tube preamp frequency comparison

Response plot comparison between a selection of old broadcast and recording tube preamps. 1 kHz point matched for comparison. Line level pink noise through 50 dB Shure line-to-mic pad, output loading checked to see if it made any response differences. Output loading will kill a bit of the highs with some types, no change with others. Loading clearly more complex than that once you get into max output levels and harmonic content. One dimensional measurements like this don't tell you much about sound.

debate and contrast....

I'm very surprised that none of them really made it flat past even the 8khz point! While I don't know when each of those preamps were designed (ie. what was the "flat" of their day), those curves certainly point out how far we've come from mere tape or vinyl.

Hey, 10 kHz is severely over-rated. Really. One can hardly tell anything is missing. I could show you plots of 1970's Quad Eight, Melcor and API stuff done in the same manner, and they will look the same or worse on top. Some of this old stuff looks better in a plot than newer stuff. You see what you want to see; you think it shows how far we've come, I think it proves how little has changed. Context context context.

THE IMPORTANT PART: This is a comparison of a 50 dB line pad input, which introduces greater high frequency losses due to series resistance. If I ran these same tests with a true 150 ohm source I would get entirely different plots. And even another set of different plots with mics. Etc etc etc. None of this is gospel; no such thing.

Does anyone have a schematic for a Collins 356A?
Thanks
Sorr

[quote author="emrr"]THE IMPORTANT PART: This is a comparison of a 50 dB line pad input, which introduces greater high frequency losses due to series resistance. If I ran these same tests with a true 150 ohm source I would get entirely different plots.[/quote]
Pad can of course still present 150 to next stage and most likely the pad you used does just that.

[HP-Network Analyzer-lingo]A tasty way of determining the transfer* of a certain box would of course be to be able to work with a Signal (or Send ?)-output, Test-input & Reference-input.
The Reference-input senses the actual input of the DUT (Device Under Test), so here after that 50dB pad.
Test goes to the output of the DUT.

T/R is then determined & displayed, giving the actual transfer of the DUT under the chosen circumstances. This removes any influence from cabling, padding etc.

Even better would be to be able to add a 'THRU'-test to that, in order to derive the transfer of the setup without the signal going through the DUT and then compensate the later T/R-result for that. Easy for testing unity-gain boxes; less so for many-many-dB mic-pre-gain. A known very neutral alternative mic-pre could be substituted for the THRU-test, or some tricks with calibrated padding but it'll become more cumbersome then..[/HP-Network Analyzer-lingo]

Regards,

Peter


*: and yep, FWIW indeed, but can be interesting

Yes, U pad here presents proper shunt loading resistance. You can compare 20 and 50 dB pads with identical shunts, and see differences in high frequency response, which I assume are due to series resistor effects (affecting power/current of signal?).

My audio interface measures (with the trusty old General Radio power meter) at between 150-200 ohms matching output; I can lower the pink noise signal level and drive a preamp under test directly, and get yet another response curve. A 20 dB mic pad will change the upper response of many preamps as compared to a direct feed.

A related and more extreme bit to consider is the curve samples you get when comparing a tube amp output with varying types of loads; test one as a line amp with a purely resistive load and you get a flat line, test with different speakers attached and you get differing curves that represent the speakers interaction and the resulting speaker response.

The great unknown for us hacks is the varying interactions with a pile of different mics and the same preamp.

UTC output curves are similar-RCA
Langevin bottom is great, that C core is superior to lams.
McCurdy needs more pri inductance.

I would guess that the units that vary most with different input pad resistances is going to be the transformers with the least turns.

DCR gets combined with X-L at a 90 degree angle to form the resultant, which is Z, or impedance.

This transformer DCR interacts with the pad DCR, so you will get different curves due to different impedances, but the variation will be slight, hopefully.

its always great to see demonstrations and comparisons like this. I'm actually a little amazed and shocked at the Collins amp nice range!

[quote author="CJ"]UTC output curves are similar-RCA
Langevin bottom is great, that C core is superior to lams.
McCurdy needs more pri inductance.

I would guess that the units that vary most with different input pad resistances is going to be the transformers with the least turns.

DCR gets combined with X-L at a 90 degree angle to form the resultant, which is Z, or impedance.

This transformer DCR interacts with the pad DCR, so you will get different curves due to different impedances, but the variation will be slight, hopefully.[/quote]

Good theory on the pad interaction. I could easily post input DCR's for consideration. The Langevin is the highest, followed by the RCA.

All transformers are custom part #s, rather than publicly available stuff.
Collins is ADC. - the high rise disappears without resistive shunt terminated input. On paper the Altec 458/459 does this too, but I've not looked at them yet. They had Collins built all the communications gear for the moonshots for some very good reasons.
Gates is UTC LS grade in CG cans. - the low and high bumps here are feedback tailoring, and you get more typical UTC curves without it.
Langevin is Langevin.
McCurdy is Hammond iron I believe.
Both RCA pre-date UTC branding, and are of undetermined origin. I am trying to determine if they are in-house, Ferranti, or UTC. The terminal boards are unlike any ever used by UTC. (CJ, I sent you one of this type output awhile back)