What Tube for a Differential Amplifier?

[Macleod]

Hi there,

I am looking for some help to define the most appropriate tube to use at the input stage of an hybrid amplifier.

Let me give you the topic : I want to make an amplifier whose input stage is a differential stage using two triodes, ouput stage will be all bipolar.

I want to have a 40Vpeak on one of the triode plate for a single input of 1V peak.

So I need at least 40 for amplification.

The issue is that I don't want to use high value load because I may have some loading from bipolar buffer. If I could stay with something like 50k, I would be happy.

Do you have any idea how to struggle with those spec :
- B+=150V
- Vp with no input signal = 100V (to be able to have +/-40V...)
- Ip=5-10mA max (I don't want to add cost on power supply, made from 50V Transformer with multiplier...)
-µ>xx
-Ra>50k
- Input signal may be 1-2V peak max (I will reduce gain if 2V, but no grid current for this signal!)

Don't want you to work to much! Perhaps you already have the ideal solution?

I am evaluating now 12AT7 on PSice, it looks quite good.

But I have been advised that it has no microphonie protection.

Ohh... last but not least, I would prefeer a tube that is always on production.

All that seems easy?

Thank you for any help!

Regards

Laurent

[Svein_B]

Amplification of 40x with one input driven is tough.

A diff-amp (LTP with a CCS tail) will have a gain of 1/2 of compared to a single common cathode stage.
With 12AT7 I can get about 30x, which leaves you 3dB short of your target.
12AX7 could get the gain you want, but not happy with low impedance loads.

SveinB

[tomchr]

As you're mixing sand with your glass anyway, you could consider using a bipolar cascode as a telescopic cascode load for the tubes in the diff pair. That way you can get gains very close to the rated amplification factor (mu) of the input tubes. This also loads the triodes with something that's very close to an ideal current source so they should operate quite linearly.

6J5 will give you a gain of about 20. I think the ECC83 (12AX7) would give you a gain of about 100.

If you run a tail current of, say 10 mA, you set up the cascode as a current source delivering 5 mA. There's a blurb about BJT cascoded differential pairs in Morgan Jones' book.

~Tom

[Michael Koster]

A pair of D3a in triode with choke or gyrator loads and a CCS tail will get you a voltage gain of 40x per side or 80X side-to-side when driven from one side (also maybe 12HG7 in triode)

There is an interesting hybrid inverter discussed in a recent thread
Phase splitter idea
Michael

[Eli Duttman]

Laurent,

The 150 V. B+ limit combined with the halved gain inherent to driving only the inverting I/P of a differential gain block makes things very tough. I think you are going to have fits trying to "shoehorn" a triode into this role. However, all is not lost. The EF86 small signal pentode will provide the gain needed, with some help from "sand". Buffer the resistively loaded anodes with DC coupled FET voltage followers. The capacitances of the ZVN0545A are quite small and you may be able to avoid a cascode in the follower.

I hope you are already contemplating a bipolar PSU. You will need a negative rail to energize the necessary tail CCS and provide compliance with the I/P signal.

[Macleod]

Thank you All, I like that web site with all possible contributions!

12AX7 : Yes, I LSPICEd it and I could see what spec curves shows : input voltage will be limited to less than 1V if high current is to be expected.

Cascode with BJT : you know what? I tested it. But I can't understand if this is equivalent to load Tubes with current source or not. In fact, as load impedance is low, we are not in best case for distorsion, we need high impedance compared to Ri (Ra/Ri should be high for low distorsion). Thus even if that provides high gain, even if Tube is forced in constant voltage current source, I don't know yet if distorsion figure is better or not. To be investigated!

Penthode : yes I agree, gain will be higher. But before commited myself to accept using other thing than the sacred triode, I would prefeer to finish with the triode. How would I face real audiopile with a penthode as input stage??? ;-)

Didn't have yet time to follow the link you provided. Some homework to do!

Thank you for your help. Any comment welcome!

regards

Laurent

[SY]

As a practical matter, I haven't had any issues with microphony of 12AT7 in the input stages of my amplifier.

If the input stage has another stage following it, and your source doesn't mind some input capacitance, a 12AX7 run at 0.75-1 mA per side will get you what you want. The issue will be bandwidth- the 12AX7 will have a rather high output impedance, and this will interact with the input capacitance of the next stage.

[ray_moth]

Have you looked at Allen Wright's PP-1C? It uses 6922s in a cascode LTP. Plenty of Gm, low OP impedance, 25k plate loads and that tube will run with low plate voltage.

[Eli Duttman]

Penthode : yes I agree, gain will be higher. But before commited myself to accept using other thing than the sacred triode, I would prefeer to finish with the triode. How would I face real audiopile with a penthode as input stage??? ;-)

Many of the past's "giants" used pentodes at the amp's I/P. We are NOT smarter than Hegeman, Recklinghausen, Hafler, and other luminaries were. You are boxed into a corner by that 150 V. B+ rail limit. Differential gain blocks should be resistively loaded, with a constant current sink (CCS) in the tail. Even the previously mentioned 6922 cascode needs more Volts than you are willing to use. Remember, a voltage drop occurs across the load resistors. The EF86 is both highly linear and very good sounding. Quite a few triodes can't come close to the sonics the EF86 produces.

IMO, the EF86 is as close to perfect as you will get, for this particular job. It's in production, sounds good, works well at modest anode voltages, and draws a low I_B.

Have you given thought to the GNFB loop OTL BJT "finals" imply? Are you planning on MCPs? Are you certain that a DC offset will not be present, at the amp's O/Ps?

[tomchr]

Originally Posted by Macleod

Cascode with BJT : you know what? I tested it. But I can't understand if this is equivalent to load Tubes with current source or not. In fact, as load impedance is low, we are not in best case for distorsion

The anodes of the tubes "look into" a collector. So they should see a load resistance that's fairly high. However, if the diff pair is not loaded evenly by the subsequent circuit, the distortion will suffer.

If I only need one of the two outputs from the diff pair, I find that putting in a "dummy load" on the unused output will be beneficial to distortion performance. Usually I use 220 nF + 470 kOhm as coupling cap, grid leakage resistor on a following stage. All that's needed is a replica of the 220 nF + 470 kOhm to keep the diff pair evenly loaded.

In PSpice, I had to use actual BJT current sources as the ideal ones created a fairly unrealistic DC operating point. You may have to do the same in LT Spice.

~Tom