Not quite audio, but we all have to calibrate things, or at least verify calibration, and having an accurate 0.1X and 0.01X divider goes a long way towards that effort. A KVD is nice, but out of range for many people. Here's the first draft of a new page that tells you how to build a Hamon resistive divider good to a few PPM if you use good resistors.
A Hamon Resistive Divider
Comments welcome-
Conrad
A Hamon Resistive Divider
Comments welcome-
Conrad
That's more cells than I want to contend with! At least the Hamon resistor in its simpler forms can be calculated with about a dozen cells. Not sure the configuration has any level control applications as I've always been able to do plenty accurate ratios with just a cheap DVM. OTOH, for calibrating the ranges of that cheap DVM, the Hamon is the way to go.
Conrad
Conrad
Okay, got a chance to read your web page and will reread with updates. I haven't determined how I can apply this yet, but it's interesting. Plan to read your published articles . . . perhaps this weekend.
-
Actually, have a pair of LTZ1000s and Vishay bulk foil resistors/trimmers (AD797s, MAT0x's) that could be used to build a nice mini-lab. However, only have inexpensive handheld DVMs. This could be interesting.
.
-
Actually, have a pair of LTZ1000s and Vishay bulk foil resistors/trimmers (AD797s, MAT0x's) that could be used to build a nice mini-lab. However, only have inexpensive handheld DVMs. This could be interesting.
.
Last edited:
The LTZ1000 is excellent but IMO it's hard to determine if the associated circuitry is up to the task unless you have something else very good to compare it too, possibly an ovenized reference bank of at least 3 standards. As far as the Hamon resistor, I used very good wire wound resistors, but my values are low enough that there's a few PPM drift if I put more than a couple volts on it. If/when I build another, I'll probably go for about 5X the impedance. The Vishay bulk metal foil resistors are better than anything out there, so they'd be the logical choice.
Conrad
Conrad
Darn, I'd like to use the LTZs in something. I printed the page for the main circuit (Fig. 2) and look forward to reading more. I'm trying to think of a way to use this to tune the selection for resistors for the stepped attenuators. What that is, hasn't fully struck me . . . lol. What kinds of things do you use a mini-lab for? There must be a few non-obvious uses.
Remember that quite a few years have passed since I wrote that. At the time I compared the reference to an ovenized standard cell bank for about a year. The only way to know what the long term drift is is to wait for it! The chip I used isn't made in a metal can anymore and the performance of plastic isn't nearly as good. I'm not sure what I'd use today, but usually if you take advantage of nulling and adjustment pins, the performance is degraded. Simplicity is usually your friend, so I wouldn't use any trims or padding. I want to know the exact value of a reference but it's not that important it be anything specific. Same thing with resistance and capacitance standards. They need to be rock stable but I don't care if they're some round number.
Today I'd probably spring for the LTZ1000 and just take super care with the passive parts. If feeling a bit poorer I'd try to scrounge up an LM399, as that has served very well in many decent instruments.
My boards were serviceable but not great works of art as they were hand made. Best bet is to roll your own.
CH
Today I'd probably spring for the LTZ1000 and just take super care with the passive parts. If feeling a bit poorer I'd try to scrounge up an LM399, as that has served very well in many decent instruments.
My boards were serviceable but not great works of art as they were hand made. Best bet is to roll your own.
CH
The first figure on the link How to Build a Hamon Resistor Divider Network
Resistance in parallel (R1*R2*R3)/(R1+R2+R3)??? No?
It's (1/R1+1/R2+1/R3+...)^-1
Resistance in parallel (R1*R2*R3)/(R1+R2+R3)??? No?
It's (1/R1+1/R2+1/R3+...)^-1
The Hamon "trick" referred to by Conrad Hoffman is indeed powerful.
If one uses +-1% resistors the total spread from lowest to highest "in range" is 1part in 50.
To find the divider ratio accuracy one simply squares that: 50² = 2500
The 0.1 divider formed from 1% resistors will have an accuracy better than 1 part in 2500 (see Finally)
Using +-0.1% resistors and squaring the worst case spread of resistance we end up with a divider than is better than 1 part in 250,000 or <4ppm
If one uses a 2000count DMM to select resistors to 0.05% i.e. a total range of 1part in 1000, the divider accuracy comes out at better than 1ppm.
At this level of accuracy selected resistors with a tolerance of 0.05% we should also look at other effects than just resistance tolerance.
Three of the biggest effects on resistance are temperature variation, age stability and soldering heat variation.
These are usually quoted in the low ppm
eg. tempco is usually specified at 5ppm/C to 100ppm/C this indicates that keeping temperature close to the ambient when resistor selection was occurring will give lower errors.
age variation is also in the low ppm. This will require recalibration of the parallel:series periodically, maybe as often as 6monthly but provided divider accuracy of 1ppm is good enough then annual recalibration should do.
soldering temperature effects are unknown to me even though many resistor manufacturer have some specification numbers in the low ppm.
Taking all three of these effects into account for a 0.05% resistor requires that they add up to less than 1000ppm of range i.e. +-500ppm.
This would indicate that temperature change must be minimised and in particular that the tempco of the individual resistors should be close for the batch from which they are chosen. Same manufacturer, same type, same power dissipation, same ambient conditions will all help in maintaining the RATIO of the resistors.
Periodic recalibration should take care of age stability. This requires that the wire ends are not soldered. Some interchangable mechanical connection of consistent low resistance would seem to offer an advantage.
Finally, the squaring of the tolerance range gives a good indication of the divider ratio tolerance.
When calculated one finds that the worst case tolerance never exceeds ~2/3rds of the squaring prediction.
i.e. using 0.1% resistors (including ALL the tolerances and variations) gives a worst case divider ratio of <0.6ppm
If one were selecting resistors that when cold are within 0.01% (1 part in 10000) then divider ratio tolerance could easily be < 0.1ppm, if one can "manage" temperature variations.
If one uses +-1% resistors the total spread from lowest to highest "in range" is 1part in 50.
To find the divider ratio accuracy one simply squares that: 50² = 2500
The 0.1 divider formed from 1% resistors will have an accuracy better than 1 part in 2500 (see Finally)
Using +-0.1% resistors and squaring the worst case spread of resistance we end up with a divider than is better than 1 part in 250,000 or <4ppm
If one uses a 2000count DMM to select resistors to 0.05% i.e. a total range of 1part in 1000, the divider accuracy comes out at better than 1ppm.
At this level of accuracy selected resistors with a tolerance of 0.05% we should also look at other effects than just resistance tolerance.
Three of the biggest effects on resistance are temperature variation, age stability and soldering heat variation.
These are usually quoted in the low ppm
eg. tempco is usually specified at 5ppm/C to 100ppm/C this indicates that keeping temperature close to the ambient when resistor selection was occurring will give lower errors.
age variation is also in the low ppm. This will require recalibration of the parallel:series periodically, maybe as often as 6monthly but provided divider accuracy of 1ppm is good enough then annual recalibration should do.
soldering temperature effects are unknown to me even though many resistor manufacturer have some specification numbers in the low ppm.
Taking all three of these effects into account for a 0.05% resistor requires that they add up to less than 1000ppm of range i.e. +-500ppm.
This would indicate that temperature change must be minimised and in particular that the tempco of the individual resistors should be close for the batch from which they are chosen. Same manufacturer, same type, same power dissipation, same ambient conditions will all help in maintaining the RATIO of the resistors.
Periodic recalibration should take care of age stability. This requires that the wire ends are not soldered. Some interchangable mechanical connection of consistent low resistance would seem to offer an advantage.
Finally, the squaring of the tolerance range gives a good indication of the divider ratio tolerance.
When calculated one finds that the worst case tolerance never exceeds ~2/3rds of the squaring prediction.
i.e. using 0.1% resistors (including ALL the tolerances and variations) gives a worst case divider ratio of <0.6ppm
If one were selecting resistors that when cold are within 0.01% (1 part in 10000) then divider ratio tolerance could easily be < 0.1ppm, if one can "manage" temperature variations.
Last edited:
If you are working with AC signals (especially lower voltage AC), a metrology grade ratio transformer is hard to beat. Stability over time is more or less the most solid you can get. This is because the output stability is based on a physical constant (the ratio of turns in the transformer). This makes it more or less impervious to drift and temperature fluctuations. Models with ratio accuracies up to 0.0001% (1ppm) are common, though 0.001% is more common. You can even find interesting models with dual back to back ratio transformers (allowing you to set a ratio of the input and feed it to the output ratio transformer).
The biggest drawback with these is there are two separate max voltages that you have to be aware of. The insulation max voltage (usually 350V or 1kV) and the maximum voltage determined by frequency (common ratio are 0.35F or 2F, where F is the frequency). Whichever is lower is the one that applies. I'm not sure why this is but if you are working with low frequency AC, then it can make the max voltage lower than anticipated.
Most of the ratio transformers are made by Gertsch or one of the subsequent companies that acquired them (Singer, Teagam, ESI, IET Labs, etc.) and sold under a large variety of model names. As with most specialized test equipment companies, it was big fish eat little fish as a combination of the Cold War winding down and the analog to digital transition. As an aside, Ratio Transformers were one thing the soviet engineers were VERY VERY jealous of. Apparently, Gertsch had invented an amazing machine(s) for producing extremely high and reproducible winding ratio accuracies. It was something the Soviets could never match, theres were typically orders of magnitude worse. There are some reports that even today, some of the original manufacturing machines are still being used.
Most people don't know these devices even exist so if you keep an eye on ebay for a while, you can get a nice 4-7 digit device for moderately cheap. With these devices, due to the physical nature of the ratio; an over voltage incident either destroys them or they just keep right on going there is no real in between. So if it works, there is a 99.999% chance that the ratio is entirely unchanged. The 00.001% chance is if someone was digging into the transformers themselves (there would be no need but people do crazy stuff).
EDIT:
GERTSCH RATIO TRANSFORMERS
This page is a wealth of information on the older Gertsch ratio transformers (there are a LOT more models than mentioned here and some mentioned on this site are VERY specialized, like the BRT1 which was a primitive Digital to Analog converter).
The biggest drawback with these is there are two separate max voltages that you have to be aware of. The insulation max voltage (usually 350V or 1kV) and the maximum voltage determined by frequency (common ratio are 0.35F or 2F, where F is the frequency). Whichever is lower is the one that applies. I'm not sure why this is but if you are working with low frequency AC, then it can make the max voltage lower than anticipated.
Most of the ratio transformers are made by Gertsch or one of the subsequent companies that acquired them (Singer, Teagam, ESI, IET Labs, etc.) and sold under a large variety of model names. As with most specialized test equipment companies, it was big fish eat little fish as a combination of the Cold War winding down and the analog to digital transition. As an aside, Ratio Transformers were one thing the soviet engineers were VERY VERY jealous of. Apparently, Gertsch had invented an amazing machine(s) for producing extremely high and reproducible winding ratio accuracies. It was something the Soviets could never match, theres were typically orders of magnitude worse. There are some reports that even today, some of the original manufacturing machines are still being used.
Most people don't know these devices even exist so if you keep an eye on ebay for a while, you can get a nice 4-7 digit device for moderately cheap. With these devices, due to the physical nature of the ratio; an over voltage incident either destroys them or they just keep right on going there is no real in between. So if it works, there is a 99.999% chance that the ratio is entirely unchanged. The 00.001% chance is if someone was digging into the transformers themselves (there would be no need but people do crazy stuff).
EDIT:
GERTSCH RATIO TRANSFORMERS
This page is a wealth of information on the older Gertsch ratio transformers (there are a LOT more models than mentioned here and some mentioned on this site are VERY specialized, like the BRT1 which was a primitive Digital to Analog converter).
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.