Rx For the Magnavox 8800 Series

The Input jack, 470K resistor, 1Meg resistor, and negative of the 1000 uF cap all connect to ground but at different points in the unit. Maybe I should draw a dot on the schematic to indicate that where the lines for those leads cross each other, that is also a connection. The input jack and 470K resistor are grounded at the input jack in the unit, and the 1Meg and 1000 uF cap are grounded at the chassis ground spike between the driver and output tubes.

I also note that I forgot to label the supply voltage that the OPT CT, screen grids, and driver tube plate resistors connect to. In the unit, they connect to the same points in the filter system as in the original design, but the OPT CT lead (after the choke) now represents a 305 volt supply point, the screen grids (after the 470Ω dropping resistor) now represents a 298 volt supply point, and the driver stage (after the 10K dropping resistor) now represents a 275 volt point.

Dave
 
Dang!!!
I just sold one of these on ebay----but I had two! The one I kept Ive already replaced the Can Cap with a JJ 40/20/20/20.
Does anyone have a good line on 50 ohm pots?
Dave if you need a couple or three 6eu7's let me know I will ship them too you. Your willingness to share is greatly appreciated!
 
Dave, I have another question if I may about the paraphase inverter used in this design.

I notice there is no resistor in the cathode circuit connected to ground, so I assume the DC path to ground is provided solely by the global feedback path? And yet you still use a quite large cathode bypass cap of 1000 uF. Is a bypass cap actually needed here? What is it bypassing?

Just noticing your 9300 series mods did use a cathode resistor to ground on the inverter section with "only" a 100 uF bypass cap.
 
Hi Kevin -- You are correct that in this case, the NFB resistor serves as not only one leg of the NFB attenuation network, but also as the common bias resistor for both section of the driver tube.

Now normally, in a more traditional circuit, there would be a bias resistor between the NFB insertion point and the top section's cathode, and the bottom end of the 100Ω resistor would be grounded. But in this case, there is no bias resistor as described, because the NFB resistor is acting in that capacity. But the bottom end of the 100Ω resistor -- which is the other element of the NFB attenuation network -- still must be grounded with respect to AC, for it to effectively complete its circuit with the secondary of the OPT. So, this cap not only acts to couple the ground side of the NFB attenuation network to ground, but it also at the same time acts to ensure maximum gain in the bottom half of the inverter driver tube (the actual inverter section), which in turn allows for maximum application of NFB around the section.

I hope this helps!

Dave
 
6 -- Very similar indeed! I just chose to take the inverter drive signal from the plate side of the coupling cap, since the DC balance circuit for the output stage places the grids of those tubes at about 1 volt above ground. With Dynaco's circuit, that 1 volt would have upset the bias of the bottom inverter stage. Of course, the Dynaco circuit would have worked just the same if a coupling cap were placed in the bottom half grid feed as well.

Dave
 
Ah, realizing that the 100Ω resistor is the bottom half of the global feedback divider network suddenly makes this circuit make sense. So the 1000 uF cap creates a virtual ground with respect to the feedback AC signal. Yup, makes sense now.

I'm warming up to this floating paraphase a bit more. I think I might be able to design one of these bad boys on my own now. Might even try it in one of my own builds.
 
6 - Correct -- the 1.0 volt designation is only an approximate figure, and will vary as the needs to balance the tubes varies. Adjusting for 0.0 vdc between the Bal Test Points is always the goal.

As for the range of the DC Balance control, it is limited to about a maximum differential of 1.3 volts. Therefore, the lowest value on either grid would be 0 vdc, while the maximum would be about 1.3 vdc. I have always intentionally limited the range of these controls, as they are best used to tweak a decently matched pair of tubes to begin with -- not allow two grossly unbalanced tubes to operate with a balanced quiescent current draw. The use of matched tubes that are tweaked by the controls always produces optimum results, as opposed to well UNbalanced tubes that are force-balanced under quiescent conditions, and at a stated level of power output by an AC balance control. Yes, they're balanced at those points. But the use of decently matched tubes -- that are further tweaked by the controls -- also ensures that they will be matched throughout the entire dynamic range (power output) level that the tubes are capable of producing, rather than just at the two points mentioned.

Dave
 
Since this is a very similar driver design to the 9300, I'll just ask it here.

I'm messing with a 9300 with the modified inverter arrangement, and I find that the Russian tubes bias a fair bit different, roughly 1/2 the cathode voltage develops vs old production. At high signal levels I've had the output of the tube just cut out. I presume the tube is being driven to cutoff and the second or so of no output is while the voltages are returning to normal so things conduct again. Old production tubes do not do this. If I change the cathode resistor to fix the bias issue with the Russian tubes, will that cause problems running old production, or can I just do a happy medium that works with both types?
 
Sounds like the tubes are pretty well of the original characteristic. I would think that you will need to make an adjustment for those tubes only. Ultimately, all you can do is experiment to see if a mid-range value will in work for both sets of tubes. Or, install an appropriate adjustable Bias and DC Balance network along the lines offered with this project, and then have the best of both worlds.

Dave
 
Two more theoretical observations here concerning the floating paraphase.

1) The output impedance of the upper tube looks like it is going to be significantly higher than the output impedance of the lower tube because the upper tube has no local feedback around it's plate-grid like the lower tube does. The local feedback applied around the lower tube gets its output impedance 5x lower or so. The issue to me is not that the output impedances are different between the two outputs, but that the upper's impedance could be large enough so that it may interact with the Miller capacitance of the following stage it's connected to. That doesn't seem to be a concern though if connecting the paraphase outputs directly to an output stage as is done with these Magnavox's. But it may indicate you don't want to connect the paraphase to another high mu gain stage.

2) The other thing I'm seeing is the floating paraphase seems like it will work best with a high mu tube like a 12AX7 or equivalent precisely because of the larger amount of local feedback that can be applied around its plate-grid to get its gain down to unity. Higher levels of local feedback applied would mean more feedback available to keep its output level constant over a wider range of frequencies and when substituting in a different tube sample without needing to retune.

But as I say I've not ever designed one of these bad boys, so these comments are just theoretical observations.
 
All correct observations, Kevin. I used the floating paraphase design because within the paraphase family, it has clear performance advantages over a straight paraphase design, and importantly, keeps the design within the realm of what the original Magnavox circuit represented.

With these types of projects, it's easy to start changing everything for improved performance, but then at some point, the unit ceases to maintain any identity of its former self. One AKer suggested that once you start changing the transformers, then a modification clearly starts turning the unit into something it never was in the first place, such that it simply becomes a chassis to house a different design. I think that's a pretty good defining point. In that spirit then, with this modification, I kept the original tubes (except for the rectifier) and circuit topology, making the most practical "best" of it as could be done, and thereby maintaining as much of the original 8800 design topology as possible.

Dave
 
Congrats on another well thought out modification tutorial. Of course, for those of us that have little or no chance (non-US based) of getting these units on the cheap, it's moot...
 
Dragon Slayer? Well.......Kinda Sorta, PT 1

Based on my own experience as well as observations of the measured results from others who've posted their work on AK, I have never been really impressed with the performance of the general replacement push-pull high fidelity output transformer offerings from the usual go-to transformer sources used. This is not to demean those companies or their products. Rather, it is simply to suggest that the results achieved are usually well off of that which the main line transformer manufacturers of yesteryear rather routinely achieved. We're talking the likes of Peerless, Acro, Dynaco, McIntosh, Triad, Chicago, Stancor, and others. Oh, today's transformers look impressive enough with plenty of core area and long, hi temp Teflon leads -- and some models in fact produce good low end performance right down to 20 Hz. But with the typical specification for today's transformers being "frequency response at rated power output", this can lead to less performance than such a specification implies. Consider the ratings given for yesteryear's high quality offerings, which were typically given as "maximum undistorted power output from 20 Hz to 20 kHz", with some manufacturers even specifying the amount of global NFB under which their products will remain stable. The maximum undistorted power response over a specified band for which a transformer is capable of is very different from simply stating the frequency response capability of a transformer at rated power output, as no reference to distortion is given in the latter specification. When today's general replacement high fidelity transformers are measured against the much tougher specifications of yesteryear's manufacturers, the differences quickly become apparent.

The usual issue is HF and supersonic performance, where interleaving and minimizing winding capacitance become critical design elements for maximizing power output up to 20 kHz and maintaining complete stability into all possible loading conditions in a NFB amplifier design. Whereas the transformers of yesterday were well capable in this regard, today's high fidelity transformers most often (honestly) perform little better than replacement guitar amplifier transformers do. And there's another point as well:

For some reason, today's transformer manufacturers all seem to think that optimum UL performance is produced with screen taps placed at 40% of the winding, or at least will generally work well in all applications. This is absolutely untrue, and not by a little bit. Optimum tap placement can be anywhere from 25-50%, which varies by tube and circuit design. When this is coupled with the fact that the window of optimum tap placement where UL operation becomes really effective is rather narrow, then using a transformer with an inappropriate tap location for the application can actually hurt performance, over that of not using any taps at all.

Against this backdrop then, I started searching for a good replacement OPT for the 8800 project, that could move it significantly forward in the performance department. With the usual suspects either eliminated, deemed not practical, or not available, I settled on a rather new (to me) high quality offering, that being the Raphaelite model OP10K15AB, which is a transformer I don't believe I've seen used in any projects on this forum. It is not an inexpensive piece, costing (and looking very much) like that of the excellent Dynaco Z-565 replacement transformers currently being manufactured. With a 10KΩ primary impedance, 15 watt rating, plenty of core, UL taps, a tapped 4/8Ω secondary, quality construction, and all the high accolades for high fidelity performance as given in its advertisement material, this could be the DS (Dragon Slayer) piece I was looking for.

When the transformers arrived, basic tests were run in pentode mode using the 8Ω secondary, and performance looked very encouraging. At that point however, work on the modified amplifier using the stock OPTs was not yet finished, so the Raphaelite transformers were set aside until their turn at bat came up for some serious testing. That time has come.

End PT 1.
 
Dragon Slayer? PT 2

These transformers are very interesting in that the things they excel at they do very well, while the things they are lacking in are glaring.

Static measurements revealed that when properly loaded, an 8Ω load on the 8Ω tap reflects 11,654Ω back to the output tubes, while a 4Ω load on the 4Ω tap reflects 12,913Ω to the tubes. This increase in primary impedance results in about a 0.5 watt loss in power output on the 8Ω tap, and just over a 2 watt loss on the 4Ω tap, all reference to 10 watts. On the primary side, it was quite pleasantly surprising to find that the UL taps on this transformer are appropriately located at 25.0% of the winding, which is ideal for tubes of the 6V6GTA class. Referencing my comments above, this is very different from the oh-so-standard 40% that most manufacturers offer. While the taps are well positioned, they do none the less produce a loss of power output, producing just over 7 watts of power on the 8Ω tap. This is getting close enough to the stock power output capability that the efforts using the taps were set aside for potentially a different project where more B+ is available to counter the power loss produced.

Dynamically, in pentode mode, the transformer produces very fine results.......... on the 8Ω tap. Power, as previously mentioned, is about 9.5 watts, with distortion never exceeding 1% at that power level from 20 Hz to 20 kHz. Frequency response throughout the audio band is very flat, being down just 0.15db at 20 kHz, and down 1.0 db at 46 kHz. Stability, is in fact utterly amazing. This is the most unflappable transformer I've ever worked with in this regard. Not only is it absolutely stable with any amount of capacitance only load applied, the unloaded square wave presentation changes very little when the capacitance is applied! Such stability always translates into a very high level of detail reproduction in my experience. HF transient response is plenty good as well -- not the best I've seen, but still quite good overall. In typical use, the elevated primary impedance will likely work in favor of performance, since many speakers tend to average somewhat lower in impedance in operation than that of the rated nominal impedance.

In the listening room, the transformer made itself immediately apparent, with a very balanced presentation, solid, deep bass, precise high frequency detail, and balanced mid-range presentation, that Ch 2 (modified as presented using the stock OPT) simply could not match. By comparison, Ch 2 -- against the Raphaelite transformer in Ch 1 -- sounded weak on the low end, and loud in the mid-range -- and Ch 2, already sounded much better than the original stock design. The performance with the Raphaelite transformer was really very, very good. But there is a catch.

The disappointment came with performance on the 4Ω tap. Mid and LF power output is reduced to the 7-8 watt range as the reflected primary impedance is even greater on this tap, and most importantly, HF performance simply falls apart. To produce a good square wave on the 4 Ohm tap, it takes a completely different type of NFB and HF stability scheme that is connected to that tap (the 8Ω tap requires its own NFB and HF stability network as well), and can only produce a 2 kHz (!) square wave with any accuracy. What's more, power output at 20 kHz on the 4Ω tap is a measly 3.1 watts RMS! Granted, little power is needed at this frequency in actual use, but for the cost of these transformers (about $80 each, delivered from the most inexpensive vendor), this is very poor performance indeed. The need for different FB arrangements for different taps is hardly uncommon with aftermarket hifi transformers. But for all practical purposes, this transformer should only be considered as having an 8Ω secondary winding, as performance at the 4Ω tap is so poor.

So the Raphaelite transformer comes off as a Dr Jekyll and Mr Hyde piece. It would be hard for me to recommend it unless you absolutely knew that you would never be using the 4Ω tap. As a result, I do not plan to even draw up a schematic for its inclusion in the modified design. Once again then, this really only leaves the new manufacture Z565 Dynaco transformers as (to my knowledge) the only truly high quality transformers that could properly work in this application on BOTH the 4Ω and 8Ω taps -- and that would take some significant design changes to accomplish at that. Still, I am committed to seek out the best performance of which the basic design is capable, so a pair of the Dynaco transformers have been ordered. That puts this project on hold for a while until the transformers come in, and significant design changes can be made. But as before, I will get to it in time and report back. For now then, a few pics to offer.....

BELOW: A fine looking piece indeed. It even includes an internal shield with it's own dedicated lead for grounding.
SAM_2434.JPG

BELOW: All mounted up for testing in Channel 1. It definitely is intimidating to the original stock OPT in Channel 2.
SAM_2430.JPG

BELOW: 10 kHz square waves. Top is Channel 1 with the Raphaelite transformer loaded on the 8Ω tap, while the bottom is Channel 2 with the stock OPT in the modified design loaded with 4Ω. Note the significant improvement in rise time with the Raphaelite transformer.
SAM_2431.JPG

BELOW: Oops. Nothing has changed except that now the scope is displaying the Raphaelite transformer's 4Ω tap with a 4Ω load connected to it. As stated, it can be made to show an excellent square wave on the 4Ω tap, but only to 2 kHz, and the load stability is at best average.
SAM_2433.JPG

BELOW: Underside of the amplifier showing Channel 1 with the Raphaelite transformer installed.
SAM_2432.JPG

So that's about that. If the Raphaelite transformer proved anything, it's that the stock OPTs can definitely be improved upon, but that the Raphaelite transformer only gets you part way there. Quite a bit of work to do to ready the 8800 for the Z565s, but as the quality of that transformer is an known value, it should be well worth it -- unless, anyone has any other suggestions.......

Happy Listening!

Dave
 
Last edited:
Dave, Will the 8k primary impedance of the Dynaco transformers still be a good match with the cathode bias arrangement of the output tubes or will EFB make more sense in that instance?
 
Dave....you asked for suggestions...:eek:

As I know very little about building amps.... here is a guestion/suggestion.

Would the stock OPTs do better if their job was to only produce a portion of the listening frequencies.

In other words bi amp or tri amp.

Each amps OPTs only tasked to handle low and high.... or low, mid and high ranges?

I know this would require an active crossover prior to each amps input....but am curious if stock OPTs would do a better job this way.

Far a field question from a noob...
 
Dear Mr. Gillespie,
Thanks for all the time and energy you devote to the readers of Audio Karma.
Your Magnavox amplifier rebuilds are an inspiration to all of us.
Up to this point most of my serious-listening speakers have been 4-ohm. I
recently purchased some very efficient 16 ohm speakers so I would ask that you
give passing mention to 16 ohm performance as you zero in on your final
transformer selection.
Again, thanks for all you do.

Regards to all,
Johnmil
 
Leland -- The original OPTs would work extremely well in the mid-range position of a tri-amp system, and even quite well in the treble portion of a bi-amp system as again, the amount of power required at the highest audio frequencies rarely exceeds even 1 watt of power under very loud listening conditions. The original OPTs shine from about 100 Hz to 5 kHz -- smack in the middle where most audio "intelligence" occurs, making them great for a mid-range amplifier. By limiting the frequencies above and below this range, the transformers would then always be working at their optimum capability.

Kid -- We'll see. With the 6V6 being a Beam Power tube, it's a whole different ball game versus the characteristics of the 6BQ5 Power Pentode used in the 9300 amplifiers. I have my suspicions as to how it will play out, but it will be neat to let the exercise prove them out -- or not.

John -- Thanks for the kind words, and welcome to AK! It's always great to add another voice here, and your comment in your first post is well taken. Many of the vintage Klipsch and Altec speakers (as but two examples) employ 16Ω drivers, and do in fact deserve a performance mention at this load impedance as well, so I thank you for bringing that point up.

Dave
 
I was eventually going to try a pair of these. Thanks for doing the testing!. They look pretty decent for 8 ohm loads, non UL, and the stability specs with feedback seems very promising if you can live with the limitations.
 
Back
Top Bottom