I noticed that some of my 1% metal film resistors are very far out. I calculated them and they fall under 1% but it is very close to the limits.
How accurate do I need to have them? Some of them are precisely the value they should be. I suppose if Dr. Leach specified 1% values then using a resistor which are 1% out is ok? right?
How accurate do I need to have them? Some of them are precisely the value they should be. I suppose if Dr. Leach specified 1% values then using a resistor which are 1% out is ok? right?
SY said:
before I address Sy's comments -- if you are truely going to measure resistance you need a meter with 4-wire capability -- the auxiliary leads compensate for the resistance in the leads and clips. this doesn't have to be very expensive if you look around for Fluke and Keithley, the Genrad RLC meters also test very acurately -- the HP meter Circlotron refers to is probably a HP3478 or HP3456 (or in that family). The HP3456 has Guildline resistors in the current source for very acurate, lab-type measurement.
1% is "OK" -- but you should be made aware that in cascaded filter designs the requirements for higher tolerances goes up, much more so for Sallen Key than MFB -- you can see this demonstrated with either the FilterPro software from TI, or Analog Devices interactive java applet -- they both allow you the option of seeing your results given varying component tolerances. Of course they assume that you are getting making a whole bunch of filters and not testing the individual components. In this case the manufacturer is supplying some statistical parameters -- either a histogram or a mean, standard deviation table -- so that the filter designer knows what tolerances they can spec.
Capacitors are going to be more problematic than resistors -- this is where a good bridge is helpful (and for $50 bucks, well that's the price of a decent Bordeaux so what the heck.)
btw, a couple decades ago TRW and GE made capacitors with 0.02% tolerance -- you can still find these in surplus shops.
Wagener said:
Yes, using a resistor within the specified tolerance is fine, everywhere. There are no resistors in this design that demand higher than 1% precision. There are a couple of cases where I might suggest matching even the 1% resistors (see below).
I suspect it is easier to just spec 1% resistors everywhere than to minimize cost by using the more typical 5% carbon film resistors. Remember this design is aimed at neophytes, so he tries to keep it simple. You probably only need 1% metal film resistors in a few places, some for precision and some for noise reduction.
First, the 300 ohm emitter resistors should be "matched", using the same meter, as close as you can get them. That is, try to find two that measure the same value, say 301.2 ohms, for each pair. The absolute accuracy does not matter, since it is swamped by the beta differences of the NPN and PNP types. This will help to minimise distortion and is significant.
It is also a good idea to match the 22k input bias resistor with the sum of the two 11K feedback resistors, to equalize input current to the pairs. I suspect this is less important, since again the differences may be swamped by beta differences between types.
I think the rest of the amp can be safely built with 5% resistors, but 1% is a safe overkill, not too expensive.
The meter was a HP34401A. It does have 4-wire resistance measurement facility but I wasn't using it at the time. It also has a null button for zeroing out the ~0.038 ohms (YMMV) of the test leads. It not mine - it's a work one.jackinnj said:
An externally hosted image should be here but it was not working when we last tested it.
Wagener said:
In most cases, it is not the tolerance that matters, but the matching value between certain pair of resistors. (So you know which pairs should go to the other channel/PCB)
So if you get 4 resistors that ALL fall at exactly +10% that is probably okay, especially if the 4 are used to divide the current in the differential stage.
Circlotron said:
The meter was a HP34401A. It does have 4-wire resistance measurement facility but I wasn't using it at the time. It also has a null button for zeroing out the ~0.038 ohms (YMMV) of the test leads. It not mine - it's a work one.
An externally hosted image should be here but it was not working when we last tested it.
the HP34401A sells for around $500 on EBay.
the HP3478A and HP3468A will sell anywhere from $70 to $200. Of these the 3478A has an additional digit. the advantage of the 3468A is that it connects to the HP-IL loop in case you want to connect it to a HP Calculator.
The 3478A is also a very good RMS meter -- while it uses the AD536 chip for true RMS it is almost as acurate as my thermal RMS reading meter. The bottom meter shown below should be reading current -- I was using it to match some transistors yesterday:
Attachments
The tolerance required in a particular amp (or other device) can span the range from very critical to surprisingly uncritical. Sometimes it is the ratio between two resistors rather than the absolute value that matters. Where you hace an RC network and are using caps with a 20% tolerance, it seems dubious that either 1% or even 5% resistors are critical.
On the other hand, in differential inputs and current mirrors (though I don't think the Leach uses CMs), matching better than 1% can be useful. I think Self mentioned that even 1% de3generation resistors could have a difference of up to 2% which was enough to degrade performance.
All that said, I have on occasion, deliberately use only 5% resistors and selected transistors for CMs and differential pairs without any attempt at matching, just to see what one could get away with. The results were a functional amp. It probably just didn't perform quite as well as one might like. I say probably, because you probably couldn't hear anything wrong unless you did a very well controlled bind listening comparrisson.
Despite this, I suggest you follow the advice for the Leach and even match some resistors closer if possible. It might be more sensative to exact values somewhere and you will feel better afterwards. This last is not intended to be silly, for amps I intend to listen to and not just expiriment with, I use all 1% resistors, match CM and differential pairs, and even try to sort caps to a better tolerance than claimed. Then, at least I know that there is now issue with component tlerance. Besides it feels good - like driving on Pirelli's even if you are only going to buy groceries!
On the other hand, in differential inputs and current mirrors (though I don't think the Leach uses CMs), matching better than 1% can be useful. I think Self mentioned that even 1% de3generation resistors could have a difference of up to 2% which was enough to degrade performance.
All that said, I have on occasion, deliberately use only 5% resistors and selected transistors for CMs and differential pairs without any attempt at matching, just to see what one could get away with. The results were a functional amp. It probably just didn't perform quite as well as one might like. I say probably, because you probably couldn't hear anything wrong unless you did a very well controlled bind listening comparrisson.
Despite this, I suggest you follow the advice for the Leach and even match some resistors closer if possible. It might be more sensative to exact values somewhere and you will feel better afterwards. This last is not intended to be silly, for amps I intend to listen to and not just expiriment with, I use all 1% resistors, match CM and differential pairs, and even try to sort caps to a better tolerance than claimed. Then, at least I know that there is now issue with component tlerance. Besides it feels good - like driving on Pirelli's even if you are only going to buy groceries!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.