My parallel stereo single ended triode is now ready :)

[Gabe001]

Thank you for all the kind comments about this amplifier. If anyone would like to know more, is tempted to build this and wants advice, or wants more pics (eg underchassis) feel free to drop me a PM. I have some spare PCBs too as the minimum order is 5 of each.

I've got a Philco 444 to work on next which will be a welcome break from amplifier building. I also need to look for/build better speakers. Maybe a future project. WAF big factor here.

Finally, I'd like to thank Jerry (cathoderay57) for providing 2 high quality tested 12au7 valves.

[kalee20]

So far, nobody's commented on the frequency response (Gabe's post #2).

It looks to me, as though 3db points are 8Hz and 11kHz, which is over 3 decades! That's really rather good - a testament to choice of output transformer as well as the rest of the circuitry.

I looked up the link given - £99.21 for TWO! - the claimed frequency response is 10Hz - 65kHz (3db points), which is very good indeed. Now Gabe's 8Hz is actually lower, so it could be the transformer slightlyoutperforms, or it could be there's a bit of LF positive feedback (via HT rail?), so although it looks good, it could be that full power isn't achievable at the extreme LF without distortion.

As for the top end, Gabe's amplifier is rolling off well below 65kHz, and I'm wondering why? Maybe the first stage or the driver stage has an earlier roll-off and dominates the overall response.

I'm interested to know..

[GrimJosef]

Gabe is using an attenuator and I suppose it's not out of the question that its output impedance is on the high side. If it's then feeding a long cable with more than usual capacitance it might be possible that that could account for some apparent HF roll-off I suppose. A quick sum should confirm or refute that though.

Cheers,

GJ

[kalee20]

The scope photo of the square-wave response is consistent with a 11kHz roll-off - the measurement needs to be above reproach.

My gut feeling is the roll-off due to the EL34 input capacitance (mostly Miller!) is the dominant roll-off, but I'd be interested to know if I'm right.

[Gabe001]

I don't have professional measuring equipment so there is going to be some error. I think in part down to the capability of the sound card etc

The Chinese claims on their transformers should be taken with a pinch of salt.

I'm sure there is no waveform distortion at high and low frequencies, as I did a frequency sweep and monitored the output on a scope. However, there is a bit of power loss at high frequency. I'll quantify this directly with a scope tonight and report back.

This is an interesting discussion. The miller capacitance of el34 in triode mode can be an issue, hence the tiny grid stoppers. In the past when I built a single ended amplifier the best frequency response was obtained by using a 6sL7 in an srpp configuration, which I kept, despite the slightly higher distortion c/w 6sn7 cascade. This was also noted by Claus in the pdf referenced earlier.

Ps: I can only hear up to 13.5k

[kalee20]

Originally Posted by Gabe001

I don't have professional measuring equipment so there is going to be some error. I think in part down to the capability of the sound card etc

Doubtless! But it might still be 'good enough.' Can you prove the sound card - feed it with a sinewave from a signal generator, for instance, while monitoring the signal with your 'scope for constancy of amplitude, and get its response characteristics? It (with your attenuator!) might be the dominant roll-off - though I'd honestly be surprised if it rolls off much below 20kHz.

Originally Posted by Gabe001

The Chinese claims on their transformers should be taken with a pinch of salt.

Absolutely, and 65kHz does seem too good to be true. All the same, the (verified!) response extending below 10Hz is astonishingly good, as is the absence of apparent resonances. I'm quite interested therefore if the HF roll-off is due to the limitations of the transformer (primary winding capacitance, and leakage inductance, will both hurt the response), or if it's the drive stage. Certainly, the square wave response does have definite slew between its two levels - a superfast snap between them would be almost invisible.

Originally Posted by Gabe001

I'm sure there is no waveform distortion at high and low frequencies, as I did a frequency sweep and monitored the output on a scope. However, there is a bit of power loss at high frequency.

I'd believe you! With no negative feedback, there'd just be loss of output at high or low frequencies - there's nothing in the middle which would be working harder (and thereby running into distortion) to make up for the shortcomings of other stages to give an overall flat response, as would be the case with overall feedback.

Originally Posted by Gabe001

This is an interesting discussion. The miller capacitance of el34 in triode mode can be an issue, hence the tiny grid stoppers.

I think the tiny grid stoppers will have unnoticeable effect on the response, till you get to several MHz! (Good practice to allow for them in your layout, just in case you have parasitic ultrasonic oscillation - it would be difficult to model).

Rough calculations... Capacitance grid-cathode of EL34 needs to be known, as does capacitance grid-screen-grid, which are the significant capacitances... unfortunately I couldn't find them as the data sheets lists Cin (with unspecified conditions!). So I assumed 10pF for the first, 8pF for the second, and a voltage gain of 10... this gives input capacitance in triode mode (where the screen-grid swings with the anode) of 10pF + 11x8pF = 98pF.

With your 12AU7, assuming ra = 15kΩ under your operating conditions, and you have 220kΩ load resistance and 220kΩ grid-leak, the effective stage output resistance will be 13.2kΩ which in conjunction with 98pF gives a roll-off of 123kHz, way above what's being seen, and at odds with my gut feeling in post #44. So unless my assumptions are wildly out, something else is curtailing the upper response - and I certainly can't see it being second-stage loading the first stage either.

Originally Posted by Gabe001

Ps: I can only hear up to 13.5k

So it may be academic from your point of view, but something is causing the roll-off, and it would be good to know what it is! After all, you have an amplifier which, although heavy and power-hungry, is completely free from crossover nasties and many other artefacts which plague sophisticated circuits.

I'm interested in the LTSpice file now, I'll PM my email address later, if you can send it.

[Gabe001]

Of course pm me your email address. Gmail is blocking the .lib file for the transformer and having difficulty with Greg it happy to send them

Does the fact that there are 2 el34s in parallel change the calculation?

[kalee20]

Done! (If files don't send, just change the extension to something else, I can change it back on receipt).

Originally Posted by Gabe001

Does the fact that there are 2 el34s in parallel change the calculation?

Well, ahem, yes it would, I completely overlooked that!

But it would only halve the frequency, and that still puts the roll-off at 61kHz and we're seeing 11kHz... still quite a discrepancy!

[Richardgr]

With parallel tubes, is the capacitance and resistance of the tubes in parallel, so Miller capcitance would be doubled, and resistance halved? If so, it would seem to be quite a limitation of that topology?

[kalee20]

Input resistance is virtually infinite, so halving it is still pretty infinite (Gabe has 220kΩ grid-leak for the pair, and that's completely dominates it).

But capacitance is doubled as you say. It's a limitation, doubling the power does come at a price! Though it doesn't appear to be the operative limit. If Gabe had put 10 EL34's in parallel, it would be a different matter - though there's no reason why there couldn't be a few ECC82 / 12AU7's in parallel to drive them, for a bit more HF grunt...

[G.Castle]

Originally Posted by Gabe001

Of course pm me your email address. Gmail is blocking the .lib file for the transformer and having difficulty with Greg it happy to send them

Sorry Gabe, nothing has arrived. I asked my nieces to forward a spice .asc and associated files to that email: works both raw and with .zip containing folder. No problem with file extension changed (to .txt), as I suggested either so

The file sizes would be small so could they be put in a .zip folder on the forum?

[Gabe001]

Ok, so here are some proper real life measurements. Digital files do odd things - I think there may be some brick wall filters coming into play, or sampling/compression issues, and I'm struggling with 16khz sine wave upwards from digital files. So definately there is a problem doing this measurement reliably on computer.

A proper old style signal generator gave best results. Readings via a scope measured on reticule

For ease of measurement I chose a 16v p-p 1khz reference measurement into an 8ohm resistive load. This means +/-8v or 5.657Vrms which equates to a 4w output. No waveform distortion was visible at any frequency.

The readings were as follows
20hz 16v p-p =4w
100Hz 16v p-p =4w
500Hz 16v p-p =4w
1kHz 16v p-p =4w
5khz 16v p-p =4w
10kHz 14.4v p-p =3.24w
13khz 13v p-p =2.64w
16kHz 12v p-p =2.25w
20kHz 11.2v p-p =1.96w

I think this means that the FR (20hz to 20kHz) is -3.1db. If this is correct I'm not at all displeased considering the amount spent. Given the parallel triode topology and the implications on miller capacitance I think the OPT may perform close to spec in a standard single ended pentode or ultralinear design.

[Gabe001]

files attached for ltspice. Attached transformer lib and symbol but you need to use your own valve models

the two asc files are identical but the ones for frequency response has a dc supply

[G.Castle]

Thanks for posting the files Gabe. I'll load them later and have a gander.

Your tests do suggest that you've nailed a pretty decent performance from the set up.

[Gabe001]

The most disturbing thing out of all this is that when I tried replicating this FR with digital test tracks, it seems that a significant number of digital tracks meant for streaming have brick wall filters at 15k. Oddly, this also includes the Qobuz frequency test tracks - also noted by other users online. What a shame. Maybe they are catering for users with celestion ditton 15 speakers 😉

+1 for analogue sources

[Diabolical Artificer]

I use PC software with an outside source having had issue's with REW's signal source too. Also all the free PC based test software I've come across is limited to a 20hz - 20khz range with difficulties experienced at the extremes of the range, more a resolution issue. For freq response I use a HP8903B,scope and paper graph, see attached.

Lastly with the SE amps I've built with average OPT's HF response can be flat up to 30khz +. the last one I re-built for a friend had a 3dB @ 25khz, that was open loop, no NFB.

I'll send you the graph G.

Andy.

[Gabe001]

Thanks Andy, very useful graph. I'll use it.

Measured FR with a signal generator matches the LT spice predicted FR exactly. Most of the HF loss seems to be happening around the v1a gain block. It looks like it can be made flat (in theory) past 20khz with a 330pf cap (C14) to ground as shown in the pic, but I haven't experimented too much to see if there is any downside. To be honest I'm not too fussed about -3db at 20khz. One day if I decide to go back in for a repair etc I may try it. If anybody else can suggest any better solutions I'd be happy to simulate them.

[Diabolical Artificer]

Would moving the 27k "lift" resistor to the second stage make a difference? I agree about not worrying about a HF response to infinity as we can't hear it. That said I saw a video about one of Sansui's super dupa amplifiers that used transistors with a very fast (forgot the correct term) high frequency parameter where the listening outcome was superb. I'm afraid I didn't understand most of it but the gist was even though we can't hear much over 20khz, an amplifier capable of a flat response up to 300khz does matter.

Maybe David RW can enlighten us.

Andy.

[G.Castle]

I suppose that whatever is said about extended frequency response, it may just boil down to good old slew rate.

The problem trying to design this into an amplifier, particularly a valve amplifier, is that it turns into a game of spurious HF 'wackamole' or just a downright classic oscillator.

[kalee20]

Originally Posted by Gabe001

Ok, so here are some proper real life measurements....
A proper old style signal generator gave best results. Readings via a scope measured on reticule

...

20hz 16v p-p =4w
100Hz 16v p-p =4w
500Hz 16v p-p =4w
1kHz 16v p-p =4w
5khz 16v p-p =4w
10kHz 14.4v p-p =3.24w
13khz 13v p-p =2.64w
16kHz 12v p-p =2.25w
20kHz 11.2v p-p =1.96w

I think this means that the FR (20hz to 20kHz) is -3.1db

I'd agree - and I think you are entitled to be pleased!

All the same, at the upper end I can't help feeling it should be better. I was really hoping to run the LTSpice sim and have a play, probing intermediate points - but without the valve models I can't do this, and I'm not sufficiently LTSpice fluent to find or create my own. (Plus, time's at a premium right now!)

Originally Posted by Gabe001

It looks like it can be made flat (in theory) past 20khz with a 330pf cap (C14) to ground as shown in the pic,

Quick calculations indicate that the 330pF, 27kΩ network gives an HF lift with turnover point starting at 17kHz - it certainly won't give a droop.

Originally Posted by G.Castle

I suppose that whatever is said about extended frequency response, it may just boil down to good old slew rate.

Slew rate and frequency response are related, but looking at the square-wave performance Gabe put up in Post #2, the amplifier isn't slew-rate limited.

SR limiting happens where there just isn't any extra current available (a valve goes into cutoff, for example) to charge or discharge a capacitance any faster. It's almost always asymmetric, in an asymmetric circuit like this (oh, for that PNP valve!).

The rising and falling edges of the square wave look the same, so while the rise and fall times are definitely not instantaneous, nothing looks as though it is running out of puff.