My first valve amp - Luxman A3600 running KT88s

[VanArn]

One end of each 68k ohm resistor needs to be dis-connected from the biasing circuit so that its resistance can be accurately measured.  Pin 7 of an octal valve is a heater  connection  and you need to inspect the wiring again, as this end of the 68k ohm resistors  must be a ground point.

Edited September 10, 2022 by VanArn

[Luckiestmanalive]

  On 10/09/2022 at 9:23 PM, VanArn said:

One end of each 68k ohm resistor needs to be dis-connected from the biasing circuit so that its resistance can be accurately measured.  Pin 7 of an octal valve is a heater  connection  and you need to inspect the wiring again, as this end of the 68k ohm resistors  must be a ground point.

Expand  

 

Apologies - it is pin 1 not pin 7 - my bad. Thanks, for the tip, I hoped it would be something like that - I think I'll leave well them alone, then. 

[Luckiestmanalive]

Replacing the 40 year-old coupling capacitors for the output tubes (C107-C110) seems to have sharpened the focus around the instruments and voices and brought out the sound reflections in the recording space. Just waiting on my other replacement parts now.

[Luckiestmanalive]

I've replaced the old bias pots and the 20pf NFB ceramic capacitors (C115-116) with new Mica ones. I didn't try the amp without them because I decided I wanted to stay true to the Luxman spec and I love the way the amp makes music already.

Edited October 20, 2022 by Luckiestmanalive

[Luckiestmanalive]

Now I'm very happy with how this beauty sounds I want to make sure it stays that way. That means I want to reassure myself the more powerful KT120 output tubes don't overtax the power supply or output transformers. The A3600 was marketed as a 50W per channel amp and the plate dissipation of the 8045Gs was rated at 45W. I've read translations from Japanese blog posts that indicate the power supply was good for 70W and the output with four 8045Gs biased at 76ma might have been as high as 62W! However, I have installed KT120s capable of plate dissipation of 60W. I have biased each of them to take 75ma plate current and with 480V B+ it looks like this equates to 36W per tube (60%) but in an ultralinear push-pull configuration I'm not really sure what it means for total output per channel? The output transformers never get more than a little warm but the power transformer is slightly warm while in idle until I start playing music and then it gets steadily warmer and after an hour it is very warm - I can keep my hand on the top of its cover for 10 seconds but it feels uncomfortable.

 

[Luckiestmanalive]

Update: I've rebiased the KT120s down to 60ma for summer, which has reduced the temp of the power supply transformer and doesn't appear to have negatively affected the sound so, bonus! I've also replaced some overtaxed 2W resistors for 3W ones and replaced the 8.2K cathode resistors for the phase inverter circuit, adding more space between the PCB and these resistors to dissipate more heat (successfully).

 

Now I'm upgrading/replacing a few other important passive components in the signal path - one pair at a time on a weekly basis - and listening in between to see what, if any, difference they make, replacing:

• C111-112 0.22uf 40-year old (negative feedback?) mylar capacitors with Wurth metalized PP film (box) caps (done yesterday - see red boxes in pic below)
• C105-106 mylar (grid leak?) capacitors (see brown lozenges below) with Wima metalized PP film (box) caps
• C101-102 0.1uf 40-year old mylar input coupling capacitors (see green lozenges below) with Wurth metalized PP film (box) caps
• R101-102 470Kohm first stage grid leak resistors with different readings (507Kohm & 490Kohm) with closely matched replacements
• C103-104 old electrolytic 100uf cathode bypass capacitors (see brown caps between bottom two tubes below) with aluminium poly caps.

Edited November 28, 2022 by Luckiestmanalive

[Luckiestmanalive]

I've researched how to work out max RMS output from my amp with 4x KT120s. According to the tube chart and load line, it looks like the voltage swing of one KT120 tube is 490V @-60V less 90V @0V = 400V so double that (800V) peak to peak for a pair of tubes in push-pull config multiplied by 0.707 = 565.5V RMS. RMS power output = 565.5V squared / 3600ohm (output transformer impedance) = 89W per channel, if my logic and calcs are correct!

[muon*]

I'd get someone with better math than myself to check that, just doesn't sound right to me for PP with two KT120s each channel..

[xlr8or]

[Luckiestmanalive]

  On 30/11/2022 at 11:58 AM, muon* said:

I'd get someone with better math than myself to check that, just doesn't sound right to me for PP with two KT120s each channel..

Expand  

 

Yeah, seems scarily high. Here is the source of the formula:
https://www.vtadiy.com/book/chapter-6-step-by-step-design-of-a-push-pull-tube-amplifier/6-1-design-of-the-power-stage/

[Luckiestmanalive]

  On 30/11/2022 at 12:15 PM, xlr8or said:

Expand  

 

Thanks, Kirk - I refer to this handy calculator a bit for biasing the KT120s but is it as simple as working out the watts from your bias level? If it is, then I just enter 60ma in the form at the bottom and it spits out 29.4W per tube, so 58.8W per channel? And is that at idle, so it is higher when playing music?

[RoHo]

  On 30/11/2022 at 11:18 PM, Luckiestmanalive said:

 

Thanks, Kirk - I refer to this handy calculator a bit for biasing the KT120s but is it as simple as working out the watts from your bias level? If it is, then I just enter 60ma in the form at the bottom and it spits out 29.4W per tube, so 58.8W per channel? And is that at idle, so it is higher when playing music?

Expand  

Your looking at two different things.  The above calculation is for the tube's PLATE dissipation at steady state which is simply plate voltage x plate current.  This should be below the max plate dissipation as specified. 

The RMS output power is related to the work the tube is doing by pushing current through the output transformer.  Your initial formula relates to that.  

[Luckiestmanalive]

That makes sense, thanks, R. 

 

Edited December 1, 2022 by Luckiestmanalive

[Luckiestmanalive]

The other angle I've been coming at it from is from a tube and load line chart (see below). The red dot shows the 60ma set point in my amp. The dotted line denotes max output of 60W and I understand that injecting an input signal will cause the dot to rise and fall along the load line. Being a push-pull amplifier, a music signal causes the current to rise and voltage to fall along this load line in one tube while the other in the pair for that channel does the opposite (current falls and voltage rises) - hence the reference to push and pull. 

 

The kink in the load line refers to the operating point where the current in the other tube in the push-pull pair has sunk so low it is essentially 'off' and this reduces the impedance provided by the output transformer. Doesn't this logic suggest that, in practice, each pair of KT120s in push-pull operation can't really produce anywhere near the 89W per channel in my original theoretical calculation as I just doubled the voltage swing on one tube? In reality, a pair together will produce more than the 60W at idle but might peak at about 70-75W in push-pull configuration?

 

 

[RoHo]

I don’t quite follow your calculation but remember the “magic” of Class AB push pull is that the power is calculated from the total voltage swing across the pair of tubes which will include each “end” of the swing where the opposite tube is in cut-off.  Yes, the impedance halves but designer uses this kink to tuck around the max plate dissipation curve.  I dunno how accurate the sim is - to really get down and dirty with the theory check out Patrick Turners (RIP) Turner Audio website.  The section on Load Matching, Push Pull amplifiers will keep you occupied for weeks.

Here is my  “homework” from a few years ago!

 

[muon*]

Great source of info Patrick has left us.

 

I should read it more.

[Luckiestmanalive]

Thanks, guys! I've been to a few sites and downloaded a spreadsheet template to check the calculator on vtadiy.com but I'll definitely check out Patrick Turner's website, too.

[Luckiestmanalive]

I love your homework, M, btw! I've been reading whatever material I can get a hold of and tinkering with the VTAdiy model to get a better idea of the operating conditions of the amp. Patrick Turner's website is an amazing resource. There are parts I follow but much of it assumes knowledge and experience beyond me, which I guess shows what a loss he is to the community...

[Luckiestmanalive]

Carrying on with my logic using the load line, if we assume a fixed bias of 60ma (red dot below), then sending a 140V peak-to-peak sinusoidal signal to the pair of KT120s causes them to move in opposite directions along the load line - one tube increases current and decreases voltage and the other vice versa, following the green arrows I've drawn until they reach the black lines (one 70V below the set point and one 70V above it). At each of the black lines, one tube is operating at roughly 74% (45W) of max plate dissipation while the other is operating at only 31% (18W), the sum of which is 63W a side. I did the same calculations using 78ma bias (winter setting) and plate dissipation increases to 26W + 49W = 75W.

 

[xlr8or]

Might it help to change the tubes to KT88's or 6550's to keep the bias adjustment fixed for both winter and summer listening sessions? Changing the bias from 60mA to 78mA for the same plate voltage will also change the tonal characteristics. Also, the KT120 looks to be cold biased when it's set at 60mA. If heat dissipation is your real concern then perhaps reducing power output by changing the tube type might be the best option to not impact the tonal characteristics between winter and summer. Just a thought ....

[RoHo]

  On 04/12/2022 at 9:20 AM, Luckiestmanalive said:

Carrying on with my logic using the load line, if we assume a fixed bias of 60ma (red dot below), then sending a 140V peak-to-peak sinusoidal signal to the pair of KT120s causes them to move in opposite directions along the load line - one tube increases current and decreases voltage and the other vice versa, following the green arrows I've drawn until they reach the black lines (one 70V below the set point and one 70V above it). At each of the black lines, one tube is operating at roughly 74% (45W) of max plate dissipation while the other is operating at only 31% (18W), the sum of which is 63W a side. I did the same calculations using 78ma bias (winter setting) and plate dissipation increases to 26W + 49W = 75W.

 

Expand  

In terms of plate dissipation you only need to look at the individual tubes the combined value isn't of any real significance.  What is interesting is looking at your load line and following it up to the left into the Class B area after the kink the plate dissipation does go above the specified max between about Va of 180-320 V.  As this is not continuous ie it only goes into this area with a large peak in the grid voltage (input signal) it is OK but you don't want the tubes in this area for long periods.  This is one of the things you consider when setting your bias point and transformer impedance.  Do you want an amp with lots of power but that also burns through power tubes every 6 months?

[Luckiestmanalive]

  On 04/12/2022 at 9:39 PM, RoHo said:

In terms of plate dissipation you only need to look at the individual tubes the combined value isn't of any real significance... 

Do you want an amp with lots of power but that also burns through power tubes every 6 months?

Expand  

 

I have been looking at the combined output because I was concerned about overtaxing the power and output transformers but now I'm more worried about heat dissipation and shortening the life of my tubes. It looks like the transformer impedance (3600 ohms) is too low and the B+ is too high to curve around the max plate dissipation curves.

 

Thanks, K, for the suggestion re KT88, but the curve for that tube looks worse...

 

 

Edited December 5, 2022 by Luckiestmanalive

[RoHo]

  On 05/12/2022 at 1:50 AM, Luckiestmanalive said:

 

I have been looking at the combined output because I was concerned about overtaxing the power and output transformers but now I'm more worried about heat dissipation and shortening the life of my tubes. It looks like the transformer impedance (3600 ohms) is too low and the B+ is too high to curve around the max plate dissipation curves.

 

Thanks, K, for the suggestion re KT88, but the curve for that tube looks worse...

 

 

Expand  

When thinking about the suitability of transformers first think about current.  With the power transformer the steady-state bias current of the tubes is the main marker.  I wouldn't stray too far from the current drawn by the original tubes.  A quality amp will have a reasonable tolerance for increased current but if you are going outside the design parameters be careful - you really don't want to damage the power transformer.  Heater current is also all a primary concern.  Do the KT120s draw more than the original tubes?  If it's more than 10% or so above then think hard.

What output transformer impedance do KT120s generally like? Are there schematics floating around?  EL34/KT88 types tend to use around 5K.

[Luckiestmanalive]

Thanks, R! Heater current is fine - according to the tube datasheet, the original 8045G tube drew 2.5A while a KT120 draws 1.9A. On the OPT front, I've seen 4.3Kohm and 5Kohm impedance trannies in a google search. My Audio Physic speakers are a nominal 4ohm but average of 6ohm load. Would plugging them into the 8ohm speaker terminal help? Even pulling the load impedance up from 3.6Kohm to 4Kohm brings the load line below the max plate dissipation curve.

Edited December 5, 2022 by Luckiestmanalive

[RoHo]

No, but……..

The impedance spec of an output transformer is worked out using the likely load connected to the secondary and is determined by the number of turns of wire.  The transformer itself doesn’t have an intrinsic impedance.  Changing the load changes the impedance seen by the valve.  So you have an approx 6 ohm load connected to the 4 ohm tap?  That means the impedance seen by the output valves will be a corresponding 50% higher than the specified 3600.  Result. Can you run the sim again using a 5400 ohm load line?

Remember these numbers are calculations and approximations as any are when you try to characterise the complex impedance of a speaker as a single number.   But to dig down deeper is moving away from what the average hobbyist like us can do armed only with limited knowledge and a multimeter!