Alternative to 6386 tube

I know that other limiters have used other tubes, I know the 6386 is crazy expensive and that it was pretty well made just for this application but has anyone tried alternative tubes such as for instance vari-mu RF valves such as ECC85 to find the one which has the greatest gain swing. Or is it that there just is not any substitute for the ubiquitous little buggers.
Steve

Nope, no direct substitute.

Modern frame-grid remote-cutoff types (e.g.ECC189) can be used - but at a greatly reduced working voltage.

Jakob E.

BTB in germany has 6386's at around 10?. They list it as an equivalent to 5670:

http://www.btb-elektronik.de/pdf/de_gesamtpreisliste.pdf

Manley uses pairs of 6BA6 tubes in a triode connection, and if you get 20 tubes you can get 10 to match up really well, but the tubes will take about double the filament current. A Fairchild copy looks pretty impressive with all those tubes:

http://www3.telus.net/~dulan/images/VariMuTop.JPG

One thing to watch for is that the 6386 tubes that are left may be mismatched from one half to the other which make them useless in a compressor/limiter. I've also used 5670's and they work but the compressor ends up being a bit noisy because the control voltage is less, and you can feed in less signal before the onset of distortion.

Thanks for the info.
PRR made it clear that his project was a "prove that the idea works" exercise and not really a complete design. The principal is simple enough
ie. detecting an audio signal into a negative voltage and using it to control the gain of a vari-mu stage, not so different from AGC in radio/TV.
I think I will buy a few 6386 and experiment a bit. The iron may be tricky
especially the output.
Steve

whats in the gates and ba6a?
is 6ES8 a remote?

Semi-remote or semi-sharp by the looks of the plots. -12 volts cuts it off. It kind of wanders down nonlinearly but it does look like it cuts off pretty sharp when it hits. I think it would work as a compressor but only in very low compression ratios.

6386 cuts off at a very, very low voltage (-70 volts or so with a 250 volt plate). 6BA6 cuts off somewhere around -70 to -100 in triode mode (from my tests). 5670 looks like about a -12 volt cutoff but it is gradual.

For sure the 6BA6 can be used to at least 20 dB of gain reduction with only the pleasant tube sound as distortion.

One caution - any imbalance between the halves of a variable-mu compressor in feedback mode will cause thumping and oscillation. I spent hours balancing the 6BA6 tubes in my creation.

Try 6BC8. It is an un-sexy TV tuner tube. Cutoff voltage is only around -12V, so it won't take the thumpin input voltages that 6386 will swallow. But at idle (high gain, no GR) it has pretty high transconductance (possibly low noise voltage), and pretty high current (so it will have some current left when in GR). Also significant variation in Mu, about 2.5:1, which eases some design problems a few dB compared to a nearly-constant Mu tube like 12AU7.

It isn't "better". But 6386 are expensive, and while the 6ES8 is fine, it is rare (I think it was a last-gasp attempt to keep tubes ahead of transistor encroachment). 6BA6 is an obvious choice, but finding pairs that match over the whole range is tough. With 6CB8, you know both halves were made in the same place, probably on the same jig by the same worker. And 6BC8 is so cheap, you can experiment with parallels. 8 or 16 of them -might- work without a makeup amp, like the 660 does, for way under $100.

The 6BC8's Ip/Gm is about 1.6V at 10mA, and that is a guideline for the maximum input voltage. In single-ended audio GR, the level must be much less than 1.6V. In push-pull, the maximum input can approach 1.6V before distortion won't cancel and distortion gets gross. The noise voltage may be a couple μV, so the total input dynamic range is around 120dB. Subtract the maximum GR to find the output S/N: 90dB-100dB. Good enough to master vinyl, and not a problem for 16-bit work. And since you only need a volt or so peak input at max GR, and many sources can supply several volts on peaks, you don't need step-up in the input transformer.

> 6386 cuts off at a very, very low voltage (-70 volts or so with a 250 volt plate).

6BA6 is a fairly normal late-1930s tube. Since then most tubes worked for higher transconductance at a given current (to raise gain and reduce power supply cost). That's actually all wrong for an audio limiter: we want low to very-low transconductance, and don't care (much) how much current it eats. The 6386 goes the other direction: it is a "lame" tube as a "cheap amplifier": pretty hungry for the gain it gives. That's what makes it good for GR. You need huge grid voltage to change its Gm, and it still has considerable current at large grid bias and low transconductance, so it can make some power even when far down the curve.

As a totally over-the-top GR tube: 6L6. It is pretty non-linear. Made to come clean in push-pull. Compared to tuner tubes, it has much larger grid voltages and much higher idle current. So it will eat a good slug of audio and a bigger slug of CV, yet not starve for current when in GR. Power tubes are the other class of tubes that didn't go overboard with Gm/Ip optimization (since Ip is "good" in a power tube).

6L6's Ip/Gm is about 5V at 50mA. OTOH, 12AX7 is down around 1V or less.

If you have a 6L6 amp apart: wire a 10K:10KCT or 600:10KCT transformer to the grids, leave the plates loaded in about 8KCT, with a load (try 16Ω on the 16Ω tap). Feed it up to 10V-20V audio. Bias the grids from the nominal setting for an amp (around -25V-35V depending on screen and plate voltage) down to -70V or further. You may not be able to get more than a volt or so output at 20dB GR, but that is in 16Ω. If you can prove that works, a 10KΩ:600Ω output iron should give 6 times higher output voltage, or about 10dB more GR at the same output voltage.

You might also try lowering screen grid voltage a lot. Note that on 6L6, a series resistor won't work: aligned-grid tubes have unpredictable screen grid current, which can even flow backward. Use a voltage divider with at least 10mA-20mA flowing. (For accuracy, the screen voltage must be regulated or the GR will drift with line voltage.)

Very unique idea!

I think I am going to give these 39/44's a try after all. I just don't see me rounding up 8 6386's for the 670, as it seems that matching is very important. I mean, how many 6386's would I need to get 8 matched? A hundread, maybe? But I guess since they are in parallel, some of the errors might cancel out.

I have 25 39/44's, and they are $2.50 a piece when on sale. Tube sockets are hard to source and the tubes are very old, so there could be some problems with the heaters being fragile. But the savings would be enormous if they worked. I guess I should do a load line on a single tube first so I don't waste all that time and money.

Can you imagine 16 of these on a single chassis! I won't need a heater this winter.
cj :razz:

I got 8 out of 10 6BA6 tubes to match pretty well - or at least to have compensating errors. They were all manufactured at the same place. Probably 8 out of 20 would be matched pretty well. I haven't played matchmaker with duals though.

Anybody ever fiddle with the metal can type 6K7 in pairs? 300v plates, & similar in room warming capabilities as those nice lookin' 39/44s! I'm thinking to try & twiddle PRR's 12AU7 circuit with these.

[quote author="cjenrick"]Very unique idea!

. I just don't see me rounding up 8 6386's for the 670, as it seems that matching is very important. I mean, how many 6386's would I need to get 8 matched? A hundread, maybe? But I guess since they are in parallel, some of the errors might cancel out.

cj :razz: [/quote]

i was under the impression that the fairchild used 8 tubes to even out differeces between the tubes. remeber it was designed to be stereo

Unmatched you can get maybe 5-10dB of reduction. For any more than that you get thumping and oscillation. The 'thump' as a tube cuts off gets amplified in the control amp, and applied to the grids so you get a 10 Hz oscillator. Been there, done that, bought the T-shirt. If you match them you can get 20 or so dB of reduction, and after that the distortion just rises, it doesn't thump.

-Dale

Toby, the 6386 is a dual triode, whereas the 39/44 is a single pentode(tetrode?), so you have to use 16 instead of 8.

Could thumping also be caused by the tubes getting unmatched and allowing a quick shot of dc into the transformer which would lower it's inductance and therefor resistance?
I was wondering if that was the reason for the BA6a resistors that got added to the circuit as an after thought/mod.
cj

isnt 6BC8 used in a old commercial compressor design, or am i thinking of 6BC6? :roll:

altec 436/438

This thread may be a good place to post about my first experiments with the 6BC8.

I try to build something similar to an UA175. So far, I have only built the PSU and the variable gain stage, but the results are very promising.

My test setup is like this:

6BC8 dual triode connected like the UA175.
Interstage transformer UTC A-19 connected to the output of this stage, feeding a monitor amplifier (and a scope). Input transformer is a small Lundahl 10k:10k transformer (LL1540).
Common mode voltage to both grids comes from a 15V lab PSU, adjusted by a 50k pot from 0V ... -15V.

CV of 0V means full gain, something like -12 or -13V turns the signal completely off.

I had no idea how large the input signal may be, so I tried.
I ran music from the HP output of my stereo amp into the input - this provides almost 10Vp with a good mix of various frequencies. (Playing Sylvian & Fripp Damage Live CD (;->) )

I found that this stage alone has a wonderful sound even at heavy overdrive. (I think this is important, because every limiter should have a pleasant behaviour even when it's limiting action is not working.)

Also, looking at input signal and output signal in X-Y mode, i noticed that the control voltage (i.e. amount of compression) significantly alters the way of distortion:

At high gain (CV = 0) the transfer courve is getting flat for high input (x) levels: not unlike the typical "S"-courve of a transitor differential pair without NFB.
At very low gain (CV = -10 ... 11V), the transfer courve shows the opposite kind of nonlinearity: Its an "S" with its ends bent _upwards_.
For some CV in between (i.e. a certain amount of attenuation), both effects seem to cancel and the courve is more linear than in euther extremes.
I don't know enough about tube circuits yet to interpret this - I just report what I've measured. (It's not the input transformer - I've checked this. But it might be some compined effect of the tube stage and the interstage transformer - or it might just be inherent to the tube at different grid voltages.)

Anyway - these preliminary results have been very encouraging, in a sense that this will be a tool with "character".

I haven't tried more than one tube yet. Had to turn the cathode trimmer almost completely to one side for decent CV rejection, so I guess that tube's triodes are not that well matched. I have 4 more of these in the drawer to try.

JH.

hi J!
glad to hear that your experiments are going well! :green:
From my limited experience so far the fun comes when the CV is rapidly varying, as in certain situations you get thumps/ LF oscillations etc, but I think some of this is due to mismatching within the valves.
I just got 2 LL1660 through the post, one of which I want to try as an interstage as i am just running straight cap coupled at the moment.
The funny thing with valves is that even if you build things completely wrong they still kind of work! :shock:
that can be a good thing, or a very bad thing depending on how long it takes you to spot the mistake :roll:
All good fun though :green:

chef

[quote author="Swedish Chef"]hi J!
glad to hear that your experiments are going well! :green:
From my limited experience so far the fun comes when the CV is rapidly varying, as in certain situations you get thumps/ LF oscillations etc, but I think some of this is due to mismatching within the valves.
chef[/quote]

The 175 has a switch that will apply 50Hz (from the heater supply) to
the center of the input transformer's secondary. So you can trim out the feedthru with a rather fast signal. (The 50Hz taking th eplace of the CV).

I haven't measured the exact CMRR after trimming, but I can already say that the matching would have been _very_ bad without the trimmers, just relying on the symmetry inside the tube.

JH.

I agree!
I built a 436 to begin with without trimers and it sucked hard! :grin:
Sorry I omitted to say that i AM using trims on each stage
Don't worry I am not depending 100% on matching...

chef