ng5921's DIY Amplifiers

[ng5921]

Well...it's unconventional, to say the least. These voltage regulator tubes are not typically used in series in the supply, they are used in parallel, which is how they regulate voltage. Also paralleling five capacitors feeding one power stage...this isn't how it's normally done. Anyway, I digress.

Getting back to the noise issue, I plugged this supply into LTSpice. Assuming I have it right, and the simulation is accurate, the ripple at the output stage is too high for a headphone amp, in my opinion. If putting in a choke really did make a big improvement, that would suggest there is too much ripple. One thing you could do is switch both of those 100uF electrolytics to 220uF, which would drop the ripple tenfold for the output stage.

Again, just going off of what you're giving me here, but that may be the issue.

I agree, I think increasing capacitances will definitely help. Thank you for simulating it! I probably should have planned and simulated before building, kind of an expensive lesson but that’s what I get for rushing

 

[Swagsupply]

I agree, I think increasing capacitances will definitely help. Thank you for simulating it! I probably should have planned and simulated before building, kind of an expensive lesson but that’s what I get for rushing

Did it help?

 

[ng5921]

So I did a bunch of experimentation yesterday, adding 47uF EL caps to the power supply since I had them on hand to see if there was any performance increase. The results were...strange, and I don't think I fully grasp how this power supply works. These were each separate experiments make on the base circuit (shown in the last diagram of the power supply I posted)

For reference, this is the base power supply circuit:

1) Adding 47uf Caps to the PSU on the input tube rail before and after the 0A3 regulator did not improve hum. Adding 47uF capacitors across the 0.1/3.3uF film caps after the 0A3 tube on the power tube rail also did not help. Having all these capacitors in at once did not help the hum.

2) Increasing the first filter capacitance to 55uF (47+8uf in parralel) helps reduce the hum a bit (yes I know this is out of spec for the 5u4g, I didn't run it for long)

3) Added the choke in as a choke-input filter, in front of the 8uF capacitor helps the hum, maybe 50% reduction

4) Adding the choke-input and 55uf (8+47uF Parralel) first filtering capacitance had the most improvement, making the hum nearly inaudible over ambient noise.

5) Choke input + 22+8uF initial filterint caps had a bit more hum than #4, but a safer capacitance within spec of the 5u4g.


I really thought adding the 47uF caps around the power supply would make some difference, but I guess I am wrong. I think this power supply might require some re-designing and I'm going to take a little break from working on this amp to do just that. I will post a new schematic here after thinking about it and trying to do some simulations in PSUD and LTSpice.

Perhaps more improvement could be achieved with a higher inductance rating choke. The one I am using is only 2Hy with an uknown current rating.

I was looking at the Lunadhl LL1685, which should be able to fit inside the chassis, and has the following configuations:

I was thinking the LL1685/100mA running in series would be sufficient. This amp pulls less than 50mA based on my math, but I will double check.

 

[L0rdGwyn]

So I did a bunch of experimentation yesterday, adding 47uF EL caps to the power supply since I had them on hand to see if there was any performance increase. The results were...strange, and I don't think I fully grasp how this power supply works. These were each separate experiments make on the base circuit (shown in the last diagram of the power supply I posted)

For reference, this is the base power supply circuit:


1) Adding 47uf Caps to the PSU on the input tube rail before and after the 0A3 regulator did not improve hum. Adding 47uF capacitors across the 0.1/3.3uF film caps after the 0A3 tube on the power tube rail also did not help. Having all these capacitors in at once did not help the hum.

2) Increasing the first filter capacitance to 55uF (47+8uf in parralel) helps reduce the hum a bit (yes I know this is out of spec for the 5u4g, I didn't run it for long)

3) Added the choke in as a choke-input filter, in front of the 8uF capacitor helps the hum, maybe 50% reduction

4) Adding the choke-input and 55uf (8+47uF Parralel) first filtering capacitance had the most improvement, making the hum nearly inaudible over ambient noise.

5) Choke input + 22+8uF initial filterint caps had a bit more hum than #4, but a safer capacitance within spec of the 5u4g.


I really thought adding the 47uF caps around the power supply would make some difference, but I guess I am wrong. I think this power supply might require some re-designing and I'm going to take a little break from working on this amp to do just that. I will post a new schematic here after thinking about it and trying to do some simulations in PSUD and LTSpice.

Perhaps more improvement could be achieved with a higher inductance rating choke. The one I am using is only 2Hy with an uknown current rating.

I was looking at the Lunadhl LL1685, which should be able to fit inside the chassis, and has the following configuations:


I was thinking the LL1685/100mA running in series would be sufficient. This amp pulls less than 50mA based on my math, but I will double check.

So you can't just use any choke for a choke input supply, you have to use a properly spec'd choke for choke input. They are more complex to design than cap input. You can put the choke after the 8uF reservoir cap for some additional filtering, but if you want to do a choke input supply I would suggest taking some time to educate yourself on it.

If adding additional filtering capacitance further down in the supply didn't improve the hum, then it likely isn't a ripple issue. The things you've done that have improved the hum are also lowering the rectifier ripple current pulses. So my suspicion is that your rectifier circuit is coupling into the signal path. If it's being picked up by the input stage, using a lower gain input tube, like I mentioned before, could potentially lower the noise floor. But if it's being coupled into the output stage, then that is a layout problem. You could try some shielding but almost always the best solution is moving the transformer-rectifier-capacitor loop away from signal path components.

 

[ng5921]

So you can't just use any choke for a choke input supply, you have to use a properly spec'd choke for choke input. They are more complex to design than cap input. You can put the choke after the 8uF reservoir cap for some additional filtering, but if you want to do a choke input supply I would suggest taking some time to educate yourself on it.

If adding additional filtering capacitance further down in the supply didn't improve the hum, then it likely isn't a ripple issue. The things you've done that have improved the hum are also lowering the rectifier ripple current pulses. So my suspicion is that your rectifier circuit is coupling into the signal path. If it's being picked up by the input stage, using a lower gain input tube, like I mentioned before, could potentially lower the noise floor. But if it's being coupled into the output stage, then that is a layout problem. You could try some shielding but almost always the best solution is moving the transformer-rectifier-capacitor loop away from signal path components.

Ah okay. I will do some more reading on this because I definitely do not have a good understanding of their differences, I appreciate the insight

I will go ahead and move the choke so that it comes after the first input capacitor for now.

I'll take another look at the amp later but I think I've separated the transformer-rectifier-capacitor loop away from the signal path components. I'll post a picture later, maybe there is something I overlooked...

 

[L0rdGwyn]

Ah okay. I will do some more reading on this, I appreciate the insight

I will go ahead and move the choke so that it comes after the first input capacitor in that case.

I'll take another look at the amp later but I think I've separated the transformer-rectifier-capacitor loop away from the signal path components. I'll post a picture later, maybe there is something I overlooked...

It's possible the proximity of the rectifier to the signal tubes is the issue. Did you ever try using a 12AU7 or rather than a 6922?

 

[ng5921]

It's possible the proximity of the rectifier to the signal tubes is the issue.

I will try the metal cookware again today and see if that has any effect! Maybe a metal water bottle just to cover up the rectifier.

Did you ever try using a 12AU7 or rather than a 6922?

I believe I have another 12AU7 on hand, I can try that later. I've been using 6n1p and a G-73R (12au7). The 6n1p has worse noise than the G-73R which I believe supports your suggestion.

 

[g0ldl10n]

Maybe try subbing in that SS rectifier?

 

[L0rdGwyn]

I will try the metal cookware again today and see if that has any effect! Maybe a metal water bottle just to cover up the rectifier.



I believe I have another 12AU7 on hand, I can try that later. I've been using 6n1p and a G-73R (12au7). The 6n1p has worse noise than the G-73R which I believe supports your suggestion.

You could try shielding the rectifier, but I'm not confident that will work. It's more that the rectifier is positioned where it is, has the rectifier ripple current pulses are traveling through the rectifier as well as the wiring from the rectifier to the capacitor, so it's on both sides of the chassis top plate, just shielding the rectifier bulb itself will not completely isolate the noise.

Oh if you already tried a 12AU7 equivalent, then that will spell things out. 12AU7 has around 2/3 the gain of the 6N1P, so that's why the noise floor is lower. But yes that means the noise is being coupled into the input stage.

If you get to a point where you feel confident the noise is coming from the rectifier loop, the best thing you could do is redo the chassis top plate and layout and move the rectifier away from your signal tubes. Not appealing, I know, but that might be the solution if everything else is ruled out.

 

[ng5921]

You could try shielding the rectifier, but I'm not confident that will work. It's more that the rectifier is positioned where it is, has the rectifier ripple current pulses are traveling through the rectifier as well as the wiring from the rectifier to the capacitor, so it's on both sides of the chassis top plate, just shielding the rectifier bulb itself will not completely isolate the noise.

Oh if you already tried a 12AU7 equivalent, then that will spell things out. 12AU7 has around 2/3 the gain of the 6N1P, so that's why the noise floor is lower. But yes that means the noise is being coupled into the input stage.

If you get to a point where you feel confident the noise is coming from the rectifier loop, the best thing you could do is redo the chassis top plate and layout and move the rectifier away from your signal tubes. Not appealing, I know, but that might be the solution if everything else is ruled out.

Totally okay with redesigning it! I really enjoy the sound of the amp so I'd like to get it to a place where I am totally happy. I jumped the gun on this amp, got too excited and skipped to the building instead of prototyping it and solving some of these issues ahead of time Live and learn.

Really appreciate you taking the time to help Keenan!

 

[Swagsupply]

Totally okay with redesigning it! I really enjoy the sound of the amp so I'd like to get it to a place where I am totally happy. I jumped the gun on this amp, got too excited and skipped to the building instead of prototyping it and solving some of these issues ahead of time Live and learn.

Really appreciate you taking the time to help Keenan!

Let us know what you decide doing with it!

 

[ng5921]

I've been doing a bit of reading about how a choke-input power supply needs to be spec'd and thought it might be helpful to write out my thoughts so others can learn from my mistakes.

A choke-input power supply needs the choke to meet a critical inductance required for the power supply to properly function. A proper choke-input filter wth a large enough inductance will output DC volts out = (secondary RMS voltage)*0.9 (and you must also account for voltage drops across the rectifier and choke DCR) after the choke. Below this this critical inductance, the power supply will instead behave like a cap-input supply with a DC voltage out of (secondary Vrms)*1.4.

For a perfect choke with no loss, the choke will pass 0.9 of the RMS input voltage from the secondary. But in reality, at full load, you will see more like 0.85. The tubes in my amp draw around 20-30mA (10mA per power tube, around 5mA per triode in the 12au7). We will assume 20mA since this conservative estimate will give us a larger choke size (larger choke => lower minimum necessary current). These seem low to me, but the ZMF OTL circuit is biased this way.

We can calculate the critical inductance as follows:

• 120Vrms wall voltage @ 60hz, Full wave tube rectification with 5u4g, Secondary Voltage 600Vrms
• Choke DCR 112ohm
• Vdc (DC output from Choke) = (600*0.85) - (0.03A*112ohm) - (35V 5u4g drop) = 510 - 4 - 35 = 471V
• Critical Inductance (L) = Vdc/(I*1200) for 60hz mains (according to https://audiokarma.org/forums/index...-power-supply-test-build.834386/post-11893984)

L = 471/(0.02A*1200) = 19.6 Henries

This is just one of the equations I have seen floating around on diyAudio and AudioKarma about how to calculate this value., but if I have done this math correctly (and please someone tell me if this is wrong), that means the tiny 2H choke I have been trying to use is no where near large enough, and fails to meet the critical inductance for a proper choke-input power supply! I would need to measure the current in the amp for a more realistic value, but I think this estimate is enough to see that this 2HY choke is not enough. For this reason, I will move the choke to after the first filter capacitor, for a capacitor-input supply.


Now according to this calculator: https://web.archive.org/web/20060106000537/http://www.hagtech.com/theory.html

The value is similar to the one I got (I think I am using the calculator properly). It appears that by adding some series resistance before the choke, we can lower the required value of the capacitor after the choke. According to this, I have 24V of ripple coming out of the LC filter (24/471 = 5% ripple).


I am a little confused as to why the choke helped with the hum despite being below the critical inductance. I don’t have an answer for that.

-----
Now, lets consider the capacitor-input power supply that I am planning to switch to (for the time being).

The input capacitor is 8uF, and the choke is 2H, so we can calculate the resonance frequency:
fo = 1/(2*pi*sqrt(L*C))

fo = 1/(2*3.1415*sqrt(2*8*10^-6)) = 1/(2*3.14*0.004) = 1/0.02512 = 39.8Hz

I do not entirely understand the significance of this value, so I'd appreciate some input here.
-----


Sorry for the long post, but wanted to get all my thoughts out on paper.

Sources:
https://ken-gilbert.com/choke-input-power-supplies-part-1-henry-pasternack
https://el34world.com/Forum/index.php?topic=622.0#google_vignette
https://audiokarma.org/forums/index.php?threads/tube-amp-power-supply-test-build.834386/
https://www.diyaudio.com/community/threads/a-handy-table-of-common-rectifier-specifications.283863/

 

[Swagsupply]

Any luck with resolving the hum? Did you try changing the power supply to be CLC rather than LC as the first filter?

 

[ng5921]

The hum has been fixed! Changing to CLC filter was the solution @Swagsupply. Now there is absolutely no hum.

It ultimately was an issue of insufficient power supply filtering, as @L0rdGwyn and others suspected.

Here's some PSUD simulations showing the improvement (for the input tube) (I accidentally measured voltage from resistor R1, but it still demonstrates the improvement).

With the 2 henry choke-input:

This results in almost 10V of ripple going to the resistor that feeds the voltage regulator tube.

A with the new CLC filter in power supply:

<1v ripple into R1, much better!

This could have probably also been solved by adding a CRC filter instead, but I did not want to significantly change the voltages in the power supply.


The amplifier is essentially complete at this point, last thing to do is install a hi/lo gain switch on the output, like this example from Goldpoint:

Right now I can't listen past 11o'clock on the volume because it gets WAY too loud, and I have a very limited usable range on the volume dial. This switch should fix the issue.

-----------------------------
Now for some information on the next project!

The idea: a high quality, sonically transparent switch box with 4 inputs (3 RCA, 1 XLR), and 5 outputs (4 RCA and 1 XLR). The XLR output will actually be a bypass directly to the XLR input, so that a fully balanced signal is available from a DAC that provides it. There will also be line input transformers inside to conver the balanced to unbalanced signals, allowing you to get an RCA output with a higher line-level voltage from the XLR input.

Parts I am thinking of using include Goldpoint selector switches (https://www.goldpt.com/selector.html), and a stepped attenuator. I may also include a toggle-able EQ/bass boost circuit into the preamp, but that is still up in the air. I will likely get a chassis from Landfall Systems and have them cut all the holes in it for me.

Here's a basic schematic of the idea (one channel shown):

What I have not shown is the grounding for the RCA inputs because I am still debating is how to handle the ground signals from each RCA input. Do I pass them directly from input to output? Or do I tie them all to the chassis ground in the switch box? I could definitely use some input here.

I am thinking of using a 3 position rotary switch (rather than having 4 inputs as shown). The number of poles will really depend on how I decide to handle the grounds. If they are passed straight through the box to the output, I will need more poles.

Another question is whether to add capacitors to the RCA inputs, in case (for some reason) the DAC being used does not have caps on its output.

 

[brunosanii]

Pretty good circuit!