The absolute tolerance matters very little. The feedback resistor sets the gain and most don't really care if the gain of their power amp is 20.0 dB or 20.1 dB.
The resistors will heat up and cool down as the voltage across them goes up and down. That causes the resistance to change, which changes the gain of the amp. So you will get a signal-dependent error voltage in series with the feedback resistor that is some function of the TCR. "Signal-dependent error voltage" is just another way of saying "distortion".
But note that the TCR of the resistor does not tell the full story. The TCR is measured by measuring the resistance of the resistor at two temperature extremes. It is not measured as function of the self-heating of the resistor. Vishay has introduced the Power Coefficient of Resistance (PCR) to account for the change in resistance as function of self-heating. Ideally you'll want low TCR and low PCR.
Tom
The resistors will heat up and cool down as the voltage across them goes up and down. That causes the resistance to change, which changes the gain of the amp. So you will get a signal-dependent error voltage in series with the feedback resistor that is some function of the TCR. "Signal-dependent error voltage" is just another way of saying "distortion".
But note that the TCR of the resistor does not tell the full story. The TCR is measured by measuring the resistance of the resistor at two temperature extremes. It is not measured as function of the self-heating of the resistor. Vishay has introduced the Power Coefficient of Resistance (PCR) to account for the change in resistance as function of self-heating. Ideally you'll want low TCR and low PCR.
Tom
Without any questions the YR1B20KCC gives great results, but if anyone wants to try Dale Ptf56 1% 10ppm ( on sale for 0.25$/piece)
Take a look here:
https://www.arrow.com/en/products/ptf5620k000fyeb/vishay
Take a look here:
https://www.arrow.com/en/products/ptf5620k000fyeb/vishay
Hi Tom,
Just been looking at the Mean Well LOP series specs. Mean Well seem to have managed to shrink the 300W LOP-300 into the same sized footprint (102x51x25) as the old EPP-200 series, and added some useful voltages - 30 and 36V. So maybe a pair of LOP-300-30 for a stereo Modulus 86 setup or 2x LOP-300-36 for one channel of a Modulus 286 based amp? Also the price appears to have dropped (UK). I am sure you have spotted this - I just thought I would raise it as it seems to offer an alternative to Connex with the new voltages being available.
Dave
Just been looking at the Mean Well LOP series specs. Mean Well seem to have managed to shrink the 300W LOP-300 into the same sized footprint (102x51x25) as the old EPP-200 series, and added some useful voltages - 30 and 36V. So maybe a pair of LOP-300-30 for a stereo Modulus 86 setup or 2x LOP-300-36 for one channel of a Modulus 286 based amp? Also the price appears to have dropped (UK). I am sure you have spotted this - I just thought I would raise it as it seems to offer an alternative to Connex with the new voltages being available.
Dave
WOW!! That is cool!
Two LOP-300-30 would be perfect for a Modulus-86 amp. $89 for a pair of them at Mouser versus $118/each for the Connex (including "express" shipping, and assuming you buy two of them like I did a month ago). The Mean Well is certified out the yin-yang. The Connex .... isn't. The ones in stock at Mouser will arrive tomorrow. The ones at Connex will arrive in 3+ weeks even with express shipping.
That's a solid find. Thanks for sharing. I'll have to include a pair of them next time I order from Mouser.
Tom
Two LOP-300-30 would be perfect for a Modulus-86 amp. $89 for a pair of them at Mouser versus $118/each for the Connex (including "express" shipping, and assuming you buy two of them like I did a month ago). The Mean Well is certified out the yin-yang. The Connex .... isn't. The ones in stock at Mouser will arrive tomorrow. The ones at Connex will arrive in 3+ weeks even with express shipping.
That's a solid find. Thanks for sharing. I'll have to include a pair of them next time I order from Mouser.
Tom
No problem Tom, the only real change in the spec (I can see) of the LOP-300 compared with the EPP-300 is the convection spec which has reduced from 200W to 180W - is this a problem with the Modulus 86 stereo set up? Presumably this is maximum constant current dissipation before the over-temperature sensor kicks in, so I guess unless you like listening to sine waves very loud or running a massive class A amp it should be OK?
Dave
Dave
That caught my eye as well. It's not a problem.
Some math:
Supply voltage: ±30 V
Modulus-86 maximum power draw (sine wave at clipping, 4 Ω load): 134 W
Modulus-86 power draw, average (music, 14 dB CF, peaks at clipping, 4 Ω load): 40.0 W
Modulus-86 peak output current (±30 V supply, ±2 V drop-out, 4 Ω load): 7.0 A
Now that I have my calculator spreadsheet open, let's look at the Modulus-286:
Supply voltage: ±36 V
Modulus-286 maximum power draw (sine wave at clipping, 4 Ω load): 199 W
Modulus-286 power draw, average (music, 14 dB CF, peaks at clipping, 4 Ω load): 61.3 W
Modulus-286 peak output current (±36 V supply, ±2 V drop-out, 4 Ω load): 8.5 A
The 300 W supplies will deliver 10 A at 30 V and 8.3 A at 36 V. The LOP-200 and LOP-300 can sustain quite a bit of overload as long as it's kept short (3 seconds). The data sheet seems to imply that the supplies can provide 150 % of the rated output current for three seconds, but read a bit closer and it turns out it's 105-150 % overload.
Anyway. Long story short. It looks like the LOP-300-30 would be an excellent candidate for the Modulus-86. The LOP-300-36 would work well with the Modulus-286 as long as the two amp channels are operated in opposite phase. That's easy to do by swapping the two input wires in the differential pair and swapping the output wires.
My personal experience with the Mean Well supplies has only been good. Others have not been so fortunate. About half of builders who tried the RPS-400-36 in their Modulus-686 builds had issues with mechanical whine from the power supply. I've asked Mean Well if that's the case for the LOP-series as well and their response was that the LOP is based on the same resonant topology as the RPS-series, so it might be an issue. "Try it and see" is basically what they said. That said; in the cases where the RPS-400 whined, the whine was squelched by turning the output voltage up to the max.
I have not heard about any issues with the RPS-200, so I'm thinking the LOP-200 and LOP-300 are probably fine.
Tom
Some math:
Supply voltage: ±30 V
Modulus-86 maximum power draw (sine wave at clipping, 4 Ω load): 134 W
Modulus-86 power draw, average (music, 14 dB CF, peaks at clipping, 4 Ω load): 40.0 W
Modulus-86 peak output current (±30 V supply, ±2 V drop-out, 4 Ω load): 7.0 A
Now that I have my calculator spreadsheet open, let's look at the Modulus-286:
Supply voltage: ±36 V
Modulus-286 maximum power draw (sine wave at clipping, 4 Ω load): 199 W
Modulus-286 power draw, average (music, 14 dB CF, peaks at clipping, 4 Ω load): 61.3 W
Modulus-286 peak output current (±36 V supply, ±2 V drop-out, 4 Ω load): 8.5 A
The 300 W supplies will deliver 10 A at 30 V and 8.3 A at 36 V. The LOP-200 and LOP-300 can sustain quite a bit of overload as long as it's kept short (3 seconds). The data sheet seems to imply that the supplies can provide 150 % of the rated output current for three seconds, but read a bit closer and it turns out it's 105-150 % overload.
Anyway. Long story short. It looks like the LOP-300-30 would be an excellent candidate for the Modulus-86. The LOP-300-36 would work well with the Modulus-286 as long as the two amp channels are operated in opposite phase. That's easy to do by swapping the two input wires in the differential pair and swapping the output wires.
My personal experience with the Mean Well supplies has only been good. Others have not been so fortunate. About half of builders who tried the RPS-400-36 in their Modulus-686 builds had issues with mechanical whine from the power supply. I've asked Mean Well if that's the case for the LOP-series as well and their response was that the LOP is based on the same resonant topology as the RPS-series, so it might be an issue. "Try it and see" is basically what they said. That said; in the cases where the RPS-400 whined, the whine was squelched by turning the output voltage up to the max.
I have not heard about any issues with the RPS-200, so I'm thinking the LOP-200 and LOP-300 are probably fine.
Tom
That's great thanks Tom! The LOP series look physically different o the EPP/RPS - made me wonder whether they operated differently and maybe the higher power units would be less prone to whining - not a very scientific analysis! I seem to remember reading somewhere about a tech guy from Mean Well suggesting a load resistor in parallel with the amp to maintain a greater load on the SMPS when the amp was idling. The LOP-300 sounds promising - my compact stereo Mod 86 design looks even better with the smaller SMPS units - I just need to get my finger out and build it!!! Unfortunately over the past few years the vicissitudes of life have screwed my attempts to kick it off
Dave
Dave
I just took delivery of about $200 worth of resistors, which is impressive given that the highest quantity of any individual part number was 10.
First, let's see what happens if we go cheap.
This is a plain vanilla carbon film resistor. Stackpole CF14JT20K0 (20 kΩ, ±5 %, +0/-500 ppm/ºC, $0.10/EA)
Upgrading to a plain vanilla metal film gets us to the Yageo MFR-25FRF52-20K (20 kΩ, ±1 %, ±100 ppm/ºC, $0.10/EA):
Now let's just try something different. This time a wirewound power resistor (Vishay RS01020K00FE73, 20 kΩ, ±1 %; 20 ppm/ºC, 10 W, $2.89/EA). This thing is a honking monster at 9.5 mm Ø x 45 mm length, so I had to mount it vertically.
And now back to the fancy (and expensive!!) resistors:
First up of the fancy resistors is the Holco (now TE) H820KBYA (20 kΩ, ±0.1 %, ±15 ppm/ºC, $1.75/EA):
.... and the Vishay PTF6520K000DZEB (20 kΩ, ±0.5 %, ±5 ppm/ºC, $4.79/EA):
Tom
First, let's see what happens if we go cheap.
This is a plain vanilla carbon film resistor. Stackpole CF14JT20K0 (20 kΩ, ±5 %, +0/-500 ppm/ºC, $0.10/EA)
Upgrading to a plain vanilla metal film gets us to the Yageo MFR-25FRF52-20K (20 kΩ, ±1 %, ±100 ppm/ºC, $0.10/EA):
Now let's just try something different. This time a wirewound power resistor (Vishay RS01020K00FE73, 20 kΩ, ±1 %; 20 ppm/ºC, 10 W, $2.89/EA). This thing is a honking monster at 9.5 mm Ø x 45 mm length, so I had to mount it vertically.
And now back to the fancy (and expensive!!) resistors:
First up of the fancy resistors is the Holco (now TE) H820KBYA (20 kΩ, ±0.1 %, ±15 ppm/ºC, $1.75/EA):
.... and the Vishay PTF6520K000DZEB (20 kΩ, ±0.5 %, ±5 ppm/ºC, $4.79/EA):
Tom
One wonders what happens if we spend even more to get even fancier resistors. More fancy, more better, right?
The fancy resistors tend to not be available in 20 kΩ, but they are often available in 10 kΩ. Two in series makes 20 kΩ. Yay! But now the dissipated power is also spread between two resistors, so that alone could provide a benefit. Lower self-heating -> less resistance shift through the signal swing -> lower distortion. So we need to compare against something.
All tests below use two identical resistors in series for the feedback network.
This is the THD with the TE YR1B10KCC (10 kΩ, ±0.1 %, ±15 ppm/ºC, $0.60/EA):
Now let's look at the Vishay/Dale CMF5510K000BEEB (10 kΩ, ±0.1 %, ±25 ppm ºC, $1.15/EA):
It's interesting to note that H3 is quite a bit lower than it was for the single 20 kΩ CMF55 that I posted a few days ago. More on that later.
Next up is the Holco (now TE) H810KBYA (10 kΩ, ±0.1 %, ±15 ppm/ºC, $1.15/EA):
And now let's just throw stupid money at resistors:
Vishay (VPG) Y070610K0000T9L (10 kΩ, ±0.01 %, ±0.2 ppm/ºC, $15.68/EA). Sorry about the typo in the graph title. The correct part number is Y070610K0000T9L. These are the naked Z-foil. Fancy, fancy. Mucho dinero.
And in the collection of resistors that my friend provided me (see Post #6330), I was able to find two TxCC (= Vishay/VPG) 10 kΩ, ±0.01 %, probably ±2 ppm/C, price unknown:
Tom
The fancy resistors tend to not be available in 20 kΩ, but they are often available in 10 kΩ. Two in series makes 20 kΩ. Yay! But now the dissipated power is also spread between two resistors, so that alone could provide a benefit. Lower self-heating -> less resistance shift through the signal swing -> lower distortion. So we need to compare against something.
All tests below use two identical resistors in series for the feedback network.
This is the THD with the TE YR1B10KCC (10 kΩ, ±0.1 %, ±15 ppm/ºC, $0.60/EA):
Now let's look at the Vishay/Dale CMF5510K000BEEB (10 kΩ, ±0.1 %, ±25 ppm ºC, $1.15/EA):
It's interesting to note that H3 is quite a bit lower than it was for the single 20 kΩ CMF55 that I posted a few days ago. More on that later.
Next up is the Holco (now TE) H810KBYA (10 kΩ, ±0.1 %, ±15 ppm/ºC, $1.15/EA):
And now let's just throw stupid money at resistors:
Vishay (VPG) Y070610K0000T9L (10 kΩ, ±0.01 %, ±0.2 ppm/ºC, $15.68/EA). Sorry about the typo in the graph title. The correct part number is Y070610K0000T9L. These are the naked Z-foil. Fancy, fancy. Mucho dinero.
And in the collection of resistors that my friend provided me (see Post #6330), I was able to find two TxCC (= Vishay/VPG) 10 kΩ, ±0.01 %, probably ±2 ppm/C, price unknown:
Tom
Yep. I like the ISS for many reasons. I use it to control the power to the Connex supplies (which do benefit from a bit of extra soft start) and the Hypex supplies (which have their own soft start). I like the LED dimmers, low voltage control, etc.
That stood out to me immediately as well. I've measured at different frequencies and have repeated the measurement after measuring some other resistors (that did show the 25 Hz + harmonics). The effect is repeatable. The big fella doesn't show the 25 Hz + harmonics. It does show a tad more 60 Hz. That's not a shock. After all it adds quite an antenna in series with the feedback loop, so it should be more susceptible to 60 Hz induction.
I also note that H2 is quite a bit lower, which is intriguing.
I'm not rushing to recommend large wire wound resistors in the MOD86. If I ever find a, say, 1-2 W wire wound one I could see maybe doing that. But they're going the way of the dodo, so maybe not... There are some options in 1-2 W wire wound SMD resistors, though. They could maybe be interesting as an option in addition to the PTH resistors. Reidon S3-10KF1 (10 kΩ, ±1%, ±20 ppm/ºC, 3 W, $2.06/EA) could maybe be interesting.
I mentioned yesterday that the amp doesn't care about the absolute tolerance of the resistors. That's not entirely true. The amp itself doesn't care. The resistor sets the gain and so what if the gain is off by a picobell? But the resistor is also part of the differential receiver, so the absolute tolerance does play into the CMRR of the input. Tighter tolerance -> higher CMRR.
Tom
And, finally... Let's take a look at the Vishay/Dale CMF55 vs RN55.
I don't normally do requests, but Dario (@ClaveFremen) mentioned in a PM that he'd heard a difference between the CMF55 and the RN55. As you can probably imagine, I was pretty skeptical of that, especially given that the manufacturer states in the data sheet, a legally binding document, that the resistors are exactly the same except for the markings on them.
But I figured I'd give him the benefit of the doubt, so I threw in an RN55 with my most recent Digikey order.
Here's the performance with the RN55E2002BRE6 (20 kΩ, ±0.1 %; ±25 ppm/ºC, $0.87/EA)
Here's the CMF5520K000BEEB (20 kΩ, ±0.1 %, ±25 ppm/ºC, $0.64/EA):
It's pretty easy to conclude that the CMF55 is much worse than the RN55, which doesn't really jibe with the manufacturer's claim of the resistors being identical except for their markings.
I only ordered one RN55 so I don't know if I somehow got a golden unit. I did order a couple CMF55, so I measured two more of those:
These both measure about the same. H2 is a quite a bit higher for Sample #3 but none of them show the much higher H3 of Sample #1.
Vishay makes no promises about the distortion in their data sheet, so I guarantee you that distortion is not measured in production. I'm not aware of any resistor manufacturer that does specify distortion other than claiming that it is "low" (whatever that means) in their marketing materials. If Vishay says that the CMF and RN resistors are identical, then I will believe them. So that means the difference between Sample #1 and Samples #2 and #3 is part-to-part variation.
I haven't seen part-to-part variation that severe on any of the other resistors I tested. Is that because the other resistors are of higher quality? Or maybe I just got lucky with the others? Or maybe the one 'bad' CMF55 was just a single random flaw that happened to be exposed in my test? I don't know.
I do know that I'll stick to the TE YR1B-series because none of the other resistors provide any significant advantage. If anything, I could consider making the feedback network from two 10 kΩ in series instead of a single 20 kΩ. That approach does show a reliable advantage of a few dB on all harmonics. That's still a pretty small difference, though. If you already have a working Modulus-86 I wouldn't rush out and modify it.
Next up will be a handful of surface mount resistors, including some fancy Vishay/VPG foil resistors. I'll use the Modulus-286 for the test.
Tom
I don't normally do requests, but Dario (@ClaveFremen) mentioned in a PM that he'd heard a difference between the CMF55 and the RN55. As you can probably imagine, I was pretty skeptical of that, especially given that the manufacturer states in the data sheet, a legally binding document, that the resistors are exactly the same except for the markings on them.
But I figured I'd give him the benefit of the doubt, so I threw in an RN55 with my most recent Digikey order.
Here's the performance with the RN55E2002BRE6 (20 kΩ, ±0.1 %; ±25 ppm/ºC, $0.87/EA)
Here's the CMF5520K000BEEB (20 kΩ, ±0.1 %, ±25 ppm/ºC, $0.64/EA):
It's pretty easy to conclude that the CMF55 is much worse than the RN55, which doesn't really jibe with the manufacturer's claim of the resistors being identical except for their markings.
I only ordered one RN55 so I don't know if I somehow got a golden unit. I did order a couple CMF55, so I measured two more of those:
These both measure about the same. H2 is a quite a bit higher for Sample #3 but none of them show the much higher H3 of Sample #1.
Vishay makes no promises about the distortion in their data sheet, so I guarantee you that distortion is not measured in production. I'm not aware of any resistor manufacturer that does specify distortion other than claiming that it is "low" (whatever that means) in their marketing materials. If Vishay says that the CMF and RN resistors are identical, then I will believe them. So that means the difference between Sample #1 and Samples #2 and #3 is part-to-part variation.
I haven't seen part-to-part variation that severe on any of the other resistors I tested. Is that because the other resistors are of higher quality? Or maybe I just got lucky with the others? Or maybe the one 'bad' CMF55 was just a single random flaw that happened to be exposed in my test? I don't know.
I do know that I'll stick to the TE YR1B-series because none of the other resistors provide any significant advantage. If anything, I could consider making the feedback network from two 10 kΩ in series instead of a single 20 kΩ. That approach does show a reliable advantage of a few dB on all harmonics. That's still a pretty small difference, though. If you already have a working Modulus-86 I wouldn't rush out and modify it.
Next up will be a handful of surface mount resistors, including some fancy Vishay/VPG foil resistors. I'll use the Modulus-286 for the test.
Tom
I bet it's sample-to-sample variability. The three parts were next to each other on the tape. The RN55 obviously came from a different tape. It performs about as well as CMF55 Sample #2.
Resistors are well understood by now. Not thereby said that they aren't complex or that there couldn't be a peach magic effect, but I doubt such an effect wouldn't have been exposed by now. Those further interested in the topic of resistors should peruse Sandman, Simon, & Szwarc, "Resistor Theory and Technology". It was published by SciTech Publishing in 2001. It's "only" 300+ pages.
The authors are two guys from Vishay and a consultant. Tom
For completeness, I decided to take a look at the sample-to-sample variability for the TE YR1B20KCC (20 kΩ, ±0.1 %, ±15 ppm/ºC, $0.60/EA). Sample #1 you've seen before. Samples #2 through #4 were measured today.
Sample #2:
Sample #3:
Sample #4:
Interesting that I happened to pick the worst of the bunch on the first try with these as well. Oh, well...
Tom
Sample #2:
Sample #3:
Sample #4:
Interesting that I happened to pick the worst of the bunch on the first try with these as well. Oh, well...
Tom