I recently did some measurements on three contemporary output transformers aimed for use in the Williamson amp.
The measurements provide a look at the primary winding inductance, as well as impedance and output frequency response over a 2Hz to 95kHz range (the limit of my USB soundcard) using REW software.
The output transformers were the venerable Partridge WWFB, plus two Australian made OPTs that came out in early 1948 to meet the Williamson diy demand (Ferguson OP25 and Red Line AF8).
https://www.dalmura.com.au/static/Partridge%20datasheets.pdf
https://www.dalmura.com.au/static/FergusonFeb1948.pdf
https://www.dalmura.com.au/static/Pages%20from%20Australasian-Radio-World-1948-06.pdf
The measurements are not for the OPT in a Williamson amp, but just of the OPT itself, but do provide an insight in to how a particular OPT could be compared to other OPT's, and to support how such an OPT could be integrated in to a Williamson amp and the required low and high frequency response tailoring that may be needed for unconditional stability.
Aspects that I thought were quite interesting were the noticeable response differences when choosing a speaker impedance setting, and the variation of measured inductance due to a variety of influences.
The write-up is linked below, and may change over time if I come across other contemporary OPT's or add to the measurements made or correct mistakes/misunderstandings. I must admit that the 'modern' soundcard/REW combination is a joy to use on a diy bench, and can but wait and hope for a new generation of soundcards that bring sub-1Hz to circa 190kHz bandwidth capability, as that would provide a nice addition to an OPT's performance assessment.
https://www.dalmura.com.au/static/Williamson%20output%20transformer%20measurements.pdf
The measurements provide a look at the primary winding inductance, as well as impedance and output frequency response over a 2Hz to 95kHz range (the limit of my USB soundcard) using REW software.
The output transformers were the venerable Partridge WWFB, plus two Australian made OPTs that came out in early 1948 to meet the Williamson diy demand (Ferguson OP25 and Red Line AF8).
https://www.dalmura.com.au/static/Partridge%20datasheets.pdf
https://www.dalmura.com.au/static/FergusonFeb1948.pdf
https://www.dalmura.com.au/static/Pages%20from%20Australasian-Radio-World-1948-06.pdf
The measurements are not for the OPT in a Williamson amp, but just of the OPT itself, but do provide an insight in to how a particular OPT could be compared to other OPT's, and to support how such an OPT could be integrated in to a Williamson amp and the required low and high frequency response tailoring that may be needed for unconditional stability.
Aspects that I thought were quite interesting were the noticeable response differences when choosing a speaker impedance setting, and the variation of measured inductance due to a variety of influences.
The write-up is linked below, and may change over time if I come across other contemporary OPT's or add to the measurements made or correct mistakes/misunderstandings. I must admit that the 'modern' soundcard/REW combination is a joy to use on a diy bench, and can but wait and hope for a new generation of soundcards that bring sub-1Hz to circa 190kHz bandwidth capability, as that would provide a nice addition to an OPT's performance assessment.
https://www.dalmura.com.au/static/Williamson%20output%20transformer%20measurements.pdf
Last edited:
Thanks Tim, that's a very comprehensive write-up. BTW I got a 404 when I tried to access your website,is it down or no longer available? I'd be interested in having a butchers at it.
Andy.
Andy.
Hopefully there is no problem with the website - all the links work afaik (I just checked with a cache refresh, and I checked when I posted) - if you continue to have hassles perhaps PM/email me. Hopefully my son put a funny 404 up.
I should take one of the Heath Williamson's with the Acrosound TO-300's back to the shop and run the analysis.
That's a great write-up. I use REW but hadn't looked into it's capabilities measuring impedance. One thing, the references in the text to the figures are all mixed up. That used to be the bane of my existance writing reports at work, until I figured out how to use Word's cross referencing feature so the text automatically updates when the figure number changes 🙂
Thanx tikiroo, I've updated the on-line doc. The links are all cross-referenced, but WORD needs manual field updating if a new figure is inserted towards the start of the doc - I guess there is a select all and update all fields process, but I seem to just keep doing a manual check.
I haven't been inclined to open up my EMU0404 yet, but REW's recent extension to below 2Hz, and the need to try and keep excitation voltage constant to get a valid spectrum may be enough incentive to crack the case (I recall that someone has swapped out the input coupling caps to extend LF response below nominal 10Hz, but I haven't seen anything on the headphone or output circuitry).
I haven't been inclined to open up my EMU0404 yet, but REW's recent extension to below 2Hz, and the need to try and keep excitation voltage constant to get a valid spectrum may be enough incentive to crack the case (I recall that someone has swapped out the input coupling caps to extend LF response below nominal 10Hz, but I haven't seen anything on the headphone or output circuitry).
Last edited:
Most of the high end VNA's allow "shaped frequency response" -- it's important, however, to use the smallest excitation voltage which gives decent results. I start at -20dBm
The main aim I thought appropriate for the tests I did was to use a constant excitation voltage, and a voltage that was consistent with vintage tests, as that then provided a comparison benchmark with respect to the OPT displaying the same primary inductance characteristic (given that inductance is quite dependant on excitation voltage level).
The EMU 0404 has some output port voltage droop below a few Hz, so I needed to define what that was to clarify where the constant excitation voltage assumption started to degrade. I added a footnote to the on-line doc yesterday about the quirk that the soundcard has in low frequency response with the selected sampling rate (eg. 44kHz sampling reduced the LF cutoff frequency performance a bit, but that then doesn't allow a single sweep to the high frequency end).
I am just starting to set up a Wavetek 111 oscillator that provides sub Hz to 1MHz range with a 10Vrms output to see if I can make some manual spot measurements outside of the soundcard bandwidth, to extend the base results.
The EMU 0404 has some output port voltage droop below a few Hz, so I needed to define what that was to clarify where the constant excitation voltage assumption started to degrade. I added a footnote to the on-line doc yesterday about the quirk that the soundcard has in low frequency response with the selected sampling rate (eg. 44kHz sampling reduced the LF cutoff frequency performance a bit, but that then doesn't allow a single sweep to the high frequency end).
I am just starting to set up a Wavetek 111 oscillator that provides sub Hz to 1MHz range with a 10Vrms output to see if I can make some manual spot measurements outside of the soundcard bandwidth, to extend the base results.
I happened to remember that EA Sowter still sells a Williamson spec'ed Transformer.
14 interleaved primary sections, 4 secondary sections! Still can be ordered. £229.47 / ea.
Yowzah!
Full measurements of leakage/pri L and a bode plot are on the data sheet:
I'd have to add - the specs are impressive!
SOWTER Type 8950
The full inventory of their OPTs is here: PUSH-PULL OUTPUT TRANSFORMERS
14 interleaved primary sections, 4 secondary sections! Still can be ordered. £229.47 / ea.
Yowzah!
Full measurements of leakage/pri L and a bode plot are on the data sheet:
I'd have to add - the specs are impressive!
SOWTER Type 8950
The full inventory of their OPTs is here: PUSH-PULL OUTPUT TRANSFORMERS
Last edited:
jrdmedford, thanks for bringing up that topic about commercial products.
The 8950 datasheet illustrates what imho is the lack of specs available. For example, there is no definition of test conditions for any of those parameters (inductance, leakage inductance, and frequency response), including what the secondary winding configuration was. Also the frequency spectrum plot doesn't align with the stated response bandwidth or the rated PP impedance, or does it identify phase or the first major HF resonance. Which is a bit sad given the price, as it boils down to caveat emptor.
The 8950 datasheet illustrates what imho is the lack of specs available. For example, there is no definition of test conditions for any of those parameters (inductance, leakage inductance, and frequency response), including what the secondary winding configuration was. Also the frequency spectrum plot doesn't align with the stated response bandwidth or the rated PP impedance, or does it identify phase or the first major HF resonance. Which is a bit sad given the price, as it boils down to caveat emptor.
Well I disagree.
Sowter uses standard procedures so FR is at 1W. They spec parameters that you NEVER get from other manufacturers, even on request, like the distortion at rated power...at 20Hz!
The picture of the FR is correct because 5K is the source impedance and 15R is the secondary connection used for that measurement. It is the same PP impedance if the rated secondary loads are used. These are not exactly 4/8/16 which is mathematically impossible if all secondary turns are to be used. Does it really make a practical difference if the secondary is 15R instead of 16R? But primary impedance in the standard audio range 20-20K and efficiency are identical in all those configurations....this is not the case with taps.
It shows the measurement with such secondary because all secondaries in series is usually the worst case. With all secondaries in parallel or anything in between it should be actually a bit better on the higher end.
The leakage inductance is rated. The resonance is easy to measure even with an oscilloscope and few other basic tools. One just need to vary the frequency and find where is the peak.....and from there the capacitance can be evaluated. If you ask they might do it for you.
P.S.
Finally primary inductance is surely not rated at max signal IMHO, otherwise you would never get 10Hz flat with 5K source impedance..so 117H has to be small signal. At worst is the same as FR.
Last edited:
45, the 8950 datasheet for frequency response has 5k:15ohms label stuck over the 'settings' info, and the settings info indicates 20dBm which is 0.1W. There is also a -10.8dBv header which could indicate a 5.5mW output across 15 ohm. The plot indicates -1.5dB at 60kHz not 70kHz. And why use a 5kohm source impedance to drive the primaries unless the amp was going to operate in pentode mode, or be used with a quad triode output stage (given the 30W capability)?
The leakage inductance is given, but not the frequency of the test, or the secondary winding configuration.
Any high frequency resonance relates to not only the frequency, but the magnitude and phase changes around that frequency, and for some transformers it can be about further resonances especially where they occur in the region of gain and phase margins.
I agree that primary inductance would not normally be rated at max signal, but the conditions of the measurement should be identified (frequency, excitation voltage, dc imbalance) otherwise 117H doesn't really mean much.
The leakage inductance is given, but not the frequency of the test, or the secondary winding configuration.
Any high frequency resonance relates to not only the frequency, but the magnitude and phase changes around that frequency, and for some transformers it can be about further resonances especially where they occur in the region of gain and phase margins.
I agree that primary inductance would not normally be rated at max signal, but the conditions of the measurement should be identified (frequency, excitation voltage, dc imbalance) otherwise 117H doesn't really mean much.
You are wrong the label 5K/15R means 5K source impedance 15R secondary. The primary impedance is 10K regardless of the secondary if 1/4/9/15 ohm secondary loads are used. This is not true a priori for transformers that have taps...it depends where is the tap which sometimes is placed not where it should be for geometrical reasons (i.e. it's a bad idea to take a tap out in the middle of a layer).
The leakage inductance can be calculated from the frequency response as explained in RDH4. There are all the necessary parameters declared. One can verify if it is plausible at least.
All transformers have multiple resonances and this is due to the different ways distributed capacitances can be combined but generally only the first high frequency resonance matters in well behaved transformers like this. The second resonance will be well above where response is basically dead...
The leakage inductance can be calculated from the frequency response as explained in RDH4. There are all the necessary parameters declared. One can verify if it is plausible at least.
All transformers have multiple resonances and this is due to the different ways distributed capacitances can be combined but generally only the first high frequency resonance matters in well behaved transformers like this. The second resonance will be well above where response is basically dead...
That Sowter Williamson clone always stayed bookmarked in my head for the impressive FR.
Here's another datasheet, but no FR graph. The Lundahl LL1620. It's a C core and lots of coil layers.
Lundahl always stuck out has having data sheets with more measurement parameters than any other vendor out there - ever. Have a look.
LL1620 Data Sheet
The point is that the Williamson OPT was specially specified for a custom wind, to get the performance the designer intended. I'm not sure every off-the-shelf transformer for sale meets those requirements....
Here's another datasheet, but no FR graph. The Lundahl LL1620. It's a C core and lots of coil layers.
Lundahl always stuck out has having data sheets with more measurement parameters than any other vendor out there - ever. Have a look.
LL1620 Data Sheet
The point is that the Williamson OPT was specially specified for a custom wind, to get the performance the designer intended. I'm not sure every off-the-shelf transformer for sale meets those requirements....
45 - to clarify, are you saying the signal generator connection to the PP primary for the purposes of the frequency response plot uses a 5kohm series resistance? And the test places a 15 ohm resistor across the 15/16 ohm secondary configuration?
The data sheet doesn't identify where the first high frequency resonance is, or more specifically what the dip level is in magnitude or in phase - as those two aspects will likely dictate the stability margins of a 'well behaved transformer' (although I don't see any statement in the data sheet that this is a 'well behaved transformer').