The Fluke 510A was an interesting exercise. I wanted to understand its level detection and AGC. The level detection is pretty interesting and its specs are very good as delivered. the level accuracy (.01% 30 days) is really hard to improve on. Most of the cost is in the reference voltage and level detection. The improvement in distortion from hacking Scott Wurcer's distortion neutralization into it really surprised me. It is worth almost 20 dB of distortion reduction in this box. Have a really accurate stable source can be very useful.
I still intend to build some based loosely on the more modern opamps. Getting that degree of stability will be the challenge.
The Boonton's are more problematic. They are good to a limit and really quick to use once you know them. However the second harmonic dominates in both the generator (and its output amp) and the analyzer. And won't be easy to reduce.
@Demian Martin -- I used to have a couple of Fluke 510As -- I set one up for 400Hz and one for 40kHz for cal'ing DMMs. When I got my HP 3458A, I discovered a couple of things about the 510s:
1) The Fluke cal procedure did not result in an exact 10VRMS output. I studied that circuitry 'til my eyes ran red and tried the various tweaks -- no love, and I never could figure out why.
2) the 400 Hz unit had less than 0.0005% THD with no mods at all, and the 40kHz unit was only somewhat worse given the increase of measuring bandwidth. This was measured with my build of Cordell's THD analyzer, which had a residual of around 0.00035% at 1kHz.
But (talking like Yoda here) hold a stable output over weeks or months they could not. My two units varied by over 500 ppm over a span of days. Very disappointing. Actually, my build of Cordell's SV oscillator has about as good stability, with no precision reference used at all. I honestly can't say what this means in terms of circuit goodness or badness, or where the variations occur -- I don't got dem nomore.
Fluke used a Motorola buried zener reference that included an on-die transistor for current control. These refs are now unobtanium, but even naked, they had pretty good stability vs. temp., and the ones in my 510s held within 20ppm every time I checked them. Fluke used these in ovens in a number of DC voltage standards, including the wonderful 5440B, and also used them without ovens in many other products, including the 8502 through 8506 6-1/2 and 7-1/2 digit DMMS, and in the 510s.
I've always wanted to revisit the 510, replacing the discrete transistors using good opamps, but I haven't seen one at reasonable cost for a long time. So the alternative is to do a layout and build one. Maybe someday.
1) The Fluke cal procedure did not result in an exact 10VRMS output. I studied that circuitry 'til my eyes ran red and tried the various tweaks -- no love, and I never could figure out why.
2) the 400 Hz unit had less than 0.0005% THD with no mods at all, and the 40kHz unit was only somewhat worse given the increase of measuring bandwidth. This was measured with my build of Cordell's THD analyzer, which had a residual of around 0.00035% at 1kHz.
But (talking like Yoda here) hold a stable output over weeks or months they could not. My two units varied by over 500 ppm over a span of days. Very disappointing. Actually, my build of Cordell's SV oscillator has about as good stability, with no precision reference used at all. I honestly can't say what this means in terms of circuit goodness or badness, or where the variations occur -- I don't got dem nomore.
Fluke used a Motorola buried zener reference that included an on-die transistor for current control. These refs are now unobtanium, but even naked, they had pretty good stability vs. temp., and the ones in my 510s held within 20ppm every time I checked them. Fluke used these in ovens in a number of DC voltage standards, including the wonderful 5440B, and also used them without ovens in many other products, including the 8502 through 8506 6-1/2 and 7-1/2 digit DMMS, and in the 510s.
I've always wanted to revisit the 510, replacing the discrete transistors using good opamps, but I haven't seen one at reasonable cost for a long time. So the alternative is to do a layout and build one. Maybe someday.
@Richiem:
I had the same confusion on setting the level. I'm not sure why the exact match between DC and AC is not happening. But in my case its more chaotic.
I have the following instruments involved: Fluke 8506A (DMM), Fluke 732A (Volt reference), Prem 6001 (DMM) and a Fluke 540B (Thermal Transfer Standard).
First the classic man with two watches doesn't know the time. . . I have several 6 digit DMM's and they don't agree. I sent the 8506 and the 732A out for cal. They came back and I finally checked them against each other to find they do not agree, not even close on the AC-DC comparison. Then I pull the certificates and find that the lab used a 5 digit portable meter to check the 732A and a calibrator no where near accurate enough for the 8506A. This all leaves me very unsure of the calibration. Needless to say they will be going back (the shop got close to $10K for calibration from that effort among all the instruments). As far as I can tell they did nothing to either except to see that they function. On DC they are 10 ppm apart when my shop is around 25C so I'll take that and build from it. I used the 540A to set the 510A (with jumping through all of the necessary hoops to make it possible) and I trust that the 732A is within 5 ppm in spite of the cal issues.
In checking the 510A with the Prema (which I trust more for AC right now) it has changed about 20 PPM from night to day, probably to be expected as its ambient has changed. I'll watch it over the next few days/weeks and see what I learn.
The distortion as received (10 years ago) was .005. All second harmonic and looking like some crossover distortion. The DN trick really made a difference. The Shibasoku has such a low floor its easy to see these things.
What I want is a modern equivalent with a few differences. I need really low distortion and really low phase noise for checking ADC's. My plan is to build three sources, battery powered and with two (11.025 KHz and 12 KHz) phase locked to crystal sources for jitter measurements and the third at 997 Hz to test every bit on the ADC. This combo will permit testing of many key parameters of an ADC, including distortion both HD and IM and jitter.
Using state variable oscillators for this is I think the best choice. The AGC and level sensing is important and Cordell's peak sensor is pretty good. I will probably use a variation on it. His analog multiplier with a FET is interesting. I'm not sure if the FET needs additional linearizing etc. Its very similar to the circuit in the Boonton except the Boonton uses analog multipliers. I won't obsess over absolute accuracy since its not really necessary in this application.
I hope to get a start on it this weekend.
I had the same confusion on setting the level. I'm not sure why the exact match between DC and AC is not happening. But in my case its more chaotic.
I have the following instruments involved: Fluke 8506A (DMM), Fluke 732A (Volt reference), Prem 6001 (DMM) and a Fluke 540B (Thermal Transfer Standard).
First the classic man with two watches doesn't know the time. . . I have several 6 digit DMM's and they don't agree. I sent the 8506 and the 732A out for cal. They came back and I finally checked them against each other to find they do not agree, not even close on the AC-DC comparison. Then I pull the certificates and find that the lab used a 5 digit portable meter to check the 732A and a calibrator no where near accurate enough for the 8506A. This all leaves me very unsure of the calibration. Needless to say they will be going back (the shop got close to $10K for calibration from that effort among all the instruments). As far as I can tell they did nothing to either except to see that they function. On DC they are 10 ppm apart when my shop is around 25C so I'll take that and build from it. I used the 540A to set the 510A (with jumping through all of the necessary hoops to make it possible) and I trust that the 732A is within 5 ppm in spite of the cal issues.
In checking the 510A with the Prema (which I trust more for AC right now) it has changed about 20 PPM from night to day, probably to be expected as its ambient has changed. I'll watch it over the next few days/weeks and see what I learn.
The distortion as received (10 years ago) was .005. All second harmonic and looking like some crossover distortion. The DN trick really made a difference. The Shibasoku has such a low floor its easy to see these things.
What I want is a modern equivalent with a few differences. I need really low distortion and really low phase noise for checking ADC's. My plan is to build three sources, battery powered and with two (11.025 KHz and 12 KHz) phase locked to crystal sources for jitter measurements and the third at 997 Hz to test every bit on the ADC. This combo will permit testing of many key parameters of an ADC, including distortion both HD and IM and jitter.
Using state variable oscillators for this is I think the best choice. The AGC and level sensing is important and Cordell's peak sensor is pretty good. I will probably use a variation on it. His analog multiplier with a FET is interesting. I'm not sure if the FET needs additional linearizing etc. Its very similar to the circuit in the Boonton except the Boonton uses analog multipliers. I won't obsess over absolute accuracy since its not really necessary in this application.
I hope to get a start on it this weekend.
A bit off topic, here, but since the zen masters of measurement are following this thread, I thought I would ask my question about measuring power supply noise.
I want to evaluate power supplies for a preamp (really more of a buffer - push pull Jfet configuration). I would like to measure noise and output impedance.
For the noise I was thinking that I could feed the supply output to my spectrum analyzer with the analyzer set to AC coupled mode. I would think this should show me the noise voltage. Is this a good way? Is ther a better way?
Now for the output impedance of the supply, I don't have the foggiest idea of where to start.
Thanks for any help.
Ken
I want to evaluate power supplies for a preamp (really more of a buffer - push pull Jfet configuration). I would like to measure noise and output impedance.
For the noise I was thinking that I could feed the supply output to my spectrum analyzer with the analyzer set to AC coupled mode. I would think this should show me the noise voltage. Is this a good way? Is ther a better way?
Now for the output impedance of the supply, I don't have the foggiest idea of where to start.
Thanks for any help.
Ken
The measurement part is pretty straightforward.
For output impedance the easy way is to use a cap, large enough to be pretty low Z, and feed a known signal back into the supply through a resistor. Don't drive the supply so hard it goes into internal saturation or you won't have meaningful numbers. say 100 mV at 1 KHz and a 100 Ohm resistor through a 100 uF cap. Measure the voltage at both sides of the resistor and you can calculate the current through the resistor and from that the impedance of the supply.
The noise really needs a very low noise preamp or your supply needs to be improved. I don't know what the noise floor of your SA is but they usually are not as quiet as a really good supply. I recently designed a really low noise supply (approx .7 nV/rtHz) and wound up building a transformer coupled preamp to even get close to seeing the noise. Unfortunately, even though the transformer is really well shielded it still picked up hum.
And then, the reality check. Make sure you are optimizing the supply for the aspects the load cares about. They are all tradeoffs.
For output impedance the easy way is to use a cap, large enough to be pretty low Z, and feed a known signal back into the supply through a resistor. Don't drive the supply so hard it goes into internal saturation or you won't have meaningful numbers. say 100 mV at 1 KHz and a 100 Ohm resistor through a 100 uF cap. Measure the voltage at both sides of the resistor and you can calculate the current through the resistor and from that the impedance of the supply.
The noise really needs a very low noise preamp or your supply needs to be improved. I don't know what the noise floor of your SA is but they usually are not as quiet as a really good supply. I recently designed a really low noise supply (approx .7 nV/rtHz) and wound up building a transformer coupled preamp to even get close to seeing the noise. Unfortunately, even though the transformer is really well shielded it still picked up hum.
And then, the reality check. Make sure you are optimizing the supply for the aspects the load cares about. They are all tradeoffs.
I'd like to add that you might want to use a power amp to drive the 100 ohm resistor -- low output Z from the source helps in getting clean data. At Tek, this device had a low resistance and a high-power pulser driving it -- it was called a "grunter" because when the pulse generator was hooked up to the output of a power supply and turned on, you could hear the supply under test literally grunt -- or squeal -- with each pulse.
FYI - super low THD oscillators for sale on eBAY
On eBAY, you can now get a 1KHz and a 10KHz oscillator on individual pcb's (no chassis available) that has spec'ed .00005% THD+N. Each pcb is only 39 dollars. Ready to go... just drop in a box and add pwr.
I bought one of each to sub for my dual domain AP's internal osc. -RNM
On eBAY, you can now get a 1KHz and a 10KHz oscillator on individual pcb's (no chassis available) that has spec'ed .00005% THD+N. Each pcb is only 39 dollars. Ready to go... just drop in a box and add pwr.
I bought one of each to sub for my dual domain AP's internal osc. -RNM
Dick did you get a chance to test those oscillators, are they as good as advertised?
jan didden
Go see Victor in Latvia for the low distortion oscillators.... viccc@ml.lv
[or hunt for him on eBAY -- would take me as long to find him again as it would for you to look.]
I didnt see this forum until long after I recieved the units or i would have kept more info on it to give to you. I just bought 2 of the 1KHz in case I smoke test one of them or it fails on its own accord someday.... and then I might not be able to find him/them again, later. After the 1KHz I asked him for a 10KHz and he made one of those too. I suppose he could do any frequency on special order.... ?
I havent got to using them yet. -RNM
[or hunt for him on eBAY -- would take me as long to find him again as it would for you to look.]
I didnt see this forum until long after I recieved the units or i would have kept more info on it to give to you. I just bought 2 of the 1KHz in case I smoke test one of them or it fails on its own accord someday.... and then I might not be able to find him/them again, later. After the 1KHz I asked him for a 10KHz and he made one of those too. I suppose he could do any frequency on special order.... ?
I havent got to using them yet. -RNM
Last edited:
I figure from what you wanted to do that telling about this would help you a lot in your R&D. especially the custom freqs you need. -RNM
With all due respect, do you really expect that any two 6.5 digit meters are going to precisely agree, even when freshly calibrated? They will not. Your 8506A in DC mode has uncertainly of roughly 20 digits (full scale, most ranges) within 24 hours of calibration, and roughly 40 digits after only 90 days. And that assumes your lab is the same temperature as the calibration lab, and that you have warmed up the instrument for 4 hours! Furthermore, once you start to get resolution down into microvolts, thermal effects can be huge. Special test methods are required when taking such measurements, you can't just plug in some probes or leads and expect accuracy.
I am well aware of those issues and the total uncertainties involved. That's why I didn't go into the calibration issues initially. As I said I go through most of the gyrations to remove as many of those issues as I can. The 20 digits you mention are 20 ppm. (actually on the 10V range the total DC is 18 ppm for 90 days) The DC differences are within those uncertainties. The AC is way out. I gave up and am using the 540 which is 100 ppm AC-DC, where the 8506 seems to be more like 6000 ppm + (the spec is 160 ppm). I have some more precision converters but this is all way beyond anything valid for audio. Its really just obsession.
So far the the Fluke 510A seems to be as stable as they claim. It fluctuates in keeping with the temperature which is to be expected but otherwise it withing the claimed tolerances.
The attenuator was just a reality check to verify the noise is in the detector, not the oscillator. On the System 1 you can set the generator impedance to 50, 150 or 600 ohms.
I found an old Tektronix 7L5 Spectrum Analyzer to look at the Reading Monitor Output. There was a strong birdie at 48.5kHz when the Monitor was on. That was with a 22kHz filter on the distortion. Monitor on or off didn't make much difference. The options can set it at 22kHz, 30kHz, 80kHz or >500kHz. Those filter selections revealed harmonics of the 48.5kHz and an increase in the residual distortion at low signal levels. (There are also LF selections and tracking filters.)
What bandwidth filter is customary for distortion measurements?
Usually distortion is reported with a 20 KHz filter (not meaningful for 10KHz source but. . ). Some even A weight the distortion which is really not meaningful. There are lots of philosophies on how best to understand the distortion measurement. First, with few exceptions its distortion + noise and with better stuff the noise dominates at real listening levels. Looking at the spectrum can be very helpful. Higher order harmonics are not good and in some cases (tubes?) low order harmonics can help sales.
The 7L5 is very useful for looking at the output of a DAC. There can be a lot more HF stuff that doesn't normally show when you stop the measurement at 100 KHz. I use mine a fair amount when developing DAC's.
The 7L5 is very useful for looking at the output of a DAC. There can be a lot more HF stuff that doesn't normally show when you stop the measurement at 100 KHz. I use mine a fair amount when developing DAC's.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.