IMO Fender does not do a great job when it comes to stability of their products, particularly when you consider these amps have a lineage going back to 1945, i.e. over 70 years ago. So far, every Fender valve amp I've owned has had marginal stability.
I had a Blues Junior that would oscillate (as you describe) unless the ribbon cables were positioned just so, against the aluminum foil. A nasty, Mickey-Mouse fix, if ever there was one.
I had (still have) a Super Champ XD (hybrid solid state/valve) that oscillated very easily; on inspection it had maybe the worst laid-out audio amp PCB I have ever seen. For example, there was no attempt made at all to separate the high output current-carrying PCB traces from the small-signal ground for the 12AX7 input stage of the power amp; speaker currents therefore "twiddle" the ground voltage at the input of the power amp, a form of unintentional feedback that is very likely to result in instability.
I heard that Bill M. briefly sold (or tested) a replacement output transformer for these Super Champ XD amps, but withdrew the product because many of the amps became wildly unstable if you simply put in a different output transformer. They were that marginal.
I also have a Princeton Reverb reissue. It becomes audibly unstable if you place certain guitar FX pedals on the floor near it - even up to maybe two or three feet away from it. (The Super Champ would do this, too.)
From what I can see, the "classic" Fender circuits tend to skimp on grid stopper resistors, and the reissues of vintage amps copy this "feature". I also see no attempt made to control bandwidth of the stages by design. Then there are additional stupidities such as running long ribbon cables to remotely-located valves far from the main PCB.
By contrast, I once looked inside an Epiphone Valve Junior. I found a beautifully laid out PCB, and a generally much higher quality of layout and construction. And this was in their cheapest valve amp!
I should add that I cannot speak for very recently released new Fender models. I've not opened any of them up. I hope they are better designed and constructed than minor horrors like the Super Champ XD. (Nice sounding amp, especially for the price, but internal design is, um, not great.)
-Gnobuddy
I haven't read everyrthing from post #1, but the response graphs rekindle bad moments in my life. Some pretty high-Q reaction round about 100KHz. My thoughts are not encouraging. Where I had similar problems (not with a Fender), I usually had to completely rebuild the works. End of trouble without any other changes.
I wish you patience and good luck!
I wish you patience and good luck!
Hi guys, I wanted to update the thread. All the good vibes sent my way must have worked so thanks!!
I wanted to replace some caps in the bias supply, g1 had screen voltage on it (tube shorted g2 to g1) so the bias supply caps got replaced for good measure even though they test fine. I added a bias pot for easy adjustment. I rerouted the OPT plate leads away from the ribbon cables and ran them closer to chassis. I also resoldered all the ribbon cables. I removed the protection diodes from the plate circuit of the EL84's since I have seen them cause more harm then good in these amps.
The oscillation is gone, even with the back off I can't get it to do it. I also put the amp back together and checked and it's still gone. I plan on running it for several hours and go through some on/off cycles. Then I will abuse it with some of my guitar playing and see how it holds up, but for the time being it seems fixed.
This is why I make my own amps, most of what I make turns out stable as a rock. I did run into a stability issue building my 5-20's w/ 30db of feedback and Edcor OPT, I had to lower the feedback to 20db and fine tune the feedback compensation networks. That was an educational experience for sure.
I wanted to replace some caps in the bias supply, g1 had screen voltage on it (tube shorted g2 to g1) so the bias supply caps got replaced for good measure even though they test fine. I added a bias pot for easy adjustment. I rerouted the OPT plate leads away from the ribbon cables and ran them closer to chassis. I also resoldered all the ribbon cables. I removed the protection diodes from the plate circuit of the EL84's since I have seen them cause more harm then good in these amps.
The oscillation is gone, even with the back off I can't get it to do it. I also put the amp back together and checked and it's still gone. I plan on running it for several hours and go through some on/off cycles. Then I will abuse it with some of my guitar playing and see how it holds up, but for the time being it seems fixed.
This is why I make my own amps, most of what I make turns out stable as a rock. I did run into a stability issue building my 5-20's w/ 30db of feedback and Edcor OPT, I had to lower the feedback to 20db and fine tune the feedback compensation networks. That was an educational experience for sure.
Congratulations!
You mentioned moving output valve anode leads; I have noticed the same thing, these can be a major cause of instability.
It makes technical sense: this is where the signal voltage is largest in the entire amp. Several inches of wire typically run from the output valve anodes to the output transformer. That wire may carry an AC signal that is 600 - 800 volts peak to peak in a typical small guitar amp. Only a tiny, tiny fraction of that signal needs to leak back into a preamp stage (with millivolt sensitivity
) to cause an unstable amp.
I have thought for years that valve output transformers should come with coaxial cable leads for the high-voltage (primary) side. With the coax shield (outer braid) grounded, there would be much less chance that the signal here would be coupled back into earlier stages in the amp.
-Gnobuddy
You mentioned moving output valve anode leads; I have noticed the same thing, these can be a major cause of instability.
It makes technical sense: this is where the signal voltage is largest in the entire amp. Several inches of wire typically run from the output valve anodes to the output transformer. That wire may carry an AC signal that is 600 - 800 volts peak to peak in a typical small guitar amp. Only a tiny, tiny fraction of that signal needs to leak back into a preamp stage (with millivolt sensitivity
I have thought for years that valve output transformers should come with coaxial cable leads for the high-voltage (primary) side. With the coax shield (outer braid) grounded, there would be much less chance that the signal here would be coupled back into earlier stages in the amp.
-Gnobuddy
Well, we certainly don't want additional poles in the frequency response if overall negative feedback is intended!
Then again: I don't know much about transformers intended for full-bandwidth audio, how much distributed capacitance do they typically have (from the closely-packed, parallel, internal copper wires?)
Most coax cable is in the region of 50 pF/ foot, so if you use a few inches of it, you'll probably be adding less than 20 pF of parallel capacitance to the transformer. With a 2k anode load (8k anode-to-anode) that gives you a pole at 4 MHz, two hundred times higher than the nominal 20 KHz upper limit of audio signals. Unless you have 46 dB or more of negative feedback, it's unlikely such a high frequency pole would cause any problem - open loop gain will have dropped below unity by then.
As you hinted at, when it comes to valves, guitar amps are my only area of interest. I prefer solid-state for everything else.
-Gnobuddy
If you have some suitably rated coax then it is likely to have an acceptably low capacitance for most amplifiers given the coax length would be under 1ft.
Most coax, if it has a voltage rating, is rated for 300Vrms - which should be ok for lower powered amps. Otherwise the coax should have a higher voltage rating.
Some form of overvoltage protection across the PP half-winding would be appropriate, such as a series connection of small MOV's with adequate total DCV rating. That would reduce the stress on the coax insulation should a transient occur, and in itself would add a similar shunt capacitance.
The coax braid would need to be carefully managed, as it represents a short circuit to B+ if a strand touches something.
A hi-fi output transformer may have a few hundred pF shunt capacitance from a plate-to-CT winding, so adding another 10-20pF is not too substantial. This would be unlikely to cause a stability issue in itself, but would be worth checking out if for example you knew the amp stability was marginal for starters.
But of course you need to firstly confirm that judicious routing of an anode wire does not achieve the same benefit as retrofitting coax. Similarly, identifying the sensitive part where signal is capacitively coupling over to, could well be managed with a screen local to the sensitive circuitry.
Most coax, if it has a voltage rating, is rated for 300Vrms - which should be ok for lower powered amps. Otherwise the coax should have a higher voltage rating.
Some form of overvoltage protection across the PP half-winding would be appropriate, such as a series connection of small MOV's with adequate total DCV rating. That would reduce the stress on the coax insulation should a transient occur, and in itself would add a similar shunt capacitance.
The coax braid would need to be carefully managed, as it represents a short circuit to B+ if a strand touches something.
A hi-fi output transformer may have a few hundred pF shunt capacitance from a plate-to-CT winding, so adding another 10-20pF is not too substantial. This would be unlikely to cause a stability issue in itself, but would be worth checking out if for example you knew the amp stability was marginal for starters.
But of course you need to firstly confirm that judicious routing of an anode wire does not achieve the same benefit as retrofitting coax. Similarly, identifying the sensitive part where signal is capacitively coupling over to, could well be managed with a screen local to the sensitive circuitry.
The HF resonance of a hi-fi OPT is already usually lower than we would like, given that we want to add decent amounts of feedback. Adding any capacitance will move it lower.
Good wire placement and physical design, combined with avoiding huge coupling caps, is the best method. Then, if still in trouble, add some grounded metal between the parts which should not 'see' each other.
Good wire placement and physical design, combined with avoiding huge coupling caps, is the best method. Then, if still in trouble, add some grounded metal between the parts which should not 'see' each other.
I believe you would be quite correct if we were talking about an inductor (not a transformer), or if there was no load on the transformer secondary.
But things are different when the transformer secondary is loaded with a resistive load: in an ideal world, the primary side would then behave like a resistor too, and not an inductor.
If, for example, you use one of those popular "8k plate to plate" transformers, you effectively "see" a 2000 ohm resistance between each end of the transformer and the centre-tap. In an ideal world, that's all you'd see, no additional capacitance or inductance.
Funnily enough, I have been thinking about making a quick-n-dirty single-frequency audio sine wave generator using a cheap audio coupling transformer in just the way you've been describing: leave the secondary open circuit (or very lightly loaded), whack a big fat capacitor across the primary to tune it to 1 KHz or so, and add just enough gain around it to make it oscillate. If you can get a decent Q out of it, it's relatively easy to get a clean sine wave.
A long time ago I used to make "bias oscillators" for cassette tape decks (remember those?
) using a similar method, using the self-inductance of the erase head itself as the oscillator tuning coil.-Gnobuddy
Shunt capacitance across the anode-CT winding of an output transformer is not a typically found datasheet parameter !
Specifying output transformer high frequency performance parameters really started in earnest as a result of the Williamson amplifier (and apparently ended with that amplifier as well) - link below is to Partridge output transformer datasheets for that amplifier - the shunt capacitance for a 10k PP half-winding was about 500-600pF.
Shunt capacitance can be measured by driving that winding with a signal generator through an added series resistance of about 5x the normal P-P loading, with all other windings open-circuit, and the CT connected to secondary and ground, and measuring the -3dB frequency across the driven winding, and assuming the test circuit provides a simple RC filter response - the maximum mid-band signal output should be identifiable in the low kHz region.
Amplifiers with just global feedback around the output transformer are likely to have an internal component roll-off that dominates, so as to ensure stability margin. The Williamson amp originally relied on the output transformer itself to ensure stability margin, but that didn't always translate to enough margin in diy constructions and using all commercial output transformers, so that amp was often slugged by an RC shelf filter on the input stage to avoid diy constructor problems.
A modern hi-fi output transformer designed for class AB operation is a very carefully interleaved and balanced construction of primary and secondary windings, and is therefore almost impossible to represent by simple lumped R, L and C components in an equivalent circuit. Even the linked Partridge datasheets indicate the rapid improvement in balanced interleaving and construction that was needed for amps wanting to extend beyond class A operation and in to class AB - see the half-primary leakage inductance levels for the original WWFB, versus the later CFB. Note that the WWFB is widely considered to have never been bettered for class A operation in the original Williamson amp.
http://dalmura.com.au/projects/Partridge%20datasheets.pdf
Specifying output transformer high frequency performance parameters really started in earnest as a result of the Williamson amplifier (and apparently ended with that amplifier as well) - link below is to Partridge output transformer datasheets for that amplifier - the shunt capacitance for a 10k PP half-winding was about 500-600pF.
Shunt capacitance can be measured by driving that winding with a signal generator through an added series resistance of about 5x the normal P-P loading, with all other windings open-circuit, and the CT connected to secondary and ground, and measuring the -3dB frequency across the driven winding, and assuming the test circuit provides a simple RC filter response - the maximum mid-band signal output should be identifiable in the low kHz region.
Amplifiers with just global feedback around the output transformer are likely to have an internal component roll-off that dominates, so as to ensure stability margin. The Williamson amp originally relied on the output transformer itself to ensure stability margin, but that didn't always translate to enough margin in diy constructions and using all commercial output transformers, so that amp was often slugged by an RC shelf filter on the input stage to avoid diy constructor problems.
A modern hi-fi output transformer designed for class AB operation is a very carefully interleaved and balanced construction of primary and secondary windings, and is therefore almost impossible to represent by simple lumped R, L and C components in an equivalent circuit. Even the linked Partridge datasheets indicate the rapid improvement in balanced interleaving and construction that was needed for amps wanting to extend beyond class A operation and in to class AB - see the half-primary leakage inductance levels for the original WWFB, versus the later CFB. Note that the WWFB is widely considered to have never been bettered for class A operation in the original Williamson amp.
http://dalmura.com.au/projects/Partridge%20datasheets.pdf
Last edited:
This will depend on the values. I was under that general impression until I discovered that with more than four secondaries, the equivalent capacitance dominated. Now that is also a value-dependant statement, but for OPTs in general.
As an example, one design was for a 120W OPT, running both class AB and with partial cathode feedback - thus the works more or less. Just for the record, I used only three secondaries giving some 4,7mH leakage reactance (1 KHz), while equvalent primary parallel capacitance was (about; difficult as mentioned elsewhere!) some 800pF. Measured as a simple transformer i.e. without NFB effects, and driven and terminated as per end use, I still obtained a low power bandwidth up to 220 KHz. With four secondaries and optimal sectionalising the h.f. -3dB point would have been lower. (Keep in mind that this is a rather large component measuring about 100mm cube, thus larger areas between sections.)
Thus my gut feeling was some surprise that anode-transformer leads could 'radiate' to a troublesome extent, as it were. But that pre-supposes that the wiring is kept neat and in keeping with proper layout practices, including laced in a loom, tied down, etc.
I would agree that modelling this will become cumbersome. I 'cheated' the engineer's way out, made the component, and then practically measured the characteristics and compensated accordingly looking at in-circuit performance! (One did of course know roughly what to expect.)
P.S: I am wondering about spread in h.f. characteristics in practice. Those would be quite dependant on physical matters. Compressing the windings can change leakage reactance and intersection capacitance notably. Such an action is hardly economical in production runs.
As an example, one design was for a 120W OPT, running both class AB and with partial cathode feedback - thus the works more or less. Just for the record, I used only three secondaries giving some 4,7mH leakage reactance (1 KHz), while equvalent primary parallel capacitance was (about; difficult as mentioned elsewhere!) some 800pF. Measured as a simple transformer i.e. without NFB effects, and driven and terminated as per end use, I still obtained a low power bandwidth up to 220 KHz. With four secondaries and optimal sectionalising the h.f. -3dB point would have been lower. (Keep in mind that this is a rather large component measuring about 100mm cube, thus larger areas between sections.)
Thus my gut feeling was some surprise that anode-transformer leads could 'radiate' to a troublesome extent, as it were. But that pre-supposes that the wiring is kept neat and in keeping with proper layout practices, including laced in a loom, tied down, etc.
I would agree that modelling this will become cumbersome. I 'cheated' the engineer's way out, made the component, and then practically measured the characteristics and compensated accordingly looking at in-circuit performance! (One did of course know roughly what to expect.)
P.S: I am wondering about spread in h.f. characteristics in practice. Those would be quite dependant on physical matters. Compressing the windings can change leakage reactance and intersection capacitance notably. Such an action is hardly economical in production runs.
While Christophe Basso's work is SMPS, he's written several papers on transformers, this one on leakage elements for OnSemi:
http://www.onsemi.com/pub_link/Collateral/AN1679-D.PDF
http://www.onsemi.com/pub_link/Collateral/AN1679-D.PDF
OT posts have been moved 
Skin Effect & Proximity Effect in Transformers
Attached refer to some correspondence I had with a long time designer of transformers several years ago. Doug Bannard is a Canadian Professional Engineer with considerable experience in regard to transformer design & electronics in general.
As always, it is wise to consider the advice & knowledge provided by others. Some here have differed & may still. But ‘hands on’ often trumps pedigree.
Attached refer to some correspondence I had with a long time designer of transformers several years ago. Doug Bannard is a Canadian Professional Engineer with considerable experience in regard to transformer design & electronics in general.
As always, it is wise to consider the advice & knowledge provided by others. Some here have differed & may still. But ‘hands on’ often trumps pedigree.
Attachments
Thank you for those. I had never heard of the proximity effect before (though it makes complete sense to me once I'd heard of it, as it will to anyone who understands Maxwell's equations.)
For me, "pedigree" in a technical field is entirely determined by knowledge and ability. I would be privileged to learn from a ten year old child with a runny nose if she understood the relevant math and physics better than I do.
-Gnobuddy
Proximity loss is the bane of smps designers, and is generally where most of the winding loss theory, modelling and even FEA is most well known over the last few decades. Large multiplier values on Rdc will make primary winding impedance increase and become more noticeable, especially for triode driven PP where plate resistance may be 10x the half-winding resistance. UL and pentode configurations are less impacted, but are in themselves restrictive configurations for HF roll-off. But winding resistance is not just additive to plate resistance, as the winding resistance is spread over the effective shunt capacitance of the winding (neglecting all the interactions with other windings).
Last edited: