The Mighty Bogen MO-200A

[dcgillespie]

Regrouping for Success

Have you ever had a project -- any project -- that got so involved and complicated that one detail starts to interfere with another, and so on until the whole game plan begins to come into question? That was starting to happen with this project, such that too many of the individual best ways to go after individual performance parameters all started colliding with one another. At those times, it's always best to take two steps back, not only to gain fresh insight and perspective, but also to remember how much this hobby we love is as much about the fine art of compromise, as it is science. For the last few days, that's exactly what I've been doing.

As part of that effort, I went back and reviewed all of my notes (nearly two dozen pages worth), and found two items I had missed. One was of little significance (using the calculated NFB figure of the original design, rather than the actual measured value, although they were pretty close), but the other was was quiet significant: With the poor power bandwidth performance observed in the original and modified 8417 versions above 10 kHz, I never even bothered to test for this performance in that region with the 6550 version. Shame on me. I would have eventually caught this in my final testing ritual, but I should have caught it sooner than I did. I also noted that because I started the modified/redesigned versions using the alternate output loading arrangement from the get-go, I never really investigated the possibilities of the original loading scheme with 6550s either (obviously until just recently), primarily because the alternate loading scheme looked so promising with the modified 8417 version.

With that revelation, I then went back and rechecked EVERYTHING regarding the output tube/output transformer interface, and the best way to apply NFB around them. Here is a summary of the rechecked and verified findings:

1. Push-pull feedback from the 70 volt winding: As was originally determined, this can only be used in small amounts (as in controlled amounts back to a possible driver stage), as larger more useful amounts of feedback from this winding (i.e. feedback directly to the output tube cathodes) promoted instability.

2. Feedback from the dedicated NFB winding was initially observed and again confirmed to be less effective in distortion reduction than feedback from the loaded winding, regardless of the loading arrangement used. However, it was also again observed and confirmed that this method of obtaining NFB produces the most stable results with respect to HF stability. Importantly, it was also found that this winding was much more effective with distortion reduction and achieving HF stability when the traditional loading arrangement was used, as opposed to the alternate loading arrangement that was used from the beginning of the modification/redesign efforts. That is the reason that efforts to use this winding were abandoned early on.

3. It was confirmed that feedback from the loaded winding is the most effective in distortion reduction, and can even be made sufficiently stable with regards to HF stability as well, but the best stability is obtained using the dedicated NFB winding with the original loading arrangement.

4. 8417 tubes provide nearly as much 1 kHz power as 6550 tubes do, but do so at less distortion. It was also noted that 8 Ohm distortion was lowest when the alternate loading arrangement was used, and is virtually equal to that of 16 Ohm distortion performance -- this since both of these taps reflect back a nearly identical impedance to the output tubes when the alternate loading arrangement is used.

5. 6550 tubes produce the most 1Khz power, but at about twice the distortion levels produced by 8417 tubes, when both use the alternate loading arrangement. However, with an 8 Ohm load using the original loading arrangement, distortion for the 6550 version drops significantly. As a result, since 8 Ohm 8417 distortion actually rises with the original loading arrangement (over that of the alternate arrangement), the 8 Ohm distortion produced by both tubes is then virtually identical when using the original loading arrangement for both tubes. Since the 16 ohm tap reflects the same impedance back to the output tubes regardless of which loading configuration is used, 8417 tubes have an edge at this frequency and load impedance.

HOWEVER, the power and distortion findings just quoted all fall to pieces in one way or another above 10 kHz. In this region, the poor power bandwidth performance -- both in the original, and the modified 8417 designs -- is largely due to use of the 8417 tubes themselves. I haven't investigated this to the point of cause determination, but may at some later date, outside of this project. Suffice to say however, that performance in this region is dramatically, and markedly improved simply from the use of 6550 tubes -- made even better by using the original loading arrangement for 8 ohm operation.

Taking all of this data into account, and taking those two steps back for reflection, significant changes have now been made to the 6550 design, which collectively and together then produce a very solid and capable amplifier for home high fidelity stereo use. Because individual specifications aren't everything, the fine art of performance parameter compromise has been used to emphasize that which (in my opinion) is most important to quality sound (flat frequency and power response, good transient response, absolute stability with no tendency towards ragged response under abnormal load conditions, and reasonably low output impedance), while other parameters (maximum power output and absolute distortion levels) have taken a very close, but none the less back seat to the important ones mentioned. The results have been so rewarding with the revised 6550 version both sonically and in the lab, that further work on the 8417 version has been halted indefinitely. In certain parameters, the 8417 can outperform the 6550 in this amplifier -- as has been noted through out this thread -- but taken in total, I can only endorse the 6550 version going forward for the re-purposed use intended.

Pics and further comments include:

1. A 10 kHz square wave of the revised 6550 design indicates a slightly slower rise time over that of the original version, but is still acceptably fast enough for an amplifier with an OPT of this size and quality level. Note that the wave top is still well damped after the slight overshoot on the leading edge.

2. Here is the same 10 kHz square wave, except with no load on the amplifier at all except for a .47 uF cap. The wave is exaggerated as would be expected, but of note, the waveform is still very smooth and controlled, without the intense and/or uncontrolled ringing that is often produced under these conditions. This waveform shows the very high degree of stability for the amplifier under abnormal load conditions, with zero tendency towards outright oscillation. This high level of stability could only be achieved using the dedicated NFB winding with the original loading arrangement for 8 Ohm operation.

3. With the 6550 version, a 100 watt RMS 20 kHz power bandwidth is readily achieved, as shown here. Distortion is about 3.75% at this frequency, which is very good given the design restrictions of the original OPTs.

4. On the other hand, the 8417 version at 20 kHz -- operating otherwise under the same conditions as that of the 6550 design, is struggling to produce anything close to a sine waveform, belching out 15%+ THD. This is with the NFB loop closed, and EFB operational -- it being far worse without the latter installed. A decent sine waveform and similar distortion reading as that of the 6550 amplifier at this frequency is not regained until power is reduced under 50 watts.

5. Lots of changes made. More data is provided now that it is becoming finalized. More voltages are provided, as is output tube current draw. Note the use of the dedicated NFB winding now. Note too that the OPT impedance values are all normalized to the 8 ohm output, to better reflect output stage operating conditions. Also note the distortion performance documented. At 1 kHz and 100 watts, .40% THD is produced. With the NFB loop providing a distortion reduction of some 7 times, that means that open loop distortion is running (and is confirmed to run at) around 2.8%. This is smack in the middle of the distortion values typically published for these tubes (2.5-3.0%) when operating in 100 watt pentode mode, with perfectly regulated B+ voltage levels. In practical economic installations such as this amplifier represents, this open loop distortion level would not normally be achievable. However, because the EFB circuits constantly work to maintain the output stage's low distortion quiescent operating point under all dynamic and power supply conditions, the theoretical open loop minimum distortion level is then realistically approached, even though regulated B+ power sources are not employed. The use of EFB then allows for the absolute minimum distortion to be produced from the NFB loop employed.

So, the 6550 wins in overtime -- but it deserves to. Comparison of the overall performance of this design to that of the original product is quite substantial. Power output rating is maintained, while power bandwidth is extended. Distortion is significantly reduced, while stability is maximized. Transient response is now in another universe, and has a decent HF response to go along with it. In short, I believe that the best overall potential of what this unit represents is achieved with this re-purposed design, and the listening room whole heartedly agrees. More listening tests are in store, and the power supply schematic still needs to be posted as well. But I can work on it while I listen and should have it posted soon.

Dave

 

[dcgillespie]

8417 Version Use

So what of operation with 8417? Well based on the original design performance and intended application, their use made and make perfect sense. They're easy to drive, have the lowest mid band distortion levels (up to ~ 5 kHz), and can be implemented with the simplest circuit.

For those with more general needs of that in line with the original design intent then, and who wish to maintain 8417 operation of these amplifiers, all of the modification work done to date on this version still results in greatly improved performance. Considering the limited frequency and transient response, and the elevated distortion levels of the original design -- not to mention the outright crossover distortion and under-biased driver circuit employed in the original design, all of the improvements made to date (including the application of EFB) are still extremely worthwhile to make to the original design.

Since the EFB development work and feedback/stability development work for this version using the dedicated feedback winding have already been performed with the recent re-evaluation effort, I will go ahead and update that schematic soon for best overall operation with these features as well.

The primary difference between the two versions then is that while the 6550 version without question produces the lowest overall distortion and best power bandwidth over the whole of the audio spectrum (making it by far the most suitable tube for the intended re-purposed use), the modified 8417 version can still provide the greatest amount of lowest distortion "grunt" power within a limited bandwidth within the power bandwidth of the original design -- with either version delivering far superior performance to that of the original product. Guitar and general PA users still wishing to use 8417s, take note.

Dave

 

[gadget73]

Just to clarify, you're using the original feedback winding again with the 6550? Just out of curiosity, how does distortion on the loaded winding compare to distortion with the original NFB winding?

 

[dcgillespie]

It is lower on the loaded winding in all loading scenarios. For example, in the 6550 version, 1 kHz 100 watt distortion typically ran around .28% when feedback was taken from the loaded winding, using the original 8 ohm loading arrangement, versus the .40% typical when using the dedicated feedback winding under otherwise identical conditions. However, the improvement in HF stability is significant enough that it trumps the slight increase in distortion produced when using the dedicated winding. In my humble opinion, the stability that a feedback amplifier exhibits is far and away more important than slight differences in distortion produced to achieve it. Where as the ability to sonically determine the difference in pure distortion levels discussed here is highly questionable (to me), the ability to hear sonic differences in stability is not.

Dave

 

[Tinkerbelle]

Dave, I apologize if I am jumping the gun as you most probably intend to do this but it might be good to tabulate your findings when this is all finished rather than having one read through a multitude of posts to find the results. However, while waiting to get to that point, perhaps it would help to capture the status results in one post that you can refer to and update your findings as you go along. Sorry - not trying to add work.

For the 8417, in post #321 (and also an earlier post I think) you indicate the 8417 performance is good up to around 10KHz but then infer ~5KHz in post #322. Can you clarify what happens between 5K and 10KHz?

Also, I would also like to better understand what happens when going above 10KHz, (I assume you will only be providing details of this for the 6550 as that is the forward path). As the transformer is supposed to be capable of response up to 50KHz (at reduced power) are you saying that the 8417 implementation specifically suffers from some compatibility issue with that particular transformer matching i.e. that is not necessarily the case with, say, the MX60A - although could suffer similarly?

 

[dcgillespie]

Hi Tink -- All great comments. Thank-you. The amount of data is something that concerned me as well, and is ultimately the product of making this a nearly real time webcast follow along development process, rather than just doing a complete but silent workup from soup to nuts, and then making a simple post of the ultimate before and after results. The former approach certainly exposes all the warts of the process (which always exist in any project like this), while the latter is obviously much simpler to do. For me however, I've always been one who equally champions both sonic and measured performance, as it take both to deliver truly high fidelity sound (by definition). Ultimately however, because so many modifications are presented with no more than a subjective try this, it sounds great documentation, I would much rather, and have always tried to allow folks to witness the entire workup process -- warts and all -- to give some validation as to the "why" an sonic improvement is perceived or made possible. When it's all finished however, I think a tabulated presentation would in fact be helpful.

In regards to your specific questions, note that in the great body of work after the initial findings, and up until my most recent posts, they have all centered on distortion at 1 kHz. That is hardly to discount full bandwidth performance, which is very important to be sure. But it is convenient to concentrate on one "easy" frequency throughout the meat of the development process, to use as a simple yardstick to see if progress is being made. As things start to gel, then the more time consuming full spectrum measurements can be made and plugged into the mix -- which is exactly what happened in the development of this unit.

With this understanding then, my earlier comments about 8417 performance being good up to around 10 kHz were simply based on scope observations of high power sine waveforms at the time, wherein the basic sine wave shape was generally maintained up to that frequency. My latest comments however are based on actual distortion measurements (when real proofing of the effort began), wherein distortion begins to deteriorate -- but is still acceptable above 5 kHz, but quickly deteriorates above 10 kHz.

As for what happens above 10 kHz, the waveform becomes highly phase distorted, with the bottom portion of the sine wave becoming increasingly triangle like. Frankly, the scope shot I posted of the 8417 version at 20 kHz is much too kind relative to what it originally looked like. With the original quiescent operating current being so low, the resulting crossover distortion, and the overloaded AF amplifier stage of the original driver design (a matter made far worst at these frequency since there is less output available to feedback), the waveforms in this region became very contorted rather quickly -- and helped to generate the comments of the initial scope observations I made.

As the 6550 version approaches full power at 20 kHz, the tendency of the waveform is to do the same thing as that produced by the 8417 -- albeit to a far, far lesser degree -- which is why the 6550 produces so much less distortion in this region. I have always equated such behavior to excessive winding capacitance -- a problem which is amplified in "PA" type OPTs, if only because of the multiple windings they employ to handle the various loads this type of equipment typically encounters. In all honesty however, the performance of the 6550 version in this amplifier is some of the best I have witnessed at 20 kHz with OPTs of this origin -- versus that which the 8417 produces, which is frankly all too typical. As a result, I have become rather impressed with the performance the transformers are in fact delivering.

As for the 8417, I simply have not had enough experience with it to make any categoric comments about it. But I can say I have witness similar performance issues when doing experiments with them in a Dynaco MK III setting. Being as how the transformers in those units were optimized for different tubes, I didn't really give it much thought at the time, other than to write it off as a mismatch. Since then (some two decades ago at least), I haven't worked with the tubes again, being rather unimpressed with them the first go around. With this effort, I certainly remembered my past experience, but hoped for better results since the tubes and transformers were matched as shoes and shoelaces in one unit. But the same basic issues are still there, which means there is a mystery to chase down. I hope to get to that at some later date, as it is perplexing. Ultimately, I still suspect some element of mismatch at the root of the problem, but it will take experimentation to suss all of that out.

Finally, remember not to confuse a given OPTs frequency response with its power bandwidth. In my work, 50 kHz seemed to be the practical limit of the transformer under feedback conditions, while still producing solid HF stability. Adequate response into this region is necessary so that when the required roll off of circuit response is added to produce the required stability, the roll off does not extend down into the audio spectrum. A transformer response to 50 kHz will allow for this to be accomplished, and is true with the case at hand.

Power bandwidth on the other hand is a whole different matter. Here, the extra winding capacitance of large OPTs with multiple windings effectively adds a huge load on the output tubes, quickly chewing up power, and increasing distortion rapidly in this region. That's why moving the loading configuration to one which provides a higher reflected basic impedance level to the output tubes improved performance notably, and enough so to allow the 6550s to in fact perform quite well, all things considered.

I hope this helps!

Dave

 

[audiodon]

While a summary is nice, I've particularly enjoyed these longer re-development upgrade and modification threads that unfold in real-time (this one and the Fisher X-1000 one as recent examples) for multiple reasons.
1. It reminds me that even an expert, and not just a hobbyist like myself, reaches stumbling blocks.
2. It educates me towards a more pragmatic method of analyzing and moving past those stumbling blocks, rather than quickly fixing the problem without really learning anything from it and moving beyond it too fast without fully analyzing the issue that caused the stumbling block in the first place.
3. It educates me about how to test, what to test, and how to balance subjective listening results against the more objective test results.

A particularly illuminating series of posts on this last point (buried now in I don't know what thread) was Dave's description of why a non-perfect square wave rise and settle might be preferable at certain frequencies to a perfect reproduction of the input square wave.

I can think of more reasons but won't, and I seek your indulgence for this OT post.

 

[buglegirl]

Hiya,

While a summary is nice, I've particularly enjoyed these longer re-development upgrade and modification threads that unfold in real-time

Don well put. I would rather read a thread like this and learn a heck of a lot more than reading some book on the subject that for authors egotistical reasons minimizes the pratfalls that occur. Mind you Dave's "Pratfalls" are typical minor things 95% of us mortals MISS.

Dave it really is refreshing to know there is a person out there (You) who has no ego and would rather tell the truth about the bad than gloss over it leaving a bad taste.

Bravo!!

Frannie

 

[Tinkerbelle]

Yes Dave, that helps, thank you.

I was aware of the power bandwidth vs. OPT frequency response. My reasoning was that I have a number of M120 and MX60A amps and was considering similar updates to this but using two MX60A OPTs (alternative two M120 OPTs) on a M120 chassis with the M120 PT for a ~50-60W stereo amp. The frequency response and power bandwidth quoted by Bogen are both up to 20KHz (although the power bandwidth suffers at the low freq end). However, if the issue is with the 8417, then I might be again better with 6550 tubes.

 

[Tinkerbelle]

While a summary is nice, I've particularly enjoyed these longer re-development upgrade and modification threads that unfold in real-time (this one and the Fisher X-1000 one as recent examples) for multiple reasons...

Don, I agree with you on having the full development thread as an enjoyable read-along, as you can share the exploration and learn during the process. The suggestion for the summary was as a follow-up reference when doing updates.

 

[buglegirl]

Hiya,

The suggestion for the summary was as a follow-up reference when doing updates.

Yep but these forums don't lend themselves well to summary's unless its locked.

Although I fully agree that a summary is a good coda to this symphony.

Frannie

 

[Tinkerbelle]

... Although I fully agree that a summary is a good coda to this symphony.

Frannie

Oooh, nice way to tie the music into this thread.

 

[borus]

I would like to chime in on the summary aspect of this project. When I get to the bench to build my MO100's I would love to have the summary of this build in front of me. I know that some where in that build something will spark my interest and I'll go wow why did he do that? Or worse I'll be working to duplicate what I see on the schematic and it just will not go. I liken this to when the guys on the woodworking shows measure diagonaly from corner to corner to check if it is the same insuring some thing is square and it always is. When I do it the measurments are different and I do not know what to do to fix it. With this project's follow along development process, I can go back and find the approach that was used to get it to fit.
I love that.

 

[audiodon]

borus,

In this case, at the end of the project, I'd bet someone (OP?) could write a single post with links to certain aspects of the redesign and implementation.

When one refers to that post, they'd have a summary. Between that post and the schematic, summarization is covered.

 

[buglegirl]

Hiya,

Did Michelangelo finish the Sistine Chapel roof in a day ??

Patience grasshoppers .. Dave will get it done.

Frannie

 

[audiodon]

I was kinda trying to say that Dave shouldn't have to do it.
I thought the OP might hop on the suggestion when all is said and done.

 

[buglegirl]

Hiya,

I was kinda poking fun ..

Frannie

 

[gadget73]

I personally am in no hurry. I've had the pair of 100a's for a year now, and they haven't been touched other than to move them around. Its not like they're half apart on the bench waiting for the next installment. I'm just very glad that someone as knowledgeable and patient as Dave is taking the time to tell us how to make one of these monsters sing, and why it works as it does.

 

[dcgillespie]

Final Amplifier Schematics

The past week has been spent finishing out the 8417 version for those who wish to continue using that tube.

Both versions use the original loading scheme now, as the plate-to-plate load that an 8 ohm load reflects on the original 25 volt tap produces the best overall performance across the full spectrum of which these units are capable, with both output tube types.

For those wishing to maximize low end performance (i.e., below 500 Hz) -- for use with bi-amp systems, sub-woofers, or guitar service, using the alternate 8 ohm connections (i.e., the GRN and OR leads) for 8 ohm loads will increase power output and lower distortion in this region. However, this will be had at a notable performance expense above 1 kHz, so for re-purposed home music reproduction service, the original 8 ohm loading scheme (using the 25 volt tap leads) for 8 ohm loads is recommended.

Mid-band performance is nearly identical between both versions, relating to distortion, power output, and internal output impedance, as are both low and high frequency stability characteristics as well. The frequency response of both units is also identical. At the frequency extremes however, the 6550 version clearly shows its superiority, producing half the distortion of the 8417 version, while also maintaining a 100 watt RMS power bandwidth rating across a 28 Hz to 20 kHz range. The 8417 version can go as low, but runs out of gas at 20 kHz before reaching 100 watts output. The only other significant difference between the two designs is that the 6550 version is twice as sensitive as the 8417 version -- a specification which would certainly appear to be backwards based on the output tubes employed, but is not. Performance documentation is provided on the schematics, to help summarize the performance differences between the two versions.

Frankly however, at the power levels typically used in the home, you would be hard pressed to know which channel was which when listening critically to the unit now. With a mono signal presented to both channels, and the outputs equalized to an exact balance, a simple switch was installed to reverse the speakers between the channels at will. Operating the switch then, after several reversals, I haven't found anybody yet that could accurately identify which channel was playing through which speaker, other than with occasional dumb luck. The sound is really quite enjoyable now, with the measured and sonic performance of either version representing a vast improvement over that of the original design. Overall then, from a purely measured performance standpoint, the 6550 version is the clear winner between the two, with the 8417 version presenting no credible sonic advantage to counter with.

Next up, I will draw up the revised power supply schematic, and present it along with relative comments to complete the effort, and put a wrap on this Bogen re-purpose, redesign project.

Dave

 

[dcgillespie]

8417 Version -- Additional Comments

Those interested in this version may note that the 20 kHz distortion documented on the modified schematic provided in the previous post is somewhat lower now than previously mentioned. As such then, some additional comments regarding this change is warranted.

Applying NFB to any amplifier is always a balancing act: Too much can actually hurt measured performance -- as obviously can too little. The balancing act also extends to the different various parameters it targets to improve, such as stability, response, and distortion performance to name a few. If any one parameter is targeted by itself, its performance can be enhanced, but always at the expense of the other parameters. Therefore, the balancing act involved in applying NFB in the most effective manner becomes crucial to the success of the outcome, making the process one of as much art, as it is science. The trick then, is to achieve balanced improvement from its application, such that the improvements noted in the lab, correlate to improvement noted sonically.

This process is made all the more difficult when using OPTs, whose original purpose was (obviously) designed for the intended application of the MO-100/200 amplifiers. With that in mind, Bogen then never even tried to coax high fidelity like power bandwidth/distortion performance out of the unit above 10 kHz, because, frankly, it was of little concern. All they cared about, was that the feedback that was applied was stable, and whatever effect the measures in achieving that end had on performance above 10 kHz was, well, what it was. In other words, stability was heavily targeted, meaning that other performance parameters suffered. A quick recheck of the original 10 kHz square wave and frequency response performance of the original designed will punctuate that point. Much of the improvement to be gained in re-purposing this unit then centers on re-balancing the performance parameters targeted by the NFB/HF stability networks, for more high fidelity like performance.

So what of the 8417 tube and all of this? Obviously, since the same OPT is used in both versions, the 6550 tube is operating in the same environment as the former tube is, so what gives?

High frequency performance issues (virtually) always come up with OPTs used in this type of equipment, because of the high winding capacitances created by all of the various windings needed in such transformers. This extra capacitance places a tremendous load on the output tubes, causing all manner of phase distortion and power limiting, producing the contorted waveforms shown when elevated power is attempted at high frequencies. Importantly, when a portion of this phase distorted output is then fed back to the input as a corrective NFB signal, it too gets phase distorted in the OPT, causing even more problems. This point cannot be stressed enough, such that the 8417 version will actually produce the lowest distortion and highest power output at 20 kHz, WITH NO NFB OR STABILITY NETWORKS APPLIED AT ALL! However, if you do that (here comes that balancing act), then other performance parameters, and specifically performance in the lower region, goes to pot. So it truly is a balancing act as explained above, and one which in total, can have a critical impact on the sonic performance of the unit -- most particularly so when it is being re-purposed for high quality reproduction, as it is here. In working extensively with these networks then, I believe the best balance has now been achieved in the 8417 version, relative to the ultimate capability the OPTs have with these tubes, when the needs of all the performance parameters are considered, relative to the newly re-purposed application. This is the same exact effort that was applied to the 6550 version, to achieve the best overall performance from that version as well.

As for the tubes themselves then, the improved performance of the 6550 version comes from the apparent ability of those tubes to dump more ultimate peak current into a high capacitance load than the 8417 can, in the operating conditions afforded in the MO-100/200 environment. This allows the 6550 then to produce a wider 100 watt power bandwidth, and at lower overall distortion. For the needs of the original application however, the 8417 was still the perfect choice -- for so many reasons. Even for practical home use as re-purposed, the modified 8417 design is still a very considerable step up from the original offering, with sonic results that are basically indistinguishable from that of the 6550 version at even insane listening levels. Still however, for those who want the most this unit can offer, the best overall performance will be had by installing the 6550 version into the amplifier.

On to the power supply drawing.

Dave