Magnavox Flea Power: Getting More Out Of The 8600 Series - A Lot More!

Po vs. Load

This must be further adjusted down to about 81% of the plate power figure to account for OPT losses.
I'm surprised about the 81% figure since you mentioned earlier that the stock OPTs were not lossy. Are the new ones a lot better?

The poor operating point caused the asymmetrical clipping, however, it failed to explain why an impedance mismatch (8R load on the 4R tap) caused the output power to be cut in half, in fact, the higher load seem to be better centered on the load line as shown in the figure below, what do you think really happened in the actual amplifier to cause this to happen?

6BQ5%20Load%20Line%20Comp%20-7V.jpeg

The chart is used for illustrative purpose only and may not correspond to
the actual operating condition of the amplifier.

Here is the chart showing Po vs Load from GE's 6BQ5 datasheet, you can see the drop off in output power as the load impedance is raised, but the drop does not appear to be large enough to account for the Po to be "cut in half".

6BQ5%20Po%20vs%20Load.png
 
Last edited:
So you want to see 243.5V/29mA (from Dave) instead of 238V/27mA, it's not much of a difference, is it?
 
I do not claim to know the actual conditions - only Dave can tell us what they are, I merely used something hopefully close to the actual conditions to illustrate that the power could not get cut in half. With the figures shown on the chart, the Po for the 5k load is 3.15W and the Po for the 10k load is 3.07W (before OPT loss) - showing there is very little drop in the Po.
 
Hi Jaz -- Great questions!

Regarding output transformer losses, in my experience, I've found that 81% is really a very real and honest power efficiency figure to use for an output transformer, and can easily be demonstrated in on any typical amplifier in which the primary impedance (and of course secondary impedance) is a known value: Simply place the rated load impedance on the output of the amplifier, and drive it to just below the onset of clipping using a 1 kHz sine wave. Measure the secondary winding voltage and calculate power based on the load resistance connected to it. Now measure the AC voltage across the primary winding as well, and calculate the power at that point using the known primary impedance of the transformer as the load. You will typically find that the voltage on the secondary is about 90% of what the turns ratio of the transformer suggests should be there based on the voltage measured across the primary winding. You will also find that the power actually delivered into the secondary load is about 81% of the power developed at the plate of the output tube(s), since a 90% voltage transfer efficiency translates into an 81% power transfer efficiency.

This is born out time after time, and shows why this figure can be used as a good rule of thumb in determine real world power output levels from power output figures quoted in the tube data sheets. Those figures do not account for output transformer losses, being calculated simply by subtracting the plate power dissipated under full power conditions, from the plate power consumed under that condition. The difference is the power output available to be applied to the load -- minus any losses through the output transformer.

I went and grabbed a random project from my files. It was from two decades ago, and involved operating 6L6GC tubes as closely as practically possible to the classic 55 watt power output model published by RCA for those tubes: 450 volts plate, 400 volts screen (both regulated), and 122 ma of balanced quiescent current total. An Acrosound TO-340 50 watt output transformer was used, which is hardly a slouch by any means. With a 5000 ohm primary, its 8 ohm secondary was loaded with a 9 ohm load to reflect a 5,625 ohm load back to the tubes, satisfying the published 5600 ohm required load. RCA black plate 6L6GC tubes were used that were of known as new performance capability. The results? 44.25 watts RMS at 1 kHz at the onset of clipping. If anything, this shows the very real losses that must be accounted for regarding output transformer efficiency.

Regarding the load line issue, the load lines drawn illustrate a couple of basic points. To get more exacting data regarding the cause and effect of my measurements, I'd have to go back and reinstall one of the original transformers to generate more complete data. But the load lines presented are still useful for illustration.

On the family of plate curves, the slope of the 5K light blue (normal) load line represents a plate power output of 4.93 watts based on the RMS value of a sinusoidal waveform. On the other hand, the slope of the dark blue 10K ("mis-matched") load line produces a plate power output of 3.06 watts, which means that based on the slope of the lines, the 10K loading produces a power output that is 62% of the 5K load line, or closer than not to a 50% figure.

Of course the 5K load line can't fully produce 4.93 watts due to the asymmetrical operation produced by the operating point. Actual power would be more on the order of about 3 watts.

While this implies a rather consistent power output regardless of loading as the GE data suggests, correlation between the GE data and the load line data is "distorted", as the GE data is based on a standing plate current of 48 ma, while the load lines are based on a standing current of only 26.6 ma. Therefore, the GE data is almost surely based on symmetrical operation under all conditions, while the load lines clearly are not.

Relating all of this back to the measured data, with a 4 ohm load on the original OPTs, I measured a maximum power output of 2.3 watts. Others on AK have measured a similar amount of power as well. Correlation with the load lines presented is slightly skewed however, because the load line conditions are slightly different from the actual operating conditions in the unit, most notably in the area of screen voltage. I also want to check the actual impedance of the original transformers as well. As stated too, measured power with an 8 ohm load was less. No doubt this could be resolved with more detailed testing, as this was hardly done with distortion measurements or at various power levels to really develop an accurate apples to apples comparison. The capability of the OPTs themselves may play a part as well, although loading them more lightly should have theoretically helped their performance. This was all basically done with an eyeball on the scope as I wasn't looking to develop such detailed data for what originated started as a simple project with OPTs that were going to be replaced anyway. I also didn't want to get too bogged down in the details for the sake of the newbies either. However, the question is interesting, and with the degree to which the project has taken on, is one that deserves further investigation once the project is finished. I thank you for raising it.

What is really interesting though, is that by using an 8 ohm load on the original OPTs, symmetrical operation is achieved. But with the load they were using, Magnavox was clearly operating the unit then with the output stage operating quite asymmetrically, as the load lines show. So the question is, why would they do that? An interesting question in deed!

Dave
 
On the family of plate curves, the slope of the 5K light blue (normal) load line represents a plate power output of 4.93 watts based on the RMS value of a sinusoidal waveform. On the other hand, the slope of the dark blue 10K ("mis-matched") load line produces a plate power output of 3.06 watts, which means that based on the slope of the lines, the 10K loading produces a power output that is 62% of the 5K load line, or closer than not to a 50% figure.
Thanks again for taking the time to respond, I just have a follow-on question on how you got the 4.93W for the 5k load, I use the formula from RDH4 to estimate the output power, which is Po = Imax^2 x RL / 8 = 3.15W, ditto for the 10k load, so the Po drop is (3.15-3.07)/3.15 = 2.54%, which is why it baffles me when you measured a 50% drop in the Po. I must be missing something still...
 
Jaz -- That explains part of it. The formula you are using is for a triode output stage. The formula for a pentode stage is significantly different. If you can't find it on line, it appears on page 32 of the RCA RC-30 manual, which the website you are using might supply, or you should easily be able to access on line.

For me, I'm old school and determine plate power right from the load line information. In the case of the 5K load line, I extrapolated what the full length of the line would be with the operating point centered as is. On positive grid swings, the load line swings 222 peak volts, which it also would do IF IT COULD on negative swings as well. Based on that then, the slope of the line represents a voltage swing of 444 peak to peak volts, or 157 volts RMS. Using that voltage and load value then produces the plate power mentioned.

But note I also stated "Of course the 5K load line can't fully produce 4.93 watts......."

That's why I started the conversation off by describing what the "slopes" of the load lines meant in terms of power (rather than what the actual power of the line itself represented), and finished the comment above with the point that the actual power of the 5K load was more on the order of 3 watts -- this because in reality, the load line cannot extend below 0 ma.

Dave
 
What is really interesting though, is that by using an 8 ohm load on the original OPTs, symmetrical operation is achieved. But with the load they were using, Magnavox was clearly operating the unit then with the output stage operating quite asymmetrically, as the load lines show. So the question is, why would they do that? An interesting question in deed!

Dave

Maybe simply that the left hand and wasn't talking to the right hand? Wouldn't be the first nor last time for that sort of thing.
 
I'd guess this 'symmetry' at 8 ohms was never intentional....but if anything, they perhaps figured in the idea that some users might also be hooking up 'extension' speakers...which was a big thing back in the '50s as well. However, that would have implied that they assumed those users would 'series' in the extension speakers...which was not so common a process. I dare say most 'Harry Homeowner' types back then would just to the 'black to black and white to white' thing when wiring up any external speakers...resulting in a Parallel set up. Based on all this....that must have sounded terrible! Now I'm amazed to see my little Maggie survived all this.

BTW, I'd note that of the two of these I have, one came in 'with controls' condition, and the other was a straight power amp. The one with the tone, volume, and balance control was from a straight record player console. The other was from a console with both a record changer AND the tuner Dave mentioned. But, both are essentially identical when the tone (eq) section is removed. I may have some different chassis markings and ID....but the chassis itself, and the OTs are, I believe, the same. The PT may have a different part number on it. I'll have to check and get back on that issue. WC
 
Jaz -- That explains part of it. The formula you are using is for a triode output stage. The formula for a pentode stage is significantly different. If you can't find it on line, it appears on page 32 of the RCA RC-30 manual, which the website you are using might supply, or you should easily be able to access on line.
That's it, and thanks again for pointing out my error. Actually the formula from RDH4 could be used for the pentode as long as the distortion is fairly low, which isn't really the case with the Magnavox design. [Edit] In fact, the distortion is so high with the original 8600 design, that even the formula in the RCA RC-30 manual isn't applicable, so an even more complicated formula would be required to calculate the output power, which is hardly worth the effort at this point...
 
Last edited:
Power supply upgrade?

Hi Dave, will you be upgrading the PS for your amp ? I found that in rebuilding these little amps using an inductor for a "pi" filter and even CLCLC if I can squeeze that in the chassis makes a big difference in sound quality. I also like to use all film caps in the power supply. The size and cost of film caps have really made them much more affordable than before with performance that is unmatched by electrolytics. Keep up the good work, tc&br, primo
 
AES output transformers arrived today. Waiting for a 500VCT Edcor PT (I'll use a 5AR4 to bring voltages up a bit - octals FTW) to arrive. Many of the other parts are in my parts bin. Just need the final schematic....

:lurk:

-D
 
Output Stage Modifications -- Pt 1

With the issues surrounding the output stages laid bare, correcting them then becomes rather easy once the new OPTs are installed. They are key to implementing the rest of the improvements, as improving the performance of the stage requires increasing the standing (quiescent) current through it. Since plate current passes through a single ended output transformer in an unbalanced fashion, it leaves the core subject to saturation if more standing current flows than the transformer is designed to handle. If that happens, that leaves no magnetism left to transfer the audio portion of the plate current in the primary winding into the secondary winding. Since the standing current needed to be increased in the neighborhood of 150%, this was almost certainly going to outstrip the capabilities of the original smaller transformers. So besides correcting the impedance mismatch of the original units, the new transformers also allow for more power to pass as well -- necessary if the little 8600 is really going to sing.

With the transformers replaced then, here are the remainder of the changes necessary to add some kick to the output stage:

1. In the power supply, the 4.7K 1W resistor should be change to a 1.5K 1W resistor. This resistor primarily sets the relationship of the screen grid voltage to the plate voltage. With the modified amplifier, the plate to cathode voltage is ~ 250 volts, while the screen to cathode voltage is now just 5 volts below this. The screen grid ultimately determines how much current can flow in a pentode tube, so it acts in a big way to match the plate load requirement of the tube to that offered by the OPT. Operating the screen grid at 245 volts optimizes that relationship in the modified design.

This change also increases the B+ voltage to the AF amplifier stage which is of no concern, as the bias for the stage is automatically adjusted to account for it. The net result is that the stage is simply capable of greater undistorted drive to the output stage than in the stock design.

2. The common cathode resistor should be removed and the cathode connection between the two output tubes separated. Also, the connection over to the common cathode bypass cap in the power supply can needs to be removed as well. Locating the cathode bypass cap in the power supply can cap was a major source of hum in this amplifier, since the can cap was grounded at one end of the chassis, and the CT of the HV winding and the common cathode resistor was grounded at the other.

This is a terrible arrangement, allowing significant power supply ripple current to be conducted through the chassis, which of course caused a slight voltage drop to occur across the face of the chassis. With the output tube common cathode resistor and HV winding CT lead at one end, and cathode bypass and power supply caps at the other, this meant that the voltage drop across the chassis was then injected into the cathode circuit of each output tube via the cathode bypass cap, amplified by the tubes, and reproduced as hum in the output. This is the essence of what a ground loop is. To resolve this, extend the HV CT lead, and ground it directly at the can cap. This prevents any ripple current from flowing through the chassis. Then, use the ground lug between the two output tubes for all output stage ground connections for both tubes.

3. Each output tube cathode terminal (pin #3) should have a 120 ohm 1/2 resistor connected between it and the ground lug. There should also be a 100 uF 16V cap connected between each pin #3 and the ground lug as well. The negative terminal of these caps should connect to the ground lug.

This is a much lower single tube cathode resistor value than most followers of these type amps are used to, where normally, a single tube resistor value would normally run in the 200 to nearly 300 ohm range. Such values typically over bias the tubes, and cause the unequal clipping previously discussed.

There is a common wisdom prevailing in this hobby that reducing output standing current draw is always good. It extends tube life, doesn't hurt sound quality, helps with today's higher line voltages, etc, etc, etc. That wisdom is often very sound (pun intended) when it comes to push-pull amplifiers. But when it comes to single ended flea tube amps -- where every portion of every watt counts -- that line of thinking couldn't be more wrong. The correct amount of standing current draw is crucial in obtaining maximum power output and minimum distortion. If you under or over bias the tube, the operating point shifts from the optimum position, so that the plate signal swing is cut off on one end or the other, neither of which is good.

A value of 120 ohms properly biases the output tubes in the modified amplifier. This value of resistance produces a plate current draw of 44 ma under quiescent conditions, and produces exactly equal clipping as maximum power is approached and exceeded. It also promotes Class A operation throughout the entire power range: As power output is advanced from zero to maximum, the cathode voltage of the output tubes barely moves if at all. This indicates near perfect constant current operation over the full power range of the amplifier, which is the very essence of Class A operation by definition. It is also why the typical thinking of making for large power supply reserves in the modified amplifier will produce little benefit. The current draw by the amplifier from the power supply is virtually constant, regardless of what level of power is being produced. The total current drawn by each tube is 47.5 ma.

As for the tubes themselves, with a plate to cathode voltage of 250 volts, this equates to a plate dissipation level of exactly 11 watts. Hello more power, at lower distortion. When this is tag teamed with correcting the impedance mis-match via the new OPTs, this is where the 300+% increase in power comes from in the modified amplifier. This dissipation level is well within the capabilities of the 6BQ5 tube family. As judged against the more modern Design Maximum standards which tubes developed after about 1960 were rated under, the tube is operating a 83% of its dissipation rating, or very close to the 80% "standard" that is often used as the benchmark today in setting quiescent dissipation levels. In short, when combined with the voltage levels applied to the tubes, it is a very safe condition to operate the tubes under -- and is required for the Maggie to be all it can be.

Now the power supply really takes this all in stride matter of factly. Total current draw of the modified amplifier is just under 100 ma -- or exactly what the power supply was designed to deliver in the first place.

Pt 2 continues in the next post.

Dave
 
Output Stage Modifications -- Pt 2

Continuing on:

4. Misc:

A. B+ FILTERING: As has been suggested by others and in other threads, the 100 ohm 2 watt dropping resistor in the power supply should be replaced with a Hammond 156R choke -- and for good reason. In properly biasing the output stages, they have been made more sensitive in the process. Therefore, they are now more sensitive to less than adequately filtered B+ supplying them. With the resistor left in place, 120 Hz hum is evident in efficient speakers connected to the modified amplifier. Adding the choke makes for a very quiet output. At 56 ohms, it also provides a few more B+ volts than the resistor does as well.

B. HEATER BIAS: The two 100 ohm resistors connected between the heater terminals and ground at the output tube socket closest to the power transformer should both be disconnected from the ground lug they were connected to (which will now have nothing connect to it), and then connected to pin #3 of the same tube socket. This will provide about 5.7 vdc of bias for the heater winding to help minimize any hum generated across the heater/cathode elements within the AF amplifier tube.

C. AC VOLTAGE ADJUSTMENT: Also suggested by others, the unused heater winding for powering a tuner should be connected to buck the AC line voltage applied to the primary winding of the power transformer. This can be accomplished by connecting the black primary lead and the brn/yel heater winding lead together and then isolating them. AC power from whatever fuse and power switch accommodations provided should then be applied to the blk/red primary and brn heater leads.

This connection provides near perfect compensation for the typical 121 vac line voltages of today. With that level of AC voltage powering the unit with the windings configured as described, the tube heaters are provided with 6.36 vac, and the B+ voltage at the output of the rectifier tube is 273 vdc in the modified amplifier. Having a proper heater voltage is another reason that the output tubes are perfectly happy in the new environment of the modified amplifier. The amplifier should be fused with a 1A Slo-Blo fuse.

D. LF RESPONSE: The coupling cap into the output tubes should be raised to .1 uF. Using any lower value than this will cause a LF bump around 30 Hz due to the loss of NFB that lower values, as well as the response of the transformer itself will cause at this frequency. Resist raising the value above this value, as that only serves to couple power supply voltage aberrations into the audio path.

Pics include:

1. Here, the amplifier is producing a 1 kHz 3.65 watt RMS sine wave (maximum power) into an 8 ohm load. In the listening room, the difference between the power level of the modified amplifier and that of the stock configuration is striking.

2. Now, the amplifier is producing 4.0 watts at 1 kHz. Note that the clipping is smooth and even -- a far cry from the over biased condition of the stock design.

3. Even with time lapse photography, there is no hint of color to be seen in the output tube plates. They have shown themselves to be quite content in their new operating conditions.

4. An underside shot of the finished amplifier.

5. A finished project shot.

The effort of this presentation has been to address design deficiencies and modifications for the re-purposing these amplifiers for quality listening in today's flea audio environment. As such, it has purposely addressed design changes and/or modifications only. For those of you with your favorite brands of this or that tube or passive component, have at it. The purpose of this exercise was to address sound quality from the standpoint of design/circuit modification only.

COMMENT: In researching all that has gone before me with these amplifiers -- and there's been a lot -- (including the much revered RH editions), I am somewhat amazed that (to my knowledge) the particular combination of modifications and operating conditions as offered here have yet to be applied -- either that or I just haven't found it yet. That is, other have used this output transformer but did not adjust the operating point of the tube. Or some have a proper operating point, but still have an impedance mismatch. Still others may have done both of these things, but stability was not addressed -- and on and on with the various combinations possible.

Producing maximum performance from any amplifier is like a chain made up of various links. However, if any one of them is weak, it really doesn't matter which one it is: The chain breaks. With flea power, this was never more important. With this effort, I've tried to address all the issues I've observed in the various renditions of these amplifiers, and bring them together into one all inclusive design that addresses all of the issues. Along the way, I must admit to being quite amazed myself at the sound quality they are capable of.

Next up will be some final thoughts, final performance results, and a schematic -- although those who have followed along can virtually make their own from all the details given.

Dave
 

Attachments

  • SAM_1452.jpg
    SAM_1452.jpg
    45 KB · Views: 939
  • SAM_1450.jpg
    SAM_1450.jpg
    44.2 KB · Views: 907
  • SAM_1454.jpg
    SAM_1454.jpg
    17.9 KB · Views: 984
  • SAM_1457.jpg
    SAM_1457.jpg
    122 KB · Views: 1,285
  • SAM_1456.jpg
    SAM_1456.jpg
    81.2 KB · Views: 1,187
Excellent stuff. I don't own one of these, but I think I understand a bit more about how SE amps in general are designed now. As always, your thorough yet easily comprehensible explanations are very much appreciated.
 
Very interesting thread. In reading and researching transformers for this amp, I have found some 5K UL SE iron. What benefits or changes to the circuit would be needed to make it UL. I am seriously thinking of a scratch build, I have great power transformers and all the rest of the parts to build this over the weekend!! Thanks!!
 
Hi-Fi -- To recommend any changes for use with your transformers and UL operation would require physically modeling and building a circuit up around them. In essence, that's what was done here with the new transformers used. I would therefore resist offering would could only be speculation without first having any direct hands on experience to work from.

So many of these amplifiers are already the result of speculation or purely subjectively based observations. I am hardly trying to be critical of that approach -- but it is to say that I was intentionally trying to go after improved performance by way of using classic engineering approaches -- something which seems to have largely disappeared in this day of anything goes audio.

For example, I have no idea if 5K remains the ideal plate load when converting to UL operation -- or what standing current produces the lowest distortion in that mode -- or screen tap position does for that matter, either. So again, my approach would require having the actual transformers to work with. However, I wish the very best with your build!

Dave
 
Dave - I'm assuming a Dynaco C354 should be OK in the power supply as it is virtually identical to the Hammond 156R with regard to resistance and inductance (not to mention almost every other critical measure). I've got a spare in the parts drawer. :D

-D
 
Back
Top Bottom