Jor-EL34 SE Amp Project

[Curious_George]

 

 

 

.

[captainbeefheart]

I think there is some confusion about these kits and the output transformers. The output transformers are not made in Japan, they are made in China which is why they are so cheap (cheap labor and loose regulations). It's the Z11 lamination's that are sourced from Japan that they use to make the transformers. Don't get me wrong, the transformers tested good so they are sufficient at winding them there in China. Quality core material is very important for transformer performance and so they are advertising they use high quality Z11 lams from Japan which is a good thing.

 

In one test using the same bobbin, same lamination thickness and core size the M6 Silicon Steel gave over 8H of inductance where the Z11 gave 12H. The higher permeability in the core material given the same amount of copper gave better results. I'd like to see some differences in eddy currents and hysteresis plots to further compare core material.

 

Don't lose too much thinking about needing to have this core material over that core material, really it's the application of part that matters. It's the engineers job to use the part in hand in the best way possible, so if you have slightly lower inductance it's not a game changer if you change to a lower source impedance. A quick explanation is we want to load the tube with a specific impedance, say 2500 ohms. The speaker reflects back via impedance ratio the impedance to the tube, so you ask  yourself okay what's the problem? The impedance of the primary inductance is in relation to frequency, impedance drops with decrease in frequency. At low frequencies if source impedance isn't sufficiently low, or if inductance isn't sufficiently high, the impedance of the primary inductance may drop far below the reflected load of the speaker and dominate loading the tube down creating excessive distortion.

 

So for 2500 ohms to 8 ohm speaker, gives us an impedance ratio of 312:1

 

Let's now assume the plate impedance of a triode strapped EL84 is 1800 ohms, which from memory sounds about right if I'm wrong just go with it and please let me know in comments what it actually is, I'm too lazy to look right now.

 

Now lets assume the primary inductance of the 2500:8 output transformer is 10H. Will this transformer give satisfactory low frequency performance? Let's see.

 

The impedance of 10H at 35Hz is 2199 ohms. As we can see at 35Hz the primary inductance is below the target impedance of 2500.

 

Down further at 20Hz the 10H drops to an impedance of 1256, that's half the impedance that we want to load the tube with. Since the inductance impedance and reflected secondary load are in parallel things are actually much worse than it appears. At 35hz the load the tube sees is 2500 | | 2199 which is a total of 1169. So at 35Hz we are actually at half the load we want for the tube.

 

The greater the inductance the better the low frequency performance will be. So why don't we just make the primary inductance as high as possible? It's a tradeoff, typically the more inductance you get the more copper is wrapped around the core to make the large inductance value so then we end up with higher amounts of leakage inductance and interwinding shunt capacitance which kills our high frequency bandwidth.

 

Yikes what a pickle right?

 

For all you smarties out there looking for more information here we go. The plate impedance of the tube, called source impedance and the load impedance create a voltage divider, just like a volume pot.

 

If the plate of the EL84 is 1800 ohms, and we have a 2500 ohm load then we can assume a divider ratio of .58, which means 58% of the signal will get through. What happens if we now have a load of 1169 like at 35Hz? Well we now have a divider ratio of .39 or 39%. We can see that we will be 19% lower at 35Hz vs mid frequencies like 1kHz.

 

Make sense?

 

So to get the best high frequency response you want to choose an output transformer with the right amount of primary inductance that's just enough to work well but not too much to reduce high frequency bandwidth.

 

This is why I prefer to choose a tube that has a low plate impedance, or source impedance because it changes the divider ratio. Say we have a tube with a plate impedance of 600 ohms. With a 2500 ohm load we have a divider ratio of .8 or 80%. With only 10H and a combined impedance at 35Hz of 1169 the divider ratio is .66 or 66%. Much better right? So with the same small 10H primary inductance choosing the right tube to work with it can be a major difference in performance.

 

I also like to choose low impedance tubes because they require a lower load impedance, with a lower load impedance you have less of a turns ratio. This means less copper is needed and so leakage inductance is less and so is interwinding shunt capacitance. A 5k:8 output transformer is guaranteed to have higher leakage inductance and shunt capacitance than a 2.5k:8 transformer, typically. This is why the McIntosh output transformer has such high bandwidth, they split load the tube so impedance needed is halved greatly reducing the leakage inductance and shunt capacitance. The fact they also bifilar wind them practically makes leakage inductance non existent.

 

Hopes this helps some of the people questioning the mystique around output transformers and what all the specs mean to the application.

[captainbeefheart]

17 minutes ago, Curious_George said:

Here is some data using Svetlana EL34's. A bit more power. 

 

When you take your measurements are you also taking bias readings? I doubt it's the brand of tube getting more power, it's more likely the Svetlana has a higher bias current allowing it produce more output power. The more current through the tube the higher the cathode voltage is, in respect to grid they become more negative which allows greater input swing also.

 

Cathode bias is tricky, a tube that biases up hot (more current) will produce more power given roughly the same transconductance health.

[Curious_George]

4 minutes ago, captainbeefheart said:

 

When you take your measurements are you also taking bias readings? I doubt it's the brand of tube getting more power, it's more likely the Svetlana has a higher bias current allowing it produce more output power. The more current through the tube the higher the cathode voltage is, in respect to grid they become more negative which allows greater input swing also.

 

Cathode bias is tricky, a tube that biases up hot (more current) will produce more power given roughly the same transconductance health.

I did not measure the bias this time, but I will make a spreadsheet of the various tubes I have (EL34/6L6) for comparison. 

[captainbeefheart]

Conversely if you have the same bias current through the EL34 then one can assume the transconductance of the Svetlana (and or it has lower plate impedance at given operating point) is much higher than average.

 

Another reason for higher output power is the divider ratio the tube plate and load make together. As bias current increases plate impedance typically is reduced getting more signal out.

 

The simulations I made yesterday assume a tube with an average transconductance (11mA/V) and a very specific operating point or bias current. I could increase the current and produce more power. It's all one composite circuit.

[Racer X]

Thank you very much for taking the time to measure, plot, and share, very much appreciated.

 

So, I see 1W at 1.25% THD,  5.5W at 5% for a fair amount of headroom, and a very nice looking frequency response extending both low and high. 

 

Really want to try a similar EL34 SE amp in my system and compare to my reference amps.

[captainbeefheart]

1 minute ago, Curious_George said:

I did not measure the bias this time, but I will make a spreadsheet of the various tubes I have (EL34/6L6) for comparison. 

 

That would be excellent. If you have a mutual transconductance tester that info would also be very useful but just bias current of the power tube should be sufficient and we can assume an average gm for the tubes. I have found some modern tubes tend to be higher/lower in gm than datasheet suggests so they aren't the exact same tube as in the tube days but close enough to call them it by name and function in said circuit.

 

A graph of power out and THD along with bias current would be great. You'll probably see that Svetlana tube is running hotter than the others you tested. If not it could be on the high side for gm or have low rp or both, in that case making it a nice tube choice if they are consistently like that.

[captainbeefheart]

Another reason I tend to gravitate towards using negative feedback is because it makes differences in tubes less critical to amplifier performance. i.e. swap tubes and you are guaranteed to get the same results.

[Curious_George]

20 minutes ago, captainbeefheart said:

I have found some modern tubes tend to be higher/lower in gm than datasheet suggests so they aren't the exact same tube as in the tube days but close enough to call them it by name and function in said circuit.

I have suspected this for a long time, especially with certain Chinese type tubes. For example, New Sensor Corporation who makes tubes (New Sensor uses a Russian factory) under the following brands: EH (Electro-Harmonix), Tung-Sol, Sovtek, Mullard, Gold Lion, etc cross markets all these tubes such as 5881, 6L6GB/GC. I suspect they are the same tube with different price points. same comment for EL34's. 

 

I have some EH300B's and they are nice, but the exact same tube sold under Gold Lion are twice the price. If you do your homework, you can get great tubes, but at a lower cost. Just have to poke around and ask questions. 

[Curious_George]

48 minutes ago, captainbeefheart said:

Another reason I tend to gravitate towards using negative feedback is because it makes differences in tubes less critical to amplifier performance. i.e. swap tubes and you are guaranteed to get the same results.

Yep. Although I have been on the SE train, I go back and forth between SE, Push-Pull (with feedback) and solid state. Then I'll start messing with speakers, etc and the cycle just goes round & round. 

[Curious_George]

50 minutes ago, captainbeefheart said:

Another reason I tend to gravitate towards using negative feedback is because it makes differences in tubes less critical to amplifier performance. i.e. swap tubes and you are guaranteed to get the same results.

I think you would like (sonically) my version (mono amps) of the Williamson's that I made. 

[captainbeefheart]

4 minutes ago, Curious_George said:

I think you would like (sonically) my version (mono amps) of the Williamson's that I made. 

 

The Williamson circuit I have found almost always sounds good, it's one of the greats especially in it's original Triode output form. The later UL or Pentode versions can be good also but not as good as the triode output original circuit.

 

 

I too have been on a single ended kick, I was doing the no negative feedback thing and had mixed results. With fully horn loaded speakers reaching 50% efficiency like the Khorn or Lascala etc.. they work fine enough and sound great but with your average 5-10% efficient speaker not so much. Even the Heresy's and Cornwall's can be a tough load for them and it's imperative to use very linear triodes, and the lower the plate impedance of the output triode the better I found the sound was.

 

Lately I switched from SET amps into getting more power out of single ended and going pentode operation with feedback. The downside is you typically need 20db of feedback to get the distortion to low levels and  with an output transformer one needs to fully understand control theory and compensation networks to achieve excellent results. I find when feedback is added without analysis of stability the sound can suffer greatly. Square wave testing will show overshoot and ringing especially if you add a little reactance to the bench load like for the real world. Another reason I feel amateur tube amp builders stray away from feedback amps, they don't like the sound of the amp because they didn't go through the effort of providing exact compensation lead/lag networks. They can make all the difference in the world. If you have great OPT, only two stages say using triode output stage, you can get away with 6-10db of feedback and it may not require any compensation. So using a little feedback with triodes is much easier to get good results.

[captainbeefheart]

@Curious_George would you post your Williamson circuit? If not in email or PM would be great. I love a good Williamson design.

[Curious_George]

On 8/27/2022 at 4:33 PM, captainbeefheart said:

 

.

 

[Curious_George]

.

[captainbeefheart]

3 minutes ago, Curious_George said:

With the SS power supply, I have not seen any of the purported "motorboating" or other instability issues occasionally reported with the Williamson amp. This power supply is my own design, not copied from the web. 

 

The original schematic used only 8uF decoupling capacitors and had a tolerance of +50%, typically they were probably much more than 8uF. Modern copies may use precision film caps 8uF that have a low ESR that won't have the dampening ability of a high ESR electrolytic. Another issue is the original had a massive 30H choke and if it's not at least 30H you'll end up with a low frequency resonance too high up that causes the audible motorboat.

 

If one increases the decoupling/filter caps you can use a lower choke and still keep the resonance below 10Hz. Small caps and small chokes don't play nice together, you need either large capacitance and low value choke, or a large value choke and smaller capacitance. Of course high capacitance and high inductance will just push the resonance towards DC.

 

I have never had any issues with any Williamson style amp I've made either but have heard others complain also. It's just from people with little experience in design mixing up the values and circuit without much thought.

 

I see you used 1625 power tubes, fancy!! I wondered about using them but was thinking pentode due to the 3.5watt screen dissipation limit in ultra linear. They are also 12v heaters so 12v front end tubes need to be chosen to keep one filament transformer, 12AU7 would work fine.

[Curious_George]

2 hours ago, captainbeefheart said:

I think there is some confusion about these kits and the output transformers. The output transformers are not made in Japan, they are made in China which is why they are so cheap (cheap labor and loose regulations). It's the Z11 lamination's that are sourced from Japan that they use to make the transformers. Don't get me wrong, the transformers tested good so they are sufficient at winding them there in China. Quality core material is very important for transformer performance and so they are advertising they use high quality Z11 lams from Japan which is a good thing.

 

In one test using the same bobbin, same lamination thickness and core size the M6 Silicon Steel gave over 8H of inductance where the Z11 gave 12H. The higher permeability in the core material given the same amount of copper gave better results. I'd like to see some differences in eddy currents and hysteresis plots to further compare core material.

 

Don't lose too much thinking about needing to have this core material over that core material, really it's the application of part that matters. It's the engineers job to use the part in hand in the best way possible, so if you have slightly lower inductance it's not a game changer if you change to a lower source impedance. A quick explanation is we want to load the tube with a specific impedance, say 2500 ohms. The speaker reflects back via impedance ratio the impedance to the tube, so you ask  yourself okay what's the problem? The impedance of the primary inductance is in relation to frequency, impedance drops with decrease in frequency. At low frequencies if source impedance isn't sufficiently low, or if inductance isn't sufficiently high, the impedance of the primary inductance may drop far below the reflected load of the speaker and dominate loading the tube down creating excessive distortion.

 

So for 2500 ohms to 8 ohm speaker, gives us an impedance ratio of 312:1

 

Let's now assume the plate impedance of a triode strapped EL84 is 1800 ohms, which from memory sounds about right if I'm wrong just go with it and please let me know in comments what it actually is, I'm too lazy to look right now.

 

Now lets assume the primary inductance of the 2500:8 output transformer is 10H. Will this transformer give satisfactory low frequency performance? Let's see.

 

The impedance of 10H at 35Hz is 2199 ohms. As we can see at 35Hz the primary inductance is below the target impedance of 2500.

 

Down further at 20Hz the 10H drops to an impedance of 1256, that's half the impedance that we want to load the tube with. Since the inductance impedance and reflected secondary load are in parallel things are actually much worse than it appears. At 35hz the load the tube sees is 2500 | | 2199 which is a total of 1169. So at 35Hz we are actually at half the load we want for the tube.

 

The greater the inductance the better the low frequency performance will be. So why don't we just make the primary inductance as high as possible? It's a tradeoff, typically the more inductance you get the more copper is wrapped around the core to make the large inductance value so then we end up with higher amounts of leakage inductance and interwinding shunt capacitance which kills our high frequency bandwidth.

 

Yikes what a pickle right?

 

For all you smarties out there looking for more information here we go. The plate impedance of the tube, called source impedance and the load impedance create a voltage divider, just like a volume pot.

 

If the plate of the EL84 is 1800 ohms, and we have a 2500 ohm load then we can assume a divider ratio of .58, which means 58% of the signal will get through. What happens if we now have a load of 1169 like at 35Hz? Well we now have a divider ratio of .39 or 39%. We can see that we will be 19% lower at 35Hz vs mid frequencies like 1kHz.

 

Make sense?

 

So to get the best high frequency response you want to choose an output transformer with the right amount of primary inductance that's just enough to work well but not too much to reduce high frequency bandwidth.

 

This is why I prefer to choose a tube that has a low plate impedance, or source impedance because it changes the divider ratio. Say we have a tube with a plate impedance of 600 ohms. With a 2500 ohm load we have a divider ratio of .8 or 80%. With only 10H and a combined impedance at 35Hz of 1169 the divider ratio is .66 or 66%. Much better right? So with the same small 10H primary inductance choosing the right tube to work with it can be a major difference in performance.

 

I also like to choose low impedance tubes because they require a lower load impedance, with a lower load impedance you have less of a turns ratio. This means less copper is needed and so leakage inductance is less and so is interwinding shunt capacitance. A 5k:8 output transformer is guaranteed to have higher leakage inductance and shunt capacitance than a 2.5k:8 transformer, typically. This is why the McIntosh output transformer has such high bandwidth, they split load the tube so impedance needed is halved greatly reducing the leakage inductance and shunt capacitance. The fact they also bifilar wind them practically makes leakage inductance non existent.

 

Hopes this helps some of the people questioning the mystique around output transformers and what all the specs mean to the application.

Even though I say I can't hear a difference between M6 and Z11 (which is Hi-B or high permeability core), I keep going back to my James OPT's that are Hi-B. They do sound great. Tango and Hashimoto use Hi-B as well. 

 

Z11 will allow you to use a smaller core with eveything else being equal. Apparantly, Hi-B is used by power companies extensively to reduce core losses in industry and utility pole transformers. Makes sense. 

[Curious_George]

.

 

 

 

 

[Curious_George]

.

[captainbeefheart]

Any core material that allows for less copper to get the same amount of inductance is in my book great as it will improve high frequency response.

 

The biggest difference will be in zero feedback amps where you have no negative feedback to increase bandwidth. Of course it's optimal to start out with the best performance before feedback is applied, I think it's less critical with feedback amps, adequate inductance is always important but often with high inductance transformers the top end may roll off earlier than you would like.

 

My buddy made a zero feedback amp with a transformer that had sufficient inductance, low frequency performance was excellent but he was already sloping down well before 20kHz, memory I believe it was around 14kHz where it started to slope down. Of course 20db of feedback will bring things up to above hearing range but he didn't want feedback for whatever reason. I've seen this in other threads of builders also.

 

So yes zero feedback amps definitely don't short change the output transformer.