Lundahls for me and friends. Still have the LL1627 and LL1635. High frequency response is not their strong point for sure, in general.
I would never buy a Lundahl for a classic Williamson (i.e. with tons of feedback). Their single geometry with multiple configurations doesn't really work with all their transformers.
Unfortunately this is independent of the cost. Two extreme cases are the small LL1682 (which should be a "budget" transformer but has wide FR up to 70-80 KHz) vs the LL9202. The latter is one the worst output transformers I have even seen. In SE mode it has a clear resonance at 17KHz driven by low-to-moderate plate resistance triodes and it doesn't go beyond .....it starts to roll off earlier and then the resonance brutally comes back.
That I cannot tell. The EL 34 in triode mode at 420-430V plate voltage with 50 mA to be used with fixed bias for 15W in class A will result in about 3.5-4K source resistance depending on the brand. Taken the real triode curves and drawn the composite tube. This with some current production EL34's.
So it would be 1.75-2K for parallel push-pull.
Lower plate resistance for the EL34 only happens if the tubes are biased with 70-80mA which I would not recommend at 400V anode voltage with fixed bias with current production. It was not even recommended with the originals, it has to be self bias.
I prefer the fixed bias and can get 1% THD @15W without feedback with a single driver stage. I also have a Williamson with EL34's but it's UL delivering 40W at 0.3% THD. It uses 2xECC82 and 2xEL34 per channel. I bought this as kit from a renewed electronics magazine when I was a student. It has been unused for a least 20 years now. The worst amp I have despite it measures well, true to its claimed specs.
I prefer the fixed bias and can get 1% THD @15W without feedback with a single driver stage. I also have a Williamson with EL34's but it's UL delivering 40W at 0.3% THD. It uses 2xECC82 and 2xEL34 per channel. I bought this as kit from a renewed electronics magazine when I was a student. It has been unused for a least 20 years now. The worst amp I have despite it measures well, true to its claimed specs.
Last edited:
I've been able to take manual measurements of frequency response at the extreme frequency ends (<2Hz and >100kHz) and updated the on-line doc.
It's not that simple a process but luckily I have a digital scope with isolated input channels so could just use a nominal signal generator and had one available that could output from 0.1Hz to 1MHz at 10Vrms.
I was thinking I would need to identify and construct a balanced power amplifier with suitably low distortion and sufficiently high output voltage (eg. 300Vrms) with a wide bandwidth to go the next step in OPT measurement. But I may be able to get away with a more common amplifier (ie. no need for a balanced output) by using the isolated scope inputs to allow the OPT to 'float'. And I don't think there is any benefit to test at high power and at high frequencies above say 10-20kHz, given the manual testing just done.
It's not that simple a process but luckily I have a digital scope with isolated input channels so could just use a nominal signal generator and had one available that could output from 0.1Hz to 1MHz at 10Vrms.
I was thinking I would need to identify and construct a balanced power amplifier with suitably low distortion and sufficiently high output voltage (eg. 300Vrms) with a wide bandwidth to go the next step in OPT measurement. But I may be able to get away with a more common amplifier (ie. no need for a balanced output) by using the isolated scope inputs to allow the OPT to 'float'. And I don't think there is any benefit to test at high power and at high frequencies above say 10-20kHz, given the manual testing just done.
Yes, no reason to measure FR at high power if the main purpose if to get magnitude and phase. And of course, I forgot to point out that Partridge used a more realistic rp of trioded EL34's.
I have a fully balanced amp. Lynn Olson Amity Clone. 3575 Input TX, LL1660S Interstage, Some Magnequest Nickel OPTs that were "laying about".
I also have access to a 200 MHz dual channel scope and a 100 MHz Siglent Generator.
How would I setup the measurement parameters? I assume an 8 ohm resistive dummy load. Would I setup the signal generator such that I have 1 W of power at the output and then sweep?
I could also borrow a set of differential probes that can work up to 1 kV. I'm wondering if it would be useful to take measure of magnitude and phase at each grid, or would the probe too heavily load the grids.. What would be the experimental setup conditions to harvest the most interesting data?
-- Jim
I also have access to a 200 MHz dual channel scope and a 100 MHz Siglent Generator.
How would I setup the measurement parameters? I assume an 8 ohm resistive dummy load. Would I setup the signal generator such that I have 1 W of power at the output and then sweep?
I could also borrow a set of differential probes that can work up to 1 kV. I'm wondering if it would be useful to take measure of magnitude and phase at each grid, or would the probe too heavily load the grids.. What would be the experimental setup conditions to harvest the most interesting data?
-- Jim
Jim, that seems to be a very open-ended query. it may be easier to target a specific measurement for a specific DUT, and clarify everything about that first, and then work out what options are available to make the measurement.
Are you wanting to make some measurements of each of the 3 transformers identified, or of the amp itself? There are plenty of threads on plenty of forums about complete amp testing, including the sub-forums on diyaudio.
Are you wanting to make some measurements of each of the 3 transformers identified, or of the amp itself? There are plenty of threads on plenty of forums about complete amp testing, including the sub-forums on diyaudio.
Yes Partridge used 3k3 generator resistance. At the start of the on-line doc I discussed the topic of KT66 Ra=1.45k, which would indicate a generator resistance of 2.9k would be more appropriate. Partridge released that WWFB datasheet well before the introduction of the EL34 in circa 1954.
I'm mulling over what could be done next. I can see that measuring the OPT for distortion as excitation voltage increases and extends in to the region of 'max' core flux would be valid, as OPT's have used a variety of core materials and so have different design levels for peak operational flux and frequency and different levels of intrinsic distortion based on their core material choice.
That type of testing can be at low to moderate frequency, but doesn't need to extend outside of a typical soundcard's bandwidth, and hence the full complement of automated distortion testing capabilities of REW can be applied. But using a soundcard test setup would still require balanced driver amp (my comment in post #25 was a bit premature as that wouldn't allow automated REW style measurements).
Patrick Turner went down the arduous path of setting up up a balanced power amp configuration a few decades ago that achieved low distortion with a very wide bandwidth out to 1MHz and high signal voltage output. I may have to go that way, but I think not with such a large bandwidth, but the distortion level would need to be quite low as it would need to be negligible compared to the DUT.
I am not sure that measuring the OPT distortion is worth all the trouble. You can measure the total distortion of the amplifier and will still be able to tell to a good degree which is the main source of distortion. Especially because transformers, unless badly made or defective, do not contribute to distortion at the typical 1KHz frequency unless you start reaching 0.0X levels. Such low distortion figures at all modulation levels was the trend that started with Williamson, then was typical in 70-80ies with SS amps but has never produced great results. It might have some relevance where a good fraction of the total distortion has significant high harmonics content but this doesn't happen with tubes + zero-to-moderate levels of feedback until the amp starts clipping.
Last edited:
To test in a right way each OT trafo (only) the best way is to drive the secondary with a good ss amp wiht in series the 8 ohm ( or 4 ohm); the ss can delivery around 20 Vrms (50 watt rms/ 8 ohm)
Then close the primary with a proper R that can simulate a Rp of tube or a nominal Z
With SE there are two way, one with Ibias and othe without ( this second test give you a decent results)
For the Ibias I am preparing a complete test set, now under test. It seems very interesting.
I hope to send the specs of the test set soon.
The use of this method give you also the possibilty to test the THD vs Frequency ( using ARTA p.e. and a good sound card with a proper attenuation for the signal)
This test give you the THD related to the OT only.
And it is them ost important test ( wiht Freq. resp, of course)
I spoke about this one old thread.
OPT Characterization
Walter
Then close the primary with a proper R that can simulate a Rp of tube or a nominal Z
With SE there are two way, one with Ibias and othe without ( this second test give you a decent results)
For the Ibias I am preparing a complete test set, now under test. It seems very interesting.
I hope to send the specs of the test set soon.
The use of this method give you also the possibilty to test the THD vs Frequency ( using ARTA p.e. and a good sound card with a proper attenuation for the signal)
This test give you the THD related to the OT only.
And it is them ost important test ( wiht Freq. resp, of course)
I spoke about this one old thread.
OPT Characterization
Walter
Walter, perhaps it is your english translation but you can't validly say that driving the secondary is the 'right way' or 'best way' as you have not presented your own testing to show a 1:1 match with testing that drives the primary (as far as I could tell from the linked diyaudio thread) although it appears you did try to prepare a test jig but couldn't get it generate sufficient drive voltage.
Also this thread is about OPTs operating in push-pull in class A, so your comments on SE are off topic, and perhaps threadjacking.
Also this thread is about OPTs operating in push-pull in class A, so your comments on SE are off topic, and perhaps threadjacking.
Christophe Basso's method of measuring leakage inductance.
Basso points out that the method of shorting the secondary and measuring the inductance has a shortcoming. The leakage inductance of the secondary is in parallel with the magnetizing inductance of the primary!
To calculate the correct primary leakage inductance:
Compute the turns ratio by injecting a voltage into the primary, measuring both primary and secondary voltages.
Measure the primary inductance with secondary open at a "lowish" frequency (he uses 1kHz, probably 100Hz better for audio), will call this L_ps_open, measure the inductance at a higher frequency (he uses 10kHz, mbe 1kHz for audio), call this L_ps_short
Coupling coefficient is: k = SQRT(1-(L_ps_short/L_ps_open))
Primary leakage inductance is: L_leak_primary = (1 - k) * L_ps_open
Secondary leakage inductance is L_leak_secondary = [(1 - k) * L_ps_open]/ (N^2)
Basso points out that the method of shorting the secondary and measuring the inductance has a shortcoming. The leakage inductance of the secondary is in parallel with the magnetizing inductance of the primary!
To calculate the correct primary leakage inductance:
Compute the turns ratio by injecting a voltage into the primary, measuring both primary and secondary voltages.
Measure the primary inductance with secondary open at a "lowish" frequency (he uses 1kHz, probably 100Hz better for audio), will call this L_ps_open, measure the inductance at a higher frequency (he uses 10kHz, mbe 1kHz for audio), call this L_ps_short
Coupling coefficient is: k = SQRT(1-(L_ps_short/L_ps_open))
Primary leakage inductance is: L_leak_primary = (1 - k) * L_ps_open
Secondary leakage inductance is L_leak_secondary = [(1 - k) * L_ps_open]/ (N^2)
Walter, can I suggest you use the other thread you linked to to continue discussion on driving the secondary winding, as well as presenting data from OPT's that are not aligned with the Williamson amplifier.
jackinnj, the 'correct' primary leakage inductance perhaps aligns itself with the equivalent model being used, but in this situation there is also a consideration needed for historical measurement practice so as to aid measurement comparisons (which is the path I have initially taken).
From comparison and measurement perspectives I'd suggest there is no need to compute a more refined value of a particular leakage inductance parameter, as long as the details of what is being measured are provided.
From a modelling perspective, now that many use LTSpice etc, the primary side leakage inductance model component can be calculated quite easily from the measured values, as can the secondary side leakage inductance model component. A quick look at the WWFB/0.95 measurements (which indicate measured primary inductance of 88/80H and leakage inductance of 14/16mH for the 3.8/8.5ohm secondary configurations) show:
3.8 ohm secondary config: Lps_sht/Lps-op=0.00016, k=0.99992 and model Llk_pri=7mH
8.5 ohm secondary config: Lps_sht/Lps-op=0.00020, k=0.99990 and model Llk_pri=8mH
There would also be a concern about how much influence the excitation voltage had on the leakage inductance measurement, given that excitation voltage has a very large influence on primary inductance which would then complicate the use of any modelling that use primary and secondary side leakage inductance model components based on the primary inductance value.
jackinnj, the 'correct' primary leakage inductance perhaps aligns itself with the equivalent model being used, but in this situation there is also a consideration needed for historical measurement practice so as to aid measurement comparisons (which is the path I have initially taken).
From comparison and measurement perspectives I'd suggest there is no need to compute a more refined value of a particular leakage inductance parameter, as long as the details of what is being measured are provided.
From a modelling perspective, now that many use LTSpice etc, the primary side leakage inductance model component can be calculated quite easily from the measured values, as can the secondary side leakage inductance model component. A quick look at the WWFB/0.95 measurements (which indicate measured primary inductance of 88/80H and leakage inductance of 14/16mH for the 3.8/8.5ohm secondary configurations) show:
3.8 ohm secondary config: Lps_sht/Lps-op=0.00016, k=0.99992 and model Llk_pri=7mH
8.5 ohm secondary config: Lps_sht/Lps-op=0.00020, k=0.99990 and model Llk_pri=8mH
There would also be a concern about how much influence the excitation voltage had on the leakage inductance measurement, given that excitation voltage has a very large influence on primary inductance which would then complicate the use of any modelling that use primary and secondary side leakage inductance model components based on the primary inductance value.
Last edited:
@TR
I was being pedantic, per usual. That de minimus leakage inductance on the secondary is facing a complex load on the other end!
Jack
I was being pedantic, per usual. That de minimus leakage inductance on the secondary is facing a complex load on the other end!
Jack
Jack, never too pedantic to bring up relevant details that someone may want to use (eg. for modelling) 🙂
I sort of started this thread as a way of clarifying the lack of detail related to OPT measurements, and perhaps even venturing in to unchartered waters.
I sort of started this thread as a way of clarifying the lack of detail related to OPT measurements, and perhaps even venturing in to unchartered waters.
Last edited:
Plot below is for OP25/15 (new) OPT with 10Vrms sinewave from balanced line driver and measured harmonic distortion level versus frequency for 15 ohm secondary configuration/load.
The 3rd harmonic dominates the THD, and is above the noise floor out to about 2kHz. At 20Hz, the 3rd HD is about 0.3%, and 5th is about 0.1%.
I can't presently go higher in generator source voltage. Going lower in level (eg. using the REW stepped sine measurement feature) doesn't show up anything abnormal. The odd-order HD levels from the OPT are significantly clear of the measurement noise floor, and discernible from the even-order levels, as shown by the other plot.
The test setup uses the same configuration as I used for frequency response measurements, and is just another automated reporting result from REW. The loopback calibration was applied to the harmonics as well as fundamental, and the loopback residual distortion levels are way below the measured HD levels from the OPT.
I'll test the other 3 OPT's I have been measuring to see how comparable they are.
https://www.dalmura.com.au/static/HD%20at%2010Vrms%20generator%20OP25-15%20new.png
https://www.dalmura.com.au/static/HD%20spectrum%20at%2010Vrms%20generator%20OP25-15%20new.png
The 3rd harmonic dominates the THD, and is above the noise floor out to about 2kHz. At 20Hz, the 3rd HD is about 0.3%, and 5th is about 0.1%.
I can't presently go higher in generator source voltage. Going lower in level (eg. using the REW stepped sine measurement feature) doesn't show up anything abnormal. The odd-order HD levels from the OPT are significantly clear of the measurement noise floor, and discernible from the even-order levels, as shown by the other plot.
The test setup uses the same configuration as I used for frequency response measurements, and is just another automated reporting result from REW. The loopback calibration was applied to the harmonics as well as fundamental, and the loopback residual distortion levels are way below the measured HD levels from the OPT.
I'll test the other 3 OPT's I have been measuring to see how comparable they are.
https://www.dalmura.com.au/static/HD%20at%2010Vrms%20generator%20OP25-15%20new.png
https://www.dalmura.com.au/static/HD%20spectrum%20at%2010Vrms%20generator%20OP25-15%20new.png
Well it's not negligible for the standard you want to reach but it's not high as well. Apart from telling that is clearly core distortion because it's odd only one cannot say much.
You could measure the distortion with that set-up (same conditions) just varying the frequency up to 100-200 Hz, in steps of 10Hz. The distortion might not drop at all and that means the core is not best quality (regardless if this is EI or C type), if the distortion drops you might improve but it will cost you in size and other parameters relevant for HF response.
IME, if the distortion drops at 30Hz already there is no practical advantage (musical performance) in trying to reduce that distortion at 20Hz.
You could measure the distortion with that set-up (same conditions) just varying the frequency up to 100-200 Hz, in steps of 10Hz. The distortion might not drop at all and that means the core is not best quality (regardless if this is EI or C type), if the distortion drops you might improve but it will cost you in size and other parameters relevant for HF response.
IME, if the distortion drops at 30Hz already there is no practical advantage (musical performance) in trying to reduce that distortion at 20Hz.
45, the first link in post #xx is to a HD level spectrum plot from 10Hz to 1kHz. The HD levels certainly fall with frequency.
Sorry I did not look in the link as you only mentioned 20Hz.
Then one has two options: 1) accept distortion as is ( the results are not that bad); 2) change something in the design. If have no control on the actual quality of the FeSi core, assuming there is nothing else wrong (especially the natural air-gap), the best option might be using higher permeability core.
Then one has two options: 1) accept distortion as is ( the results are not that bad); 2) change something in the design. If have no control on the actual quality of the FeSi core, assuming there is nothing else wrong (especially the natural air-gap), the best option might be using higher permeability core.