Beefy output stage for a DAC

[Biasrocks]

I just recently purchased a DAC for use with my outboard gear and my DAW.

It was designed to drive modest loads and seems to be working fine into modern inputs. When it comes to driving some real 600 ohm loads like my LA2A or 1176's it's loosing almost 10db of level into that load.

I've sketched the output section, and have modified the gain to get the levels up by putting around 2K in the feedback loop. Other than that, it's stock. The opamp is an OP275 and the power supply is +/-12v.

The datasheet indicates that it will drive 600 ohm loads out to 10V RMS, but it's definitely not coming close in this application.

http://www.analog.com/static/imported-files/data_sheets/OP275.pdf

Any ideas on how I can modify this to drive 600 ohm loads?

Mark

 

[ruairioflaherty]

Hey Mark,

You don't want to listen to me but the values of R76/77 and R79/80 seem very high to me.  It seems to me that if you scaled those back and brought the cap values up to compensate you'd have a far lower source impedance. Are you sure they're 1K?

Waiting for an expert..... 

Cheers,
Ruairi

 

[Biasrocks]

You may be on to something. I've noticed that other output circuits between the opamp and XLR are much simpler, going into maybe 100R and then off into the world.

All the resistors are marked 1001. SMD code for 1K, verified by measuring as well.

I noticed I didn't put a value on C97, I believe it's 100nf. Measured in circuit, so that could be off.

Would I cause any potential problems stripping out R76/77/C97 and changing R79/80 to 100R?

Is the R76/77/C97 network there to prevent oscillations or some other function?

Mark

 

[Andy Peters]

I just recently purchased a DAC for use with my outboard gear and my DAW.

It was designed to drive modest loads and seems to be working fine into modern inputs. When it comes to driving some real 600 ohm loads like my LA2A or 1176's it's loosing almost 10db of level into that load.

I've sketched the output section, and have modified the gain to get the levels up by putting around 2K in the feedback loop. Other than that, it's stock. The opamp is an OP275 and the power supply is +/-12v.

The datasheet indicates that it will drive 600 ohm loads out to 10V RMS, but it's definitely not coming close in this application.

http://www.analog.com/static/imported-files/data_sheets/OP275.pdf

Any ideas on how I can modify this to drive 600 ohm loads?

Mark

For each side of the output (hot and cold), the two 1k resistors in series give you a source impedance of 2k, which is way too high, especially if it's driving a true 600-ohm input impedance. It's nothing more complicated than voltage divider action. The 1nF caps hanging on the XLR output are also too big, so combined with the 2k you've got a nice low-pass filter.

It looks like that circuit might have been taken from an ADC input driver. It's a weird output/line-driver.

With OP275s driving, replace R76 and R77 with something like 24.9-ohm or 49.9-ohm resistors, replace R79 and R80 with jumpers or zero-ohm resistors, and replace C97, C100 and C101 with something on the order of 100 pF. Since C100 and C101 are for EMI mitigation, the pins that connect to GND really should connect to the XLR Pin 1 which is chassis ground, not signal reference.

-a

 

[Biasrocks]

Andy,

Wow, is all I can say. I stripped the parts you suggested and replaced with 24.9 resistors. Major difference, it can now drive anything in the studio with ease and sounds 100% better IMO. It's like night and day. I also verified the capacitors value after removing the resistors, it looks like they are 300pf not 100nf as I measured in circuit, I removed them completely. I'll replace them with 100pf at some point, but so far no issues with RF.

Not sure why they chose to saddle the output stage that way, it just choked the life out of those OP275's. The only thing I can think is that this is aimed more at the consumer side of things and they don't need to drive anything lower than a 10K input. Not sure how that works in Stereophile land.

Regards,
Mark

 

[abbey road d enfer]

Not sure why they chose to saddle the output stage that way, it just choked the life out of those OP275's. The only thing I can think is that this is aimed more at the consumer side of things and they don't need to drive anything lower than a 10K input. Not sure how that works in Stereophile land.

Regards,
Mark

The reason is that they form a 2nd-order LPF that removes the sampling frequency residues and aliases that are left at the output of the DAC's. Generally there is a balanced buffer after it, but there, it's a perfect example of cutting corners.
The presence of these residues now that you have eliminated the filter may or may not create problems, depending on bandwidth, EMI protection, linearity...

 

[JohnRoberts]

The LF412 will not be very happy driving 600 ohm loads... Did you also swap opamps like Andy suggested?

Also their original output circuit includes a one pole LPF perhaps to catch HF artifacts above the opamp's gain bandwidth.

You might be better served replacing the 1k resistors with 100 ohm resistors and scaling up the C to keep that passive output filter in place. 

JR

 

[ruairioflaherty]

The LF412 will not be very happy driving 600 ohm loads... Did you also swap opamps like Andy suggested?

Also their original output circuit includes a one pole LPF perhaps to catch HF artifacts above the opamp's gain bandwidth.

You might be better served replacing the 1k resistors with 100 ohm resistors and scaling up the C to keep that passive output filter in place. 

JR

I think the output driver is an OP275, I'm not sure if that was what came stock or Mark modded it.  I'm not sure why the schemo is marked with the LF412.

I'd be inclined to keep the filter and like you my first instinct was to scale down the resistors to 100R and multiply the caps by a factor of ten but it seems that many rookie mistakes were made in the design of this so I wouldn't assume that the value was ever correct.  So with that in mind what cutoff corner frequency should we be aiming for here assuming that this DAC is 192KHz capable and so has a bandwidth up to 96KHz?

Also Mark, I'm guessing that you can revert to 1K in the feedback loop now that you're getting better voltage transfer?

Cheers,
Ruairi

 

[Biasrocks]

Still needed an extra boost in the output stage, ended up with a value of 1K4 which got it very close to my reference for +4db, about 2/10's of a volt over it but close enough for contact karate.

Yes, the output is/was an OP275. It's not in the schematic because I didn't have a symbol for it. Sorry about the confusion.

So with the 49.9 24.9 ohm output resistors, what would be a good filter value?

Mark

 

[JohnRoberts]

It would be useful to look and see what kind of HF content is there to filter out.

While modern opamps are better for HF gain bandwidth, at high enough frequency the active filters may run out of gas, so a passive filter will always get it done.

Everything in engineering is a simultaneous equation with multiple parts, and while making the series output resistors smaller improves drive capability, it works against you for decoupling output capacitance from the opamp to insure stability, and in the case of the passive pole, if high edge rate signals are present, the scaled up larger cap can draw significant instantaneous currents.

So the general tradeoff is larger resistor values, help 2 out of three and hurt one. Clearly two 1Ks in series is too high, for a studio with one foot still in the old school, 600 ohm world (why! rhetorical.. don't answer).

I am not a fan of 600 ohm terminations without some practical, functional reason, but i will try not to veer off into a different discussion.

The simple math for loss is ohms law divider so R source impedance will be divided down into 600 ohm input impedance. 2K source was losing a bunch o' dBs, 24 ohm ea, or 50 total, < 1dB loss.

regarding a practical pole frequency, since this is in series with the poles across the feedback path, these will combine for an over damped response (starts rolling off early), so needs to be tuned well above the audio passband. Tuning it high will help keep cap value modest. In general I would be pretty aggressive here with a LPF, but this will combine with the rest of your path, so if there is already marginal bandwidth you don't want to over do it.

So answer is it depends.

JR

PS: For some 600 ohm terminations, they actually expect a nominal source impedance in the hundreds of ohms, not tens or Ks, but I don't know the inner workings of your 600 ohm gear. Nor am I the right person to vet the optimal interface design, since I avoid such things.

 

[Biasrocks]

Hi John

Unfortunately, no scope here. It's on the list. I suppose I could hook up some analyzer software and see if that tells me anything, but I suspect the hash would be above the sampling frequency in most cases. I'm not hearing anything out of the ordinary such as hash or ringing, but I suspect the only way to be sure is to hook up a scope and have a look.

I've got a lot of old school gear that has transformer based 600 ohm inputs that need to be driven properly, LA2, 1176's, Lang PEQ2, etc. So there's no way, nor desire to convert this equipment to lighter loads. I suspect it's part of the sound. ;0)

Mark

 

[JohnRoberts]

If it sounds good, it usually is good.

Perhaps just assume their initial design was right, wrt to LPF pole tuning, and just scale that.

Regarding output resistor values, sourcing a good characteristic (NPO?) cap to fit there may be the hard part. Using closer to 100 ohm resistors should make the cap smaller and easier to find/fit.

JR

 

[Biasrocks]

Thanks John,

I'm a bit week on the math of filters, could you suggest a resource that I can look at?

I ended up measuring the cap values out of circuit and they were 300pf, not 1nf as I stated earlier.

Does that make sense for a LPF value jumping off point?

I've included the updated schematic of the original circuit.

Mark

 

[Biasrocks]

And the modified version.

 

[Biasrocks]

Some tests on the modified DAC.

Frequency Response @ 96kHz sampling rate

 

[Biasrocks]

Spectrum Analysis

 

[Biasrocks]

Impulse response

 

[ruairioflaherty]

Hey Mark,

What's the DAC chip?  You might want to look at the datasheet to see what the suggested active and passive filter values are.

Cheers,
Ruairi

 

[abbey road d enfer]

The original schemo shows three poles:
one defined by R70/C94 and R74/C95
two defined by R76/R77/C97 and R79/80/C100/101
After mod, there's only two poles left:
the 1st one (unchanged) and a second one that would bedefined by R76/77 and an hypothetic cap placed across pins 2 and 3 of the XLR.
So scaling the cap would not re-create the original response.

In order to get it, the proper mod should include increasing the output gain of the opamps by increasing R70/74 and decreasing accordingly C94/95.
Then R76/77/79/80 should be 24.9R (for a total output Z of ca. 100R) and C94/95/100/101 scaled to 15n.
But would it be optimum? I don't know.
Ruairi is right: what is the manufacturer's recommended implementation?

 

[Biasrocks]

The DAC chip is an AD1955.

http://www.analog.com/static/imported-files/data_sheets/AD1955.pdf

It looks like the output stage filter is a copy of the datasheet implementation, althought there are some differences in the opamp choice, feedback loop and how they are biasing the opamps.

Datasheet says -3db @ 100kHz.