building your own hybrid amp

[ionomolo2]

A couple of months ago I was given more than one hundred VHF amplifier (triodes)/mixers(pentodes) at the local part store. The guy said they were to dispose them, so this was my chance.

For these two months I've been working in my own not-lethal-voltage hybrid amplifier, with little success (before that I had worked with 'gainclones' and chip amps with discrete power stage and had much more luck). And was quite surprised that there wasn't a general-purpose tube amp (general-purpose meaning discussing techiques and active parts rather than construction/modding of a yet-designed circuit). So I decided to post here the problems I came across while doing my own amp and to encourage others to share their design-prototyping experience.


So here I come with a few questions:

How critical the ccs in low voltage tubes? I have been using a jfet ccs (BF244 + 2K trimpot) and, being this the only difference between the tube stage of my amplifier and millet hybrid/millet max/soha I assume that this my not work well. How good are diferent ccs topologies? How high must be the transconductance of a jfet amplifier to get nice results?

Can a catode follower drive high impedancy headphones with admisible distortion? May this be improoved using some solid-state trick to keep constant the voltage across the tube?

Can tubes be paralleled to increase drive current?

Does AC-coupling between tube and power stage have a negative effect on sound? Has anyone done a simple (for example jfet source-follower) buffer between tube and power buffer to allow both load-relief to the tube stage and coupling trough 1u cap (auricap or so) to allow better sound quality than with (most of) electrolytics when placed in series with the headphone?

I've named the forum "building your own hybrid amp" since I expect that your wisdom will help anyone experimenting with his own, and not only me.

 

[ionomolo2]

I've missed a question

There is a link in the pete millet web site telling that some pentodes can operate as space-charge triodes with similar voltage ratings that those in millet's amp. Since television pentodes are really easy to adquire (most part stores have some unsold stock and no hope to sell it anytime soon), has someone done something arround it? It seems curious that there is no "established" amplifier using that techique, since 12AE6/12FK6/etc are much more hard to find that those.

 

[tomb]

I'll try a couple of these, but you'll probably need Dsavitsk to answer most of these. I believe Morgan Jones' tube book is the best source for most of your questions:
http://www.amazon.com/Valve-Amplifie...9985991&sr=1-1
Quote:

Originally Posted by ionomolo2 /img/forum/go_quote.gif A couple of months ago I was given more than one hundred VHF amplifier (triodes)/mixers(pentodes) at the local part store. The guy said they were to dispose them, so this was my chance. For these two months I've been working in my own not-lethal-voltage hybrid amplifier, with little success (before that I had worked with 'gainclones' and chip amps with discrete power stage and had much more luck). And was quite surprised that there wasn't a general-purpose tube amp (general-purpose meaning discussing techiques and active parts rather than construction/modding of a yet-designed circuit). So I decided to post here the problems I came across while doing my own amp and to encourage others to share their design-prototyping experience. So here I come with a few questions: How critical the ccs in low voltage tubes? I have been using a jfet ccs (BF244 + 2K trimpot) and, being this the only difference between the tube stage of my amplifier and millet hybrid/millet max/soha I assume that this my not work well. How good are diferent ccs topologies? How high must be the transconductance of a jfet amplifier to get nice results?

There are more differences in the amps you site than the CCS's. The SOHA, in particular, has about 5 combinations of CCS designed for it. It also de-couples the output stage from the tube as answered in one of your questions below. Besides being a "tube hybrid", the SOHA resembles little of the Millett and Millett MAX.

The higher the impedance presented by the CCS, the better controlled will be the tube. There is more inherent PSRR, which helps prevent noise and resists external effects. The tolerance desired is close enough from what I understand that even a high-grade trimmer may not be a good choice. You see trimmers used to adjust bias, but not the output current of the CCS.

Quote:

Can a catode follower drive high impedancy headphones with admisible distortion? May this be improoved using some solid-state trick to keep constant the voltage across the tube?

I think the solid-state trick to which you refer is a solid state output buffer.

Quote:

Can tubes be paralleled to increase drive current?

Yes.

Quote:

Does AC-coupling between tube and power stage have a negative effect on sound? Has anyone done a simple (for example jfet source-follower) buffer between tube and power buffer to allow both load-relief to the tube stage and coupling trough 1u cap (auricap or so) to allow better sound quality than with (most of) electrolytics when placed in series with the headphone?

What you describe is the SOHA design. It has limits because there is nothing to block the inherent offset produced by the output stage itself. An opamp is used in a unity gain, current buffer mode. This prevents problems with offset, which can be severe when other types of stages/added buffers are applied. Steinchen seems to have had success applying Amb's Jisbos buffers, but this has not been done yet on a wide scale.

The Millett does just the opposite. There is little, if any, load on the tube from the output stage (because of the CCS), but offset is prevented with the large output coupling caps.

Quote:

I've named the forum "building your own hybrid amp" since I expect that your wisdom will help anyone experimenting with his own, and not only me.

Quote:

Originally Posted by ionomolo2 /img/forum/go_quote.gif I've missed a question There is a link in the pete millet web site telling that some pentodes can operate as space-charge triodes with similar voltage ratings that those in millet's amp. Since television pentodes are really easy to adquire (most part stores have some unsold stock and no hope to sell it anytime soon), has someone done something arround it? It seems curious that there is no "established" amplifier using that techique, since 12AE6/12FK6/etc are much more hard to find that those.

Don't know about this one, either, although it sounds like an interesting idea. However, I don't think the Millett tubes are that hard to find from established tube vendors. They are hard to find in the local antique store or flea market, where TV tubes predominate because no one wants them.

 

[ionomolo2]

Thank you for this great reply, I will pick that book as soon as I can.

What I said is to use a single field effect transistor buffer between tube and buffer in the soha, I havn't seen any schematic doing so.

This schematic with 1k & 10k bias resistors (choosen randomly between reasonable values and with no optimization techiques) & BF244A shows 0.02% THD (mainly 2n harmonic) when simulated and should not place *any* load to the tube since jfet has very high input impedance and input capacitance is bootstraped in the source-follower layout. Should be nice but it sounds bad, although I think there must be a way to improove it.

The solid-state trick to reduce distortion in a cathode-follower stage I took from a tube-buffered gainclone page:

http://www.customanalogue.com/tubes_...gainclones.htm

I don't know if i can cite anything copyrighted here so i'll will go the safe way and ask to read the paragraph starting with "Something else may have caught your eye as well" in the "So What About Tubes & Gainclones Together?" section. I wonder if the tube does anything in this schematic, or if it's a way to shift the price of the amplifier.

 

[tomb]

SE Borbely Tube-MOSFET hybrid headphone amplifier

 

[ionomolo2]

Wow that's impressing. I have seen that schematic before but hadn't realized that it was operating in that voltage range.

I will start experimenting with tubed differential inputs as soon as I get good results with single-ended ones.

At the moment I've got better results with 20K resistors than with jfet ccs. Maybe variations in bias current translate into variations in bias voltage thus being feedback into the amplifier at high gain in a positive-feedback fashion (I found the tube to oscillate sometimes but always blamed the LM4562). This would lead to a 3.01% current-induced voltage swing that would be much higher than the 0.3 suggested from pspice simulations.

I will build a current source similar to those used in millet max (which show a 0.03% current swing @ 1Vrms) to see if that explains bad sound. If I success I would start working with the jfet buffer, since I suppose its small "Class-A" 2nd harmonic distortion won't hurt too much the overall sound, assuming better bias and transistor choice.

Another question, can a tube operating at low voltage able to provide enough current to drive a power mosfet at a decent slew rate? May that topology be extremely prone to oscillations?

 

[ericj]

Quote:

Originally Posted by tomb What you describe is the SOHA design. It has limits because there is nothing to block the inherent offset produced by the output stage itself. An opamp is used in a unity gain, current buffer mode. This prevents problems with offset, which can be severe when other types of stages/added buffers are applied. Steinchen seems to have had success applying Amb's Jisbos buffers, but this has not been done yet on a wide scale.

It's also done in all-glass designs, like the morgan jones / earmax amps.

 

[ionomolo2]

While working on my amplifier I've found that the LM4562 buffer tends to oscillate and needs carefull layout techiques to archieve good sound. Anyone knows of a solid state buffer with *decent* sound and very easy to use?

decent = enough to judge the tube stage, not to be placed on a definitive build

 

[ericj]

Really? The LM4562 in my SOHA seems to behave. It just doesn't like low-Z headphones. Maybe i got it right by accident.

And fwiw it's an ultra-low-noise bipolar opamp with fairly low input offset current (for a bipolar opamp).

The buf634 is a bit on the harsh side but has been used in a lot of headphone designs - try that. Also opa551. Note that these are single-channel devices.

If you want to keep the layout you've got, drop in some other dual channel opamp, but try a well-behaved one - opa2134, opa2227, opa2107, ad823, ad8620 (on browndog adapter), etc.

 

[ionomolo2]

LM4562 with its 20V/us slew rate is great when proper source decoupling is applied, but that's hard to do in a protoboard, since parts must have longer legs and placing too much decoupling caps difficults experimentation. I was looking for a "just enough good" stable thing to listen for different tubes and bias points.

These days i've been trying different bias points but not enough to draw final conclusions. What about a diode instead of a resistor at the cathode?

Another nice thing to experiment would be to force a constant voltage across the tube and use current to drive the next stage. Has someone listened to anything like that? How good does it sound?

 

[ionomolo2]

I've forgot to say the tubes i'm working arround are PCF80's from mullard.

It seems that these tubes can supply as much as 20 mA at 70V. Linearity drawn from graphs (assuming constant current load) seems good and gets better with higher currents.

The graph discourages operation at constant voltage, but used with an output transformer and seeing such a high current capability makes an all-glass non-lethal tube headphone amplifier seem feasible.

Could it be possible to make a single-tube per channel all-glass amplifier?

 

[ionomolo2]

I've built one millet with the ccs from millet max and LM4562 driving a bjt class A push-pull and it sounds really good. I expect that to be the problem with the other designs. I'm not already satisfied with the results, but this is probably due to low quality cathode bypass and output cap. I'll give a try to television tubes the next weekend. There are nine types at the local store and i've found that they produce dramatically different sounds so that may be very interesting.

Has somebody listened to various television tubes and bias currents? How do they sound?

 

[00940]

Here's a preamp with the PCF80: http://www.diyaudio.com/forums/showt...threadid=62681 . Add a servoed discrete current stage as suggested by Kevin Gilmore here : http://www.head-fi.org/forums/showthread.php?t=80748 and AMB DC protection module.

Should sound good and 160V is not much more dangerous than 80V.

 

[ionomolo2]

Nice links, despite i'm not planning to use a pentode as css, but rather the LME49810, which should be able to supply 20 mA at 150V with relatively high precision. I'm planning to use a pair of normal general purpose diodes as voltage reference and targeting a differential impedance of about 1MOhm. Current will be set with fine and coarse trimpots and plan to build it into a case with the buffer (ac-coupled with the ccs the soha way), so i can use it to try diferent pin-layout tubes without having to build the whole thing each time. I've also planned to use a voltage source as cathode bias, but don't know if it will lead to good results, despite being quite confortable for the experimenter to have independent grid voltage-anode current settings.

 

[ionomolo2]

I've been thinking on the optimal way to bias many different tubes without having to change layout and this is the sketch of the thing i'll work around.

I'm posting it as-is because i'll be busy for the next couple of months. If someone is interested is allowed to copy/modify/sell for any purposes.

Warning! Untested schematic, do not build. shown for ilustration purposes only:

(Current source driver is LME49810. Both input and output buffers should be mounted on socket to change them if broken due to accidental short. Source decoupling caps are omited for clarity)

The main aim is to provide precisely-adjustable bias point for a variety of tubes and to isolate both source and headphones/power amplifier from the experimental region so anyone can try different bias points for various tubes without worrying about damaging expensive gear (I do not encourage to use expensive gear for experimenting, some of we have learned not to do that the hard way). Don't know if it will operate with voltages as low as the ones on the millet hybrid/millet max but sure for the soha and upper (national datahseet for the LME49810 claims it can be supplied with up to 200V).

This shematic operates as current source + voltage sense. It would also be interesting to do the opposite (And also provide an adjustable bias resistor), but it seems that it's extremely hard to change within one topology and the other in the same board.