Magnavox 169 BB Output Transformers
- Thread starter Rudiger
- Start date
To set the feedback compensation cap, a 10kHz square wave at low amplitude is used. Output loaded, feedback resistor set, dial in a square wave output of about 2V RMS. (You use low amplitude so the waveform is not "slew rate limited" by the amp.) You generally will see ringing on the upper and lower plateau of the square wave as shown below. (This is the ringing of the output transformer at it's resonant frequency.) As you add more and more capacitance in parallel with the feedback resistor the ringing will be reduced, so the waveform looks more and more square with a flat top. Disregard the cap values shown in the animated gif below, they are from a different amplifier. You can see that as cap is increased further, the square wave starts to get blunted and rounded over at the rising and falling corners- this is too much, and you are reducing the audio bandwidth of the amplifier.
Don't get too fussy about eliminating all ring and rounding off the corner- some overshoot and a slight ring that resolves within one resonant cycle is OK- that actually provides the best, flattest, widest bandwidth. In the above sequence, I consider the waveform as shown at the 600pF value to be about right- any more and that corner starts getting rounded. Let me know if you have any questions, and let's see how it looks.
I still have the 470+1247 installed with the cap cut out. Does that matter? I haven't changed it back. If I did change it back would that effect the feedback resistor value?
The first pic in no cap. Second pic is a 150pf ceramic. I could try a 100pf? I have plenty of caps but right out the box that 150pf looks good. Third picture is an 82pf ceramic. That looks even better. Good enough?
The first pic in no cap. Second pic is a 150pf ceramic. I could try a 100pf? I have plenty of caps but right out the box that 150pf looks good. Third picture is an 82pf ceramic. That looks even better. Good enough?
Attachments
Last edited:
You want to leave that cap out. By removing that capacitor, you have reduced the open loop gain down to a sane enough value that it is in a very good place and we can proceed with setting feedback.
Based upon your pictures, (I would zoom up a bit so you fit one cycle in the window and see more clearly what the details look like) I agree with @GordonW - I think 82pF is a bit too much. The ideal waveform will have as "square" a front rising edge corner as possible- and to do that you must have just a little overshoot that resolves quickly.
However- There is a better way to determine the "optimal" value. It is by characterizing frequency response, not just transient square wave performance. When you are looking at the square wave, you are in essence looking at a graphical demonstration of the frequency range of the amplifier. In order to produce sharp square corners, the amplifier must be able to reproduce very high frequencies- 5 to 10 times the fundamental 10kHz frequency of the square wave. If your frequency response only extends to 20kHz, your square wave is going to be rounded badly. The square waves I showed in my animated .gif were from an amp that was flat to 50kHz (with 600pF).
The resonant frequency of the transformer usually produces an exaggerated peak in the output at that frequency, and your high frequency reproduction will tend to rise as you approach that resonant frequency. Consider the bandwidth plot produced at 1 watt output from another push-pull amplifier below. The transformers shown here are not nearly as good as yours, so the effect is exaggerated. The blue bandwidth trace was for open loop- no feedback at all. Notice how the frequency response sucks at both the low and high end. Negative feedback is a wonderful thing! The orange frequency response trace with feedback enabled but with no compensation is what was causing the ringing at the transformer resonance that you saw in the square wave. As you increase the compensation capacitor, you are causing the frequency response to drop at the high end. The optimal capacitor is one that exactly compensates for this resonant rise, resulting in the flattest passband all the way up close to transformer resonance. In the plot below, it's the green one. Too much cap, you begin to roll off your high frequencies, and you get your rounded square wave.
To measure your frequency response manually, you start with a 1kHz sine wave. Dial the generator until you achieve 1 watt output into rated load, which for 8 ohms is a 2.83Vrms amplitude. (I just use anything around 2V, it's not that critical.) Whatever voltage you choose there, write it down, that is your "reference" value, which, being 1kHz should represent the middle of the pass band. Without touching the amplitude knob, begin setting the signal generator to different frequencies, and write down the output amplitude measured for each frequency. (Hopefully your signal generator is nice and flat, and produces the exact same output amplitude at each frequency like the digital ones do- if not, you have to measure the input too, and make sure it's constant.)
I will generally do a LOT of points down around the lower roll-off- like 20, 30, 40, 50, 60Hz, and so on, just a few points across the pass band from 100Hz to 10kHz because it's usually flat, the more points at the upper roll-off. I created a simple spreadsheet where I "normalize" the voltage measured in the pass-band to be 0dB, then plot the data points in dB to generate a plot like that above. For adjusting the feedback compensation capacitor, you really only need to do a few points- the reference at 1kHz, then a couple of high frequency points above the pass band, but below resonance- like 30kHz, 40kHz, 50kHz, etc. If your compensation capacitor is too small, the voltage amplitude will be higher at those high frequencies than at 1kHz. If the capacitor is too large, they will be lower and roll-off at the higher ones.
A flat passband up to resonance is the best compensation you can choose- and when you go back and look at the 10kHz square wave, you will see what I have described- a slight overshoot, that quickly resolves into a nice flat.
Last edited:
I paralleled two ceramic caps that measured 70pF total. Scope picture is below. The frequencies from 20 hz to 1 khz all stayed at 2vrms. The below frequencies is where they started to change. I have ordered a 62pf, 75pf, and 82pf mica caps. When they arrive I'll do this again with the 62 and 75 and choose the best one.
freq - voltage on scope at 125mv input
20khz - 2.08
30khz - 2.14
40khz - 2.25
50khz - 2.41
60khz - 2.46
freq - voltage on scope at 125mv input
20khz - 2.08
30khz - 2.14
40khz - 2.25
50khz - 2.41
60khz - 2.46
Attachments
Last edited:
Dialing it in! Now you have bounded the problem- you see what is too high, and what is too low, and you know what you are doing. You can see from the way the amplitude is rising for the the higher frequencies above the 1kHz value that the current capacitance is too small. However, keep in mind that not all capacitors are created equal. The quality of the cap matters here- you can see how just a few pF make a difference. You could replace these ceramic caps with a quality silver mica of the same value and see a somewhat different result. Silver mica caps are used frequently in RF circuits because they is effective at very high frequencies, and are very stable across both time and temperature. They're a bit more expensive, but worth it.
Great work by the way!
Great work by the way!
worth a mention but the cap across the feedback resistor isn't the only means of limiting response. Looks like yours is rising past 20khz, which isn't too surprising. Can you find where it peaks and then starts to come down? this may benefit from a step network off the plate of the pentode to limit the bandwidth a bit in order to flatten that out.
This evening I'll see where it peaks and report back. I do intend to use one of the mica caps I ordered so I will be doing this again in a few days when they arrive.
When it comes time to play with the UL taps, will that make a difference to the feedback resistor and cap value?
When it comes time to play with the UL taps, will that make a difference to the feedback resistor and cap value?
I was not certain whether you intended to play with UL or not- Often this decision is made up front and setting feedback and the requisite compensation and stability tuning proceeds with that decision baked in. You now know how to set the feedback level and compensate it, so if any adjustment is necessary you will be able to adjust. The 25% ultra linear ratio is quite a nice fit for the 6V6GT- it is a "shallow" ratio, providing much more of a pure pentode level of output power than with a much deeper 40% tap, yet providing more linearity and less distortion. It is really a matter of taste, your experience with it will be affected by the speakers you use, the volumes, and music selections you prefer to listen to. It's nice to have the option, and be able to make any necessary adjustments.
You are on the right path with compensating the feedback. To a first order, the 82pF from before, or the 70pF you have now are both in the ball park and would result in a most enjoyable amplifier. (I hope you are listening to it now.) Being detail oriented in dialing this in now is providing an excellent teaching example on how this is done for your future builds, and I have every confidence that you will achieve a nice flat pass-band with just a slightly higher value than the 70pF currently.
I was approaching the tuning process in an ordered fashion, in optimizing the feedback path and compensation first, then we were going to proceed with stability tuning. This process was laid out in expert fashion in a conversation that I had with Dave Gillespie in which @gadget73 was included. First, set the amount and compensate feedback, then, proceed with stability tuning. It is done in this order because the amount of feedback, and the solution used to compensate it have significant impact on the degree of stability tuning necessary afterwards- decrease feedback, and decrease the amount of compensation (so you have more ringing) and it will actually be easier to achieve stability. That is counter intuitive, but true.
The shelf network that @gadget73 wisely indicates may be needed, is part of the frequency limiting done to achieve stability and we would explore that next. It is composed of a small value capacitor in series with a resistor that is placed in parallel with the 470K plate resistor of the 6AN8 pentode. I prefer to connect this between the plate and the 300V supply because it is usually easier and there is less voltage drop across the cap, rather than plate to ground- but either work identically as both the 300V and ground are "AC ground". Another method of adding stability margin is a zobel network directly across the speaker output terminals- also a shelf network of series capacitor and resistor. This is less favored by Dave Gillespie, as it is a bit of a brute force approach, but it can provide necessary margin in higher gain or higher power situations, or where the shelf network may already be impinging upon the audio band.
You might be wondering, why all this concern about stability? I don't see any oscillation in my amplifier. Well, any vacuum tube amplifier should be able to tolerate being powered on while there is no output speaker load connected. This is why I had you convert your load so you can still scope while it is switched open. That does not mean any significant signal should be applied- many output transformers have been burned open when the speaker connection is lost while the the amplifier is putting out power. If output load is lost while driving a large plate signal into the primary of the transformer the inductive backlash can easily arc across a winding. This test is done with very little signal, and that is not even strictly necessary. The ability of the amplifier to remain stable with no load is further tested by determining it's margin- by placing a small amount of capacitance across the open output, and increasing that capacitance in steps to determine how much pure capacitance load the amp can tolerate before it goes unstable. This is actually important because sometimes a speaker cable may be connected with no speaker connected at the far end. The parasitic capacitance of the cable itself can cause oscillation if you don't have some margin. But, overall, this is not even really about stability only under those conditions. An amplifier that has very little stability margin with no load can actually be partially unstable WITH load. Dave and I both saw this observing low frequency response with amps operating with too little margin with full rated load. On the left was mine at 10Hz on a 6BQ5 push-pull, on the right is Dave's on a Magnavox 8800 6V6GT push-pull at 90Hz.
So, once you are content with your feedback compensation, (after finally choosing ultra linear or not), the amp should be connected to the test setup with rated load enabled, and a very small amplitude square wave driven to the output. In the example below I used a 10kHz at ~2Vrms. You generally want to use a high enough frequency square wave so your scope is operating at a time resolution that enables you to clearly see the oscillation. Then, switch the load to open. The shape and amplitude of the square wave will change, but you should not see oscillation. If you do, turn the load back on and you know you have work to do. If the waveform is basically stable with no oscillation, you can then try a series of small capacitors across the load terminals.
The steps in the picture below are with 0uF, .01uF, .015uF, .018uF, .021uF, and .031uF:
You will quickly find a clear demonstration here that not all capacitors are equal. Two different capacitors of equal value can be used, one may cause oscillation and the other not- revealing the ESR and high frequency quality of the cap. Film caps are generally better than ceramic, mica generally better than film, but use whatever you have and play with it. Dave Gillespie does this margin analysis on every amplifier that he builds, and indicates the results on this schematic. From his work, he and I do not consider the amplifier to be acceptably stable until you can tolerate from 0.02uF to 0.03uF across the load. The amplifier above is considered stable to 0.02uF.
How difficult this stability is to achieve, is highly variable depending upon the particulars of the design, wire dress, and quality of the output transformer. Frequently, when using real Dynaco Z-565 transformers (that these are supposed to be a reasonable clone of) Dave is able to achieve "absolute stability", meaning any reasonable amount of capacitance across the load causes no oscillation. We'll see how these actually do.
Once you have a feel for what margin this amplifier has, if any, that will determine next steps, and as Dave has indicated, that is as much art as it is a science. The final stability tuning solution will ultimately trade off bandwidth versus distortion, and will represent a compromise that fits your individual preference. Let us know how this goes, and we can suggest solutions as needed.
You are on the right path with compensating the feedback. To a first order, the 82pF from before, or the 70pF you have now are both in the ball park and would result in a most enjoyable amplifier. (I hope you are listening to it now.) Being detail oriented in dialing this in now is providing an excellent teaching example on how this is done for your future builds, and I have every confidence that you will achieve a nice flat pass-band with just a slightly higher value than the 70pF currently.
I was approaching the tuning process in an ordered fashion, in optimizing the feedback path and compensation first, then we were going to proceed with stability tuning. This process was laid out in expert fashion in a conversation that I had with Dave Gillespie in which @gadget73 was included. First, set the amount and compensate feedback, then, proceed with stability tuning. It is done in this order because the amount of feedback, and the solution used to compensate it have significant impact on the degree of stability tuning necessary afterwards- decrease feedback, and decrease the amount of compensation (so you have more ringing) and it will actually be easier to achieve stability. That is counter intuitive, but true.
The shelf network that @gadget73 wisely indicates may be needed, is part of the frequency limiting done to achieve stability and we would explore that next. It is composed of a small value capacitor in series with a resistor that is placed in parallel with the 470K plate resistor of the 6AN8 pentode. I prefer to connect this between the plate and the 300V supply because it is usually easier and there is less voltage drop across the cap, rather than plate to ground- but either work identically as both the 300V and ground are "AC ground". Another method of adding stability margin is a zobel network directly across the speaker output terminals- also a shelf network of series capacitor and resistor. This is less favored by Dave Gillespie, as it is a bit of a brute force approach, but it can provide necessary margin in higher gain or higher power situations, or where the shelf network may already be impinging upon the audio band.
You might be wondering, why all this concern about stability? I don't see any oscillation in my amplifier. Well, any vacuum tube amplifier should be able to tolerate being powered on while there is no output speaker load connected. This is why I had you convert your load so you can still scope while it is switched open. That does not mean any significant signal should be applied- many output transformers have been burned open when the speaker connection is lost while the the amplifier is putting out power. If output load is lost while driving a large plate signal into the primary of the transformer the inductive backlash can easily arc across a winding. This test is done with very little signal, and that is not even strictly necessary. The ability of the amplifier to remain stable with no load is further tested by determining it's margin- by placing a small amount of capacitance across the open output, and increasing that capacitance in steps to determine how much pure capacitance load the amp can tolerate before it goes unstable. This is actually important because sometimes a speaker cable may be connected with no speaker connected at the far end. The parasitic capacitance of the cable itself can cause oscillation if you don't have some margin. But, overall, this is not even really about stability only under those conditions. An amplifier that has very little stability margin with no load can actually be partially unstable WITH load. Dave and I both saw this observing low frequency response with amps operating with too little margin with full rated load. On the left was mine at 10Hz on a 6BQ5 push-pull, on the right is Dave's on a Magnavox 8800 6V6GT push-pull at 90Hz.
So, once you are content with your feedback compensation, (after finally choosing ultra linear or not), the amp should be connected to the test setup with rated load enabled, and a very small amplitude square wave driven to the output. In the example below I used a 10kHz at ~2Vrms. You generally want to use a high enough frequency square wave so your scope is operating at a time resolution that enables you to clearly see the oscillation. Then, switch the load to open. The shape and amplitude of the square wave will change, but you should not see oscillation. If you do, turn the load back on and you know you have work to do. If the waveform is basically stable with no oscillation, you can then try a series of small capacitors across the load terminals.
The steps in the picture below are with 0uF, .01uF, .015uF, .018uF, .021uF, and .031uF:
You will quickly find a clear demonstration here that not all capacitors are equal. Two different capacitors of equal value can be used, one may cause oscillation and the other not- revealing the ESR and high frequency quality of the cap. Film caps are generally better than ceramic, mica generally better than film, but use whatever you have and play with it. Dave Gillespie does this margin analysis on every amplifier that he builds, and indicates the results on this schematic. From his work, he and I do not consider the amplifier to be acceptably stable until you can tolerate from 0.02uF to 0.03uF across the load. The amplifier above is considered stable to 0.02uF.
How difficult this stability is to achieve, is highly variable depending upon the particulars of the design, wire dress, and quality of the output transformer. Frequently, when using real Dynaco Z-565 transformers (that these are supposed to be a reasonable clone of) Dave is able to achieve "absolute stability", meaning any reasonable amount of capacitance across the load causes no oscillation. We'll see how these actually do.
Once you have a feel for what margin this amplifier has, if any, that will determine next steps, and as Dave has indicated, that is as much art as it is a science. The final stability tuning solution will ultimately trade off bandwidth versus distortion, and will represent a compromise that fits your individual preference. Let us know how this goes, and we can suggest solutions as needed.
At this point I plan to wait until the silver mica's come. I ordered 62, 68,75, and 82. Then I will run them and see which looks best on the scope. While I wait I will be drilling out and painting the other mono block, plenty of work there yet.
If I want to use the UL taps, then I need to start back at finding the feedback resistor value? Then continue to where I am now? Once the capacitors arrive I think I'll be going the UL route. The taps would connect to pin 4 of the 6v6's?
I have not actually hooked a speaker up yet. Maybe tonight or tomorrow.
I have the information that @gadget73 asked for. I would switch to the next frequency and then adjust input voltage to 125mv.
Frequency - output voltage
20k - 2.08
30k - 2.11
40k - 2.23
50k - 2.20
60k - 2.31
70k - 2.11
80k - 1.83
90k - 1.50
100k 1.25
If I want to use the UL taps, then I need to start back at finding the feedback resistor value? Then continue to where I am now? Once the capacitors arrive I think I'll be going the UL route. The taps would connect to pin 4 of the 6v6's?
I have not actually hooked a speaker up yet. Maybe tonight or tomorrow.
I have the information that @gadget73 asked for. I would switch to the next frequency and then adjust input voltage to 125mv.
Frequency - output voltage
20k - 2.08
30k - 2.11
40k - 2.23
50k - 2.20
60k - 2.31
70k - 2.11
80k - 1.83
90k - 1.50
100k 1.25
so it looks like your resonant peak is around 60khz. Thats far enough out that it can be tuned without seriously affecting anything within the audio band. Its also not a massive peak which helps considerably. The lower in frequency and the sharper the peak, the more difficult it gets to strike a compromise that keeps things stable without also having serious HF roll-off.
It may not change anything much- once you make the switch you can re-evaluate.
Yes- the two screens (pins 4) would be disconnected from each other, the 68 ohms would go away, the "325V" tap of the power supply would be unused, the green/white wire would connect to pin 4 of the 6V6 that has the blue/white wire at it's plate, the green wire would connect to pin 4 of the 6V6 that has the blue wire at it's plate. Your B+ and 300V supply voltages would shift around a bit, but not enough to correct. That's basically it.
I understand. What a pain. I used to have a signal generator like that. Now, I have a really cheap Chinese digital unit like this one that is far easier (but more tedious) to use. Microcontroller based, so it synthesizes the waveform in a rudimentary DAC, but has a fixed amplitude across any frequency.
Yes - I agree with @gadget73 - you can see clearly in the data that 60kHz is the resonant peak, and it always drops off like a stone after that. I will generally look at just one frequency near or before that peak, here 40 or 50kHz, and change capacitance until that is closest to the 1kHz amplitude.
If you are interested, as part of the full amplifier performance characterization suite, in addition to "Frequency Response" that is measured at 1 Watt, you can also measure "Power Bandwidth", that is measured at maximum power minus 1dB. The maximum power here is just a bit back from what you were showing in your earlier pictures- Maximum power should be measured at the biggest "good looking" sine wave just before any flat appears at the top or bottom. (It's up to your good conscience to determine where that is- I've seen plenty of fudging in the past to get to the nearest round number). Then, you calculate the RMS voltage for 1dB less than that, which is about 89% of that previous power. (I'll let you calculate voltage.) That 1dB back-off is to allow any small resonant peaks or response bumps to have the headroom to occur, and not be obscured because clipping has leveled it off.
The ability to reproduce the full bandwidth at this near full power is what really distinguishes a quality transformer like these Dynaco clones from the stock iron. At the low end, it takes great iron (and windings done well enough to utilize it) to conduct the huge amount of magnetic flux required to reproduce extremely low frequencies without saturating. At the high end, the windings must be designed to minimize stay inductance and capacitance to be able to reproduce the very high edge rates required to reproduce high frequency at large amplitude- lesser transformers "slew rate limit" and show high distortion and lower amplitude at these frequencies while under power. The actual Dynaco Z-565 are quite able to reproduce 20Hz to 20kHz at full rated power within 1dB. Others, not so much.
In the plot below, taken from a 6V6GT push-pull monoblock from a Lowrey organ, the frequency response at 1 W is pretty decent- only down 1dB from 20Hz to 45Hz. But, when under power (This is a 10 W amp, so at 8.9W) the 1dB down points are only 35Hz to a sucky 9kHz. I have to quote 3dB down points to indicate it's a 25Hz to 20kHz amplifier. (This is why -3dB is used so much in advertising- that's the "half power" point, which "sounds" about 15% to 20% quieter.) Let's see what yours can do . . .
I have it hooked to a schiits dac and my phone playing retro cocktail hour. Sounds pretty good. Fresh tubes arrived today but haven't installed them yet. I'm thinking building the second with the UL taps and then I can hear one of each and see if I can tell a difference.
Surprise! I couldn't wait. I got a 68pf silver mica in today. I wired it to the UL taps, went through the feedback setting again, to get 125mV to 2 vrms I had to up the resistance. Currently has 29.2k as the feedback resistors with a 68pF in parallel. The scope picture for this arrangement is below. Do I go any lower? I think I have a 62pF coming.
It looks like 8.75vrms is right before it starts to flat top. So maybe 8.5vrms is a good top power voltage. Looks like 9 watts. If I have the right formula 8.5x8.5/8=9.03.
It looks like 8.75vrms is right before it starts to flat top. So maybe 8.5vrms is a good top power voltage. Looks like 9 watts. If I have the right formula 8.5x8.5/8=9.03.
Attachments
Last edited: