• 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.

PP 6V6 amp from RCA RC-19 manual -- Thoughts?

I know that putting a 4 ohm speaker on an OPT's 8 ohm tap is not a good thing to do -- in general -- but the specific problem is that I have an 8k:VC OPT with only 8 and 16 ohm secondary taps. However, practically all contemporary speakers have large areas on their impedance curves that dip down to 5, 4 or even 3 ohms.

This graph is for a single EL86 (6CW5, actually) with plate voltage and screen voltage at 170V. I'd be using the 6P43P with Vp = 265V and Vg2 = 245V.

1726451292553.png


The above graph shows the deepest distortion null for a single EL86 pentode at about 2100 ohms. For push-pull class A operation, you'd double that. So now we're at 4200 ohms.

For highest power output, the peak for a single EL86 pentode is at about 1750 ohms. Again, double that for PP class A, and now we're at 3500 ohms.

So a PP primary load somewhere between 3500 and 4200 ohms would seem to be the sweet spot.

Push-pull EL86 into a 4k p-p primary load seems like a pretty good spot. I would want it a little on the high side with the higher plate and screen voltages. That's why I was thinking of using the Z565 with a nominally 6 ohm speaker on its 8 ohm tap. That speaker's impedance curve does dip down to 4 ohms, but they all do these days.

Is there literally no way to use this Dynaco Z565 OPT with a contemporary speaker system?
--
 
I was also thinking about the profile of harmonic distortion vs. OPT primary impedance.
In this graph from the data sheet for the old metal 6L6, the primary impedance is varied with power output and the various levels of individual harmonics charted, for pentode operation with 250V on both plate and screen:

1726452667933.png


The optimal load impedance for single ended pentode operation is shown as 2500 ohms.
Lowering that to 1500 ohms reduces the power output by a third, but H3 is now practically absent, with H2 up a bit. In push-pull operation, H2 from the output tubes would be at least partially cancelled in the OPT, would it not? That would argue for going lower than optimal for cleanest operation open loop.

This is the opposite of the situation for triode operation. For power output triodes, raising the primary impedance reduces THD along with reducing power output, while lowering primary impedance increases both power output and THD.

Keep in mind that the above graph (and the one for the 6CW5/EL86) were drawn with no NFB applied.

So here's a question. With the feedback scheme in the RCA SP-10 amplifier, are the output tubes still acting as pentodes? Doesn't the local NFB make the output tube act like a triode? (According to Schade's infamous paper, it certainly lowers the tube's apparent plate resistance by a lot.) Remember the triode input stage, phase splitter and OPT are not within the feedback loop. All the NFB is between the output tube's plate and the driver tube's cathode.

I point this out because in simulations, I get more power and higher H3 when I reduce the OPT primary impedance, and less power but with lower H3 when I increase the OPT primary impedance. That's very similar to what I'd expect from a push-pull triode output stage.

Does the NFB arrangement create a sort of composite tube out of the output pentode and driver pentode?
 
Last edited:
Maybe the best way to think about NFB is that it reduces gain and distortion in the same ratio at any particular frequency and level. In a flat amplifier it cannot change the distribution of harmonics or the clipping level.

In this feedback arrangement, the reduction of source impedance to the OPT does reduce the odd (predominantly third) order distortion arising from the OPT core.

All good fortune,
Chris
 
Rongon,

1. Thanks for posting a graph of the JBL Studio 530 in Post # 153 !
It has a nominal impedance rating of 6 Ohms.
You measured 4.8 Ohms DCR.

The woofer operates up to at least 1 kHz.

Notice the very low AC Impedance Ohms of the woofer which are extremely close to the DCR you measured of 4.8 Ohms.

2 Hz 4.5 Ohms
10 Hz 5 Ohms
44 Hz 5.5 Ohms
140 Hz 5 Ohms
180 Hz 4.5 Ohms
250 Hz 5 Ohms
And, In the enclosure, the woofer impedance peaks are at 50 Hz and 73 Hz.

In my Post # 150, I described those characteristics for a woofer in a ported enclosure:
Two resonant woofer impedance peaks
Impedance below the first resonant peak
Impedance null between the first and second resonant peaks
Impedance null above the second resonant peak.


2. Unfortunately, with an output transformer that only has an 8 Ohm tap, the performance will not be optimum when used with a loudspeaker that has minimum impedances as low as 4.5 Ohms.
 
Last edited:
with an output transformer that only has an 8 Ohm tap, the performance will not be optimum when used with a loudspeaker that has minimum impedances as low as 4.5 Ohms.

Yes, I knew that. But will it be good enough if I'm using those low rp EL86 tubes with a Dynaco Z565?

My thinking was that EL86 has much lower rp than EL84.
The Z565 OPT was designed for use with EL84 tubes.
What happens if I use that same transformer but with the lower-rp EL86 tubes? Wouldn't that work better for driving lower-than-intended impedance loads?
 
Last edited:
Once you put a check mark in the boxes for safety of the speakers and amp, what have you got to do but try it? The thing is, is this the nostalgic, vintage amp you are trying to build, or now just a compromise to fit the parts you have? It'll be a one-off amp you'll have to explain and then get the usual polite affirmations for its qualities. Since you have some 6-7 ohm speakers, what is the goal of the build? Just to use some EL86's?
 
what is the goal of the build? Just to use some EL86's?

Actually, that is part of the goal. Also, just to use the Z565 OPTs. Both.
I have two pairs of actual 8 ohm speakers, Snell Type C (a large 3-way from the 1980s) and a cosmetically scruffy pair of Sonus Faber Concerto (a ca-2000 6.5 inch 2-way bookshelf speaker). But my two favorite speakers are more like 4 to 6 ohm speakers (JBL Studio 530 and Snell Type E-III).

So I'm rethinking trying to use the Z565 OPTs, because this circuit looks so promising. Why compromise it?

is this the nostalgic, vintage amp you are trying to build, or now just a compromise to fit the parts you have?

I'm always looking for ways to use the cache of parts I've collected over the years. But I get the point about using those parts in optimal ways, rather than pushing square pegs into round holes.

I'm really curious about this old RCA design. It simulates so well, yet it looks so wrong (current-starved 6AU6s, weird biasing of the driver stage, etc.).

I also have a pair of Dynaco A470 OPTs (about 5k:VC), which should work well with EL86, so I guess I'm back to those. There are worse things, right?

I just need to decide what I'm going to do before I start the dreaded metalwork. I don't enjoy metalwork at all, unfortunately. I do have an aluminum chassis drilled and punched for a stereo PP amp w/ 9 pin mini sockets. That's another reason why I was thinking of going with EL86s. Perhaps PP EL86 with Dyna A470 OPTs.
 
I would lower the screen voltage to the max rating of 200V, ideally regulated, since idle-to-max screen current is 1:6 ratio. UL or triode is risky with EL86 at 250V. The ideal Z for AB1 would be around 3K, so again, using a 4 Ohm speaker on the 8 Ohm tap might be best - or keep it class A with 5K. The A-470 has less copper loss and less leakage inductance than the Z565 - win-win. I have a pair of 4K "50 watt" EL34 transformers that I've thought of using for EL86s. And a pair of Sherwood 6Ks that would be terminated at half impedance.
 
Yes, the American data sheets for 6CW5 and the Philips data sheet for EL86 both show max Vg2 = 200V.

On the other hand, the only data sheet I could find for 6P43P-E states that the max Vg2 is 250V (which is why I chose 245V in the current design idea).
https://vintageguitaramprepairs.co.uk/data-sheet/6P43P.pdf
(I have 6P43P-E tubes, not actual EL86s.)

There seems to be no solid consensus on max Vg2 from the various Euro manufacturers.

Mazda EL86F data sheet says max Vg2 is 250V:
https://frank.pocnet.net/sheets/020/e/EL86F.pdf

The Mullard EL86 data sheet says max Vg2 is 250V:
https://frank.pocnet.net/sheets/129/e/EL86.pdf

Yes, the A470 would be better than the Z565 with a pair of 6P43P-E.
But the Z565 is generally thought to be a better transformer than the A470. Is that just people saying the Z565 'sounds better'? Or is that because there are technical and objective reasons why the Z565 is more highly regarded?

Maybe I should go back to the idea of using a pair of 6JC5. That has a max Vg2 of 300V, so would be quite safe, and I have a few different OPTs I could use with it (choice of 6.6k or 8k primaries), all with 4 ohm taps, in addition to the A470 (5k primary).
 
My gut feeling tells me that a PP pentode output stage with heavy local feedback, like this one, probably is one of the better ways to build a tube amp that can handle large variations in the load impedance?
Re. the 6CW5/EL86 G2 voltage I've done some reading since I bought a couple dozen Zaerix PL84 (the euro 300mA heater version of this tube) a while ago. I seem to remember that the key to success for a long tube life was to keep the screen grid voltage low (170V), especially when the plate voltage is on the high side to maximize output power.
 
Just putting this out there, for a data point.
It's the insert/data sheet from a brand-new-in-the-box 6P43P-E, USSR era.
I believe it lists the max plate voltage as 300V, and the max screen voltage as 250V.
The max plate dissipation is shown as 12W; max screen dissipation as 2W.

6P43P-E_Reflektor.jpg


Here's a translation to German I found somewhere, years ago...

1726533749100.png


That also shows the max Vg2 voltage as 250V.

The German author seems to think the 6P43P-E is similar to 6P18P, so he shows the specs from a 6P18P data sheet. Strangely enough, they line up with the Soviet data sheet in the first graphic.

Of course these things are relative. I think if you use an EL86 with a plate voltage >350V, you'll need to keep the screen voltage down under 200V. But if you keep the plate voltage at 265V? Wouldn't that allow you to safely raise the screen voltage?
 
Perhaps the russkies are a bit sturdier than their western equivalents, I don't know?
The discussions I read were about western brands and I don't remember the exact numbers but the general conclusion was that these tubes are happier at lower G2 voltages. 245/265V is probably quite safe, I ran some of my PL84s at around 240V on both plates and screens for some time in a project but it had a hum problem that I couldn't fix so I it didn't live long enough to get any indications of tube life.

Nice tubes though, I'm contemplating a PPP project with 4x PL84 per channel and a pair of huge mains toroids as OPTs, I have a pair that measures and sounds quite good when used as 2K:8R transformers with 24% CFB from a 37-0-37V secondary winding.
 
I've finished a bunch of simulations of the proposed circuit, some using a 6JB5 model for the output tubes, others using an EL86 model.
Remember, what you see below is from simulations, meant for comparisons only. I don't expect THD to be this low in real life

For 6JC5/6JB5, with Vp = 295V, Vg2 = 290V, Ip+g2 = 52.5mA, I've found that:
  • Lowering the plate-plate primary impedance (Zpri) of the output transformer (OPT) increases output power (24W at 1.1% THD), and increases H3 in the output by almost 10dB. Using a model of a Dyna A470 OPT (approx 5k:VC) with 1kHz sine wave input, THD at 1W output into 8 ohms is 0.02%. H2 -76dB, H2 -75.5dB.
  • Raising the Zpri to 8k:VC decreases output power (17W at 1.1% THD), and also decreases H3 in the output. Using a modeled Acrosound A310 OPT, with a 1kHz sine wave, THD at 1W into 8 ohms is 0.014%, H2 -78dB, H3 -85dB. The level of H3 at 1W out is -10dB down from the same circuit using the 5k:VC OPT.
For EL86 run with Vp = 270V, Vg2 = 210V, Ip+g2 = 45.5mA, and using an Acro TO-310 OPT with a 4 ohm load on its 8 ohm secondary tap (to simulate using the Dyna Z565), the simulation predicts mind-boggling performance.
  • At 1W into 4 ohms (from the 8 ohm tap), THD = 0.008%, H2 -82.5dB, H3 -89dB. Max power out is 21W at 1.15%. Doesn't that seem too good to be true?
I don't have a model of the Dyna Z565, so I used the Acro TO-310 instead.
I also tried using a model of the Hammond 1650F (7.6k:VC). That yields similar output power, but H3 is quite a bit higher, which I'd like to avoid.

Again, I have no idea how well these simulations will carry over into real world results. My experience with designing phono preamps, line stages and an SET headphone amp this way is that the simulations predict DC conditions very well, and AC conditions like max signal output fairly well.

So, how about those numbers from this circuit? Pretty amazing, no? Too good to be true? Probably. But you can see why I got excited about using EL86s. The circuit seems to work really well with that type for the output tubes.
 
Last edited:
I've never learned to use any (modern) simulation softwares so I can't say anything about their accuracy. I guess they can only be as good as the input data, ie the component models. The plate curves in 70-80 years old datasheets are probably so-so and output transformers are probably very complex to simulate accurately.

What I think is happening here is that the strong local feedback makes the output tubes behave as very linear, low Ra triodes similar to what happens in circuits with "schade" feedback.
 
"Schade" feedback is actually plate-to-grid feedback around the output tubes only. In other words, you wire up your output pentode as an "anode follower".

The local feedback in this RCA SP-10 design is from output tube plate to driver tube cathode. I think it's basically making the output stage into a single block. The combination of the driver stage and output stage yields very little gain (due to 23dB or so of NFB). By the time the output waveform is stepped down through the output transformer, the gain from the driver stage inputs to the speaker output is only about 1.5X. Practically all the gain in this design comes from the first stage voltage amplifier (the phase splitter has a little less than unity gain).

So I would very much like to agree that the strong local NFB makes the combination of the driver pentodes and output pentodes behave like a pair of low rp triodes with very low distortion. That's what Schade described in his old paper about the 6L6. With the plate-grid NFB applied, a curve tracer showed curves from the 6L6 that looked exactly like a triode's, but with much higher power output than from simply triode-strapping. Pretty cool, I think.

As for the models... For the tubes, I've collected the best-regarded models I could find. Many of them are courtesy of diyAudio member Adrian Immler, who I believe has done some excellent work. Adrian traces curves himself from multiple examples of a particular brand/type, and compiles his model from a composite/average of those examples, or from a single example he considers an average representative of the type. The previously best-regarded method for modeling vacuum tubes was put together by Ayumi Nakabayashi. A lot of members here have used Ayumi N's method to create models, which I have also collected. However, the output transformer models are the weak point. They're quite rudimentary. Modeling a real transformer will be complex. So many parasitic elements...

I do want to add that I made a simple SET headphone amp based on results from simulations. The DC voltages were very close to reality and the performance was also very close to reality. I think the THD is higher in real life than in the simulation, but I only tried once to do a proper FFT with my cheap digital 'scope, and I don't think I got useful results. I'll have to try that again sometime.

At any rate, the entire reason I'm driving myself crazy with this ancient RCA design is that in simulation results it's looking like the best-performing tube amp I've come across. Not only is the THD incredibly low, but the H3 is lower than H2, which should make it sound less 'push-pull'. If it works out that way in real life, I think I'll really like the result.
 
Last edited:
"Schade" feedback is actually plate-to-grid feedback around the output tubes only. In other words, you wire up your output pentode as an "anode follower".

The local feedback in this RCA SP-10 design is from output tube plate to driver tube cathode.
Yes, and it appears to me as two different ways to achieve the same results: Triode-like behavior but with pentode efficiency. Schade-style feedback was very popular here on Diyaudio some years ago, and lately I've seen this output plate to driver cathode feedback being discussed in various threads.
As for the models... For the tubes, I've collected the best-regarded models I could find. Many of them are courtesy of diyAudio member Adrian Immler, who I believe has done some excellent work. Adrian traces curves himself from multiple examples of a particular brand/type, and compiles his model from a composite/average of those examples, or from a single example he considers an average representative of the type. The previously best-regarded method for modeling vacuum tubes was put together by Ayumi Nakabayashi. A lot of members here have used Ayumi N's method to create models, which I have also collected.
Sounds good, probably much more reliable than old data sheet curves.

Practically all the gain in this design comes from the first stage voltage amplifier (the phase splitter has a little less than unity gain).
Considering the amount of feedback in the output stage, I guess it's safe to assume that much of the sonic "flavor" comes from the input triode?
This reminds me a bit of an OTL project I read about many years ago. The feedback was wrapped around the output stage, drivers and phase splitter to make a unity gain power follower, then a single triode stage was added at the input to provide the actual voltage gain. I seem to remembed that the guy who built it was happy with the results and had some fun testing different input stages.

By the way, I'm currently working on an amp that will use a little bit of output plate to driver cathode feedback. It's an SE class A2 design using T20 triodes as outputs and 6SJ7 as gain stages. The prototype sounded quite good using only CFB from the 16R taps for feedback but it lacked a bit of bass control so I'm adding a little bit of this "special sauce" in the real version.
 
  • Like
Reactions: rongon
Interesting idea to add a little plate-grid feedback on top of the existing CFB. I hope it works out well.

Yes, I think the sonic flavor will most likely come from the input voltage amp. A 6CG7/6FQ7 might have too little character. The original design uses a 12AU7 with less than a mA of plate current, which probably sounds soft and tubey. Maybe that's why people have liked this design so much...

I wonder why all the power amp designers abandoned plate-grid feedback by 1962, except possibly the H-K Citation II. It seems all the amps changed to using a global NFB loop only, and things stayed that way for decades.

An amp that got discussed a lot a few years ago was the RCA 50 Watt High Fidelity Amplifier, which employed three nested NFB loops. It's basically similar to this SP-10 design, but combines feedback from the output 7027A plates to the 6AU6 driver cathodes, plate-grid feedback around each 7027A, and a global feedback loop from the speaker tap to the cathode of the voltage amplifier input stage (the pentode part of a 7199). It's a complicated amp that uses hard-to-find/expensive tubes, so I'm not sure who has actually built it. It's too large and complicated for me to tackle right now, but it does look interesting.

The Pete Millett DCPP Engineer's Amplifier looks extremely good in simulation. I might buy a pair of the red boards with the empty spaces for chassis-mounted tube sockets. It's similar to this RCA SP-10 design except that it uses plate-grid NFB (anode follower output stage) with a pentode LTP phase splitter/driver, and adds refinements like CCS loads for the driver tubes. I think it doesn't employ as much NFB as this old RCA design, so it has higher gain and doesn't need the voltage amp 1st stage for adequate sensitivity.
 
Interesting idea to add a little plate-grid feedback on top of the existing CFB. I hope it works out well.
Thanks! As I'm designing this thing more or less empirically I'm aiming to find a combination of feedback and cathode resistors that sets a good DC point for the tubes and then adjust the level of feedback by connecting the resistor to different taps on the OPT primaries. No idea yet how well it's going to work but since I'm aiming for only a few dBs of extra feedback it should work. There is a choke loaded cathode follower between the gain stage and the output tube (for A2 drive) and I'm afraid the choke inside the loop might cause troubles if there is too much feedback.
The original design uses a 12AU7 with less than a mA of plate current, which probably sounds soft and tubey.
A current-starved 12AU7, exactly the kind of thing that keeps some audiophile purists awake at night 😀
I wonder why all the power amp designers abandoned plate-grid feedback by 1962, except possibly the H-K Citation II. It seems all the amps changed to using a global NFB loop only, and things stayed that way for decades.
One can only guess. Perhaps the output transformers became good enough to allow large amounts of global NFB and the manufacturers saw an opportunity to cut costs by getting rid of a few resistors.
One possible scenario is that the speakers at the time started to rely more and more on the amplifiers damping factor and global NFB had an advantage here by including the output transformers. Just guessing.
RCA 50 Watt High Fidelity Amplifier
Yes, the RCA 50-watter is an interesting one. It should be doable to build something similar using currently available tubes and transformers but I think I'll have to leave that to someone who has access to modern test equipment.
Pete Millett DCPP Engineer's Amplifier
Also a good one. I have a pair of mono blocks built p2p but very much inspired by mr Milletts design. 75W from a pair of PL519 (~40KG6) per channel with Hammond 1650T OPTs, with proper power supplies they could probably produce 2x that power. Good sounding amps but they don't see much use due to buzzing power transformers, the plan is to rebuild them completely with better PSU iron and use them as subwoofer amps.
 
1725464422957-png.1352433


The main thing I'd be concerned about is the totally foxxed up design of the power supply. 82uF across a 5Y3 is way, way, way too much. The Isurge= 440mA. When doing a design that used the beefier 5U4GB (Isurge= 1.0A) even 47uF was a spec buster. Your 5Y3's aren't gonna last very long. I've seen this done with commercial rigs, these excessive filter capacitors. It's cheaper than a proper PS with an RLC LPF, and they were counting on the average consumer's not being suspicious about the frequency they were replacing power diodes back in the day where nearly every pharmacy, convenience, and grocery store had those tube testers.

Get that reservior capacitor to a reasonable size that doesn't bust the Isurge spec, and replace that RC filter with an RLC LPF. You'll get better ripple suppression and improved voltage regulation.
 
  • Like
Reactions: rongon