Original UA console - 100D preamp, EQ, 101D program amp

[Brian Roth]

Best case it works and you have much higher distortion. Worst case it burns out the transformer wiring. 10mA is a lot to ask of an ungapped transformer, do the comparisons on the expected current as designed, consider the primary wire size and the primary resistance.....ask the designer.....

It's rated 6W = 37dBM, that suggests you might get away with it but I'd rather order the right custom part or pick another design versus burning down the part. There's no provided datasheet with info for the 9600T that I've ever found. This is an Ampex 351 output replacement right? I read elsewhere in that circuit it's PP 12AU7 drawing 13mA. So you're wanting it to draw almost as much current across twice the wire resistance, before considering imbalance. If it's like the Sowter it's 1500 ohms, they tend to use big wire and have 'lower than' DCR's, that suggests the wire can take it but it's also mighty low for a non-power tube output stage. I suspect it's actually slightly gapped to account for imbalance at it's current/power levels anyway.

44dBm is 25W. That's gotta be with reduced low end and it needs a current measurement to go with it.

All the RCA stuff is direct coupled output transformers for higher signal levels with lower current requirements. Keeps consoles cooler and running longer. OP-6 remote amp is an outlier having a parallel feed approach, same result.


RANT - holy cow the number of things autocorrect wouldn't let me type in this post, and the number of others I had to find on review after posting, and correct! The robots are out of control!

Oddly, I was recently looking at the 351 schemo for a different reason ....answering a question for someone. Here it is, in all it's glory!

Bri

 

[systemtruck]

Well....change those things, it's a different preamp. No point copying the rest, really. The existing NFB and C10 is more about the EQ than anything else. You really only pull NFB from the secondary when it's a speaker.

I suppose I could give it a shot with just no NFB at all, for fun. I’d expect the current draw to jump and tax the power supply, and for maybe those poor little 12AY7’s to glow up. But if it’s stable, maybe it’ll sound terrific after some attenuation on the output. I’ve had great luck with that in another project here.

I don’t know what I’m doing regarding calculating dB loss/gain with the NFB in relation to the rest of the gain, but if it operates as a standard voltage divider, i can say that the signal going back into the input in reverse phase is about 10dB less than the signal it’s pulling from the tertiary. That comes from a 8300 / 3790 voltage divider. Then, with the modded schematic here, which brings the top of the divider way up to 24000 by adding two 12k resistors in the EQ jump path, the NFB signal should be an additional 7.5dB less. So in my pretend understanding that says the signal will jump 17.5dB without NFB. There are also a couple capacitors strapped in there and i don’t know the impact on NFB.

Anyways, maybe I’ll reach out to David at Cinemag about a custom output transformer that is gapped, and has center tap of primary, and has a tertiary winding. I’m guessing he’d just take the 9600T and add the gap and center tap. That might be helpful for anyone wanting to do 100D builds going forward. However, I’m really not sure what the turns ratio and impedance of the TT 5471 is.

 

[ruffrecords]

Very interesting stuff.
Makes me want to ask a meat-head questino…. Other than saturating the transformer and losing audio performance like low to high frequencies, what happens with SAFETY / DAMAGE when i go right ahead and toss 10mA of a HV DC current through a non-gapped 15k:600 transformer running in single ended mode? Do i burn the house down? Do i destroy the transformer? Do Tesla lightning bolts shoot my eyes out? Just wondering in case somehow it sounds good.

It is a conventionally biased class A stage. Essentially this means the power it draws over time is constant most of which will be emitted from the plates. You are unlikely to get that stage to draw 10mA unless you change its bias and then you will simply cook the tube. The only power dissipated in the transformer will be due to the plate current and the transformer primary dcr. Even if the dcr is 3K, then 10mA running through it will only generate 300mW of heat

Cheers

Ian

 

[systemtruck]

An interesting and pertinent update about the transformers for the 100D…
I spoke with David on the phone, great guy.
Firstly he thinks that 9600T can handle the up to 10mA DC current of the 100D without batting an eye. However, the ratio of just the tertiary winding of the “stock” 9600T is 26.5 : 1, so pretty small, which he says he can’t always guarantee will be strong enough to provide enough NFB for any given task.
He did make a custom 9600T for someone who wanted a much stronger NFB signal, so they made it 5.5 : 1 : 1. The secondary and tertiary are just as strong as each other. He saved this design as an actual Cinemag part# so that others can order it, and named aptly the CM-351. I’m assuming this is THE transformer that is shown in the example schematic on the Cinemag website, but i didn’t confirm that with him. So if someone wants to order that specific transformer, it’s probably the CM-351.

Now, I’m not certain that the 15K:600 impedance setup of the 9600T is best for the 100D, but i suppose i can’t hurt anything by trying. They are around twice as expensive as the 30K:600 CM-27101, which doesn’t have the tertiary. Certainly this will be an influence on using 8 of them.

I’m looking forward to getting this stuff active in test builds…

 

[emrr]

Given R5/R13 values, one could probably sort the Z range of the tertiary based on what it takes to make the EQ do what it's quoted at. 220K is pretty big also, suggesting room to play, possibly. Or simplify, and change the 220K until the gain is as it should be.

The 'rule' on NFB most followed is none or 20dB+, anything in between is noted for making additional harmonics without attenuating them sufficiently. There are outliers out there.

My biggest problem the the 100D is a cascode front end with pentode-like gain, and no interstage gain control. You will virtually always have a 20dB 'U' pad in front of this amp unless you only use ribbons with moderate volume sources.

 

[systemtruck]

Given R5/R13 values, one could probably sort the Z range of the tertiary based on what it takes to make the EQ do what it's quoted at. 220K is pretty big also, suggesting room to play, possibly. Or simplify, and change the 220K until the gain is as it should be.

The 'rule' on NFB most followed is none or 20dB+, anything in between is noted for making additional harmonics without attenuating them sufficiently. There are outliers out there.

My biggest problem the the 100D is a cascode front end with pentode-like gain, and no interstage gain control. You will virtually always have a 20dB 'U' pad in front of this amp unless you only use ribbons with moderate volume sources.

The lack of gain control is a killer. Pad would be needed, and the idea of being a good recording engineer using proper “gain staging” in sessions is not an option.
What if we took R7 470K and turned it into a pot? Feed both pins 2 and 7 with the one wiper. Does that screw up, or accomplish anything?
Or, what if R7 is left alone and a 1M dual pot is placed just before pins 2 and 7, so that both inputs to the output triodes are attenuated together.

 

[emrr]

NFB 101 - can’t put a gain pot (voltage divider) in the middle of a NFB loop without changing the NFB per knob setting. It’s self defeating. It makes the knob throw super weird.

There’s pretty much no time I ideally want pentode-like gain after a high gain input transformer, even if there’s a voltage divider pot right after it. It’s almost always too much. Transformer/triode/pot is the greatest flexibility.

 

[ruffrecords]

It should be possible to work out some basic figures from the schematic, the transformer ratios and the specification.

First, the only specification for the 100D I can find days its gain is a modest 40dB. If we assume that, on the 600 ohm winding, the input transformer has a 1:10 ratio then that accounts for 20dB of the gain. So from the grid of the first tube to the 600 ohm output we need another 20dB gain. The 30K:600 output transformer drops the level by 17dB which means the electronics needs to have 37dB of gain.

The overall gain of the system is determined by the tertiary winding ratio and the negative feedback network consisting of R2 (820) and R13 (6800) plus R8 (1500). This means the nominal gain from the grid of the first tube to the tertiary winding should be close to (6800 + 1500) / 820 which, surprise surprise, is almost exactly 20dB.

From this I conclude that the tertiary winding is almost certainly simply another 600 ohm winding.

As far as overloading the first stage is concerned, the NFB takes care of that. All you need to remember is the overall gain is 40dB which is pretty modest unless you are micing a kick drum with a sensitive condenser mic in which case you will need a pad.

Cheers

Ian

 

[systemtruck]

It should be possible to work out some basic figures from the schematic, the transformer ratios and the specification.

First, the only specification for the 100D I can find days its gain is a modest 40dB. If we assume that, on the 600 ohm winding, the input transformer has a 1:10 ratio then that accounts for 20dB of the gain. So from the grid of the first tube to the 600 ohm output we need another 20dB gain. The 30K:600 output transformer drops the level by 17dB which means the electronics needs to have 37dB of gain.

The overall gain of the system is determined by the tertiary winding ratio and the negative feedback network consisting of R2 (820) and R13 (6800) plus R8 (1500). This means the nominal gain from the grid of the first tube to the tertiary winding should be close to (6800 + 1500) / 820 which, surprise surprise, is almost exactly 20dB.

From this I conclude that the tertiary winding is almost certainly simply another 600 ohm winding.

As far as overloading the first stage is concerned, the NFB takes care of that. All you need to remember is the overall gain is 40dB which is pretty modest unless you are micing a kick drum with a sensitive condenser mic in which case you will need a pad.

Cheers

Ian

Couple ideas and questions…

1. What about, as a test to complete the circuit without having a large custom expensive transformer built, a readily available 1:1 + 1 transformer is tacked on at the end and one of those two outputs becomes the NFB?

2. And if NFB is successfully in place, what about simply adjusting that upper portion of its voltage divider for some gain control? I don’t know who did this, but this pdf has a mod for adding 8dB.. https://funkwerkes.com/web/wp-content/techdocs/MixedProAudio/UA-100D.pdf , and all they did was manipulate the voltage divider. So can’t that upper R value just keep going up or down to change gain up or down?

3. If a pot installed somewhere / anywhere within a NFB system gives bad performance behavior… be it at that output tube entry point or be it at that NFB divider network.. is it just the taper that is then bad? Can a few customized steps on a switch give some coarse gain adjustment options?

4. Lastly, I should be able answer this just based on what you’re saying @ruffrecords , but if the NFB is deleted, what is the gain of total circuit including all transformers? 60dB now instead of 40dB?

 

[C.B. - Boudio]

3. If a pot installed somewhere / anywhere within a NFB system gives bad performance behavior… be it at that output tube entry point or be it at that NFB divider network.. is it just the taper that is then bad? Can a few customized steps on a switch give some coarse gain adjustment options?

Wasn't the NFB in the V78 added by a pot (pot as a cathode resistor, wiper connected to NFB)?

 

[emrr]

As far as overloading the first stage is concerned, the NFB takes care of that. All you need to remember is the overall gain is 40dB which is pretty modest unless you are micing a kick drum with a sensitive condenser mic in which case you will need a pad.

Sorry, I’ve been seeing R8 as R5.

As someone who’s run 40dB fixed gain preamps in a studio setting for years, no to the above. 90% of the time you’ll have a 20dB pad in front making them 20dB preamps, and you’ll wish for 25-30dB, which you’ll get from compressor make-up gain. Kick with a low output dynamic will need a pad. A loud singer may need a pad, even with a ribbon. Sounds fantastical, but it’s true. A U67 into an RCA BA-2 needs a 20dB pad on the average singer, and the gain pot is in the 90% range. Several Avett Bros records I made with Collins 356A’s needed pads on acoustic guitar and banjo with M160 ribbon mics. At that time my converters were +18 input and you’d be clipping them without one, note +18 is also the rated output of a 40dB 356A.

In comparison, even an API pre needs an input pad on kick drum with a dynamic mic most of the time - also big step-up gain in the transformer.

Yes, many things use a rheostat for NFB gain variance. At some point you usually find where response falls apart and have to set limits. You will tonal variance with NFB change; it’s not the same amp at any setting. What’s acceptable?

If the tertiary is same as secondary, it’s possible it’s a split 600 being used as 2 windings and spec’d by the console load resistances. You’d have to prove this ‘misuse’ worked ok for a given transformer.

 

[ruffrecords]

Couple ideas and questions…

1. What about, as a test to complete the circuit without having a large custom expensive transformer built, a readily available 1:1 + 1 transformer is tacked on at the end and one of those two outputs becomes the NFB?

That should work.

2. And if NFB is successfully in place, what about simply adjusting that upper portion of its voltage divider for some gain control? I don’t know who did this, but this pdf has a mod for adding 8dB.. https://funkwerkes.com/web/wp-content/techdocs/MixedProAudio/UA-100D.pdf , and all they did was manipulate the voltage divider. So can’t that upper R value just keep going up or down to change gain up or down?

That is precisely what the add on EQ does but in a frequency selective way

3. If a pot installed somewhere / anywhere within a NFB system gives bad performance behaviour… be it at that output tube entry point or be it at that NFB divider network.. is it just the taper that is then bad? Can a few customized steps on a switch give some coarse gain adjustment options?

The problem with negative feedback is stability. Certain components and parasitics in the circuit conspire to turn it into positive feedback either at very high or very low frequencies. As long as the gain in the loop is less than 1 at these frequencies the circuit will be stable. Changing the negative feedback changes the gain at these frequencies and can ultimately make the amp unstable. It is relatively straightforward to ensure stability over a small range of gains so a switched coarse gain control is quite possible.

4. Lastly, I should be able answer this just based on what you’re saying @ruffrecords , but if the NFB is deleted, what is the gain of total circuit including all transformers? 60dB now instead of 40dB?

Good question and the answer is the circuit needs further analysis to determine that. As a rough guess the first stage almost certainly has a gain in excess of 40dB. The output stage has local NFB which probably limits its gain to no more than 20dB. So with the 20dB of the input transformer and the 17dB loss in the output transformer we get 63dB total without NFB. It may be more than that but it is unlikely to be less.

Cheers

Ian

 

[systemtruck]

Great stuff. Really great.

Another thought for gain control earlier in the circuit…
I’ve noticed some Collins and Fairchild designs use a pot immediately after the input transformer and before hitting the grid of first tube.
I believe that would be isolated from the NFB completely.
It’s not quite a good spot for noise control since in the theory the transformer is perfectly quiet and the noise is all after that, but still it’s a fine input control that isn’t a cumbersome pad on mic.
Seems like that could work smoothly?

But stepping backwards and planning this test build, my plan is to just make the 100D have no NFB. Use the 30K:600 transformer I feel confident in, and let it rip. I’m looking for tube tone anyways, not perfect sound. Who knows.

Then I’ll work backwards. If it sounds like garbage, or is unreasonably noisy, I’ll try adding NFB via a good splitter output transformer.

BUT, if it sounds good already without NFB, I suppose that opens up all options for putting a pot, or two, in the midst of the circuit and have good gain staging.

 

[ruffrecords]

A gain control before the first tube is perfectly feasible. You just need to ensure the leads to and from it are properly screened to prevent interference getting in and also are as short as possible to minimise HF losses. Sounds easy but in practice it can be difficult to achieve.

One thing to remember with no NFB is that the high gain means the noise level at the output will increase by an amount equal to the increase in gain (probably at least 20dB). This may or may not be acceptable depending on what you are recording..

I highly recommend you build it and play with it. That is the only way you will learn the importance of good layout and screening. With a mic pre, implementation is as important as design.

Cheers

Ian

 

[emrr]

If you want zero loop NFB, build a BA-2. Use triode connected 5879’s if octal and grid caps are a concern. You could even do it with a 12AU7. All that strikes me as more reasonable than building an esoterically designed feedback based amp w/o the feedback.

 

[systemtruck]

If you want zero loop NFB, build a BA-2. Use triode connected 5879’s if octal and grid caps are a concern. You could even do it with a 12AU7. All that strikes me as more reasonable than building an esoterically designed feedback based amp w/o the feedback.

Shoot, a BA-2C looks pretty good. Seems like a super low part count too.
I happen to have 4 good RCA 12AU7’s.
What’s the adjustment to use those instead of 1620’s? And the sonic / performance change?

Also from poking around build threads of it, it seems like I could get away with using my 30k:600 output transformers.