UA175 variable mu compressor output transformer calculation

I made some calculations about the UA175 output stage and transformer, and I don't know if I got everything right, so I better ask on th eforum before I go shopping.

Problem is, I don't know what output transformer type was used, and calculating backwards gives some surprising (to me!) results.

Here's the outline of the math:

Output stage is 12BH7 pp with ca. 300V plate voltage.
Each triode has a bias current of ca. 17mA, both triodes sharing a 330R cathode resistor.

Transformer output 600R.

Calculating winding ratio / primary impedance:

Manual says level at the onset of limiting is +13dBm.
That would be 3,5Veff or 5Vpeak on the seconday.

On the primary, the diodes start to conduct at ca. 45Vp.
I guess this must be the onset limiting.

Calculating 45V/5V = 9:1 for one side, so the transformer
would be

18(c.t.) : 1

200kOhm : 600 Ohm

This looks like a rather high ratio to me - I had expected something
around 5, but little do I know.

Is the 200kOhm p-p value reasonable?

Of course this means almost no load to the triodes running at high current.
An almost horizontal load line ?!

Then the next info from th emanual is the point of clipping when limiting is turned off: +24dB.
That would be 12Veff or 17Vp on the secondary, and 153Vp (306Vpp) at *one* plate of the output tube.

Can the 12BH7 go down to 140V? Spice simulation says no, but I'm pretty sure that Spice is wrong here. The triode shoud go down that far with no problem, given that light load, shouldn't it? And the whole pp stage should even stay in class A all the time, shouldn't it?
(I really don't know what's wrong here. Spice simulated the WCF in my LA2A-based limiter quite ok. Also using 12BH7 models.)

That's what I have cobbled together so far. Please give me some confirmation or correction before I go shopping for a 18:1 transformer.

Hints where to buy such a transformer are welcome, too. (Must be 250mW - 24dBm - or higher).

Hints about the *original* transformer would also be appreciated!

JH.

For reference.

http://www.waltzingbear.com/Schematics/Urei/UA_175.htm

This version has 270R cathode resistor, but I've seen 175 schemos that have the same 330R as the 176:

http://www.waltzingbear.com/Schematics/Urei/UA_176.htm

Guess the 270R version will give a little more headroom, while putting more stress on the tubes.

But for my transformer question, not that much difference.

JH.

Would this be a good one?

http://www.lundahl.se/pdfs/datash/1689.pdf

JH.

:green: I rang Canford 20mins ago about the 1680 and 1689 for our own Vari mu which we are working on here!
I think those are the correct ones. Sowter sells a tx for an Altec 436 clone which is a 10:1 so I think we're definately in the ballpark with these! :thumb:

chef

[quote author="Swedish Chef"]:green: I rang Canford 20mins ago about the 1680 and 1689 for our own Vari mu which we are working on here!
I think those are the correct ones. Sowter sells a tx for an Altec 436 clone which is a 10:1 so I think we're definately in the ballpark with these! :thumb:

chef[/quote]


Ah, interesting!
Is your vari mu also based on the UA175 ?

JH.

[quote author="Swedish Chef"]Sowter sells a tx for an Altec 436 clone which is a 10:1 so I think we're definately in the ballpark with these! :thumb:

chef[/quote]

Well, that one is 5+5:1 (not 10+10 : 1), isn't it?
And even then it's only specified for 150R load in that case.

JH.

Normally, you would want the transformer primary P to P Z to be about 2 times the plate impedence of 5.3K or 10.6 K. then we solve for the ratio needed to bring this down to 600 ohms. Divide 10.6K/600 = 17.66. Take the square root of that to get our ratio of 4.2:1. So a 5:1 ratio would be ok. Better yet, a 4:1.

RonL

[quote author="rlaury"]Normally, you would want the transformer primary P to P Z to be about 2 times the plate impedence of 5.3K or 10.6 K. then we solve for the ratio needed to bring this down to 600 ohms. Divide 10.6K/600 = 17.66. Take the square root of that to get our ratio of 4.2:1. So a 5:1 ratio would be ok. Better yet, a 4:1.

RonL[/quote]

That's why I expected something like 5:1 (see above).

But then you will "never" reach the threshold of the rectifier (45V) with the (ac-coupled) plate voltage!

Well you will reach it sooner or later, but with a 5ct : 1 that would mean:
primary one side reaching the threshold at 45V. Vprim_eff = 32V (one side). Secondary Vsec = 32V/2.5 = 13V. That's 280mW or +25dBm at the threshold of limiting. Add 11dB headroom, and you're at +36dBm output level!

I think the goal here wasn't to get the maximum power out of the tubes. Maybe they designed such a big mismatch on purpose, to get better linearity and frequency response? Anyway, when I'd build my clone with a 5:1 transformer, that would als mean this transformer must deliver 4 Watts.

Does this make any sense? If this limit threshold level / max level info is wrong, then of course all my conclusion are wrong as well. (They might be wrong anyway. (;->) )

JH.

> Calculating 45V/5V = 9:1 for one side, so the transformer would be 18(c.t.) : 1, 200kOhm : 600 Ohm

I get 38.2V bias, 4.9V peak, and arrive at nominal 146KΩ plate-to-plate impedance.

> Is the 200kOhm p-p value reasonable?

Whether 146K or 200K: not really. I was quite startled by your result, but basically agree with your calculations.

As RonL says, conventional wisdom is 2*Rp for best clean power up to 5*Rp for low distortion at slight drop of power. 10K to 25K per side or 40K-100K nominal plate-to-plate.

However, the big load here isn't that 600Ω terminated line, but the rectifier. +13dBm, or +20dBm in heavy GR, is only 20mW to 100mW. Recall that the very fast rectifiers in boxes such as Fairchild 670 and the big GE use 10 Watt (10,000mW!) amplifiers just for the rectifier. The 175 does not have such fast goals, and saves a big buck by combining an about-1-Watt rectifier amp with the 1/10th Watt output amp.

So when not smacking the rectifier, the audio output runs really "easy": low current swing in the triode for low distortion, large impedance mismatch for good damping of the transformer and a predictable low-Z output (modified by R21 and output pad to 600Ω in most situations).

FWIW, the early H-P oscillators also used very hi-Z loading. For 1-Watt output, they used an 8-Watt pair of tubes, then loaded very high-Z for low distortion, low output Z, and better tolerance of leakage inductance.

[quote author="PRR"]> Calculating 45V/5V = 9:1 for one side, so the transformer would be 18(c.t.) : 1, 200kOhm : 600 Ohm

I get 38.2V bias, 4.9V peak, and arrive at nominal 146KΩ plate-to-plate impedance. [/quote]

There's 45V written in most variants of the 175/176 schemos, so I always wondered if the barely readable 82k in the divider actually is a 62k. But then again I wonder how many errors there are are in the schematics anyway.

> Is the 200kOhm p-p value reasonable?

Whether 146K or 200K: not really. I was quite startled by your result, but basically agree with your calculations.

[...]

However, the big load here isn't that 600Ω terminated line, but the rectifier. +13dBm, or +20dBm in heavy GR, is only 20mW to 100mW. Recall that the very fast rectifiers in boxes such as Fairchild 670 and the big GE use 10 Watt (10,000mW!) amplifiers just for the rectifier. The 175 does not have such fast goals, and saves a big buck by combining an about-1-Watt rectifier amp with the 1/10th Watt output amp.

Click to expand...

You are right! I completely forgot about the rectifier. Will only draw current during a small portion of the phase, but then it can be heavy.

So when not smacking the rectifier, the audio output runs really "easy": low current swing in the triode for low distortion, large impedance mismatch for good damping of the transformer and a predictable low-Z output (modified by R21 and output pad to 600Ω in most situations).

FWIW, the early H-P oscillators also used very hi-Z loading. For 1-Watt output, they used an 8-Watt pair of tubes, then loaded very high-Z for low distortion, low output Z, and better tolerance of leakage inductance.

Click to expand...

Thanks for this - I think now I can really go shopping for a transformer!

JH.

I won't quibble 38V or 45V. As you say, the schematic is hard to read and may not reflect reality. These things were hand built, and if they didn't work right, they messed with things until they met spec. If they got a crate of transformers that were a slightly different turn ratio, it would be easier to change the bias resistor than return the crate of transformers.

I suspect what you really want is like 20K to 60Ω, or something in the 10K to 40K range with the appropriate ratio, good for about 200VAC across the primary at your lowest desired full-power frequency. (Lectric bass only goes to 42Hz, and will usually be more overtone than fundamental, so you seldom need full power to 20Hz.) The source impedance is around 11K plate-to-plate, a spec noted on some transformer datasheets. There is no harm in putting a 600Ω load on a "60Ω" winding. It may be inefficient, but not horribly so, and is probably "part of the sound".

Anyway you are not going to find 200K windings, certainly not at this power level.

Alternatively, you could use a lower-ratio transformer and change the bias voltage to set a reasonable limiting level. But if the bias voltage is much lower, the compression ratio goes down. The ratio of bias voltage to the grid voltage needed for significant GR tells how much the output rises after hitting limiting level. Probably 40V diode bias, 10V of grid voltage for significant GR, the output level rises from 40Vpk at threshold to 50Vpk at heavy GR, and proportional on the other side of the transformer, a 2dB rise. If you set bias to 5V, and needed 10V on the grid, you have 5V to 15V rise, a 10dB rise for the same GR. If 10V on the grid causes 20dB GR, then the first case gives 20-into-2 or 10:1 compression ratio, the second case gives 20-into-10 or 2:1. (And also a much lower output level, which is why it is so hard to implement a useful compression ratio control in this type of limiter.)

[quote author="PRR"]I suspect what you really want is like 20K to 60Ω, or something in the 10K to 40K range with the appropriate ratio, good for about 200VAC across the primary at your lowest desired full-power frequency. (Lectric bass only goes to 42Hz, and will usually be more overtone than fundamental, so you seldom need full power to 20Hz.) The source impedance is around 11K plate-to-plate, a spec noted on some transformer datasheets. There is no harm in putting a 600Ω load on a "60Ω" winding. [/quote]

Ah, I see. I was hesitant to use a transformer with too low impedance, but as the source is *vastly* lower than required, the transformer still sees a low impedance when it's a nominal 20k transformer. Now that you say it, it's clear.
So this one looks fine:
http://www.lundahl.se/pdfs/datash/1689.pdf
(They even specify a source impedance, not a nominal / load impedance.)

Alternatively, you could use a lower-ratio transformer and change the bias voltage to set a reasonable limiting level. But if the bias voltage is much lower, the compression ratio goes down. The ratio of bias voltage to the grid voltage needed for significant GR tells how much the output rises after hitting limiting level. Probably 40V diode bias, 10V of grid voltage for significant GR, the output level rises from 40Vpk at threshold to 50Vpk at heavy GR, and proportional on the other side of the transformer, a 2dB rise. If you set bias to 5V, and needed 10V on the grid, you have 5V to 15V rise, a 10dB rise for the same GR. If 10V on the grid causes 20dB GR, then the first case gives 20-into-2 or 10:1 compression ratio, the second case gives 20-into-10 or 2:1. (And also a much lower output level, which is why it is so hard to implement a useful compression ratio control in this type of limiter.)

Click to expand...

Yes, this is clear. The tube that make the gain reduction, and the compression ratio pretty much define the required voltage swing.

Thanks again,

JH.

And BTW, I bought that push-pull calculator yesterday. Quite impressive!

I tried the 175 output, but the upper limit for the pp impedance in the program is 100k.

No problem: use two 12BH7's in parallel, work with the 100k setting, and divide all current and power results by two!

JH.

LL1689/PP, 9+9 : 2, Alt O, PP to Line Out seems like a good plan.

Unless you get quoted £74 per unit LL1689 like I just did from Canford... :sad:

They are expensive, but seems to be the best for this job.

Where else will you get a 50k+50k:600 = 200k:600 output PP iron?

Will the ALTEC 436 from sowter make it? Is it any cheaper?

[quote author="Swedish Chef"]Unless you get quoted £74 per unit LL1689 like I just did from Canford... :sad:[/quote]

Approx. 100 Euros. A lot of money indeed.
Do you have any cheaper options?

JH.

[quote author="rafafredd"]They are expensive, but seems to be the best for this job.

Where else will you get a 50k+50k:600 = 200k:600 output PP iron?

Will the ALTEC 436 from sowter make it? Is it any cheaper?[/quote]

As I understand it, the 436 is only 5 + 5 : 1.

JH.

[quote author="rafafredd"]Where else will you get a 50k+50k:600 = 200k:600 output PP iron?

I'm really curious what's in the original 175.
The Lundahl transformer seems to be a little oversized.

JH.

Click to expand...

I have the Sowter one already so I'll try it when I get the unit built to see what happens...
100Euro is too much for a single o/p tx!

cherf