matching resistors

[sbelyo]

I want to match some PRP resistors to 0.10%  so I read tangent's blog on this and I need a meter with 10,000 counts or better.  Is this worth doing or am I wasting my time?  I was wondering what everyone else's experiences were?

[mwl168]

It may help to know if the purpose of the exercise is to only match same spec'ed resistors to each other or if the absolute accuracy of the resistor value is also critical. 

[sbelyo]

It's for the TPA IVY-III output stage.  It's my understanding it's the accuracy of the resistor value is the critical part

[G600]

Another route would be to identify the critical resistor, and if they are few enough, buy texas components metal foil tailored to your value and tolerance ?

May be a plus soundwise too...

Edited May 14, 2019 by G600

[Craig Sawyers]

Good way to match to that accuracy is to make four of the resistors into Wheatstone bridge. Feed with a DC voltage that makes sense (say 10V), and check for a null between the arms. If the resistors are worst case 0.1% all the wrong way, you'll get a voltage of 0.2% of 10V across the arms, or 20mV. So you now mess with your resistors until you get a voltage of less than 20mV. Then you know they are matched to better than 0.1%

In an odd quirk of history, Wheatstone also designed musical instruments, the most famous of which is the concertina. 

[sbelyo]

6 hours ago, G600 said:

Another route would be to identify the critical resistor, and if they are few enough, buy texas components metal foil tailored to your value and tolerance ?

May be a plus soundwise too...

I looked at those briefly, anywhere from $10 - $15 each.  There's 28 total resistors needing 0.1%  the schematic is here http://www.twistedpearaudio.com/docs/linestages/ivy3_schematic_1_0.pdf   

The 0.1% resistors are

R1-R4

R5-R12

R13-R16

R17-R24, R32-R35

 

 

5 hours ago, Craig Sawyers said:

Good way to match to that accuracy is to make four of the resistors into Wheatstone bridge. Feed with a DC voltage that makes sense (say 10V), and check for a null between the arms. If the resistors are worst case 0.1% all the wrong way, you'll get a voltage of 0.2% of 10V across the arms, or 20mV. So you now mess with your resistors until you get a voltage of less than 20mV. Then you know they are matched to better than 0.1%

In an odd quirk of history, Wheatstone also designed musical instruments, the most famous of which is the concertina. 

I will look this up and give it a try Thanks

Edited May 14, 2019 by sbelyo

[randytsuch]

Just looking at the circuit, I find it hard to believe the actual values are that critical.

My hunch is since its balanced, they really want sets of resistors to be close.  For instance, I could see how R1-R4 should all be close to each other, but I don't think it would be critical they are all close to 750 ohms.

And in the end, the op amps may be what limits performance, not the resistors, but that is also just a hunch.

But because I'm anal, I will usually buy extra resistors and hand match, unless I'm using expensive exotic resistors.

 

Randy

[Pars]

I would just use some IRC/TT RC55 0.1% resistors such as 756-RC55Y-750RBI (Mouser) and be done with it.
The Texas Components resistors will run on the order of $15/ea...


Sent from my iPhone using Tapatalk

[sbelyo]

1 hour ago, randytsuch said:

Just looking at the circuit, I find it hard to believe the actual values are that critical.

My hunch is since its balanced, they really want sets of resistors to be close.  For instance, I could see how R1-R4 should all be close to each other, but I don't think it would be critical they are all close to 750 ohms.

And in the end, the op amps may be what limits performance, not the resistors, but that is also just a hunch.

But because I'm anal, I will usually buy extra resistors and hand match, unless I'm using expensive exotic resistors.

 

Randy

I'm building this to see if I can improve on the sound of the first one that I built using 1% PRP's.  The original kit uses yageo and when I swapped them for PRP the sound improved so I figure why not try to match to 0.1 to see if it makes a difference.  You're probably right but the PRP's are only $0.35 a piece so I figure it's worth a try.

39 minutes ago, Pars said:

I would just use some IRC/TT RC55 0.1% resistors such as 756-RC55Y-750RBI (Mouser) and be done with it.
The Texas Components resistors will run on the order of $15/ea...


Sent from my iPhone using Tapatalk

That was my other thought as well.  I priced out a full set from TC and it was like $382 so I'm not gonna do that route

I bought this meter since it seems to fit what tangent recommended for matching resistors and wasn't expensive  https://www.ebay.com/itm/113744202815

I'll order 10 resistors for each position to see how close I can get

[mwl168]

1 hour ago, Pars said:

I would just use some IRC/TT RC55 0.1% resistors such as 756-RC55Y-750RBI (Mouser) and be done with it.
The Texas Components resistors will run on the order of $15/ea...


Sent from my iPhone using Tapatalk

IIRC, years ago I used the IRC 0.1% resistors to build my first Dynalo. I measured each of the resistors when I received them and I was not impressed with the accuracy and the tolerance of those I received. I have since adopted the practice of spending the same (or less) money and buy large quantity of good quality 1% resistors and hand pick and match them. Personally I think I have better outcome with the latter practice.  

[Craig Sawyers]

Common mode rejection is critically dependent on resistor tolerance in a balanced amplifier like the OPA1632. I'd have to look it up but I think 0.1% resistors will mean a CMRR of -60dB. And at that level you need to make sure they thermally track too, and have very low voltage coefficient.

All that hoohah with discrete component tolerance and CMRR is why THAT have adopted everything on a chip with laser trimmed resistors and bootstrapping to make the common mode impedance huge. The InGenius devices get to -90dB CMRR, the only real disadvantages are the pretty average SNR of -105dBu because the internal impedances are in the tens of k range and 5ppm distortion.

That is where the OPA1632 with truly impressive low noise of 1.3nV/rootHz and ridiculously low harmonic distortion of 0.22ppm wins big time.

[Craig Sawyers]

Looking at the TP schematic, the noise is limited by the input resistors of 2k2. The native noise of the OPA1632 is an equivalent noise resistance of 100 ohms (ignoring 1/f noise).

Even so, with 2k2 input resistors the input circuit SNR should be ~ -120dBu, which is still pretty darned impressive.

[sbelyo]

it's a very good output stage but for me in stock form it's a tad bright.  For the SE outputs I swapped the opamp out to OPA1642.  That gave me the sound I was looking for along with amtrans filtering caps and PRP resistors.  I just want to see if building another one with 0.1% PRP's and Silmic II electrolytics makes a difference