Tube comp/limiter under development

The main similarity I see is the location of the output attenuator. As we both realized, one cool thing about a separate sidechain amp is that you can move the output attenuator upstream and not have to waste power heating up resistors!

Whats the purpose of the interstage transformer?

To keep the plate voltage from bouncing around with plate resistors, and to keep the grids of the long-tailed pair afterward from getting slammed with too much CMRR during the voltage swings.

With plate resistors, Plate voltage and current both vary. With an xfmr on t he 1st stage plate load, just current varies.

So its a B+ centertapped with the plate(s) idle current flowing through the windings?

Im not to familiar with vari-mu circuits.....

Yep, the transformer has to be gapped to tolerate the DC also.


dave, as an experiment, take a 5k resistor and feed the centertap of the 1st stage from the regulated B+ through the resistor. Then, either short it with a cliplead or wire a switch across the resistor and short it out. the sound changes a lot,! Both are useful, but the resistor in series sounds much more like a stock Altec 436 than a Fairchild.

Kit, Unlike a conventional amplifier stage that you bias to a particular current, and forget about, the mu circuit is constantly altering the bias. Therefore, the plate current is varying at the same rate. This will, of course, cause the voltage to vary across the plate resistor and this, depending on who you talk to, either enhances or detracts from the desired sonic signature. I like the fixed-voltage, variable-current sound, myself.

[quote author="Larrchild"]I like the fixed-voltage, variable-current sound, myself.[/quote]

Got to try this...time for some listening tests.

analag

Looking good! Show us more!
/Anders

The block diagram seem indicate a single ended sidechain amp. Is this so?

analag

Yes, it is. But everything is subject to change--it's still in the breadboard stage, which is why I'm not posting a schematic now.

Dave, what software did you use to draw that block diagram?

I am curious if an Edcor XSM10K/10K could be used for the interstage transformer with a light load on the secondary to make reflected impedance fairly high.

That looks pretty close to what I just finished up, but I used Sowter iron. 5670 GR tube, 12AU7 + 7044 output (with negative feedback), 6AL5 GR amplifier. Interstage transformer between the GR stage and the output. Mine has a simple front panel - input level, output level, and a 3-position switch - FAST, SLOW, DUAL. I made the output amp do double duty as the sidechain amp, just as in the UA175 and Altec.

I needed a break from building mic capsules, recording, and getting married....

-Dale

I took the plate resistors out of the MU stage and fed the B+ to the primary center tap, thus using steady voltage and variable current to attain compression. The sound is different, punchier it lacks the sudden stop it used to have, I think. The attack is still the same speed but the recovery has changed a little. In this configuration it's quite clear that the characteristics of the tube have more effect on the compression where as with the plate resistors installed this is not so. I have been able to get good gain reduction with 6N1P's and 6DJ8's and these are not remote cutoff tubes.

analag

[quote author="rafafredd"]Dave, what software did you use to draw that block diagram?[/quote]

It looks like orcad. Orcad is pretty nice.

Kiira

. In this configuration it's quite clear that the characteristics of the tube have more effect on the compression where as with the plate resistors installed this is not so.

Click to expand...

PRR, or sombody smart could illustrate how a tube swings voltage with a transformer vs plate resistors in a fixed-bias amp circuit. But clearly, when you are varying the bias and therefore, introducing voltage swings across the plate resistor that are independent of the audio voltage swing, you are adding color. I call it the "Altec Whump".

Orcad was a whale, no?

The attack is still the same speed but the recovery has changed a little.

Click to expand...

Yeah, the tubes arent folding back anymore giving you "faux compression". I bet your sidechain has to work harder now too!

[quote author="Larrchild"]
Orcad was a whale, no?
[/quote]

Yes. Orca's philandering brother.

> the transformer has to be gapped to tolerate the DC also.

Ideally, the DC cancels to zero.

Tube mis-match may cause a slight DC unbalance.

The main issue, as I see it, is that when you change from maximum current to minimum current, the two tubes generally can't be exactly balanced at all currents.

To make-up some numbers: you idle at 10mA per side, go to 0.1mA for maximum attenuation. To make decent output power at 0.1mA (which is also your maximum output condition), you will probably want like 10K or 100K windings. But as you just start to go into limiting (which can happen a lot if you just need a light trim), the tubes will be say 5mA with maybe 10% unbalance. That's a 0.5mA unbalance. That's maybe acceptable on an E-I core wound for 10K, but could distress a 100K winding.

Yes, the common-mode plate resistance has a large effect on gain-law.

> I call it the "Altec Whump".

That's sloppy common-mode rejection or bad primary balance. The Altecs were intended as safety valves, not artistic tools.

I did use plate resistors on my GR stage. (As I mentioned earlier, it's similar to that of the UA175--not identical, but similar). The plate resistors are small enough that, in concert with the regulated B+, the plate voltage variations are small compared to something like an Altec 436. As a result, the knee is harder--but as Larry points out, series resistance can be introduced deliberately when a softer knee is desired.

The main reason for these plate resistors is that they largely define the source impedance working into the interstage transformer. Without them, the xfmr would be working from a source impedance that starts out at one value and rises substantially with gain reduction. Frequency response would suffer at the low end (due to finite primary inductance) and the high end (due to winding capacitance).

If you allow the tube plate resistance alone to define the source impedance--as would be the case if you just fed the B+ to the xfmr centertap, with no plate resistors or other load shunted across the primary--it would take a very carefully-made and expensive transformer to maintain flat response throughout the wide variation of plate resistance. That's reason #226 why you can't really do a Fairchild on the cheap.

Shunting the low resistance across the plates loads the tubes and restricts output swing; but if you keep I/O levels under control and provide adequate gain after the GR stage, it's not a big problem.

You may need to operate the tube/transformer system into a "matched" load to avoid high frequency droop at higher GR values. For instance, if the differential output impedance of your GR stage is about 10K (defined primarily by two ~5K plate resistors), and you're using a 1:1 interstage, then your secondary load would also need to be 10K. I suspect that the UA175 (as an example) may have used a 1:2 ratio on the interstage, since the source impedance is about 10K and the secondary load is about 40K.

One other approach, if you desire an essentially constant plate voltage but don't have the super-duper Fairchild interstage xfmr, would be to use a lower B+ (still regulated), feed that to the xfmr centertap, then shunt a resistor across the primary to help define the source Z.

[quote author="NewYorkDave"]... I suspect that the UA175 (as an example) may have used a 1:2 ratio on the interstage, since the source impedance is about 10K and the secondary load is about 40K.[/quote]

From memory, I believe the 175 interstage is a flipped around UTC A-18.
Don't have a UTC spec sheet handy but I recall something like 15KPP:80KPP? becoming 80K:15K as used in the 175. Could possibly be wrong as it's been a while...

Edit 1: found some old (1960's) UA notes for this and can confirm it was "mostly" an A-18. I say "Mostly" because I just read a memo written by Putnam Snr. where he's suggesting that Engineering sub in an A-19 as a temporary measure because delivery from UTC on the A-18's is behind schedule!

[quote author="NewYorkDave"]... shunt a resistor across the primary to help define the source Z.[/quote]

TAB U73 & U73B is an example of this.

Edit 2: forgot to add "nice one Dave :thumb: "

Dave sez:

One other approach, if you desire an essentially constant plate voltage but don't have the super-duper Fairchild interstage xfmr, would be to use a lower B+ (still regulated), feed that to the xfmr centertap, then shunt a resistor across the primary to help define the source Z.

Click to expand...

That might be a good compromise. And you have enough downsteam gain to make up any loss. It should flatten out the matching a little when using a more average xfmr.