AD 797 Preamp

I recently purchased a preamp board based on the AD 797 ex Ebay.
The schematic provided with this board wasn't accurate so I have re-done using LTSpice and attached as a PDF.
The first problem was that the AD 797's outputs could not be zeroed using the 2 preset pots (per channel), W1 and W2. I tried NE5534s and LME49710s both of which allowed me to zero the output. As such I thought of using the LME 49710s as U1 and the NE5534s as U2. Comments?
The input has a 50K log ALPS pot which has been 'set' to 42K (apparently 50% of rotation) as a mid-point. The 2 presets have been set to 29/21K which (in the simulation) provides the closest to zero output (but not in practice).
The power lines are decoupled close to each opamp using 3 WIMA caps ('face to face' so I can't read the values without unsoldering) and a 220uF electrolytic.
Otherwise 'peculiar' resistor markings such as R1a and R1b are because the original schematic only indicated one resistor, not 2.
There are a few peculiarities in the circuit such as the Opamps offset being on the +ve input terminal and not the negative (as I've read in books).
Given that I now have this circuit board, are there any simple mods that could be made to improve the performance or any other relevant comments?
Thanks
 

Attachments

  • Preamp #1.pdf
    144.4 KB · Views: 592
Last edited:

Monte McGuire

Member
Paid Member
2013-07-28 2:58 am
We didn't get the PDF attachment - can you try to post it here?

One overall idea is that, while the 797 has a relatively low input offset voltage, its input bias current is pretty large. When the + and - input bias currents interact with large, mis-matched source resistances, it's likely that offset voltages will be generated that are much larger than the 797's offset voltage. The fact that a gain pot setting could present a variable resistance anywhere near 10KΩ to the input terminal of a 797 makes me think that the designer had little clue about how to make a 797 work well.

Sure, the 797 can be an annoying amplifier to implement well, but it is an extremely clean amplifier with a very large output drive, and very low noise. One can't just stuff it into some random, half-baked circuit without considering that the input bias currents need to be well balanced to get all of the performance that one would expect; DC offset problems are just a symptom of the larger failure of properly coddling that all-important input pair in order to get maximum performance. Starve one side of the 797 input pair by messing with the input bias balance, and linearity will probably suffer.

Post the schematic if you can, and perhaps this circuit can be salvaged. Honestly though, given this lack of attention to important details, it doesn't seem so promising. The 797 is an amazing amplifier (I'm using them on a new project extensively, and they're gorgeous if you get all of the details right), but they're far from forgiving.

Another random thought: WIMA caps used for bypasses are a really silly idea with a 100MHz op amp. Why not a quality NP0/C0G multilayer ceramic bypass? They're extremely linear, and won't have the self-resonance problems of a huge film cap.
 
Hi Monte,
Thanks for the reply.
I've added the PDF file now editing the original message - not sure why it didn't come through initially. As initally remarked, I re-drew it using LTSpice because the one supplied didn't match the actual component values in a few cases.
I think the original PCB was designed maybe for the NE5534 and just replaced with the AD797.
I have a few caveats on the design but, being a novice, thought I'd just see what others say.
 

Monte McGuire

Member
Paid Member
2013-07-28 2:58 am
Great! Much easier to make constructive comments!

So, my 'scoldings' about input bias symmetry are probably not warranted, since the 50K trim pots could be set to compensate for the bias induced voltage offsets of each stage. To adjust each pot, you'd need to measure the DC offset at the output of U1, adjust W1 to zero the U1 output offset, then measure the offset of the output of U2 and adjust W2 to zero that DC offset.

But, one problem that remains is that VR1 is part of the DC path of the non inverting input to U1, and when you change VR1, the bias induced offset will change, preventing you from compensating for it using W1 for anything but one (actually possibly two ;-) ) VR1 settings.

To fix that, one could put a coupling cap in series with the wiper of VR1 so that this pot is not part of the DC bias circuit of U1. This will also prevent small but non-zero DC from flowing in that pot, but this might not be a big deal.

The VR1 wiper is loaded into basically 10KΩ, so the coupling cap should be somewhere around 10µF to give you a flat response over the audio band. A 10µF polypropylene film cap would be nice here, since it will also have essentially zero leakage, allowing you to trim the DC offset using W1 and W2 in a stable manner. A 10µF polypropylene will be pretty large though, and it's not clear that there's room to cram that into the board. If the VR1 pot is mounted remotely to the PCB, then it might be simple to add a series coupling cap to the wiper connection. Just make sure that the cap is not physically close to any of the amplifier stages or components, since there might be stray coupling from the circuit wiring to the capacitor foil, possibly causing stability problems.

One other point is that the second stage provides a gain of 6dB, but both stages are non-inverting, so there's no need for the second stage to undo any inversion caused by the first stage - they're both non-inverting. So, since the first stage only gives you a gain of 5 or so (~14dB), you could ditch the entire second stage, and make R3 equal to 2K2 to set the first stage gain to 20dB and be done with it. You'd get your overall 20dB gain and skip the second stage - it's not helping out much here.
 
To fix that, one could put a coupling cap in series with the wiper of VR1

The entire point of offset adjustment is to avoid coupling caps. Otoh, if one removes the superfluous second stage the offset will be much easier to zero. No need to make up for lost gain either: it exceeds any reasonable needs as is.

So why the idiotic two stage topology then? Most likely they started out by emulating the MBL with an attenuator sandwiched between gain stages then changed their impenetrable minds and were left with an extra opamp.
 

Mooly

Administrator
Paid Member
2007-09-15 8:14 am
True offset null is to be performed using pins 1 and 5 of the AD797 as detailed in the data sheet.

Injecting a correction current into either of the main inputs is not 'Offset Null'. All that is doing is introducing a variable 'fiddle factor' into things, and one that is going to introduce noise and ripple from the supply directly into the forward signal path.

The circuit design shown does not lend itself to DC precision... my advice would be to dispense with the presets and simply AC couple output and possibly the input too.
 
Hi Monte,
Thanks for your input.

I didn't like the log pot connected with the wiper to the parallel 10K resistor because it looks like a 'fake log' circuit which should use a linear pot. But worst of all the combined input resistance drops as the volume is increased.
When I measured the output voltages on both stages it was with the input shorted and the pot jumpered across. Using NE5534 or LME49710s the second preset has no effect unlike using the AD797s where it did but still couldn't get the outputs to zero.
The pot was soldered to the board but I removed it as I will mount the pot separately so I can centralise it on the front panel. Thus using a couple of caps between the pot and the input of the first opamp is possible. However 10uF poly caps are scarce in this neck of the woods - not to mention expensive. 4.7uF polyester might have to suffice.
 
Thanks Mooly.
That they use the preset topology for zeroing on the input indicates the design was not for the AD797, which as you state, has pins specifically for that purpose.
Worried me too (although I'm a beginner) for the same reasons you mention.
Perhaps it's better to AC couple only as Monte has suggested and disable the preset adjustments?
BTW, for an Ebay PCB, the PCB itself is well manufactured.
 
The entire point of offset adjustment is to avoid coupling caps. Otoh, if one removes the superfluous second stage the offset will be much easier to zero. No need to make up for lost gain either: it exceeds any reasonable needs as is.

So why the idiotic two stage topology then? Most likely they started out by emulating the MBL with an attenuator sandwiched between gain stages then changed their impenetrable minds and were left with an extra opamp.

I think putting the pot on the input may have been to get around a patent infringement of some sort. Think it might be the German "Reference Line 6010D preamplifier"
 

Hans Polak

Member
Paid Member
2005-03-17 4:25 pm
Blaricum
The circuit as shown in the diagram should be working properly.
The check with other types of op-amps proved that there is nothing wrong.
The most probable reason why you can't null is because the AD797 oscillates in the high Mhz range.
I can't see how the decoupling of the op-amps has been done, but I can assure that this is very important.
Have you tried just one op-amp in both positions, and what about nulling in that case?

Hans
 
Hi Hans,

The board came 'pre-populated' with 4 AD797's (SMD's mounted on DIP pins) and I tried each one without success - before shorting the input and after shorting the input. Best I could get to zero was around 26mV on the final output.
I then tried 2 by NE5344s on one channel and 2 by LME49710s on the other. I found that I could adjust the output of U1 in each case to zero but the second opamp any amount of adjustment made no difference (unlike the AD797s where adjusting both stages made a difference to the final output).
Eventually I decided to plug the LME's in the first stage (because of their superior noise characteristics) and the NE's in the second stage. When I did this I put in the NE's first and adjusted the second stage to give zero output then added the LME's in the first stage and adjusted to get zero on the output. I tried adjusting the second stage with all the opamps installed and again there was no difference in output (OK, maybe I tried only 5 or so turns so I could 'reset' the presets to their original positions without forgetting how many turns I made).
My guess is the output of the first stage opamps 'swamps' the adjustments made with the second stage due to the 10 Meg resistor involved.
However, having bought the PCB 'complete' from Ebay there's always a possibility that the AD797s are 'clones' or just at the maximum of their characteristics?
 

Hans Polak

Member
Paid Member
2005-03-17 4:25 pm
Blaricum
Nulling the second amp has no use at all. All nulling can be done with W1.
Remove R11 (10Meg) connected to W2 and insert just U2, no U1.
With the AD797 inserted you should see a Vout of 5mV or less.
When shortening R5 (10K), Vout should be below 0.1mV.

When this is not the case, there are several possibilities:
1) There is something wrong with the PCB, 2) the AD 797 is fake or broken down or 3) the AD 797 is heavy oscillating.

Let me know how this worked out.

Hans
 
Nulling the second amp has no use at all. All nulling can be done with W1.
Remove R11 (10Meg) connected to W2 and insert just U2, no U1.
With the AD797 inserted you should see a Vout of 5mV or less.
When shortening R5 (10K), Vout should be below 0.1mV.

When this is not the case, there are several possibilities:
1) There is something wrong with the PCB, 2) the AD 797 is fake or broken down or 3) the AD 797 is heavy oscillating.

Let me know how this worked out.

Hans

Hi Hans, thanks for your reply!

OK, did most of what you asked except shorting of R5 plus a little extra...
Removed the 2 by 10 Meg resistors (R4 and R11) to take them out of circuit.
Added one chip to stage 1 and measured output of OpAmp 1.
Added 2nd chip to stage 2 and measured combined output
Removed chip from 1st stage and measured output of OpAmp 2.
Results are:
1) AD797 1st stage: -23 mV, second: -5 mV, combined: -51 mV
2) LME49710 1st stage: +0.3 mV, second: -0.1 mV, combined +0.4 mV
3) NE5534 1st stage: -0.2 mV, second: +0.5 mV, combined -2.8 mV
4) LME49710 1st Stage, NE5534 2nd stage combined: +2.2mV

This preamp is to be used for a small project so I'm not too fussed over it - just trying to understand it. As they don't (always) supply the circuit on Ebay one doesn't know what one is getting. I expected (or rather hoped) the volume control to be sandwiched between a buffer and gain stage.
Later on I intend to build a very good preamp including RIAA stage so I might just save the AD797s for the RIAA and stick with the LME's or the LME/NE combination in the first and second stage (but using their provided offset adjustment pins ...).
I calculated the gain as 6.45 and 2 for the 2 stages but am wondering if a gain of 12 isn't too much, particularly with the high outputs of CD players?
I was considering changing the first stage to 2 also (as the second gives a lower offset at this gain)?
 

Hans Polak

Member
Paid Member
2005-03-17 4:25 pm
Blaricum
Hi,

You can count on it that you do not have a AD797, let me explain.

U2 has a gain of 2.
Contribution of input offset voltage for the AD797 can be fully neglected, since it is the low uV range.
What remains is the offset current, 80nA typical and 200nA Max.
Pin 2 sees two resistors of 2K2 each, in parallel thus 1K1.
Pin 3 sees in one direction 100R + 10R + 1K2 + 220 = 1k53 and in the other direction (without R11) 100R + 10K = 10K1
The two legs in parallel giving 1K3, almost the same load as on Pin2.
Offset voltage (without U1) at the output of U2 because of the offset current will be typically: Gain x 80nA x 1k3 = 0.2mv.
The 5mV that you measure is 25 times higher !!

My experience with the AD797 is that offset current is almost always lower than the typical specs from Analog devices.
The LME49710 however conforms very good to the specs, proving that the PCB is O.K. and that your AD797 op-amps are fake.

U1 has a gain of 6.45 as you have calculated (instead of 2 for U2) and with an open input, pin3 sees 10K//42K = approx 8K (instead of 1K3 for U2). Both factors leading to a higher offset than U2 has.
But connecting U1 to U2 leads suddenly to an almost zero impedance of pin3 of U2, changing its offset voltage, that is why a measurement with a shorted R5 also makes sense, because only then will the offset of U1 +U2 be the sum of both.

But the conclusion is simple and straight, you have fake AD797 amps.

Hans
 

kozard

Member
2012-03-26 8:08 pm
The entire point of offset adjustment is to avoid coupling caps. Otoh, if one removes the superfluous second stage the offset will be much easier to zero. No need to make up for lost gain either: it exceeds any reasonable needs as is.

So why the idiotic two stage topology then? Most likely they started out by emulating the MBL with an attenuator sandwiched between gain stages then changed their impenetrable minds and were left with an extra opamp.

I have looked at a number of the MBL6010 based clones and wonder why they have two or even three NE5534 in each channel.

For example, I have seen one with the potentiometer followed by two NE5534 gain stages and then followed by another NE5534 unity gain buffer.

I assume the correct way to do it would be NE5534 gain stage followed by the potentiometer and then finally an output buffer (or second gain stage) NE5534.



Also I assume that the offset trim is purely to avoid coupling capacitors? Any other reason? If it is purely to trim offset should it instead use the op-amp balance pins? Why would offset trim have been implemented any other way?



Finally, why not use the NE5532 with the first half of 5532 dual as the gain stage before the potentiometer and the second half of 5532 dual as either a buffer or second gain stage after the potentiometer? In this configuration I suppose the offset could be trimmed in the manner shown on the clone boards. Is there any possible imaginable degradation from having the same channel go through a dual op-amp twice?
 

Attachments

  • MBL6010 Three NE5534 Per Channel.png
    MBL6010 Three NE5534 Per Channel.png
    787.6 KB · Views: 225

Bonsai

Member
Paid Member
2003-07-25 10:44 pm
Europe
www.hifisonix.com
A 100 Ohm resistor or greater in series with the + input is a must with the AD797.

I use them in my commercial pre and had HF oscillation with Av= 5.7 using a 1k and 4.7 k feedback arrangement, non-inverting.

I had a 220 pF cap across the 4.7 k feedback resistor in my case to limit the bandwidth. After removing the cap, the stage is completely stable with no HF oscillation.

I wonder if in your case you don’t also have HF oscillation. Try removing the caps across the feedback resistors and seeing what happens.

Separately, the bias currents on this device are very high (one of the tricks to help get low noise), so it might be you don’t have enough adjustment range.

I would strongly suggest that you AC couple into the AD797. It should not be fed directly from a pot because you will get quite some DC offsets as the pot resistance increases and definitely pot scratch noise (same with most bipolar opamps like the NE5532/4 BTW).
 

kozard

Member
2012-03-26 8:08 pm
I am trying to figure out the reason to add the third NE5534 (the output buffer in each channel) in the MBL6010 based preamp. Is it there because the second voltage gain stage will perform better if it is always loaded by a high impedance (the NE5534 buffer) as opposed to a potentially low impedance?


Can the buffer be omitted (without degrading performance) when driving a JLH amplifier or something like an MX50SE or L20.5?
 

Bonsai

Member
Paid Member
2003-07-25 10:44 pm
Europe
www.hifisonix.com
Just to clarify on my post 18 above, I should have probably used only a few pF across the feedback resistor (phase lead compensation) and overcooked it with the 220pF. However, my experience is you have to be careful with this type of comp and it is not the ideal way to try to limit the amplifier bandwidth - that is better accomplished with an L pad in the non-inverting input. So, to summarize limit bandwidth using an RC pad on the input and use a [small] cap across the feedback resistor for compensation should you require it.