• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

6SL7 Diffamp front end

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I've completed a fixed bias 6V6 Ultralinear plus Cathode Feedback Push Pull Amp for a niece. It runs a current source biased 6SL7 diffamp front end with direct coupled MOSFET source followers to drive the 6V6 grids.

HT + to the diffamp is 270V with the current source tail returned to -20V.

I started with 47K Anode loads on the 6SL7 with 2mA per side.
- It sounded nice, warm, round BUT a bit covered (too much 2H distortion?)

I changed to 82 K anode loads with 1 mA per side
- This sounded less covered and more detailed with some additional top end.

This encouraged me to go to 150K anode loads with 600uA per side.
- This sounded even more detailed and brighter with maybe a little edge stating to creep in (too much 3H distortion?).

Is this just a function of the anode loads getting larger with respect to rp of the tube and therefore less overall distortion?

Is the edge in the 600uA per side due to running the 6SL7 where rp curve vs Ia is larger and hence I will have higher order distortion products?

Or do I not have a clue?

Anyone - can you explain what is happing here?
Anyone - what does TubeCad or other modelling say?

Thanks,
Ian
 
The higher current and the higher the anode load resistance the better the linearity. Best way to achieve this is to use a high supply voltage.

I like to run 6SL7 with 1.3mA and 220k from a 400V supply which is the operating point I have in my differential input stage in my EL34 amp.

In my opinon ;)

Edit: I forgot the answer. You wish to place the load line in the IaVa curves as far away as possible from the curvier part. You can do this by increasing the current and/or lessen the slope of the load line (higher anode load resistance).
 
This is interesting.

All three op points give about the same plate voltage of 170V. All three loadlines are highly linear (at least for the swing range required to drive the 6V6), as one can see below.

Gain rises with rising plate resistor values, Zout falls. Zout should not have much influence due to the high Zin of the follower.

What´s left? The gain changes. Did you listen at the same SPL? Did you change NFB (if any)?

Tom
 

Attachments

  • sl7_2.jpg
    sl7_2.jpg
    34.9 KB · Views: 798
Ex-Moderator
Joined 2004
Ian, what did you have in the cathodes of the 6SL7? If you had a resistor (unbypassed) between each cathode and the tail (a CCS?) then that would effectively raise each triode's Rp by (mu+1) * valuoe of the resistor.

In addition, it's worth remembering that the 6SL7's Rp increases markedly as the plate current decreases. At 2mA it's the published 45k but at 600uA, it rises to around 60k.

I found my Sovtek 6SL7s perform very nicely with:

* 1.5k in each cathode and 4.7k across the cathodes;

* 120k plate load, shunted by 330k grid resistor of the following stage

* 220v plate-cathode voltage

* 700uA plate current (controlled by 6AU6 pentode CCS in the tail).
 
Ex-Moderator
Joined 2003
Normally, as RL reduces, H2 rises, but the CCS will enforce balance and pretty well null that out. I would say, though, that 47k is a very low value of RL for 6SL7, and I would normally look for at least 100k, and preferably 150k. The problem then is that for any sensible current you need 400V of HT and I expect you only have 300V. You're in an unstable region of operation - you want to increase RL, but that means reducing Ia, and that increases ra, which is exactly the effect you don't need. I would expect that small changes will make a disproportionate difference to the sound and that it will change with the valve.
 
Ex-Moderator
Joined 2004
Sorry, Ian - maybe my post was a bit of a red herring! I forgot to mention that my 6SL7 LTP is feeding a 6SN7 diff driver to EL34 triode-connected PP output stages using Thorsten Loesch's cross-coupled negative feedback (with 2.2Meg resistors from EL34 plates cross-coupled to 6SL7 plates). The raised 6SL7 Rp (caused by having resistors in the cathodes and low plate current ~ 0.7mA) serves the purpose of feeding the NFB loop. The 120k plate load resistors ensure that the 6SL7 can get the required plate voltage and current.
 
Hi guys,
Thanks for the responses.
Ray - I have just 56R in each cathode for a little degeneration and hopefully better balance.
Current source is cascode BC547B with a reverse connected, floating base BC547B "low noise 7.6 Volt zener" as the reference.

I'm limited to that 270V rail - probably means that the 6SL7 was not really a good choice.

In addition the cathode feedback to the 6V6 increases drive voltage requirements from about 40V pk-pk to 56V pk-pk at full power (around 7 watts).

The whole thing is a bit of a compromise as I started off with a non-technical design requirement - I wanted something that looked "arty" to suit Kristy's (the niece) personal tastes. Hence some NOS 6V6G and the Octal 6SL7. The MOSFET source followers were added to cope with the 100K Rg1 for 6V6G in fixed bias and this then allowed me to increase the anode loads on the 6SL7.

Anyway, after the weekends work in trying to set anode to screen zobel networks on the 6V6s (they are ultralinear + cathode feedback) I found that the transformer resonances (cheap Hammond 1608) are way too low because I have separated the 2 off 4 Ohm secondaries to achieve the cathode feedback where they should be run in parallel. Best results were with 1K8 and 3n3 for the zobels but high frequency response is down 3dB at only 23kHz and a 10kHz square wave looks very "rounded". Soundwise this translates to an amp which has no real attack and is just generally a bit "bland".
For these reasons, I am now about to abandon cathode feedback and rewire these amps to Ultralinear plus shunt feedback as per the "Baby Huey" design BUT retaining the MOSFET source followers. This will allow me to wire the output transformer secondaries as per manufacturer instructions and hopefully get much better high frequency response.

They fullfill their main design requirement in that they look stunning - it is just getting them to sound stunning which is causing me some grief.

What I learned - For cathode feedback you need a good quality output tranny. I'll try it again when I build my next amp - 4 x KT88 Ultralinear + Cathode Feedback using Plitron VDV2100-CFB/H Toroidal Output Trannies. For cheaper transformer selections use the "Baby Huey" scheme.

Cheers,
Ian
 
Ian,

Some years back I did a lot of work on this tube for my Glass Harmony hybrid, though I'm using NOS RCAs.

I found that indeed 170V on the plate was best, around 1.09V bias, corresponding to 660R cathode resistor and just over 1.6mA.

When I say best, I mean best for me, so this is one man's opinion = very dangerous..........:clown: YMMVALD........:D (....and likely does)

The rp of a 6SL7 is given as 44K. I generally try to run a plate load around 2.5 times the rp, so this means 110K. In fact I use 100K, npv, and generally dispense pretty much with loadlines, using only this ratio and then dialling in different B+ and currents to find the sweet spot. It seems to work very well with cathode bypassing, though input overload margin is not great, typically around three times the bias value, in this case 3.25Vpp. If you do the math, you come up with around 335V B+, and this is very accurate and clean, not veiled at all, but with the marvellous image qualities this tube is renowned for. Sopranos have mouths like Cadillac grilles, and pianos are ten yards wide, but man, it's exciting in a small room!

Cheers,

Hugh
 
Hugh,
I agree.
In the "Baby Huey" design however, where this circuit is being used, I can't get 6SL7 currents up to 1.6mA per side due to all of the tube current being supplied via the shunt feedback network and the limitation imposed by my 270V rail.

I have a spare 6.3V @ 4 Amps winding on the power tranny and kicked around the idea of voltage doubling that and sitting the HT on top of it which would give me another 15 or 16 Volts which means I could run the 6SL7s at another 100uA each BUT decided that KISS applies. WHY?
For the usual KISS reasons:
Simple Circuits have Simple Problems
Simple Problems have Simple Solutions.

You will note however that I'm running a SERIOUS current source in the tail of the diffamp. Cascode BC547B with a reverse connected 3rd BC547B used as a low noise, temperature stabilised, zener of approx. 7 volt as the reference. I haven't done the math/modelling but it would be providing an AC impedance of at least 1MOhm.

Cheers,
Ian
 
I have been experimenting with CFB for years. It works well with some transformers and doesn't do much for others. I find that it can really wake up a cheap transformer on a SE amp.

A P-P amp is a different story. P-P with CFB was sucessfully used commercially by Audio Research several years ago. That is where I got the idea. They used special transformers to make it work. If the OPT is not completely symmetrical it will cause more distortion than it fixes. Weird ugly sounding distortion. The only P-P transformer that I have found that works is an old UTC LS-57.

Another simple sucess used a toroidal mains transformer as an OPT. 230 volt CT primary is the plate winding. Each 6 volt secondary was used for CFB. A new secondary was wound using hookup wire for the speaker winding. This was a simple 10 watt P-P amp using 6AQ5's.

If you want to try a conventional OPT for this circuit, it must be wired in an unusual manner. Ground the 4 ohm winding, the 0 ohm end goes to one cathode, the 16 ohm tap goes to the other cathode. The speaker goes between 0 and 8. This is straight from the Audio Research schematic.
 
I've had similar experience- when I played with that arrangment, I could get a useful lowering of output impedance, but no real distortion improvement. I think that because driver stage distortion usually scales with current and CFB increases the required drive swing, what you gain in the output stage ends up being a wash with what you lose in the driver.

Most useful with pentodes, IMO.
 
Tubelab and SY,
My experience exactly. The Hammond 1608 is not up to use of the separated secondary windings for cathode feedback. The amp was slow and just had no attack what so ever.

With the amp rewired to Ultralinear plus shunt feedback, however, as per the schematic referenced above, the amp is super fast and in fact I kept pushing up the shunt feedback to cope with a general impression that it was "getting ahead of
itself" - if that makes sense. Image is as Hugh put it "pianos ten yards wide". It really demonstrates the value of zero global feedback. Changes to that schematic are the cross coupled shunt feedback resistor increased from 13K to 20K and the zobels on the output tranny primary have been deleted.

The speakers I've connected are some nominally 8 Ohm 3 way Philips (12" bass, 5" mid, 3" tweeter) from late 60's early 70's - I just replaced all the cross over caps with modern metalised polypropylene. To "cut to the chase" I have connected these to the transformer configured for 4 Ohm output to get the best from parallel connection of the two secondary windings (thanks EC8010 for this tip).

I've been "p*ssing" about with that feedback set resistor going between 18K, 20K 22K and 24K for about 2 weeks without being able to make up my mind as to what sounds BEST - mostly because I don't have consistent set of these values in all the same type of resistor and that particular resistor HEAVILY influences the sound. Ordinary metal film - about 20K is good, Philips MRS25 24K is good. Makes a great resistor sound test bed (27K Rikken Ohm was awfull).

Cheers,
Ian
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.