PMsix61 Limiter DIY

[alexc]

Thanks G.

Just finished the sidechain amp and feedback network. Have a few parts to wait on.

So now to complete the GR tubes. These are so simple its fun to wire up.

Then just to knock up the cathode balance bias network and that's that.

All that done now sans waiting parts.

Onto the front panel.

 

[alexc]

Buttoned up the SC bias and various bits ready for the front panel. 

Starting with the time constants network and bridge then onto the dual switch I have to make up.

But that's pretty much the gnarly part done for a first pass.

 

[alexc]

Some more progress - just waiting on a final audio cap and need to figure out how I want the Threshold dual-rotary switch  of 100K per pole : log or linear and what actual attenuation range makes sense.

For now I'll use a dual pot to help me figure out what I want there.

So ready for the 'smoke release' test. Power up and get all the rails going.

With some luck I should be passing stuff tomorrow 

Then the exciting part begins!  This is definately the high water mark of DIY as far as I am concerned.

A fairchild-e compressor with 12 tubes, 6 irons, 2 tripletts, 17kg of weight, 150W and 1000usd of hard earned  8)

I'll be comparing it with my 100usd BYOC 5 knob ross compressor. It used 0.1W and weighs 250 grams!

 

[alexc]

Wowee zowee!

All unloaded rails are working, all regulators are working

The 'analag clamp' is adjusting right as is the diode bridge reference.
The GR tubes kathode dual-locking pot is adjusting as expected.

No smoke, no nothing.  8)

This is the Poorman 670 PSU #4 for me -  two in service for years, no service issues.
One spare and now this one working fine and first go.

Now to start the load up - firstly the sidechain tubes : 6V6x2, BH7 and AX7 all unregulated heater and HV.

Really getting to the fun part now.

Firstly I'll make sure the sidechain amp is working as expected with no signal, then I'll check then dc bias circuit is good and then the signal tests.

After I note down the no negative feedback preliminary results, I'll focus on the nfb and measuring the output impedance.

 

[alexc]

Looking at the sidechain, I have a B+ of  +317Vdc  (no GR tubes installed).

The plates are at +314Vdc and screen at +311Vdc with B+ fed thru the Hammond transformer.
Fixed bias is at -14.0Vdc right now. This gives -13.5V or so at the grids.

I'll adjust that as required - it comes from my regulated reference rail.

The original has +440V on the plate (from CT fed B+), +154V on the screen (from tube regulation) and -14V fixed bias from the negative reference supply section. This gives -12V at the grids. To my semi-educated eye, that seems very hot indeed.

Sowter control amp output is specified with 10K primary. Still seems like it must be driving those finals very *** hard. Of course, the original runs low screen voltages that ensures the total current thru each of the finals doesn't run to cherry. Still, plate voltage does seem a bit high for a 6V6 (as in the original 660 - not sure about the original 670 with 6973 finals)

So I  differ in that my screen is UL fed from the Hammond, not a tube regulation section and both the plate+screen are significantly different. For that reason, I need to alter my operating point so as to get as close to original as I sensibly can.

The Hammond 1645 has a primary impedance of 5000 ohms plate to plate. Right now them 6V6 plates do seem to be getting a little 'cherry' 

So far, so good - more or less. Now to look at the dc operation in a bit more detail.

 

[alexc]

My garden buddy, however, is not impressed !

 

[alexc]

So I need to rethink the sidechain finals : UL is not good. The max screen voltage spec is 285V so I need to drop quite a bit. I'll try my spare 245V regulated rail. At present, even with my max negative bias voltage of -18.5V the 6V6 is still passing some 58mA at the cathodes. Way too high.

Part of the problem is the 5K plate-plate load of the hammond - too low.

So, now to rejig things to get that cathode current to around 37mA total plate+screen at a reasonable bias of around -14V or so.

OK - that did the trick. I now have the screens fed from +242V regulated thru some 680R resistors giving around +241V at the screen. I can now adjust fixed bias supply to -15.5Vdc giving grid of -15.25V and a cathode current of 37mA each 6V6.

Now that I'm not diming the plates to cherry, I have B+ 325V, plates +322V

That's my basic starting operating point.

 

[alexc]

Now looking at the BH7 driver.

In the original, it is also fed from same B+ going to output traffo CT feed of around +442V. With 1K cathodes at 2.6V and 2.6mA or so.

I currently have BH7 with the same plate and cathode resistors as original, fed from B1+ of  +313Vdc. Checking the operating point now.

Have now +312V B1+ and around +76V on the plate with +2.9V at the cathodes, on 1K so 2.9mA or so.

Pretty close.

Now to the AX7.

In the original, it is fed from the tube regulation section at around 240V B2+ and has some 235V on the plates for each side current of 0.05mA - virtually off! The bias is provided by the combination of a voltage divider and the dc bias pot. The divider sets the maximum bias (max hotness, min +ve voltage) and the pot reduces the bias (makes colder, increases the +ve voltage) to force the stage to only conduct the peaks of the incoming side chain signal.

That way, the limiter threshold is achieved as the stage is more and more in class B.

The voltage divider does 1.77% of the B2+ for around +4.24V. That's the 'hotest' bias
I'm seeing that  more or less when not connected to the cathodes or the pot.

The dc bias pot is one side grounded, middle with 270K to B2+ other side floating.
In the mid position, it forms a divider of 270K+50K so should have 15.6% or 37.5Vdc. That's the 'coldness' voltage max. I see that more or less when not connected to cathodes. Floating side is similar.

When I connect the pot to the 4.24V divider, I end up with the pot centre at 28V or so and the other side at 12.2V or so.

When both are connected to cathodes things go somewhat awry! I can't seem to get the complete circuit cathode voltage below around 15V or so. Should go to 4V or so as I dial out the contribution of the dc bias pot.

OK - that's it for now - the problem wins for today. I'll go and think about what I've done

Still - pretty good progress. All tubes are in, heaters are good and the basic sidechain nearly working to the first level of expectedness.

 

[DaveP]

Well done Alex,
I have attached my Fairchild circuit.
Its important to get the 3.6V cathode voltage right of the 12AX7, I had to trim the stock resistor.
keep it up
DaveP

 

[alexc]

Thanks kindly Dave 

I'll be going thru that carefully right now!

 

[alexc]

OK - found my issue.  :

First and stupidmost, I have the grids of the ax7 floating.. d'oh!

Secondly and similarly dumb is I have unregulated B2+ to the ax7 and bias network. Since this stage is set to operate at tiny currents with very large resistors, it should be fed regulated, not via a pi-filter. Adjusting the dc bias pot makes huge changes to the B2+ voltage as I currently have it. No es bueno.

Now to fix those things 

OK - that works fine now - I can adjust the dc bias such that I can dial out the locking pot 100K contribution completely and end up with the cathode voltage set fully by the 1M+18K divider. B2+ is not affected by setting of the dc bias pot or the ax7 currents.

As Dave mentioned, I need to trim that divider to get the +3.6V as in the original rather than the 4.24V I get now. At present I am getting on the ax7 B2+ of 242V and plate of 239V as a result of the bit too cold bias. Given the tiny current thru the ax7 stages, that's quite significant. Like 2 calories to a starving man!

Now to trim some and get some signal happening for open loop measurements.

OK so I can trim the 18K part of the divider down to 11K and I get the 3.6V cathodes, but there the ax7 plate voltage is still too high (not enough current, biased still too cold).

Trimming further to 8K5 gives me the expected 5V across the plate 100K resistor, meaning 0.05mA thru each ax7 half, which is what is expected. So I will go with that for now. The divider voltage is after all simply to bias the stage for 0.05mA each.

So that cathode voltage is 8.5/1008.5 * 242 = 2.04V plus the voltage contribution from the cathodes current, which is 2x 0.05mA * 8.5K -> 2.89V total approx. Which is what I see. The dc  bias pot bumps that up quite a bit, as required to make it very class B.

Now that I  have the first level of operating conditions set, I will measure some open loop ac parameters with signal injected into the primary of the control amp input transformer. I have no 'ac threshold' rotary switch in for now.

 

[alexc]

The sidechain amp or control amp depending on your terminology, is certainly working but I have some out of balance condition for the first stage (ax7) that I need to address.

*** First stage gain is is very dependant on the bias voltage.

Second stage gain is x10 or so
which seems a bit high. I would have expected some x5 or so

Finals with 5K plate-plate load  x53 and then step down to x15 which is /3.5  both in the ballpark.

These are early numbers - I need to address the imbalance, do some trimming of dc voltages and so on.
But I can easily get -100Vdc at the bridge output for an unattenuated input of 1Vpp sine.

OK - now to do some detailed tweaking and measurement.  Tomorrow. 

 

[alexc]

For the ax7, I needed to use some matched 100K plate resistors and try out a few different tubes as well as using my more accurate meter, but I have the 2 stages pretty well balanced to about 3% or so. That's about as good as I'm likely to get.

Funny - it was a cheap china tube which gave best result Quite a bit of variation with a few sovteks and such.

As I can now see, at such low currents, the gain of the stage is very, very dependent on the bias voltage.

So I do need to trim to 3.6V quite closely. The dc bias pot can increase the bias to some +16.5V. This decreases the gain of the stage and quickly shuts it off. I have 2Vpp to ground from each phase of the control amp input traffo secondary.

With the dc bias pot dialled fully out :

The spec 18K gives me 4.24V and quite a low gain.
Trimming for 9K gives me the 3.6V bias and a gain of x11.1 with a current of 0.118mA each half.

Seems like too high a current compared to the schematic I have - it says around 0.05m A each half.
Since I don't know what the required gain is, I'll leave it like this for now.

It is nicely balanced but has the positive going peaks clipped off. I need to think about that for a while.

 

[DaveP]

It is nicely balanced but has the positive going peaks clipped off. I need to think about that for a while.

Its push-pull so the other half takes care of that
best
DaveP

 

[alexc]

Of course, you're right - I'm just trying to understand the numbers a bit more carefully.
ie. correlating the clip with the bias voltage set and the effect of dialling in the dc bias.

As is often the case, there's a lot more going on than I expected 

I also need to do the matching of the bh7 100K plates - I have a bit more imbalance there I can improve.
The finals and traffo are pretty good for that. Don't think I need to do anything more there.

I'll probably stay on the sidechain for a while till I can get a more intuitive feel for how how the dc bias affects  various signal levels and such.

Still, am reasonably happy so far - not bad for someone recently described as an ignorant superstitious 'hack' 

 

[DaveP]

Alec,

"Still, am reasonably happy so far - not bad for someone recently described as an ignorant superstitious 'hack' "

Ooooh, that sounds like the kind of insult only a woman can deliver

BTW. I think the feedback from the tertiary to the cathodes will sort out any residual imbalance, the amp is only for the negative DC bias so don't lose too much sleep over it

The DC bias (on the 12AX7) gives you the classic Fairchild compression curves attached. (along with the correct caps and resistors of course)
keep up the good work
best
DaveP

 

[alexc]

Thanks for the curves and good that the bh7 slight imbalance is no biggy. I only have the one spare  bh7 left.

 

[alexc]

SO I realized I made the mistake of doing the 18K resistor trim with some signal present.

Doing it again with no signal present gives me the quiescent bias voltage of 3.6V for the ax7 for the trimmed divider resistance of 13K

Interesting thing I am seeing :

- increasing applied signal increases the bias which decreases the stage gain

eg. applying a 1KHz sine
      1Vpp -> 3.68V dc bias, x8.3 gain
      2Vpp -> 3.88V dc bias, x6.3 gain
      4Vpp -> 4.36V dc bias, x5.3 gain
      and so on

Something to do with clipped waveform current or something as the stage starts to round off positive peaks at around 1Vpp input and is fully clipped to flat > 1.3Vpp.

Anyway I don't think I've seen that before, but then I've never really looked for it either  
I've never really looked closely at tube 'class of operation' transitions before.

I would have expected each input phase to amplified with no clipping at all until the instantaneous amplitude at each grid approaches the quiescent cathode voltage beyond which it can no longer swing more positive resulting in the clipping of positive going peaks.

I would not have expected to see the clipping so early ie. 1.3Vpp applied  More like 7Vpp or so applied
I would not have expected the quiescent cathode voltage to vary so much with increased applied signal

Given that *seems*  ;D to be happening, I would expect the gain to rapidly decrease with colder bias as the stage approaches cutoff.

Anyway - food for thought. As I said, more stuff going on there than I would expect!

 

[alexc]

bh7 bias doesn't vary by much at all until it's input is fully clipped and then to  a much lesser extent at maximum input.
Maybe 0.2V max on a quiescent bias of 2.9V.

So I guess the clipped input results in a clipped output voltage waveform which has a 'dc like' component in it's current waveform. Which alters the quiescent dc current flowing in the stage thru the cathodes which alters the dc voltage at the cathode.

The ax7 being such a tiny current exhibits this effect in a much more exaggerated way.

More or less

Anyhow, onwards thru the finals and traffo.

 

[alexc]

With no feedback, the finals are pretty much as I expect, with a very high gain at the plates.

Looking at the secondary of the control amp output transformer, distortion sets in pretty quickly.

I have the bridge disconnected for now and the secondary is driving a 4K7 1W resistor.

Closing the nfb loop from the tertiary winding to bh7 cathodes using the oringinal's 560R and 2K7 network certainly cleans  up the secondary waveform substantially and results in a bh7 cathode voltage of +2.4V down from +2.95V or so.

The secondary shows a clean sine all the way up to max amplitude of around 100Vpp after which it begins symmetrical clip at that amplitude. That's for an ax7 input of 1Vpp each phase to ground.

I am seeing some weirdity on the bh7 output voltage waveform however - has gone from a nicely clipped sine with flat top to a wiggly top now.
-> this is due to clipping in the 6V6 plates causing clipping at the feedback to the bh7 cathodes

So now to think about that for a while!