• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Historic Tube Amplifiers

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Ex-Moderator
Joined 2011
Turner made a clarification on the CPI's balance later in the thread by saying:

"The Miller capacitance seen by the anode or the cathode is virtually the same to certain HF, and while the load stays the same value, including the capacitance component, the balance of the CPI also stays substantially the same. However, in practice, the stray capacitance seen by the anode or the cathode is often different, therefore there is a tendency for the anode's response to sag before the cathode response, and as I have been saying for years, to make the sag in the balanced response for both outputs above 20 kHz, all you need to do is to add a little capacitor across the cathode resistor. I mentioned 15pF in my last post; but it may be too large - the actual value has to be determined experimentally."

So I guess to be generous and not beat on a dead-horse further, we could interpret Turner's comment as - if all else being equal, then the cathode compensating capacitor would not be needed at all, which mates well with your experimental results.
 
Ex-Moderator
Joined 2011
Perhaps, even though I could not find a reference on how he tested the CPI despite his voluminous writing, and it is also hard to imagine the stray capacitances being that different in well-constructed amplifiers, then again, I have not built enough amplifiers like Turner to contradict his finding.

But the point about using two identical probes when measuring CPI has been drilled into my head thanks to your revealing article, so nowdays, any claim about CPI imbalance is automatically viewed with suspicion.:)

p.s. thank you (and the other Mods) for getting rid of coinmaster, what a pest...
 
Last edited:
I just put up a doc that tries to collate design details for the Williamson amp, and make a listing of the changes applied to the circuit over the decades.

I didn't envisage it would end up over 10 pages long, as the aim was not to wax lyrical over particular pros/cons or changes, but rather to note down the known information and keep comments curt and to the point.

I'd be happy to correct any errors and add any missed items, as I had to scrounge around my limited technical library and the internet for appropriate references.

http://dalmura.com.au/projects/Williamson%20design%20info.pdf
 
Ex-Moderator
Joined 2011
I just put up a doc that tries to collate design details for the Williamson amp, and make a listing of the changes applied to the circuit over the decades.

A great summary of the Williamson! I only skimmed through it since I am traveling at the moment... but I do have a question on the step network calculation - on p.2, A' for the 6SN7 is shown as 14.3, while on p.5, A' is shown as 10, shouldn't it be 14.3-3 = 11.3? Ditto for the HF re A' ~ 6.5...:scratch1:

Also the calculated frequency response (HF ~52kHz) of the Partridge OPT, seemed to differ quite a bit from the factory spec and measured examples, which is closer to 100kHz.
 
Last edited:
Sorry for ambiguity Jazbo. I'll try and tighten up the meaning and symbols used.

The first stage V1 internal resistance, with closed loop operation, is I think going to be influenced by the unbypassed cathode resistance. So the voltage gain of that stage is going to be as per the mid-band stage gain A' ~ x14.3 (23dB).

The mid-band gain is for R3=47k, and there is no other circuitry loading on the anode.

At some frequency, the step network would reduce the effective plate load R3' down to 19k impedance, for which the gain will fall by 3dB, from x14 to x10 (20dB), which is at about 25kHz.

At frequencies beyond about 185kHz, the stage gain will plateau to x6.5 (16dB), making the total gain step about -7dB. The references indicate what that level of step means with respect to phase change and roll-off within the region of the step. Williamson doesn't go in to detail about the step performance, and Cooper (1950) indicates it is more than 9dB - I'll have another look at Coopers assumptions when I get a chance.

The WWFB datasheets indicate a measured leakage inductance and shunt capacitance levels, which I calculated out a range of simple LC resonance at between 47-58kHz, and then took a middle position as being indicative. Williamson indicated his sample gave a resonance at 60kHz, but unsure if that was diy or WWFB prototype. I have one of the WWFB/0.95, so at some stage I was going to try and recreate the factory test conditions.

http://dalmura.com.au/projects/Partridge%20datasheets.pdf
 
Last edited:
Ex-Moderator
Joined 2011
At some frequency, the step network would reduce the effective plate load R3' down to 19k impedance, for which the gain will fall by 3dB, from x14 to x10 (20dB), which is at about 25kHz.
I see where I went wrong now, it's not "14.3-3", I mixed up the multiples with the decibels! :eek:

Looking forward to see more data on the Patridge WWFB's, even though they are basically unobtainium these days...:mad:
 
I am in a quandary about the leakage inductance and shunt capacitance related parameters that Partridge used for their WWFB and CFB output transformer specs, and how they relate to the equivalent circuit models offered up to assist with estimating high frequency performance in a class A PP output stage.

In particular, Partridge refer to a half-primary leakage inductance, which I was planning to use along with the half-primary related shunt capacitance, but they also refer to a “leakage inductance measured as a series element in the primary”. The concern is that the CFB datasheet indicates the WWFB has 900mH of half-primary leakage inductance, whereas the WWFB datasheet makes no reference to that parameter.

I just updated the on-line design doc, outlining that quandary in Section 4(a), as well as cleaning up a few bits and pieces.
 
Last edited:
Yes the Naval report showed poor CFB response, with no tangible grasp as to why. The report indicates they spent time checking what feedback circuit changes could influence the particular transformer they had swapped in, so it seems strange for such a different response performance.

That report certainly 'went for broke' on the WWFB, reporting output response to 1MHz !!

The stock Williamson amp they used for OT comparisons was subtly different to the original circuit, but I couldn't identify anything that should interact with the OT to such a degree.

http://www.dtic.mil/dtic/tr/fulltext/u2/011593.pdf
 
Did find something with the title you mentioned. However the author is different.

http://www.mif.pg.gda.pl/homepages/jasiu/eka/wil/willimp.pdf

My dad built the W3M Heathkit Williamson amp in the 1950s and by the 1960s when I was a kid the amp motorboated badly. A friend of the family also built a similar Heathkit and his did it also. I worked with my dad mostly watching as he tried to fix it, he'd swap tubes and coupling caps but it would always come back. I was certain that there was something marginal with the design. I also measured the output power in UL mode and only got about 12W before clipping. The article linked above by Wright explains that when the design was brought to the US similar tubes were substituted without adjusting the operating points and by rebiasing them he is able to get much more power. The US Heathkit version used 6SN7 drivers and 5881 outputs. He also points out that more power supply decoupling was needed to eliminate the motorboating. Mystery solved, I've not tested the mods but Wright seems to have debugged the main issues with the US design.
 
For any prospective builders May I suggest you read about the original design by DTN. Motor boating can occur and the original operating point of the driver 6SN7 leads to overload near full output; my comments are prompted by personal experience some decades ago. There are solutions and its worth perusing the (copious) literature.
 
peteki, the original operating point of the driver 6SN7 has each anode sitting about 160V which allows sufficient swing for KT66's.

I tested Wright's proposed bias change on the driver stage using a 860 ohm common cathode resistance, and noted that 2nd and 3rd harmonics at 25Vrms output signal level on each anode were worse than when using DTN's original 390 ohm.
 
The original W-3M was indeed a disaster, as Dave Gillespie documented in his "From the Frying Pan Into the Fire" thread over at Audiokarma. It was quickly replaced by the W-3AM which was a far more stable design.

Having spent a year building a reconstruction of the original "Musician's Amplifier," it's possible to achieve decent stability without going through hoops, provided the output transformer is up to the job, which the Peerless S-265-Q certainly is. When you take shortcuts like Heathkit did you run into trouble.

I have tried those Wright mods and frankly, they sound awful. They drain the amp of it's character and musicality. The only sensible (and IMO musically sound) changes I've made to the MA circuit are to 1) increase the value of the first pair of coupling caps and reduce the value of the second pair, to 1uF and .1uF respectively, to reduce the chances of LF instability, 2) add a stock Williamson shelf network across the first stage plate resistor to improve HF stability, and 3) increase the driver stage cathode resistor to 560 or even 680 ohms. These were all sensible improvements made over the course of the early '50s by Keroes, Hafler and others. There were other changes as well, but those are more dependent on the specific components and tubes used.

If you still have that W-3M and want to get it working properly and turn it into an enjoyable piece, head over to Audiokarma and read Dave's "A New Beginning" thread.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.