[drtechno]

the voltage regulator looks like a half wave, but the fets need a power supply with a lower output impedance that the uf4007 can provide. if you changed them to a lower drop diode (like a 6A10) it would work. But i would redo it so it was full wave voltage doubler...

 

[SonicTrance]

the voltage regulator looks like a half wave, but the fets need a power supply with a lower output impedance that the uf4007 can provide. if you changed them to a lower drop diode (like a 6A10) it would work. But i would redo it so it was full wave voltage doubler...

Thanks for your reply!
It's a full wave doubler like the example on the right below. With the diodes facing the other way of course since it's a negative supply.

The 6A10 is a beefy 6A diode. Way too big for my needs. This supply will draw about 15mA total. Anyway, I got a clean enough supply now with the 115 VAC PT. Will stick with that.

 

[drtechno]

Thanks for your reply!
It's a full wave doubler like the example on the right below. With the diodes facing the other way of course since it's a negative supply.


The 6A10 is a beefy 6A diode. Way too big for my needs. This supply will draw about 15mA total. Anyway, I got a clean enough supply now with the 115 VAC PT. Will stick with that.

its not the amperage size, its the voltage drop (about 1/3 of a uf4007). A much better diode to use there since you need a low impedance out. also why have that 6.8nf cap in it? because I can see that shunting some of the current potential away n the power supply. The problem is your 10A regulator is starved for current. Alternatively, if you use something that would be more realistic, like a zfp4424A for a sub 1A regulator, you wouldn't need that high current demand (which requires a lower impedance output) on startup.. also why the high value caps since if the regulator is set up correctly, you can have 10% or more ripple coming out of the doubler, and sill be mv of ripple on its output. Parts of your circuit makes sense, others seem you are putting an elephant in a Volkswagen Beetle.

 

[MrCurwen]

drtechno, absolutely no malice implied here; either you don't know what you're talking about or I don't understand what you mean in your message.

its not the amperage size, its the voltage drop (about 1/3 of a uf4007).

This is not at all an issue. Look at the percentages of voltages; a lower voltage drop diode wouldn't change anything. UF4007 is more than able to pump the caps in that doubler or any high(ish) impedance PSU / circuit.

Pumping the caps isn't the issue. It's oscillationl. It cannot be pumping the caps, since the measured ripple BEFORE the reg was higher than after. That means 100% certainty it's some kind of oscillation, most probably interaction between PT secondary (which has all kinds of nasty things) and the reg. If there were more R in series between these two, the problem would most probably go away. Just a theory based on experience, the cause of oscillation could be many things, but it is 100% certainly oscillation.

also why have that 6.8nf cap in it?

To increase stability. Shunt some MHz crap out before it gets into the FETs.

because I can see that shunting some of the current potential away n the power supply

No.

Look at the relations again. 6.8nF is a drop in the bucket when you are pumping current into the electrolytics. Absolutely no effect between these two.

The problem is your 10A regulator is starved for current.

Not at all. If you think so, please elaborate.

These kinds of regs are very easy to set up and very very stable. Useful tools. Absolutely no reason to run them at insane currents. Understand we are talking about a high impedance circuit driving 10k output transformers, not 4 ohm direct output transistor amp. The whole circuit draws some tens of mA.

Alternatively, if you use something that would be more realistic, like a zfp4424A for a sub 1A regulator, you wouldn't need that high current demand (which requires a lower impedance output) on startup

See this is where I definitely start to think you don't know what you're talking about. Again, no offence meant, just statement of fact.

A FET that is able to pass 10A of current does not NEED to pass 10A of current. It is 100% fine and useful (with lower gm) passing say 1mA.

The reg doesn't try to draw 10A at startup. If you think this, please elaborate.

Parts of your circuit makes sense, others seem you are putting an elephant in a Volkswagen Beetle

I'm not sure if you mean the circuit as a whole, or this B- supply in specific, but I can kind of agree with this! It's a bit crazy circuit until you get into the details and read my stories (earlier in this thread) on how and why I put it together like this. There is a specific reason for most things in the circuit (some are just arbitrary, like most stuff before the regs in the PSUs) and all the parts fit together beautifully (in my opinion) in the end.

Of course the specific applications in the SE project are all Sonic, I just gave some ideas to him.

 

[drtechno]

some stuff looks oversize for what it does. Sometimes going too big in certain rating can cause the intrinsic flaws with a device to show up... Just saying.

oh, did you try a 1M resistor between your grid of the tube and the grid stopper and ground? Thats how I maintain bias in my 9A small signal tubes so the grid bias doesn't flop all over the place as I adjust volume in the unbalanced circuit.

btw there shouldn't be any ground references in your signal path. This includes your attenuator, and btw, I don't really see your input circuit anywhere.

This woman needs chocolate now......

 

[SonicTrance]

drtechno is obviously a troll. I don't feed trolls.

Build update:

I've completed one channel input stage and biased the C3G tube.
Grid bias on the left and voltage drop across R3 on the right. Operating point is 150Va, 10mA Ia and -2.48Vg (the -2.248 is to ground, not cathode)

Put a 400mVpp signal in and this came out. 19.2Vpp means a µ of 48. A bit higher than the 40 that's stated in datasheet.

Nice looking sine wave though. So far so good!

 

[MrCurwen]

Looking good!

You've got some serious measument tools there. Usually measuring grid bias at the grid is pretty useless, at least with a cheap multimeter. It doesn't give very accurate results especially with tubes where 2.2 and 2.4 V grid bias mean very different Ia.

But, this is not important. Important is that Ia us where you want it to be, and that can be very accurately measured as you've done.

10mA on a single voltage amplification stage. Whew, you sure are throwing efficiency out the window. Not that there is any efficiency to talk about in any of these designs anyway.

Small correction; this SE amp is not constant current draw. The balanced amp is. When the output stage swings current, there is nothing to counteract that; in the balanced circuit the tubes (or the FETs above them) conduct in push-pull.

There are some schemes to make SE stages more or less constant current (most notably Broski), but I personally don't find them worth the trouble. Combination of simple but relatively effective B+ regulation and the gyrator offering high impedance load make the PSU relatively transparent compared to trad SE amps. Truth is, the SE amp is not as good as the balanced one, but it's SURPRISINGLY good if you've built a bunch of trad SE amps.

Also the CCS at the output also compensates this. That first 4mA portion of the wave IS constant current, the CCS acts as "counterweight" in that part of the wave. So microdetails get the "no PSU" treatment.

 

[SonicTrance]

10mA on a single voltage amplification stage. Whew, you sure are throwing efficiency out the window. Not that there is any efficiency to talk about in any of these designs anyway.

Yes, looking at the triode curves for the C3G, 10 mA @ 150 Va looked like a good place to be.

Truth is, the SE amp is not as good as the balanced one, but it's SURPRISINGLY good if you've built a bunch of trad SE amps.

Really looking forward to hear it for myself! Should be ready for first listen in the next few days.

 

[SonicTrance]

More progress! Completed one channel output stage (almost, haven't wired the outputs yet)

This is at 6BG6 gyrator output. Operating point is 180Va, 20mA Ia (+4.3mA CCS) and -16.86Vg. At this operating point current and voltage should theoretically clip at about the same point.
50mVpp signal input. This is max volume:

 

[SonicTrance]

The SE amp is now complete!

Here're some pics:

Overview of the inside. Yes! I have a thing for large Russian caps! Those large green ones in the middle are K75-10 10µ parafeed caps. Also used K75-10 caps for input and coupling.

To the right of the choke is the vu meter backlight ps. That's what I'm using the 274B rectifier for! I realize of course that it's completely unnecessary to take 115 VAC, rectify it with a tube/diode bridge and load it with two backlights @ 50mA a piece. The result is 8 VDC out and about 90mA current draw. That's maxing out the Sophia Princess "mesh" 274B! The lamps do get a very nice slow start anyway! Only doing this because it's fun and I wanted a use for the tube!

To the right of the speaker outputs is the ps for the fans. I have two 120mm fans on the bottom panel. See the ground lead going from the resistor lead there to chassis ground. I had to remove that as it caused hum to the amp. The fan ps is not very clean. It needs much larger caps than the 220µ that's there to make it clean, maybe 1mF. Good enough for fans though. I run them at 7V, very quiet. And with a floating ground no noise problem. The vu meter ps is also floating for the same reason. I've done one "circuit ground" and one "dirty ground". Was planning to have the first lytics in B+ and B- supply, HV centertap, meter ps and fan ps on the "dirty" point but meter ps and fan ps are now floating.

I have a pair of C3G's that I've stripped from their shield but can't use them. Way too microphonic. The shield actually grounds at the socket to circuit ground. Maybe the "naked" pair is naturally microphonic. Haven't used them in years.

So, how does it sound? I've listened for a few hours only. To both my headphones (Audeze LCD-3 and LCD-XC) and speakers in my lab/listening room (Monitor audio bronze 2) and I'm very impressed! First thing a noticed was the lack of noise, there's none! Dead quiet! Even with my LCD-XC's which are my most sensitive cans and also my favorite. I would say the balanced LTP is a bit faster sounding and have bit more micro detail than this SE version. But the SE also sounds very crisp, dynamic and packs a punch. Bass is very pleasing indeed! I think the LTP pairs better with my speakers though than the SE.

The major advantage the SE amp has is obviously size. Only one chassis. Also much easier to build, less components. I could definitely live with only the SE version.

I've not done any power output test yet. Right now I'm just enjoying the music!

Feel free to ask questions!

 

[MrCurwen]

The SE amp is now complete!

Congratulations! The pictures show once again first class build quality inside and out. You've really got this P2P thing under control, I'm impressed.

Overview of the inside. Yes! I have a thing for large Russian caps! Those large green ones in the middle are K75-10 10µ parafeed caps. Also used K75-10 caps for input and coupling.

Have you made any noteworthy sound quality observations? These green K75-10's are later production, late 70's and 80's (some still made in the 90's), metallized plastic film with wax if I recall correctly. No oil. I've got a couple of these I've used (10µF / 400V) but as B+ caps they don't have much sonic impact anyway so I don't know how good they are.

I'm beginning to think correctly used all caps sound pretty much the same.

To the right of the choke is the vu meter backlight ps. That's what I'm using the 274B rectifier for! I realize of course that it's completely unnecessary to take 115 VAC, rectify it with a tube/diode bridge and load it with two backlights @ 50mA a piece. The result is 8 VDC out and about 90mA current draw. That's maxing out the Sophia Princess "mesh" 274B! The lamps do get a very nice slow start anyway! Only doing this because it's fun and I wanted a use for the tube!

List of your audiophool sins is ever growing; you've used expensive high end boutique rectifier for non-audio assistance job. Whatever will be next... (I recommend small CRTs.)

It needs much larger caps than the 220µ that's there to make it clean, maybe 1mF.

Or, keep 220µF and add a "filament regulator" unit. It loses a couple of volts but should take care of any ripple.

I have a pair of C3G's that I've stripped from their shield but can't use them. Way too microphonic. The shield actually grounds at the socket to circuit ground. Maybe the "naked" pair is naturally microphonic. Haven't used them in years.

Most tubes meant for radio actually really need their metal casings.

First thing a noticed was the lack of noise, there's none! Dead quiet!

I've noticed the same. I think the reason is simply that there's less gm inside the chassis; or, in other words, less FET drains connected to the same B+ regulator output.

One possible solution would be to go true dual mono. So make full one channel amps (they can be in the same chassis) with separate PSUs and PTs and everything. Maybe separate PSU raw filter section branches and separate B+ regulators for different gain stages.

Too much work.

I think the LTP pairs better with my speakers though than the SE.

Very possible. Most modern speakers are made for lowest possible output impedance. Your balanced amp has near-transistor amp output impedance because of the output buffer.

I've not done any power output test yet.

You'll be surprised how low the RMS out is. I don't remember if we've talked about why it can be so low and yet sound so loud and clean.

 

[MrCurwen]

I would say the balanced LTP is a bit faster sounding and have bit more micro detail than this SE version. But the SE also sounds very crisp, dynamic and packs a punch.

So now you have a feel for what the balanced version sounds like, and what the SE version sounds like. The breadboard workhorse amp I'm using has a SE input stage and balanced grid drivers and output stage. No buffer. It's "good enough" to keep using every day. Sound quality is somewhere between SE amp and balanced amp, maybe leaning a bit more towards the balanced amp.

The SE amp is by far the easiest to build for a beginner, but for a bit more ambitious builder the SE / balanced 'hybrid' could be a good solution as well. It's easier to build and set up than the full balanced version, and if chassis allows, can be extended to full balanced later.

 

[SonicTrance]

Congratulations! The pictures show once again first class build quality inside and out. You've really got this P2P thing under control, I'm impressed.

Thanks! Means a lot! Glad you like it!

Have you made any noteworthy sound quality observations? These green K75-10's are later production, late 70's and 80's (some still made in the 90's), metallized plastic film with wax if I recall correctly. No oil. I've got a couple of these I've used (10µF / 400V) but as B+ caps they don't have much sonic impact anyway so I don't know how good they are.

Not made any comparisons this time. Just decided to use the K75-10's and went with them. Mine (both the huge 10µ 250V and 680n 250V) are from the 90's. Don't know if you're thinking of some other cap? These are hybrid with both paper and lavsan in oil. There's definitely oil in them as they're very heavy and solid. Think they come in 250, 500 and 750V, not seen any other voltage ratings.

Or, keep 220µF and add a "filament regulator" unit. It loses a couple of volts but should take care of any ripple.

Thought about adding some regulation but meh, no room and works like a charm now anyway. Might swap those caps at some point though.

Most tubes meant for radio actually really need their metal casings.

Might sadly be the case! You know I love my glowing tubes! C3G's are quite pretty ones naked. I might take a chance and buy another pair and strip them. If you're very careful it can be reversed!

I've noticed the same. I think the reason is simply that there's less gm inside the chassis; or, in other words, less FET drains connected to the same B+ regulator output.

One possible solution would be to go true dual mono. So make full one channel amps (they can be in the same chassis) with separate PSUs and PTs and everything. Maybe separate PSU raw filter section branches and separate B+ regulators for different gain stages.

Too much work.

Mono blocks would be a fun project! If using IDH tubes there's less work. My 6C8G / 47 amp is quite complicated but then performs like a champ.

You'll be surprised how low the RMS out is.

Will do a test for sure. I remember the output buffer breaking in the LTP when I did the power output test. This amp should not be so sensitive to clipping I think?

I don't remember if we've talked about why it can be so low and yet sound so loud and clean.

I think so, a long time ago, refresh my memory.

The SE amp is by far the easiest to build for a beginner, but for a bit more ambitious builder the SE / balanced 'hybrid' could be a good solution as well. It's easier to build and set up than the full balanced version, and if chassis allows, can be extended to full balanced later.

I agree! If you want "super high end" full balanced is the way to go. But after spending some time with this SE version I don't think it's "worse" than the balanced but rather "different flavor" if that makes sense. I really like it!

 

[MrCurwen]

Not made any comparisons this time. Just decided to use the K75-10's and went with them. Mine (both the huge 10µ 250V and 680n 250V) are from the 90's. Don't know if you're thinking of some other cap? These are hybrid with both paper and lavsan in oil. There's definitely oil in them as they're very heavy and solid. Think they come in 250, 500 and 750V, not seen any other voltage ratings.

Yes I recall completely wrong. They have toxic oil inside. I should try and find the pdf of that russian cap catalogue I downloaded off some forum years ago and just post it here.

Might sadly be the case! You know I love my glowing tubes! C3G's are quite pretty ones naked. I might take a chance and buy another pair and strip them. If you're very careful it can be reversed!

One possibility is that if you use anti-oscillation methods, the extreme microphony might get under control. Stoppers on every pin, reduced gm etc. When used in pure voltage gain mode (input tube), gm doesn't matter at all so just toss it.

Will do a test for sure. I remember the output buffer breaking in the LTP when I did the power output test. This amp should not be so sensitive to clipping I think?

Probably not, but the CCS in parallel might break. Just increase volume veeery slowly and you should be ok. Use a higher f, maybe 2k or 4k; then the flat top of clipped wave is less like DC (because it's shorter in duration).

I think so, a long time ago, refresh my memory.

Human hearing cannot really make out the exact volume relations of peaks, unless they're grossly different in volume. If they're in the same general category of volume, you cannot tell which is louder.

Your can only produce the output power it can; after that it will clip. The output power is quite small, the amp clips a lot during transients. What matters is how the amp RECOVERS from that transient.

Now if you have a RC bias (cathode bias) somewhere, the bias circuit needs to recover from that transient. This takes time, during which the amp doesn't behave normally.

Worse yet if you have large NFB loops, spanning several gain stages; this takes a long time to stabilize again, during which time the amp behaves absolutely horribly. This is why when you use gNFB you want to have enough power capability for the loudest transients, even if your baseline power need is 0.1W RMS. That 100W transistor amp really needs that 100W in reserve, even if your baseline need is 0.1W (which it about is if you think that SE amp is loud enough).

There's lots more things inside typical tube amps that require some recovering time after a clipping transient.

Anyway, what happens if these things are taken care of? Well, the amp just clips the top off the transient, and then carries on just normally.

Your ears hear the music just continuing; they don't know how loud the transient was supposed to be, and even if the transient was there in full, they wouldn't be able to say in blind A/B comparison which transient is louder, the clipped one or the full one. They are both "loud", since they rise very fast up from the baseline. Ears detect the delta of the transient, (how "fast" the music sounds) not the amplitude (how high the wave peak is).

So the key to making the amp "fast" is investing in transient response and recovery. This means driving all kinds of capacitances from low impedance sources (grid, OT), not using bias methods that can be destabilized (caps need recharging time), having a flawless PSU.

If you want "fast" sound dealing with parasitic C is very important; if the signal has to "wait" while a poor current source is trying to charge a parasitic C, the crisp edge of the transient is lost. The voltage needs to be able to rise instantly. Ears detect delta, not amplitude of transient.

The last one is the biggest difference between the balanced amp and the SE amp (when talking about "speed"); since the balanced amp is constant current draw, it has a flawless PSU. The SE amp draws varying current, so it doesn't have.

I agree! If you want "super high end" full balanced is the way to go. But after spending some time with this SE version I don't think it's "worse" than the balanced but rather "different flavor" if that makes sense. I really like it!

How would you place

1) IDH balanced amp
2) DH balanced amp
3) SE amp
4) modded LD

in relation to each other regarding sound quality. In different aspects of SQ ("speed", "realism", "microdetails" etc) the order of preference might be different of course.

 

[SonicTrance]

I should try and find the pdf of that russian cap catalogue I downloaded off some forum years ago and just post it here.

Yes, you should! Would be nice to have!

One possibility is that if you use anti-oscillation methods, the extreme microphony might get under control. Stoppers on every pin, reduced gm etc. When used in pure voltage gain mode (input tube), gm doesn't matter at all so just toss it.

Like you see in the schem I have stoppers at grid, cathode (both pins) and anode. Quite conservative values though.

Probably not, but the CCS in parallel might break. Just increase volume veeery slowly and you should be ok. Use a higher f, maybe 2k or 4k; then the flat top of clipped wave is less like DC (because it's shorter in duration).

Will try later today or tomorrow!

Human hearing cannot really...

Thanks! Excellent explanation!

How would you place

1) IDH balanced amp
2) DH balanced amp
3) SE amp
4) modded LD

I'm comparing the DH fully balanced LTP now against the IDH SE amp!

I had to rearrange things. I'm running out of space here! Anyway...

When switching between the two you can clearly hear the LTP has more "air", instrument separation, soundstage. The LTP is the superior amp, BUT the SE is not far behind! Can't remember how the IDH LTP sounded but remember I thought the DH version was a step above. All versions of this amp beats the modded LD MK6, which in it self is a great amp but this design is a step above. Also a lot cheaper as the LD required boutique caps and expensive tubes to sound it's best, the "modern" design doesn't!

Think I'll keep the SE as my daily driver and use the LTP when I just want to lean back and listen to music and not do other things at the same time. The LTP deserves full attention! If you're not concentrating on the music you're missing the advantages of the LTP.