Scratch building a RCA SP20, help with transformer and OPT needs.

[dcgillespie]

That is true in a perfect transformer, but once a load is connected to the secondary of a real transformer, losses increase through the transformer producing phase shift and response changes relative to the primary winding, that can increase significantly as delivered power is increased through the transformer. The best transformers minimize these losses, but even in the best transformers, the losses will still occur to some degree. In a true global NFB design, the transformer must be very effectively coupled internally to prevent upsetting loop stability under all operating conditions, while in a non-global design, the there is little concern for the transformer’s changing transfer characteristics under changing power and load conditions. Such designs therefore invariably do not use or require transformers of the caliber that global designs typically demand.

Dave

 

[gadget73]

The 6L6s you mentioned could also be used of course with the appropriate primary impedance.

yeah they wouldn't really thrive at ~5K, just thinking tubes more in the RCA playbook rather than them Euro jobs.

Such designs therefore invariably do not use or require transformers of the caliber that global designs typically demand.

Dave

I've occasionally wondered if maybe not pulling feedback from the secondary when using less than great transformers might actually produce better results than trying to force poor transformers to behave. I actually have a little stereo amp that has pretty lousy transformers on it that I need to mess with for unrelated reasons at some point, maybe I should use this as an excuse to test that idea.

 

[BassKulcha]

I'm confused. What's the point of changing all the parts of the SP-20 that make it an SP-20? What's wrong with 6V6s as outputs? That is a classic tube with its own interesting (and period) sound (@bandtuber is a big fan of this tube). I get that this is a scratch build, so you can really tweak it to be anything you want, but if the goal is an SP-20, why not build an SP-20? The RCA engineers were some of the best at the time (or any time, really).

Of course, this rigid position is my own bugaboo, probably, because I'm in the process of gathering all the parts right now for a WE 91B build. While I'm familiar with all the tweaks that have been done to that particular amp over the years, my particular windmill is that I am sticking to *exactly* the same values, voltages, and types of parts used—i.e no film subs for lytics, using discrete single-section CE Manufacturing electrolytic cans, no tweaking of values to add HF extension of or adding/reducing NFB for bass rolloff, etc., using WE wiring, period in-spec IRC and A-B resistors, rewound WE cores for OPTs, and period in-spec American-made PIOs for the hard-to-find and $$$$$ multi section oil cans mounted to the chassis) in an attempt to try to capture some of the magic of the original amps. This even extends to bringing parts of the 500A theater system into play by using external field coils as chokes. There seems to be a special magic to what the WE engineers did, and if trying to follow the build precisely may help capture even a little of it, I want to try.

Or not, and this turns out to be an unsuccessful attempt at capturing the magic, and it's just "meh". But I think sticking with the original build is a good place to start. Just my own (non EE/barely even DIY) two cents...

 

[Ken Boyd]

I will be following the drawing as closely as possible no tweaking, but I am sure it won't end up the same as an original as I will be using as close to the same values of the parts from the drawing but that won't account for the actual quality of the components that I can source today over from that period.

I have not built an amp from just a drawing but this looked like a fairly straight forward build other than sourcing good parts. I am sure I will find it very difficult considering my skill level, but what better way to learn.

 

[BassKulcha]

I will be following the drawing as closely as possible no tweaking, but I am sure it won't end up the same as an original as I will be using as close to the same values of the parts from the drawing but that won't account for the actual quality of the components that I can source today over from that period.

I have not built an amp from just a drawing but this looked like a fairly straight forward build other than sourcing good parts. I am sure I will find it very difficult considering my skill level, but what better way to learn that jumping in a doing it.

I'm definitely not implying you use original, period parts—that's just my own total nuttiness! I was just speaking to swapping out the 6V6s for 6L6s or EL34s etc.

 

[TriodeLuvr]

Such designs therefore invariably do not use or require transformers of the caliber that global designs typically demand.

I don't doubt that. But despite losses, significant coupling obviously remains, and the effects seen at the loaded secondary do affect the signals at the primary. What I recall from occasional measurements without feedback is that the phase shift at the plates of the output tubes - while less than at the secondary - is much greater than the simple 180º shift typical of the tubes themselves. I admit to not having ever attempted to characterize this in great detail, simply because (as mentioned earlier) I don't build push-pull amplifiers this way.

I've occasionally wondered if maybe not pulling feedback from the secondary when using less than great transformers might actually produce better results than trying to force poor transformers to behave.

The benefit probably depends on the specific characteristics that are "less than great." It seems to me that certain desirable amplifier traits, such as damping factor, will suffer from this. It might be interesting to experiment with transformers that aren't optimum for a full bandwidth push-pull amplifier and try a few different approaches. OPTs from Hammond organs come to mind as being virtually impossible to apply in a hi-fi design with 20dB GNFB. One alternative to allow keeping the secondaries in the loop might be to restrict the amplifier's bandwidth. Limiting the response to, say, 15kHz rather than 20kHz might appear truly wretched at first glance, but I suspect many of us wouldn't hear the difference. Another option is to reduce GNFB to something less than 10dB and to change the outputs to triodes or triode mode. That will minimize stability issues while still extending the response and improving damping.

Jack

 

[TriodeLuvr]

What's the point of changing all the parts of the SP-20 that make it an SP-20? What's wrong with 6V6s as outputs? That is a classic tube with its own interesting (and period) sound (@bandtuber is a big fan of this tube). I get that this is a scratch build, so you can really tweak it to be anything you want, but if the goal is an SP-20, why not build an SP-20? The RCA engineers were some of the best at the time (or any time, really).

I'm a big fan of the 6V6 as well. The issue for me regarding this particular amplifier lies in the fact that times have changed. I would find it objectionable to source two matched quads of these tubes for a stereo amplifier in view of simpler options. Of course, if this isn't a problem for someone else, by all means proceed as-is.

Jack

 

[dcgillespie]

The 6V6 is a wonderful tube, basically being 1/2 of a 6L6 within it's ratings limits, and the 6L6 being basically basically 1/2 of a 6550.

The beauty of the SP-20 circuit is that it can be used with virtually any transformer offering the the proper reflected load, and therefore work quite well with a transformer that otherwise would not work well in a design employing GNFB. That's because NFB always works to reduce the impedance at the point it is sourced from -- in this case, the output tube plates. With push-pull feedback from those plates, the design presents a much lower drive impedance to the transformer than typical GNFB pentode designs do, where no push-pull feedback is employed. The lower drive impedance always makes for the best possible performance from any given transformer, which is why triodes can perform very well with average grade transformers as well, where as pentodes otherwise cannot.

This is reinforced by RCA themselves: For their GNFB amplifier designs in the RC-30 Receiving Tube Manual, besides the usual impedance information offered, the transformers are also specified for frequency response, with even exact model transformer recommendation information provided as well. For the SP-20? Only impedance information is given as otherwise, while poor specifications for the chosen transformer's power and frequency response will definitely impact the performance that the amplifier can deliver, it won't impact the operation of the amplifier itself.

Dave

 

[gadget73]

The benefit probably depends on the specific characteristics that are "less than great." It seems to me that certain desirable amplifier traits, such as damping factor, will suffer from this. It might be interesting to experiment with transformers that aren't optimum for a full bandwidth push-pull amplifier and try a few different approaches. OPTs from Hammond organs come to mind as being virtually impossible to apply in a hi-fi design with 20dB GNFB. One alternative to allow keeping the secondaries in the loop might be to restrict the amplifier's bandwidth. Limiting the response to, say, 15kHz rather than 20kHz might appear truly wretched at first glance, but I suspect many of us wouldn't hear the difference. Another option is to reduce GNFB to something less than 10dB and to change the outputs to triodes or triode mode. That will minimize stability issues while still extending the response and improving damping.

Yep, this one has Hammond tone cabinet transformers. I can get good HF response, or I can make it stable. I can't do both. Its already fairly low feedback, somewhere around 10-12dB as I recall, any more and its impossible to stabilize at all. Its currently operating in a compromise arrangement where its got acceptable response, but will become unstable with certain unlikely loading conditions. If I could keep the same sort of response and resolve the stability issue I'd be OK with that. The real fix is obviously better transformers but its a small chassis that won't fit anything bigger.

I would find it objectionable to source two matched quads of these tubes for a stereo amplifier in view of simpler options.

Same, I own a PPP 6V6 amp, but I wouldn't be upset if it had been a pair of 6L6 instead. I'm not motivated to change it, but its not my first pick for a new build to clone either.

 

[TriodeLuvr]

Yep, this one has Hammond tone cabinet transformers. I can get good HF response, or I can make it stable. I can't do both.

So if you drive those with a lower impedance source like triodes and remove all feedback, are the transformers reasonably flat to 20kHz? If that's the case, I wouldn't expect the addition of a few dB GNFB (to improve damping) to create any issues. I realize this won't produce as much output power, but the sound of triodes is different and might be preferable if the output power is still sufficient.

Jack

 

[gadget73]

no, they roll off a decent bit without the feedback in pentode mode at least. I haven't tried it in triode mode. The GNFB is doing some heavy lifting to get the performance acceptable. It does about 10 watts a channel with the 6P15P tubes, I'd guess around half that in triode mode which honestly isn't enough to be useful to me. Same power with an EL84, so its not the tubes limiting things. I want to say the plate voltage is around 285, and its biased using an EFB cathode reg so there is maybe 270v to work with plate to cathode. Already solid state rectified too, so not really any more juice to squeeze on that front.

what I meant is that if I limit the bandwidth so its rolling off well below 20khz, it can be rock solid stable but its limited enough to sound bad. I can get nice square waves with really nice measured response thats flat out to 20khz, but sneeze and the thing will go off the rails. Its currently somewhere between those points, though precisely where I left it I don't remember without digging up the notes. I want to say its flat within 1db to about 17khz, but it will oscillate with a cap-only load of some value. If I can trade out the feedback coming from the speaker side for coming off the plates to maintain similar frequency response but not have the "not quite perfectly stable" caveat I'd be OK with that.

 

[nerdorama]

I've never tried it but I've seen examples of using a cap between primary and secondary to perk up the treble response although the examples I remember were interstage transformers. Not sure if it would work with an output transformer or interfere with the feedback behavior.

 

[gadget73]

Dynaco did it on something, not sure if it was the ST70 or the Mark III, maybe both? Its a cap from one output transformer primary lead into the feedback circuit, whichever amp it was. I have a small cap from one of the plates into the feedback line on my Rauland Borg amps to tame an HF ringing problem. I don't think I ever tried that on the amp with the Hammond iron, but its also something I could experiment with.

 

[BassKulcha]

I'm a big fan of the 6V6 as well. The issue for me regarding this particular amplifier lies in the fact that times have changed. I would find it objectionable to source two matched quads of these tubes for a stereo amplifier in view of simpler options. Of course, if this isn't a problem for someone else, by all means proceed as-is.

Jack

I'm curious—why do you say objectionable? 6V6s are easy (and cheap) to find, whereas EL-34s are much more expensive, generally (and specifically when it comes to the "best-sounding" ones like vintage Mullard XF-2s...) I think you could probably find a matched quad of good 6VGs under $100 (Brent Jessee has many 6V6s for sale anywhere from 15 bucks a pop and up...)

 

[TriodeLuvr]

they roll off a decent bit without the feedback in pentode mode

The output Z of pentodes is very high. Response will probably straighten out quite a bit with a low-Z drive that can better swamp the parasitic capacitance of the transformers.

It does about 10 watts a channel with the 6P15P tubes, I'd guess around half that in triode mode which honestly isn't enough to be useful to me.

Yes, well that's a different problem. Sounds like it can't be resolved with the existing design (limited B+ and limited tube dissipation).

Jack

 

[TriodeLuvr]

I'm curious—why do you say objectionable? 6V6s are easy (and cheap) to find, whereas EL-34s are much more expensive, generally (and specifically when it comes to the "best-sounding" ones like vintage Mullard XF-2s...) I think you could probably find a matched quad of good 6VGs under $100 (Brent Jessee has many 6V6s for sale anywhere from 15 bucks a pop and up...)

You don't need vintage Mullards to build an excellent sounding EL34 amplifier. And I think you're way wrong about the cost of a good matched quad of 6V6. That's just my opinion though, let me know if you beat the odds.

Jack

 

[GordonW]

The circuit is interesting

View attachment 3797899

No global feedback. Its using local feedback from output tube plate to 6AU6 screen.

Should make the output transformer selection less fussy since there won't be a bunch of feedback re-tuning to do, though you'll want something good here. The only obvious frequency tailoring I see is the cap+resistor across the first 12au7 cathode resistor, and that will be providing some HF boost. Can tweak those parts as needed if its got a rising or falling response with different transformers.

Actually, it's more interesting than that. The feedback is actually occurring at the 6AU6 CATHODE and INPUT GRID, Look at the way the feedback from the output plate, feeds directly into the biasing circuit for the 6AU6. That simultaneously feeds an AC feedback signal to both the input grid (through the input "ground" resistor}, and directly to the cathode.

The screen (G2) of the 6AU6 is actually not connected to the feedback. The SUPPRESSOR grid (G3) is, since it's directly connected to the cathode of the tube, like is done with almost all pentode and beam power tubes under normal use conditions. The way the diagram is written, makes it look a little like some funny business going on there, but it's actually standard cathode/G3 connections.

 

[GordonW]

The output Z of pentodes is very high. Response will probably straighten out quite a bit with a low-Z drive that can better swamp the parasitic capacitance of the transformers.

Yes, well that's a different problem. Sounds like it can't be resolved with the existing design (limited B+ and limited tube dissipation).

Jack

Speaking of output impedance- one advantage of EL34s vs even PPP 6V6s, is LOWER output tube impedance, as seen by the OPT input. Typical EL34 output impedance is about 10-15K ohms. Compare that to the 6V6, which is 45-65K or so. Even two 6V6s in parallel is going to be about 22K to 33K or so. That's about twice the output impedance of a single EL34.

A lower tube output impedance, will help the tube drive the output transformer, to flatter frequency response, even without global feedback. That is a definite advantage in this case.

Regards,
Gordon.

 

[TriodeLuvr]

Typical EL34 output impedance is about 10-15K ohms.

And if one really wants the best chance to flatten the response of a wayward transformer, I believe Rp in triode mode of both the 6V6 and EL34 is less then 2K. I'll be experimenting along similar lines later with a dedicated headphone amplifier that's currently in progress.

I have not built an amp from just a drawing but this looked like a fairly straight forward build other than sourcing good parts.

Ken, if you will be spending serious money on output transformers, I would suggest a pair of the 4.3K transformers for a Dynaco ST-70. The performance should be stellar with a quad of 6V6 in pentode mode.

Jack

 

[GordonW]

And if one really wants the best chance to flatten the response of a wayward transformer, I believe Rp in triode mode of both the 6V6 and EL34 is less then 2K. I'll be experimenting along similar lines later with a dedicated headphone amplifier that's currently in progress.

Ken, if you will be spending serious money on output transformers, I would suggest a pair of the 4.3K transformers for a Dynaco ST-70. The performance should be stellar with a quad of 6V6 in pentode mode.

Jack

I was thinking that same transformer option, yes. It's hard to do badly with a pair of Dyna A470s.

Regards,
Gordon.