First attempt at testing

[zackthedog]

Bear with me, no laughter please. ;-). My summer goal is to teach myself to tune my amps. I just got my GW Instek GOS-620 scope and Koolertron portable function generator. I set up my breadboarded Williamson amplifier with Heyboer's Peerless S-265-Q reproductions, using the 8 ohm taps with an 8 ohm dummy load, and did some preliminary fiddling. Input voltage is .6 volts. First pic is 10K square wave. Looks like a fair amount of ringing! Second pic is with standard Williamson shelf network across the input plate resistor (4.7K + 200pF) and 240 pF across the 3.3K feedback resistor. Looks better to me. Last pic is 100Hz square wave.

I have not tried any capacitance across the load, but will experiment with that. I also have a variable cap to fine tune the feedback but haven't used it yet, I just grabbed a 150pF and a 240pF I had on hand. 240pF worked best.

Am I on the right track so far? I notice that the pre-tuning square wave has straight sides but the tuned square wav curves a bit on the sides. Is that significant?

I have proper probes on the way so that should improve the accuracy of measurement, but I was excited to take a peek and see what's going on with the amps.

I'm trying to figure out how to test for low frequency stability. Reading what I can find but any suggestions would be helpful.

ETA: Just found the thread on low frequency stability testing.

 

[zackthedog]

So after further experimenting I got it to the pix above. 1K, 10K and 20K respectively. Did I overdo it? Remarkable how much more instrumental tone comes through without the ringing. Sorry about the poor pix, I'm new at this. ;-)

Tested for LF instability with a load attached. No oscillation that I can see. Dave G recommends trying it with no load--not quite ready to venture that yet, at least until I get a switch on the dummy load to reinstate it quickly.

 

[GordonW]

The curving of the corners in the readout, is just the inevitable bandwidth limitation of all transformers. To have perfectly "square" corners, would require infinite bandwidth.

Those plots look pretty good- but the real test- when the rubber hits the road- is with capacitive loading, instead of the dummy (resistive) load. I would recommend looking up Dave G.'s instructions and advice on how to do this.

Regards,
Gordon.

 

[zackthedog]

The curving of the corners in the readout, is just the inevitable bandwidth limitation of all transformers. To have perfectly "square" corners, would require infinite bandwidth.

Those plots look pretty good- but the real test- when the rubber hits the road- is with capacitive loading, instead of the dummy (resistive) load. I would recommend looking up Dave G.'s instructions and advice on how to do this.

Regards,
Gordon.

Thanks for the response, Gordon. That is my next step.

 

[zackthedog]

Okay, so I ran a stability test by placing .22uF across the load and got oscillation. So I turned to Dave G's Heathkit mods, first the W2. The shelf network (1200pF + 1300 ohms) produced a somewhat rounded leading edge at 10K, so I tried the shelf network from his EL34 triode W4 mod (1200pF + 820 ohms) and that struck me as just about right. My aim was to duplicate as closely as possible his screen shots for the W2. I tweaked the feedback cap value a bit and came up with the first pic. It maybe could still use a little fine tuning. The second pic is from the W2 thread. The third pic is my build of his EL34 amp using the W4 output transformer. I think I got pretty close. It passes the capacitor test very well. What do we think?

All my testing with the Heyboer was in triode mode. Low frequency stability appears to be good, pulses settle down almost instantly.

Later on I'm going to test it with EL34s and see what that does, adjusting the feedback accordingly.

Practically speaking, this means someone could build one of Dave's modified Heathkits with a well-made, current production OPT as a near-drop-in replacement, without having to source parts that have become almost unobtainable now. (I have both W2 and W4 OPTs, luckily.)

 

[gadget73]

Looks pretty good to me. I've got stuff that sounds good that has significantly worse square response.

Output transformers have always been the hitching point with any feedback amp. They're all a little different, so changing things requires some fussing with the feedback to make it right. Some of them just aren't very good and will not stand large amounts of feedback no matter what you do. The Williamson amps are typically in the 20db realm for the early ones at least. the more feedback, the more touchy things tend to be.

 

[kward]

Looks pretty good to me also. But you need to try different values of reactive (capacitor)-only loading from say 0.01 uF to 20 uF. And you need to ratchet up the power, and you need to try those loads at all audio frequencies.

You also should test LF stability. Sometimes LF instability looks like HF instability. Easy way to test LF stability is to use a battery. Touch the input of the amp to the battery for a split second then release and watch how the scope trace settles. I use my signal generator producing a low frequency square wave of enough amplitude to drive the output stage to slightly over full power (i.e., onset of clipping) instead of the battery, but the intent is the same. If the scope trace bounces around as it goes to zero after you remove the signal, LF stability is marginal. The trace should settle within one bounce. I've had marginally LF stable amps take 3 to 4 bounces to settle. Really stable amps will settle so quickly you cannot see a bounce

If you poke the amp and it goes into full power LF oscillation, best to turn off the power rather than disconnect the load. On a recent amp, I tried disconnecting the cap only load while in full power LF oscillation and it arced the tube across the pins as well as ruined the tube. So be careful there.

Best to test these things on the bench rather than using your expensive speakers as the test.

 

[zackthedog]

Thanks, gadget. These are, I think, fairly faithful copies of the original Peerless. They don't have the liquid magic of a real Peerless, but they are solid and musical, with smooth highs and excellent lows. And they are hefty!

First pic is the final tweak at 10K. Then I figured out that I need to set the scope to DC offset for LF measurement, so what follows are 100Hz and 60Hz. Square wave doesn't really start to tilt until 20Hz. Not bad at all.

I think if someone wanted to build a modern "original" Williamson, this is a very good OPT to use as the foundation. Affordable, readily available and essentially drops right into a known, highly stable circuit! Honestly, you can drop this into right into Dave's W2 mod and while the HF squares waves show a little rounding at the leading edge, it frankly sounds lovely. Anyone could build it.

Anyway, lots more to learn about the using the scope!

 

[zackthedog]

Looks pretty good to me also. But you need to try different values of reactive (capacitor)-only loading from say 0.01 uF to 20 uF. And you need to ratchet up the power, and you need to try those loads at all audio frequencies.

You also should test LF stability. Sometimes LF instability looks like HF instability. Easy way to test LF stability is to use a battery. Touch the input of the amp to the battery for a split second then release and watch how the scope trace settles. I use my signal generator producing a low frequency square wave of enough amplitude to drive the output stage to slightly over full power (i.e., onset of clipping) instead of the battery, but the intent is the same. If the scope trace bounces around as it goes to zero after you remove the signal, LF stability is marginal. The trace should settle within one bounce. I've had marginally LF stable amps take 3 to 4 bounces to settle. Really stable amps will settle so quickly you cannot see a bounce

If you poke the amp and it goes into full power LF oscillation, best to turn off the power rather than disconnect the load. On a recent amp, I tried disconnecting the cap only load while in full power LF oscillation and it arced the tube across the pins as well as ruined the tube. So be careful there.

Best to test these things on the bench rather than using your expensive speakers as the test.

Thank you, kward! I will follow your suggestions, since I'm getting a bit more comfortable now with the process. And I appreciate the safety advice. I'm definitely on the "bench" (kitchen counter, to my wife's dismay) and not using my speakers.

I have a pretty good little function generator. What frequency do you use to test LF stability?

 

[kward]

I have a pretty good little function generator. What frequency do you use to test LF stability?

Doesn't really matter. I use a 100 Hz square wave. If you use a battery, of course that's like 0 hertz.

 

[dcgillespie]

Super job Zack! The only comment I would offer (constructively) is that I don't recall the square wave performance of my W-4 EL34 Williamsons as having that much ringing. If I get a chance this week, I will throw them on the bench and recheck, and post my results in your thread here if that's ok with you -- but my recollection was of a very flat wave top.

I'm encouraged to see the results you're getting from a Heyboer copy of a Peerless design. To be perfectly honest, from the personal contact I've had with others who have used their rewinding service for output transformers, the results have not been particularly good. In fact yours is the first that appears to be positive -- and what's really nice is that you are also the first person (that I am aware of) to actually post any documented performance results of a Heyboer wound output transformer, so thank-you for that!

As you get some experience under your belt with this kind of work, then when you look back on some actual vintage designs to see who got it right (generally Dynaco), and who didn't (most -- but not all -- others), then you can really get a picture of how they struggled with stability, and when the idea of advertising a response to the moon quickly had to be dumped, since it was the prime culprit of most stability problems.

Dave

 

[zackthedog]

Thank you, Dave. That means an awful lot coming from someone whose threads here I've found so fascinating and instructive. I fell in love with my first crappy, oscillating Williamson experiment a year ago and have been eager to get it working properly ever since.

Now I'm embarrassed that I posted that pic of "your" amps! I'd be delighted to do some trouble shooting on them. A very keen-eared, technically-minded colleague at our audio production company was wowed by them, and they sound awfully good, but clearly they could be better. I have number-matched OPT's, and maybe I chose the wrong feedback values. I look forward to some trouble-shooting. Bear in mind that my 'scope skills are preliminary!

I'm also building a pair of Heathkit W2 clones for a friend with a pair of the original OPTs he found on eBay. Maybe as that progresses I'll post some results and ask for advice.

I'm very glad you find the Heyboers interesting. There are precious few options for a DIYer wanting to build a genuine Williamson these days. I've been tempted to drop-ship you a pair in the service of DIY to see what you could do with them! If you're game, I am!

The development of the Williamson has become a subject of great interest to me. To my ears it combines the best of vintage audio with the beauty of modern, hi-res classical and jazz recordings, if you have speakers that will oblige. Sarser and Sprinkle had a good thing going, IMO.

 

[zackthedog]

Doesn't really matter. I use a 100 Hz square wave. If you use a battery, of course that's like 0 hertz.

Okay, thanks. My wife and daughter won't be thrilled to learn that summer vacation means me dominating the kitchen counter doing testing. ;-). But I can send them off to VidCon for a week to get them out of the way. (My daughter is a *bit* of a gamer!)

 

[zackthedog]

So I've been looking at these sine waves, and I notice that my 10K wave exhibits a slight dip along the top, while Dave's appear to be straight after the initial rise. Is the shelf network perhaps too aggressive in this case? And would I be out of line if I said I could hear a loss of "air" (or HF ease) in the tweaked amp?

 

[kward]

Transformers are not wound identically from sample to sample. There can be a smidge of variation between samples. Measure frequency response first. If your tuning networks are working you should be able to duplicate the frequency response Dave reported on his thread to within a few percent. If you’re there then I wouldn’t be overly concerned about the exact flatness of the shape.

Make sure your signal generator is flat in the measuring range. If it’s not, you’ll need to manually compensate for that as you take measurements.

 

[zackthedog]

Transformers are not wound identically from sample to sample. There can be a smidge of variation between samples. Measure frequency response first. If your tuning networks are working you should be able to duplicate the frequency response Dave reported on his thread to within a few percent. If you’re there then I wouldn’t be overly concerned about the exact flatness of the shape.

Make sure your signal generator is flat in the measuring range. If it’s not, you’ll need to manually compensate for that as you take measurements.

Thank you!

 

[gadget73]

something I've started doing is hanging one scope probe on the input as well as the output. I've had funny output that I finally realized was funny input. Being able to watch both at once makes for a handy comparison of what the amp is doing to the signal going in.

only works with at least a dual trace scope of course.

 

[zackthedog]

something I've started doing is hanging one scope probe on the input as well as the output. I've had funny output that I finally realized was funny input. Being able to watch both at once makes for a handy comparison of what the amp is doing to the signal going in.

only works with at least a dual trace scope of course.

Thanks, that's a sensible idea and there's no reason not to do it. ;-). I have a dual trace function.

 

[zackthedog]

This is interesting. First pic is 10K with just the input stage shelf network (1200pF + 1300 ohms) and NO cap across the feedback resistor. The second pic is with .22uF across the 8 ohm load. Do I *need* a feedback cap? What effect will doing without it have, if any? Third pic is 20K.

On the other hand, with the stock Williamson shelf network and a feedback cap, it is stable up to .1uF. Is that good enough or should I really be shooting for greater stability? I haven't pushed the power range yet.

Is there a way to back off the first shelf a little? I wish I could do the math but I can't...yet.
:-(

ETA: Ignore that last question, it's dumb. ;-)

 

[GordonW]

something I've started doing is hanging one scope probe on the input as well as the output. I've had funny output that I finally realized was funny input. Being able to watch both at once makes for a handy comparison of what the amp is doing to the signal going in.

only works with at least a dual trace scope of course.

Man, I've had THAT happen before. Like chasing my tail for hours, until I got the bright idea of testing the input. I then found the output VERY CLOSELY matched the input. Job done, even though it took me several hours to realize that it actually WAS working correctly...

Regards,
Gordon.