A question about trim pot values

[Category 5]

I finally did some recording with the J99s and flipped out at howw good they sound. Looking at the frequency curve (sine sweep) there is about a +1db boost starting at around 6k and ending around 35k (from memory). Very smooth and present without sounding hyped.

Anyway, the gains match almost exactly (within 1/20th of a dB) at all settings when the trim pots are down, but with both up one preamp has about 1.5 dB more gain. In fact, while one has about a 6dB variance in the trim (as expected) the other has almost 8dB.

Is this something that is to be expected due to variables in the pots? ...or is it likely I botched a resistor somewere? Until I do some more recording I won't know if it effects soujnd otherwise...but I don't think so.

1.5+ dB extra gain seems awfully out of spec for just pot variances though. I did test each resistor as I stuffed the board, and other than the 2 phantom resistors being .3% instead of the commonly suggested .1% they were al well within tolerance range.

Peculiar. I think it must come down to the trim pot though, since the gains are almost exactly matched with the pot turned all the way down. I am using 990c's.

Shane

P.S. For those of you looking for clean gain, and lots of it the J99 is the best I have ever heard. Maybe it's becasue it's the newest one I have, but it even sounds better than the A12 and N72 in a lot of ways (although I still love them all!)

[tele_player]

First thing I'd check is the actual resistance of R16 and R17.

[Category 5]

Well, I am not home right now, but I just looked the pot up on digi-key's site and the tolerance is 20%. I guess being that the pot is 1k the 1.5 dB differance I am seeing is probably in line with their tolerances. I will actually measure it when i get home.

Shane

[Category 5]

Well, I did some in circuit measurements and here's what I got.

With r17 fully clockwise I measure

r17 = 641 ohm and r16 = 619 ohm on the one with lower gain

with r17 counter clockwise I get

r17 = .3 ohm and r16 = 320.6 ohm


On the second preamp i get

With r17 fully clockwise

r17 = 704 ohm and r16 = 647 ohm

with r17 counter clockwise I get

r17 = 0 ohm and r16 = 320.8 ohm

I did measure all the resistors (except pots) when I did the building, and both r16s were well within tolerance and pretty much 1k for all intents and purposes.

I did trim the shaft of the pots so that the knobs would fit all the way on (as with the other modules). I did this carefuly with a fine hack saw so I do not think I damaged the pots in he process.

Shane

[Category 5]

Should I replace the trim pots? Is there any reason not to cut them down to fit the knobs?

Shane

[tele_player]

Have you tried swapping the op amps to see if it has any effect?

With reasonable care, I can't see trimming the shafts as a problem. I did it.

I wonder what Tim thinks on this? It's not making sense to me.

[tpryan]

A few things:

1) Make sure you are really measuring such a difference. Always be skeptical about your measurements, especially if they're dependent on some complicated software you didn't write. Using a signal generator and scope, I don't measure anything like a +1 dB boost above 6K.

2) Try a null test to see if the output cancels. If it doesn't completely cancel, try to measure the amplitude of the residual. That will be the real gain mismatch at the frequency under test.

3) Worst case, you might expect to see as much as a 1.8dB difference in gain between two J99s. You'd have to be very unlucky to see such a big difference, though it is possible.

4) You can't expect to get an accurate resistance measurement in-circuit. If you want to measure the resistance of the pot, you'll have to lift R16.

5) If you really need to match the full CW gains to within a dB, you can tack an appropriately valued fixed resistor in parallel with the pot. You'll need to get an accurate and precise value for the pot first.

6) Carefully trimming the shafts won't damage the pots.

[Category 5]

Well, like I said above, if the pots are at minimum (counter clockwise) position the gains are within 1/20th of a decibel at all positions. This is pretty much the same as with the A12s and N72s. With the pots turned up I get roughly 6dB gain increase with one J99 and 7.5 dB with the other. The pots on the A12s give me just about the same amount of gain (attenuation) between modules. In fact, surprisingly everything is ultra close (like at maximum 1/10 dB difference) except this J99 thing.

With so much gain on the J99 I doubt I'll have the pot at anything except minimum in most cases, so the discrepancy isn;t what's bothering me. It's the logic that perhaps it's just an indication of something awry elsewhere in the circuit.

I am using RME's digicheck software to test the overall gain, and it is accurate to 1/100th of a dB (even though I am skeptical) so I doubt the software is at fault. I also did a null input to output measurement to remove any error due to channel inconsistencies (which was about .03 dB, and .04 dB respectively). I am sure the software has a minimum accuracy, but it certainly isn;t off by a full 1.5 dB.

I think i will try to replace the pots, this time measuring them before I install them. If they are pretty closely matched when i install them I will start to look elsewhere for the problem. I will also try swapping op-amps.

The 990cs sure are hotter than the 2520s or SC25s when running at 24volts! I am running the SC25s at 20v, but they are only a tad hotter than the 2520 at 18v.

Shane

[tele_player]

For comparing temps, it makes sense to run them all at the same voltage. Why run any of them beyond 18v?

[PRR]

> r17 = 641 ohm
> r17 = 704 ohm

I don't know the circuit off-hand, but assuming gain is about porportional to this R17 then 704/641= 1.098 which is 10% (well within the pot's 20% tolerance) and 0.8dB of gain.

Did you measure the pot resistance before installation? Can you conveniently measure it out-of-circuit now? It would not be odd for a pot to be 20% off. If it is used as a rheostat, that doesn't matter: you are going to set it to the value you need. And in potentiometer connection, it usually does not matter much and a wise designer will make sure a large tolerance pot causes only insignificant shift in performance.

Trimming the shafts can result in gross breakage (learned that the hard way) but probably not a small shift in value.

As Tim says, you have to be ultra-careful to prove 1dB errors especially at the top of the band. That level of difference could be a well controlled transformer resonance, or it could be in your test setup. It may be flat with the amp bypassed, but loading changes when you put the amp in. Also transformer response depends a lot on source impedance, have you got a good solid mike-size source impedance? You probably need a resistor attenuator, not only to cut-down the signal generator output, but to ensure the input sees 150-200Ω. Also be sure it is well balanced: transformer response can change with stray capacitance to ground, and unbalanced is not normally how we use mike amps.

> I am using RME's digicheck software

That software is perfect enough, at the mercy of your sound card which may be very-very good, but the analog interfacing between soundcard and preamp is important. You sure can't just jack a soundcard output, even a balanced output, to a mike input and get the same conditions as a microphone source.

Anyway, a 1dB top-rise will completely vanish in mixing and mastering. I might not believe it and I sure would not sweat it.

The 1.5dB-2dB maximum difference between channels is not quite enough to bother me, and I do very minimalist plain-stereo. 2dB will give an obvious image-shift, but (unless I could afford far better stuff) there is always a few-dB shift that I fix in the mastering.

And Hmmmm... this is only a 6dB trim? Then the reason it is there is SO you can trim multiple preamps to identical gains. (It isn't for adjusting to different sources: sources vary 40dB.) Set your first unit at mid-point (or to a nice round gain-number) and then set the other(s) to the same gain as the first one. I've had to do that, in days when 10% resistors were common and everything was +/-1dB.

> 990cs sure are hotter than the 2520s or SC25s

Yes. The original 990 is designed for low impedances, meaning high currents. The 2520 is aimed at more high impedances, does not need so much current. There are good reasons for each philosophy, but nothing earth-shattering; both are good amps. (Unless you live on batteries; then a 990-type amp means a very short session.)

[PRR]

> I don't know the circuit off-hand

I found it.

Obviously you set gain with the switch, in 5dB steps. And in music, you never know your levels with even 5dB precision; no reason to have a finer trim. If you set levels to -1dBfs in rehearsal, in performance the band gets exicted and plays at +4dBfs.

My opinion, this trim should not be a front-panel control. And it shouldn't be a $0.39 internal trim-pot either. For general use, I'd just either short it out or use a 1K 1% fixed resistor. There are about five 1% tolerances in the rest of the design, so worst-case gain error will be less than +/-1dB, which most users will never notice. Then a side-note for well-equipped very-fussy users: tack a 1KΩ resistor there, measure all your channels with high precision, and swap in 950Ω or 1.05KΩ to get all channels in one studio the same.

To disagree with myself: in commercial audio product design and marketing, the price of a box is the number of knobs and jacks times the value of the brand-name. And a basic mike preamp has very few knobs to justify some rather expensive guts. Adding a knob makes it seem "worth more".

And harpsichord, pipe organ, and maxed-out guitar-amp can be consistent +/- few-dB from one take to the next.

And I see all the SC models have this trim, probably in respect of one of the "classic" designs evolved when resistors were low-precision, transistors were low-gain, and every tech had calibrated test gear (sometimes built into the console).

I don't know if our hosts figured it that way, knobs/$, or as a technically useful detail for technically fussy users. My own opinion it that it invites confusion or even trouble.

(I shudder to think if the pot-wiper lost contact in the middle of the perfect take. With a high quality pot like that, it "shoudn't happen". But Murphy never sleeps....)

[Category 5]

Wow, thanks PRR. I am planning on removing he pots to test the actual resistance out of circuit. While I did test all of the resitors prior to stuffing the boards I never did test the pots. Also, the difference in the amount of trim doesn't bother me, unless it is just an innocent byproduct of a greater problem elsewhere.

FWIW, I AM using an attenuator cable I made with a pair of 25k 1% resistors (I think they are 25k) in series with each hot and cold, with a 150r resistor across pins 2 and 3 to simulate mic impedance. I am using balanced outputs from my RME interface (+4) and I am using the same cables and channels to test the gains. This has been a very effecyive, and seemingly very accurate way of measuring gains before so I expect it to work well here also.

I also did the experiment with my function generator and some XLR->aligator clip adapters. Same deal.

The gains are indeed matched with scary precision without the pot being dialed in, but using the trim is where the error arises.

I expect it is nothing to worry about. Perhaps I just can't bear to seal the case up and rack it already, building these has been so much fun!

I will report my conslusions to this board, when i arrive at some.

Shane

[Category 5]

I remved the pots and tested them at full on. They were 1.117k and 1.330k so they WERE almost 20% apart. I installed new pots which measure 1.074 and .974 so I feel like I will fare much better. I am going to restest gains tomorrow.

There was a difference of about 19% between the first two which would account for at least a dB+ so I am pretty sure the difference was up to part tolerance.

I have a voltage offset of more than a half a volt on one of the servos, so that one seems to hunt around longer before settling at 0.

I installed AD706JN in place of both LT chips, and the one pre seems to settle quite a bit faster, the one with the voltage offset takes a while longer still. The pres sound great though, despite the fact htat the op-amps get almost too hot to touch.

If you guys think all sounds well I will seal up the case until the DI gets released. I don;t know what I'll do with myslef in the mean time. I have become a DIY addict. My mics are in trouble! I forsee cap upgrades in all the cheapos (Oktava, MXL).

Shane