Help me in finding the load for a pair of PP KT88 in UL mode

[Alan0354]

Hi

I am trying to learn how to draw the load line. I am thinking about using KT88 as it's the only one that gives the graph of UL. The article I use to write this started out with RDH4 P584 in the part of drawing the composite load line of pentode. More importantly, using the Valve Wizard article: http://www.valvewizard.co.uk/pp.html in drawing the composite and class B load line.

I am using Ebb=420V supply. The steps I did:

1) Draw the composite load line of my choice with one end at ( Ebb=420V, Ib=0mA) and draw the load line below the max power dissipation curve. The Composite load Rc = 420V/0.36A=1.166K. Therefore Raa = 4 X Rc = 4.67K. The Rc is drawn in RED color.

3) I then draw the class B load impedance of 1/2 X Raa = 2.33K. I adjust the line up and down to find the idle current where the class B load line still below the max power dissipation curve and draw the line in BLUE. The idle current is 70mA as shown:




3) I calculate the output power P= 1/2 X (420V-30V) X 180mA = 35W. More realistically, P= 1/2 X (420-50)V X 0.18A = 33W as I don't expect the plate voltage to go below 50V

Lundahl 1679 primary impedance is 4.5K, it's very close to what I have here.

Can anyone verify whether I am correct in this graph and calculation?

Thanks

 

[kward]

Plate power is calculated in the normal way. P = I*E, (using RMS values of the max excursions of I and E). Max power is usually what you want to know, so use the class B load line as an estimation. Thus

Plate power = (max change in I)/1.414 * (Max change in E)/1.414.

Power output as seen on the secondary under a resistive load that is matched to the secondary impedance is usually about 75% to 80% of the power generated at the plate.

 

[Alan0354]

Yes, your calculation is the same as my calculation. I use the peak value times 0.701 to get the RMS value, so W= Vpeak X 0.707 X Ipeak X 0.707 = 0.5 X Vpeak X Ipeak.

the question is the way I am doing it correct?

Also, I read about using lower primary impedance reduce distortion. I did a few graphs of Ebb= 380V, 320V and 300V, the output power is 31.5W, 29W and 27.5W respectively. That's not too bad. The lower the Ebb, the lower the primary impedance I can use, also more importantly, the bigger the class A region I can get so the amp is running in class A at lower volume.

This is the graph for Ebb=300V and Primary Raa=2.3K ( available from Lundahl). Output power is 27.5W:



There should be sound improvement not only in distortion, lower primary impedance got to reduce the parasitics as the turn ratio is lower and less turns in the primary. It got to help.

And also by lowering the Ebb, I can use 100mA idle current per tube. That gives me 5.5W of class A power from calculation. That's is very important to lessen the crossover distortion effect.

Is this the right way of looking at things?

 

[kward]

Given the Raa of the transformer you quoted in post #1, the slopes of both load lines and the crossing points look correct. So yeah, I think the approach you are using is sound. However, given the use of KT88s I would probably lean toward the opposite end of the scale, that is, maximizing power vs. trying to minimize distortion. That's a fair trade in my book as long as distortion doesn't rocket off the scales at the frequency extremes.

That's why I was suggesting the Triode Dynakit clone transformers. They are solid performers across the spectrum. Their A431S is just an excellent transformer. Several years ago I built a dual channel PP amp using these output transformers with KT120 tubes. I used 475V B+, and squeaked out 50 solid watts/channel, both channels driven, where the output stage was connected UL, and with distortion at the frequency extremes in check (less than about 2% at 20KHz as I recall). The low end was more of the surprise. The amp was 0 dB down (reference 1 KHz) at 16 Hz at max power with distortion of < 1.4%. I think I could have reduced distortion even more by using 6550s or KT88s instead of KT120s. KT120's are not quite as linear as the previous two.

So, if you really are considering the A431 clone, I think you can push the KT88s or 6550s, or even KT120s to 70 watts output if you had a beefy stout power supply that delivered say 525V B+, UL connected. And you could do that with low distortion output across the audio spectrum to boot.

 

[maxhifi]

One thing to keep in mind, is those UL curves are only valid for one specific B+ value, in that case 425V. That's because this is the DC value on the screen grid. Any other B+ value would have different looking curves, this one becomes invalid. I suspect this is why UL curves are so hard to find, you would need a different one for each possible value of B+, so they're not very useful.

I suggest to look at manufacturers recommended operating conditions as a starting point. Will do a more detailed job checking calculations when I have a bit of free time.

-Max

 

[Alan0354]

Given the Raa of the transformer you quoted in post #1, the slopes of both load lines and the crossing points look correct. So yeah, I think the approach you are using is sound. However, given the use of KT88s I would probably lean toward the opposite end of the scale, that is, maximizing power vs. trying to minimize distortion. That's a fair trade in my book as long as distortion doesn't rocket off the scales at the frequency extremes.

That's why I was suggesting the Triode Dynakit clone transformers. They are solid performers across the spectrum. Their A431S is just an excellent transformer. Several years ago I built a dual channel PP amp using these output transformers with KT120 tubes. I used 475V B+, and squeaked out 50 solid watts/channel, both channels driven, where the output stage was connected UL, and with distortion at the frequency extremes in check (less than about 2% at 20KHz as I recall). The low end was more of the surprise. The amp was 0 dB down (reference 1 KHz) at 16 Hz at max power with distortion of < 1.4%. I think I could have reduced distortion even more by using 6550s or KT88s instead of KT120s. KT120's are not quite as linear as the previous two.

So, if you really are considering the A431 clone, I think you can push the KT88s or 6550s, or even KT120s to 70 watts output if you had a beefy stout power supply that delivered say 525V B+, UL connected. And you could do that with low distortion output across the audio spectrum to boot.

Thanks for the reply.

2% THD at 20KHz is really high!!! More importantly, when you look at the 2nd and 3th harmonics vs plate load, 3th harmonic goes up when the plate loading resistance goes up. At low resistance, 2nd harmonic is high, but it gets cancelled in PP stage, but 3rd and odd harmonics add in PP. My understanding is 3th harmonic is not good for the sound. I want to do better than 2%. Maybe I am really not used to tubes, to me 0.1% THD is really really high!!!

Thanks

Alan

 

[Alan0354]

One thing to keep in mind, is those UL curves are only valid for one specific B+ value, in that case 425V. That's because this is the DC value on the screen grid. Any other B+ value would have different looking curves, this one becomes invalid. I suspect this is why UL curves are so hard to find, you would need a different one for each possible value of B+, so they're not very useful.

I suggest to look at manufacturers recommended operating conditions as a starting point. Will do a more detailed job checking calculations when I have a bit of free time.

-Max

Hi Max

This is only a test run, I mainly want to check my work.

The curves in the first post should be very close as I use Ebb=420V. So my first graph and calculation is valid. Antek has a 400VA 320V-320V torroidal power transformer. That gives me 450V after SS rectifiers. If I put a MOSFET regulation circuit to get it down to 420V, that would be perfect.

The problem is the data sheet for other condition is just if not more limited. 6L6 and EL34 only spec Vg2=250V for tetrode/pentode mode, the graph is useless for other voltage also. 6L6GC data sheet does show plate curves of different Vg2 for Vg1=0, at least I can use that. But none other tubes show that. So it's a wild guess also. Problem the datasheet is way too old, and no body tries to update the information. It's all a guessing game.

About the only mode that is reliable is triode mode. But I drew the graph and the output power is very low, not better than parallel SE amp.

From drawing all the graphs, I notice the way the description, people really don't look at the plate curves, mainly look at the power dissipation curve and tries to make the load line stay under the max dissipation. It is really NOT possible to draw the load line so it intersect at the knee or to the right of the knee at Vg1=0 curve both in pentode or UL mode unless the Ebb is very very low. This implies the accuracy of the curve is really not that important as one doesn't even have choice unless they are willing to stay with very low power design.

I am just trying to make the slope of the load line as steep as possible to get lower distortion and try to settle with about 30W only.

Thanks

Alan

 

[Alan0354]

I was just reading RDH4 again on distortion, I find a contradiction:

In page 570 and in other places, I always see graphs on distortion vs load resistance like this:



Here, you see 3rd harmonic goes down with DECREASE load resistance, 2nd goes down with increase load resistance. We know odd harmonics get CANCELLED out in PP configuration. This means we don't need to worry about odd harmonics, optimize for low 2nd harmonic.



BUT

In p577, it is written:



See the RED lined part, saying that decreasing RL increase distortion and increase RL reduce odd harmonic distortion.

This is like speaking from both sides of the mouth totally contradict each other. Any explanation on this?

 

[kward]

2% THD at 20KHz is really high!!!

Even for a tube amp it is a little high. But that was without any aids whatsoever, such as voltage regulators. Decent tube amps produce up to about 1% THD at the frequency extremes, and maybe about 0.15% at 1 KHz, at full power output, both channels driven. Monoblocks probably do a little better because of the lesser strain on the power supply. The very best tube amps can do a little better than that, but still not down to 0.1% across the board.

 

[Alan0354]

Even for a tube amp it is a little high. But that was without any aids whatsoever, such as voltage regulators. Decent tube amps produce up to about 1% THD at the frequency extremes, and maybe about 0.15% at 1 KHz, at full power output, both channels driven. Monoblocks probably do a little better because of the lesser strain on the power supply. The very best tube amps can do a little better than that, but still not down to 0.1% across the board.

That's why I am trying to sacrifice power to lower the THD. BUT as seen in post #8, I am totally confused. Two polar opposite in the two pages in RDH4. I cannot find explanation on why the even and odd harmonics behave as in Fig. 13.30, nor reason why it said that in p577.

Problem with those old books and article, you just have to trust what they say, but if they speak from both sides of the mouth, you are $crewed. I like the Valve Wizard article, even though it's for guitar amp. It make sense, I read over and over the composite load line of PP amp in RDH4( page 584), it only gave the line from the (Ebb, Ib=0) to the knee of the Vg1=0 curve, NOT caring whether it cross the max power dissipation curve, and set that to 1/4 Raa. That's it!!! The simple Valve Wizard showed it all, draw two lines for class B ( same as composite) load line and Class A load line BOTH UNDER the max dissipation curve. Then push the class A line up to get the idle point. It's so simple and none of the old books I saw manage to explain that. It is frustrating to read those old books......Not that they are difficult to read, they are actually very simple, the math is simple, I have no issue following the math, it's the inconsistency that really giving me problem. It's like .......Say what?!!! Where's that come from. I read the MIT paper and went back and fore, helped, but not much. Then the Valve Wizzard cleared up everything with just simple graph.

 

[maxhifi]

I was just reading RDH4 again on distortion, I find a contradiction:

In page 570 and in other places, I always see graphs on distortion vs load resistance like this:

View attachment 1041490

Here, you see 3rd harmonic goes down with DECREASE load resistance, 2nd goes down with increase load resistance. We know odd harmonics get CANCELLED out in PP configuration. This means we don't need to worry about odd harmonics, optimize for low 2nd harmonic.



BUT

In p577, it is written:

View attachment 1041492

See the RED lined part, saying that decreasing RL increase distortion and increase RL reduce odd harmonic distortion.

This is like speaking from both sides of the mouth totally contradict each other. Any explanation on this?

Note in the graph, power output drops off as impedance goes down. The only fixed condition is signal input voltage. If the signal voltage were reduced as the impedance increases, so that distortion was compared at equal power output, I doubt that the drop in third harmonic would occur the same way. I think the 3rd harmonic increases because the tube is putting put more power, due to being better matched to the load, not because the load impedance is higher.

I think this explains the apparent contradiction you are seeing.

 

[Alan0354]

Note in the graph, power output drops off as impedance goes down. The only fixed condition is signal input voltage. If the signal voltage were reduced as the impedance increases, so that distortion was compared at equal power output, I doubt that the drop in third harmonic would occur the same way. I think the 3rd harmonic increases because the tube is putting put more power, due to being better matched to the load, not because the load impedance is higher.

I think this explains the apparent contradiction you are seeing.

But the distortion is in %, output power is not relevant then.

 

[maxhifi]

But the distortion is in %, output power is not relevant then.

Signal input is 14V peak, as stated under the graph. Output power is relevant because when the tube is matched properly to the load, it is putting more out more voltage and current, and thus generating more disortion.

In other words, it's no suprise it makes more distortiob at 7W than at 4W, regardless of loading.

This graph would be more useful if it showed the 1W distortion versus load, instead of varying power output and load at the same time.

 

[Alan0354]

Signal input is 14V peak, as stated under the graph. Output power is relevant because when the tube is matched properly to the load, it is putting more out more voltage and current, and thus generating more disortion.

In other words, it's no suprise it makes more distortiob at 7W than at 4W, regardless of loading.

This graph would be more useful if it showed the 1W distortion versus load, instead of varying power output and load at the same time.

Ec1 is the DC grid bias not the AC signal, the book make it very clear the "E" is DC component at -14V, they need to specify ec. I don't think they mean that's the AC signal, that will imply the GAIN falls off, which should not happen.

If the output power is lower, you put smaller signal to the input, at the peak power, you measure the % of harmonics. the % is very relevant regardless of the max power.

See, this is the kind of things that make you go around loops. The new books usually are very clear on this.

 

[maxhifi]

Ec1 is the DC grid bias not the AC signal, the book make it very clear the "E" is DC component at -14V, they need to specify ec. I don't think they mean that's the AC signal, that will imply the GAIN falls off, which should not happen.

If the output power is lower, you put smaller signal to the input, at the peak power, you measure the % of harmonics. the % is very relevant regardless of the max power.

See, this is the kind of things that make you go around loops. The new books usually are very clear on this.

Hi Alan, look again, the bias voltage is -14V, the "Peak signal" voltage is also 14V. This is actually a peak to peak voltage, and the reason they use 14V is because it's driving the tube to grid current, i.e. from -14V to 0V on the grid.

Yes, the voltage gain does vary with load. Why? The tube's plate resistance is in series with the load. This is why the power output drops below Rl = Rp, and also above it. As you can see, power output drops to zero as Rl drops to zero. This is because more and more of the voltage is dropped across Rp, and less and less is dropped across Rl. Once you get Rl > Rp, then now there's plenty of voltage, but less and less current, so less power to the load.

What p. 577 is talking about, is that if we choose an arbitrary power output, which is well below the maximum amount when RL=Rp say, 1W, that the distortion will be lower, when Rl > Rp, than when Rl = Rp.

Going for Rl < Rp is bad news for several reasons. Less power output, higher distortion, and worst of all, worse voltage regulation/damping factor.

p.577 is accurate, and a common way to take advantage of this from a practical perspective, is to put 8 ohm speakers on the 4 ohm taps of a tube amp. This effectively doubles the damping factor, and lowers distortion, although it also reduces power output. I run my amps this way pretty often if I can handle less power.

 

[Alan0354]

Hi Alan, look again, the bias voltage is -14V, the "Peak signal" voltage is also 14V. This is actually a peak to peak voltage, and the reason they use 14V is because it's driving the tube to grid current, i.e. from -14V to 0V on the grid.

Yes, the voltage gain does vary with load. Why? The tube's plate resistance is in series with the load. This is why the power output drops below Rl = Rp, and also above it. As you can see, power output drops to zero as Rl drops to zero. This is because more and more of the voltage is dropped across Rp, and less and less is dropped across Rl. Once you get Rl > Rp, then now there's plenty of voltage, but less and less current, so less power to the load.

What p. 577 is talking about, is that if we choose an arbitrary power output, which is well below the maximum amount when RL=Rp say, 1W, that the distortion will be lower, when Rl > Rp, than when Rl = Rp.

Going for Rl < Rp is bad news for several reasons. Less power output, higher distortion, and worst of all, worse voltage regulation/damping factor.

p.577 is accurate, and a common way to take advantage of this from a practical perspective, is to put 8 ohm speakers on the 4 ohm taps of a tube amp. This effectively doubles the damping factor, and lowers distortion, although it also reduces power output. I run my amps this way pretty often if I can handle less power.

I see the 14Vpeak AC signal. I agree about the gain drop off due to rp. I still need to think about the % distortion. This still means % of the peak output. The smaller the output, the smaller the harmonics.

eg. If the lower power condition has 10V output, 1% gives 100mV, For higher power condition, say it has 20V output, 1% is 200mV. So the amplitude of the harmonics change with the output voltage it is referenced to. That's the only thing that is important, is the %, not the absolutely value.


I don't see from p577 that it's for small signal, it is a general statement. Also, it's important to look at the graph in p570, if you look at the distortion graph, it has a minimum point in the middle where the 2nd is falling and before 3rd really goes up. BUT if you think of a push pull, it cancel out all the 2nd harmonic, it will not do anything to the 3rd. So still the result is you get lower distortion with lower load resistance.

 

[Alan0354]

The % of distortion vs the absolute distortion amplitude is very different. This is what make the crossover distortion so intrusive. The distortion amplitude is constant regardless of the signal amplitude. Say the distortion is 10mVpeak. If the THD is measure at output of 40Vpeak, the THD is 10mV/40V X 100%= 0.025%. But when you listen to the music at low volume, say 2Vpeak. The THD is 10mV/2V X 100% = 0.5%. That's bad. That's why Class AB amp has a "U" shape THD curve where the THD is high at low amplitude, down to minimum in the middle and rise when getting close to clipping. That's how company try to cheat people by spec the THD at close to max. That say nothing about in normal listening level. This can only be fixed by having a big class A region, so when it transition to class B, the signal amplitude is so high that the THD is low at that point.

Back to the question in %, it's the % that is important, it's relative to the amplitude used to measure, The output power level is irrelevant. It's only the %.

 

[maxhifi]

I see the 14Vpeak AC signal. I agree about the gain drop off due to rp. I still need to think about the % distortion. This still means % of the peak output. The smaller the output, the smaller the harmonics.

eg. If the lower power condition has 10V output, 1% gives 100mV, For higher power condition, say it has 20V output, 1% is 200mV. So the amplitude of the harmonics change with the output voltage it is referenced to. That's the only thing that is important, is the %, not the absolutely value.


I don't see from p577 that it's for small signal, it is a general statement. Also, it's important to look at the graph in p570, if you look at the distortion graph, it has a minimum point in the middle where the 2nd is falling and before 3rd really goes up. BUT if you think of a push pull, it cancel out all the 2nd harmonic, it will not do anything to the 3rd. So still the result is you get lower distortion with lower load resistance.

Yes, but there's no information to compare distortion at different load resistances, with the same output power. I think that graph isn't useful on that basis. I think third harmonic goes up sharply as the tube approaches maximum output power, and would be % wise much lower, if you reduced the signal voltage to produce the same power as you are getting at the lower Rl.

Eg. if you look at the attached graph (bottom image), you see d3 increases rapidly towards maximum output.

The lower Rl just prevents the tube from getting to maximum output, it doesn't actually have better performance, or lower distortion.

 

[triode17]

The number of "0.1%" THD does not tell the whole story. It depends on the spectrum of that distortion figure. If it's all odd harmonics (3rd,5th, 7th, etc), the amp. will sound bad. In musical terms, dissonant. But if they are all even harmonics (2nd, 4th, 6th, etc.), the amp. should sound good. Even harmonics are direct octave multiples - i.e. octaves of each other.

 

[Alan0354]

Yes, but there's no information to compare distortion at different load resistances, with the same output power. I think that graph isn't useful on that basis. I think third harmonic goes up sharply as the tube approaches maximum output power, and would be % wise much lower, if you reduced the signal voltage to produce the same power as you are getting at the lower Rl.

Eg. if you look at the attached graph (bottom image), you see d3 increases rapidly towards maximum output.

The lower Rl just prevents the tube from getting to maximum output, it doesn't actually have better performance, or lower distortion.

That's the reason people use % to spec so the we don't depend on the max output. We just care what is the % of distortion relative to the signal amplitude. Like the output power is lower with lower load resistance, the output voltage is lower, the harmonics are lower also.

In the sense of the output power, I agree that the amp with lower output power definitely is in a disadvantage as it might be pushing the max power and THD goes up high while the higher power amp is still working in the comfortable region at the same output power level of the lower power amp. This means if one use lower plate load resistor, you have less output power and you stress the amp more and have more distortion at the same power when pushed.

But that's where one has to make a decision, want to have lower distortion and run at low plate resistance, one might want to consider using 4 tubes instead or two to get back the power, then you have advantage of lower distortion.

I guess this is the difference between SS and tubes, we don't have enough power or drive, just parallel more transistors, no big deal. I use 9 pairs of output transistor to avoid beta droop, parasitic swamping the output resistance. Transistor is cheap, but I guess it's a big deal to add a pair of tubes. So if I use 4 tubes and with lower load resistance, I get better performance than if one optimize that single pair for maximum power.

Still, I cannot accept p577, There are a lot of graphs similar to the one in p570 that show the relation of 2nd and 3rd harmonics. We know PP amp cancels out even harmonics, isn't it common sense to run at lower plate load resistance that gives low 3rd harmonics and high 2nd harmonics and let the PP configuration cancels all the 2nd harmonics? p577 is about PP stage. That doen't make sense no matter what to me.