RCA/175 Style Limiter DIY

[alexc]

- Looks like the gains at idle are :

+19dB on the input traffo
+17dB on the 6BC8 at idle, -7.2dB at the interstage,  +13.3dB at the AX7 +22dB at the BH7
-19dB at the output traffo

So that's +45dB excluding the attenuators which do around -70dB and -85dB at max

-----

So yes - the input atten stays quite low to prevent lots of THD thru the unit

The 6BC8 cuts off at a grid-cathode difference of around -6V or so and probably sit around -3V in common usage, so that means a max input at the grids around 6Vpp or so.

For 6Vpp at the grids, that means 0.66Vpp at the input primary.

So, 0.66Vpp after the input attenuator -> means the input attenuator is operating at the attenuated end of things most of the time, for a typical strong input at the xlr of say, 8Vpp.

ie. some 20dB or so would be the minimum typical input atten. Probably more like 20-30dB.

------

Anyway - there's a lot of gain thru this unit, 45dB - even with typical 12dB or so GR, that's still a lot.

Hence, the capability for doing a lot of THD quite easily! 

I figure typical input atten  is 9'oclock and output atten at 11 o'clock for 12-15dB of GR and an output at the xlr of around +4dBu.  (with some -10dBu or so at the xlr in)

More on the input atten  and the THD shoots up. More on the output atten and the noise floor creeps up.

I see typical usage 12-15dB GR and 15dB makeup gives a no-signal noise floor of about -79dBu, which is an increase of around 5dB of noise over bypass (loopback).

That's very good in my diy experience with limiters  - absolutely well below the threshhold where I would use a noise gate to prevent .. noise .. when the unit is with no signal.

---

My worst acceptable  'no signal' noise floor would be around -63dBu.
Noisier than that and it's time to gate the silences.

No problem when there is signal (of course!) - the limiter 'GR's the noise as well as the signal.
So the noise is much reduced (of course!), as well as being masked by the presence of signal.
But with silence, well .. you get the fixed makeup gain on just the silence. Hence .. noise!

I use yamaha digital gates on tracking channels where neccessary in the monitor mix.

They are actually very, very good. Better than any analogs I have used.

For monitoring stereo stuff where gates are needed, I use the Finalizer multiband gates - best I've used by far of any kind. They really let you dial in the gating on the bottom end, 'where the noise dwells' 

----

This will be an important reference for my RCA/175 style build.

So - now this guy is in my music room

It is after my 'Vintage Channel 3' build - an API style pre + gpultec eq with tube + MNats 1176 revG

I can only really do this because the Drip 175b is so very quiet - with all that, still no gate needed.
Otherwise, forget chaining 2 vintage limiters!

I 'discovered' that a tube limiter after an 1176 is exactly right for me! Ever since then, an 1176 alone is a bit 'lonely' sounding. ;D

 

[alexc]

and another gratuitous rack shot

 

[alexc]

So - now the Drip 175b is done and I have good data as a reference, back to my dual channel RCA Ba6a / UA175 style mashup

When I left it I had all the tubes up and running, biased up properly and was starting on the front panel controls and claculating the gain staging and the required feedback factors.

Here's a top pic

 

[alexc]

And an inside pic.

This is wots known as 'a work in progress' 

It'll clean up a lot as I progress!

 

[alexc]

First order of business is to add some feedback with rough estimate of fb factor so I can focus on the input and output attenuators and the resistive-balancing center-tap networks for feeding B+ to the input and interstage transformers.

These have a big effect on performance - they set the scene for a lot of the gain staging and have a big impact on the noise performance.

Indeed, one of the mysteries from my pm660 style limiter was the big difference in unity gain noise floor I measured between 't-atten before input traffo' vs 'balanced rotary atten switch after input traffo' methods.

There was a major improvement in the 'after input traffo' case - as far as I can fathom, there shouldn't be a massive difference, generally speaking. If the gain staging is similar, the noise should be similar.

But it wasn't the case in my PM660.  I went with the 'balanced rotary atten after input traffo'.
Dead silent, whereas the other was way noisier - 'a mystery, wrapped in a riddle, shrouded in enigma'

For the Drip 175b, no problem at all with the 'atten before input traffo' case. Silent like.

Anyway, that's high on my list of things to figure in this build. First up, the stock 't-atten before traffo'. case

----

Another thing which I want to check some more, is the 'overshoot' and 'settling time' on the 'slam'.

With the pm660 build, there was a pretty big overshoot - I have to redo the view at 96Khz since the event is very fast, but yep. A massive overshoot and a certain settling time on the pm660, at least at some settings of 'dc threshold'.

Indicates either an incorrect 'dc threshold' setting or a damping issue with the sidechain-signal amp connection. ie. an impedance thing at the interconnect.

So I'll be keen to see how that pans out - will these rca/175 builds with the hi impedance sidechains massively overshoot on the transient 'slam' like the pm660?

Not likely, as the CV charging current in the hi-z case is tiddling in comparison to the massively heroic, low-Z pm660 case.

But, as always - doan assume nothin'. (it ain't over till it's over)

Doubtless more ponderoso on the issue will yield some more food for thought.

(I have the plug ins as a reference  8) )

----

So - some quick 'smoke release' tests to get back in the game and onto those attenuators and B+ feeds.

With a bit of luck I'll get the same momentum I had going before - this build was working out pretty well and was going very quickly along.

Cheers

 

[alexc]

Some good starts ...

HV      still all there at 85mA + 85mA  for the 2 seperate HV supplies.
Heater too at 5.7A

I have the output attenuators in place, and they work as expected.

Big buggers they are! I may need to put some higher power resistors around them to deal with the monster line amp stages - octal side can approach 10W max power (max power ) and the noval side is not too far behind.

Since these are my first monster line amp to 600ohm stages, I'm not quite sure what's required.
But I'll figure it out. I think these Mallory t-pads are around 2W or something. Have to check that.

For now, no probs - with no signal and input attenuator at alot (not sure yet how much), the noise floor with max output attenuation is at it's minimum  ie. -84.4dBu which is the same as my loopback.

Right now my overall gain is thru the roof at >50dB so I'll also tame that down quite abit with feedback from the finals' plates to drivers' cathodes.

My guess is I'll need 15dB of GR and 15dB of makeup typical, so I'll probably be looking for 20dB or so of feedback.

Once that gain is tamed, I can look at the noise floors again and get a better idea of what gain staging I can run.

----

Now for the input attenuator, I am trying first up a mini t-pad from CAPI on one side and a dual rotary with 25K per deck on the other.

So - time to get to it with those things.

 

[alexc]

So far it seems there isn't much difference between the two methods of input atten - I have one of each connected and the results are very similar in the unity gain noise floor.

One thing stands out - these guys can put out some juice! Have to insulate those output attenuators.
Because the gain isn't yet tamed, it's easy to SLAM those outputs - enough to give a shock. Something like 70Vpp max  ???

Not to mention upsetting the thing it's connected to   I think I'll hook up the load boxes I made for power amp testing! At this stage the relative amplitudes of the noise is the important thing, absolute levels isn't important, so it's still useful to check the the load boxes on.

Right - now to do some checks for oscillations and work on finding the best orientation for these utc a15s - as I found before, they are real noisers in the wrong orientation and on a bad day.

Anyway - I'm right in the thick of it now

So far I'm seeing unity gain noise around the -70dBu mark - not too bad considering it's the first look.

Rotating the utc is again making a big diff - looks like 6dB or so with labels facing left (seen with unit upside down).

Checking some more with the utc shield - here it gives a clear 7dB improvement so far.

And it does on both sides. Awesome.

Now getting around -83dBu or so at unity on the best side and around-79dBu on the worst side.

So that was well worth the $30usd each or so I paid.

What a relief! That low noise level at this early stage bodes well for a result at the end  8) 8) 8)

 

[emrr]

Excellent exhaustive reports.  You shouldn't need massive power rating for the output attenuators, given you don't plan on cranking up pure sine waves to high levels, just typical varying audio.  Everyone used the same Daven pots here in the USA for both input and output.

 

[alexc]

Thanks Doug

Well it looks like I've hit a major snag - my 'found object' utc 10K:5K interstage not so happy 

Has a significant peaking HPF effect - unless I can think of something, they have to go.

I guess they are either a low inductance or are having issues with dc imbalance causing loss of inductance -  even though the voltage waveforms look to be quite close at each end of the coil.

I don't have any spec so it's a try and see thing right now. It's reasonably 'weighty' and has a 0.13W power rating, 50Hz-20KHz labelling. Model is LH-519

I'll do a final confirm by replacing with caps and if neccesary remove them - I'll do some more testing on them out of circuit. A good time to practise measuring inductance

I'll find a use for them somewhere else.

I will try the utc a15, a44, a25 I have in that position and see how it goes. It's a tight space, so unfortunately the a-series is probably my only option right now.

The a15, a25 has an 8mA imbalance tolerance and a 40-20KHz, 1W spec.
The a44 some less, probably 3mA, with a 30-20KHz spec.

I would think the dc imbalance at idle would be <1mA so I'm a little disappointed - IFF that's the problem. Maybe I just have it loaded like a dingus. Don't think so, but one never knows!

Anyhow, I'm pretty sure I will find a utc to suit.

In the meantime, it will be a good opportunity to measure 'cap coupling' method.

Such is life in the heady world of prototyping!

 

[alexc]

Yep - the 'cap coupling' instead of interstage traffo restores the freq response. 

-----

I tried reversing, loading etc - no joy. Both a massive low freq rolloff from 500Hz down AND  a rather large bandwidth bump at a couple of KHz.

And yep - the utc a15 works fine as interstage, being very close to cap coupling method in response.

Imma see what happens with the utc LH-519 as input

----

Well that works great 8). For now, I'm simply using the utc LH-519 as 5K:10K input traffo, and the utc a15 as the interstage. I can have 10K or 2K5 on the primary and 2K or 500 on secondary.

I'll go with 10K : 2K which is ah 2.24:1 voltage stepdown.

The response is about the same as before in the cap coupling case - the low freq cutoff of the a15 is the dominant thing now.

As I have found before, it is quite a bit worse than a15 spec - I'm seeing -3dB at 80Hz or so.

Totally flat to beyond 20KHz. I need to do some loading things to try improve some.

I think it was more like 55Hz or so in my pm660 with the a15 as input without any particular loading improvements.

---

It's not a big problem however - this unit is destined to follow a diy dual fender style preamp I did some time ago. It's not needing to be 'mastering quality'. ;D

One channel is the 'blonde bassman' and the other is the 'brown vibrotwin', complete with 'harmonic vibrato' circuit - the one with 5 triode stages!

ps - making that 'harmonic vibrato' work in the context of a studio rack needed the addition of a 'vox style' post vibrato filter to improve the 'vibrato punchthru' into audio. Like a 4 to 5 stage HP  8)

Each channel has an la-2a style line amp with traffo balanced out tacked on the end.

Anyway, they both have strong bass response, so no problem at all to add a couple of db to compensate the a15 rolloff.

But anyway, I would like to figure a best loading for it if I can.

---

Now to wire all that up and redo the noise and freq response tests.

If I get thru that OK, I return to figuring the gain staging to run.

Cheers

 

[alexc]

I have now made those input and output traffo changes as well as moved to the dual-rotary input attenuators of 25K similar to original.

Starting to look a little more civilized 

 

[alexc]

And the front is getting there too ...

 

[alexc]

OK  - now to resume with the basic performance measures.

Here's hoping

Some mixed  result :

-----

Chn A (Octals and furthest from PSU) - unity gain noise floor -83.3dBu : 1.0dB worse than loopback

- 50Hz -99dBu, 100Hz -96dBu and rest lower

With +10dB of gain, noise floor worsens by 7.2dB

Chn B (novals, closest to PSU) -  unity gain noise floor -78.5dBu : 5.9dB worse than loopback

- 50Hz -97dBu, 100Hz -90dBu, 150Hz -85dB, 250Hz -95dB 300Hz -95dB and rest lower

With +10dB of gain, noise floor worsens by 8.5dB

------

low freq -3dB point is 155Hz for Chn A and 123Hz for ChnB
hi freq +1dB 20KHz for both and quite flat

So that's no good at all - I need to figure something to do about that. I can try some different input, interstage impedances and I can try some loading things.

The 'cap coupled' method (4.7uF) instead of a15 as interstage was more like 85Hz -3dB point.

---

Noise profile is very stable, for both, not much wiring or ground sensitivity found yet

'Ping' test shows not much microphonics at all - a very small amount at 5KHz
(much less than the Drip 175b build)

Harmonic distortion so far looks pretty good at unity gain and with only preliminary balancing.

Chn A : 2nd Harmonic 45dB below fundamental, 3rd harmonic 45dB, 4th 85dB, 5th 73dB
Chn B : 2nd Harmonic 53dB below fundamental, 3rd harmonic 47dB, 4th 63dB, 5th 62dB

-----


So - that's the challenges : improve the lower freq -3dB point for both channels and improve Chn B noise too. Both are showing a bit too much noise-under-gain.

But it's very early days yet - I don't have any feedback so far.

I have to say I'm pleased - I expected a whole lot worse than that.

 

[alexc]

OK - so I'm once again snookered.

Looks like cap-coupling or find a better interstage with more primary inductance, suitable for use in a hiZ plate  pentode circuit

Oh well - one needs to learn these things first hand!

I guess that's why the Sowter costs around 240aud for the rca model and 175aud for the 175 model.
However, with Sowter, you get what you pay for.

---

Some rechecks :

cap coupling    gives  -3dB at 95Hz  (1uF caps) and 85Hz(4.7uF)
wsm 10K:10K  gives    "        95Hz

Bypassing the input traffo and with 4.7uF cap coupling gives -3dB 82Hz

OK - so maybe there is something else which is dominant wrt the lo freq -3dB point.


Time to review!

Hmm..

 

[DaveP]

Alex,

Another option is the RCA BA 6A method of solving the low inductance problem.  Divide your plate resistors up and feed your interstage from the 4.7k plate resistor.  So your Hi Z pentode is actually giving you a "low" output Z of 4.7k to drive the interstage.  You lose gain but you might be able to make that up elsewhere.

best
DaveP

 

[alexc]

Hi Dave - thanks for the suggest - I already have that  :'(

I'm currently using the ba6a method : pentode GR tube plate to 22K to 4K7 to B+, each side.
B+ also to interstage primary CT.

I actually jumper-shorted the 22K each side, of course saw a decrease in gain, but the lf extension marginally better only.

I'm surprised a bit - the a15 is no star in the low-freq extension, but should be capable of better than I'm getting.

I'm also suprised when I remove it and the input traffo, I still only see an extension to 85Hz or so (from 125Hz)

I have the two channels to play with and both exhibit very similar behaviour, despite different tube sets (rest same).

As a pentode vari-mu 'noob' I'm baffled!

 

[DaveP]

Have you tested what's going into the pentode from your input TX?
Might be a matching problem and your losing the low end there?
best
DaveP

 

[alexc]

Yep - the frequency response is all there at the grid and at the pentode plate pin but disappears after the 22K plate resistor.
ie. at the interstage primary / 4.7K resistor junction.

I thought it was odd - I expected to lose it *after* the interstage, but no - plate good, after 22K bad.

I messed with secondary loading, cap+resist  but all it did was tailor the hi-freq response.
Reverse of interstage no effect, different interstage impedance settings, no effect.

Thanks!

 

[DaveP]

You could try varying the loading resistance on the secondary of the interstage to reflect back a higher Z?
best
DaveP

 

[alexc]

I'll give that a shot tomorrow

At present, on the secondary, I have 22K each grid to ground, so I'll have a play there and see if anything happens.

Thanks!