I not not an audiophile nor do I possess some special golden ears. Neither am I gifted with superfluous vocabulary often found in most audio reviews. So take this post with a pinch of salt - disclaimer done and out of the way..
I was intrigued as to what is the real difference between the sound quality of commercial amps vs my much loved LM3875..
My LM3875 is built using audiosector premium KIT + 20,000UF caps
so, here goes a poor man's review:
SETUP
Roksan Caspian M2 vs LM3875
Audiolab 8000S (Pre) + Audiolab 8000P (Power) vs LM3875
So, 2 of the commercial amps I tried really pissed all over my beloved chip amp. I was somewhat disappointed that LM3875 was beaten this easily. The soundstage on LM3875 in comparison was very much 2d compared to both Roksan and Audiolab. The depth and weight were also lacking. Like someone watered down a fine wine and you are drinking a weird spritzer..
So, I took a step in a different direction and found a "cheap" Rotel integrated from the days gone by. This is the Rotel RA 921, it was featured in many of the best buy articles in around 1998 and retailed for around £100.
I picked up Rotel RA 921 locally on ebay for £23, very good condition and no signs of abuse. Not expecting much from a bargain basement amp that is nearly 20 years old and sells for peanuts. Surely this has to be a match for LM3875..
Rotel RA 921 vs LM3875
If it is true that £23 amp, that is nearly 20 years old can sound very very similar (if not better); what am I doing wrong and spending almost £150 (case + kit + cables + toroidal etc) in building something that can easily be compared to a bottom end, bargain basement amp.
By all means this is not a Golden ears, ABX style or blind test. Maybe I am biased, maybe the price labels and visual aesthetics of commercial kits swun my decisions. Hence the reason for this post.
What have your experiences been in comparing your gainclone/chip amps to commercial offerings?
I was intrigued as to what is the real difference between the sound quality of commercial amps vs my much loved LM3875..
My LM3875 is built using audiosector premium KIT + 20,000UF caps
so, here goes a poor man's review:
SETUP
- Speakers: ProAc Studio 140MK2
- Dac: Musical Fidelity M1DAC
- Source: Macbook Pro (USB)
- Music files: FLAC
Roksan Caspian M2 vs LM3875
- - Roksan M2 has more seperation
- - Roksan M2 has more detail
- - Roksan M2 deeper sound stage
- - Roksan M2 Bass response was far better
Audiolab 8000S (Pre) + Audiolab 8000P (Power) vs LM3875
- - Audiolab has more seperation
- - Audiolab has more detail
- - Audiolab deeper sound stage
- - Audiolab Bass response was far better, better and deeper than Roksan
So, 2 of the commercial amps I tried really pissed all over my beloved chip amp. I was somewhat disappointed that LM3875 was beaten this easily. The soundstage on LM3875 in comparison was very much 2d compared to both Roksan and Audiolab. The depth and weight were also lacking. Like someone watered down a fine wine and you are drinking a weird spritzer..
So, I took a step in a different direction and found a "cheap" Rotel integrated from the days gone by. This is the Rotel RA 921, it was featured in many of the best buy articles in around 1998 and retailed for around £100.
I picked up Rotel RA 921 locally on ebay for £23, very good condition and no signs of abuse. Not expecting much from a bargain basement amp that is nearly 20 years old and sells for peanuts. Surely this has to be a match for LM3875..
Rotel RA 921 vs LM3875
- - Rotel RA 921 has more seperation
- - LM3875 has a LITTLE more detail
- - Rotel RA 921 and LM3875 have very very similar soundstage. Nothing compared to Roksan or Audiolab
- - Rotel RA 921 Bass response was marginally better than LM3875. Deeper and more authority.
If it is true that £23 amp, that is nearly 20 years old can sound very very similar (if not better); what am I doing wrong and spending almost £150 (case + kit + cables + toroidal etc) in building something that can easily be compared to a bottom end, bargain basement amp.
By all means this is not a Golden ears, ABX style or blind test. Maybe I am biased, maybe the price labels and visual aesthetics of commercial kits swun my decisions. Hence the reason for this post.
What have your experiences been in comparing your gainclone/chip amps to commercial offerings?
Last edited:
Keep in mind that chipamp performance is, like SE amps, fairly speaker dependent. They like a fairly flat impedance curve.
If this impedance curve is any indication -- for the model under with only 1 woofer, one could expect the twin woofer model to be similar except for being about half the impedance in the bass.
This is what one would consider a pretty ungly impedance curve. Unless specifically built to work into low impedance, ie 25V rails, your will chipamp tend to fall apart in the bass, and it is not going to be happy with the huge impedance rise in the midrange.
Plus, i don't know about using such big caps. Mine has the smaller recommended caps.
My LM3875 is a good amp, but not the best one i have, but by using the speakers you did for the test, i expect your LM3875 is at an immediate disadvantage.
dave
If this impedance curve is any indication -- for the model under with only 1 woofer, one could expect the twin woofer model to be similar except for being about half the impedance in the bass.
An externally hosted image should be here but it was not working when we last tested it.
This is what one would consider a pretty ungly impedance curve. Unless specifically built to work into low impedance, ie 25V rails, your will chipamp tend to fall apart in the bass, and it is not going to be happy with the huge impedance rise in the midrange.
Plus, i don't know about using such big caps. Mine has the smaller recommended caps.
My LM3875 is a good amp, but not the best one i have, but by using the speakers you did for the test, i expect your LM3875 is at an immediate disadvantage.
dave
Hi Dave,
Thanks for the reply. I wanted to eliminate the speaker variable and tested with my other speaker pairs and results weren't that dissimilar..
Here is what I tested with.
Proac Studio 140MK2
- Nominal Impedance: 4 ohms, 3.2 minimum
- Frequency Response:25hz to 30Khz
- Sensitivity: 91db linear for 1 watt at 1 metre
TDL RTL 3
An externally hosted image should be here but it was not working when we last tested it.
- Frequency Response: 35Hz to 20kHz
- Impedance: 8Ω
- Sensitivity: 90dB
KEF Cresta 2
- Nominal impedance: 8 ohms
- Frequency response: 48Hz to 20kHz +/-3.0dB
- Sensitivity: 90dB at 1m for 2.83V
and finally
Pioneer S-T100
An externally hosted image should be here but it was not working when we last tested it.
- Frequency response: 30-40k
- Efficiency: 88 dbl@1watt/meter
- Nominal Impedance: 6 ohms
All those speakers have XOs in them. What do the impedance curves look like?
What are the rails on your chipamp?
dave
What are the rails on your chipamp?
dave
How can i measure this? Never measured this before..
LM3875 is powered with 250VA 18-0-18 VAC. So after rectification its around 24VDC
You need some impedance measuring kit. Usually diyers use software (lots free) and a simple external circuit.
So your chip-amp is optimized for lower impedance loads.
What I am saying is that the performace of an amplifier is not independent of the speakers it is used with, one should never consider the performance of an amplifier without the speaker.
In this case, these amps may be better than your chip-amp driving these speakers, but i suspect none of these speakers have a load that the chipamp likes.
dave
So your chip-amp is optimized for lower impedance loads.
What I am saying is that the performace of an amplifier is not independent of the speakers it is used with, one should never consider the performance of an amplifier without the speaker.
In this case, these amps may be better than your chip-amp driving these speakers, but i suspect none of these speakers have a load that the chipamp likes.
dave
You aren't the first and won't be the last. The human mind is an unreliable device and needs carefully thought out and scrutinised protocols to attempt to screen out the biases. That is why pharmaceutical testing is so expensive and lengthy. In the same way, some people think they can be cured by sugar water (aka homeopathy) but physics, chemistry, and RCTs prove them wrong.
Hi,
I did not realize same review, but I have some good commercial amp, and I prefer LM3875..
I build one with Chipamp.com pcb (same as Audiosector), and with Chipamp.com snuberrized PSU in low power (18 vdc).
I change many componants and values as you can see :
http://www.diyaudio.com/forums/chip-amps/271984-lm3875-chipamp.html
The result :
I add a safety loop breaker for more soundstage. Gain is 26, best compromize, gain of 33 give "bad" mid and hight but good bass..
Cf and Rf2 (schematic post one) give better detail for example.
I try Caddock, and I found that is not a good choice for price.
For R1 (330 Ohm), I use carbon Takmann.
Grouding and heatsink are very sensitive with LM, in my case heatsink stay very cold about 28°C (inroom 20°C).
I love this amp, better than my others, but powered with Li Ion battery, it's just amazing, but not comfortable due by recharching !
What is lenght of your cables, are they not too capacitive ?
Can you post few picts of your amp ?
Phil.
I did not realize same review, but I have some good commercial amp, and I prefer LM3875..
I build one with Chipamp.com pcb (same as Audiosector), and with Chipamp.com snuberrized PSU in low power (18 vdc).
I change many componants and values as you can see :
http://www.diyaudio.com/forums/chip-amps/271984-lm3875-chipamp.html
The result :
An externally hosted image should be here but it was not working when we last tested it.
I add a safety loop breaker for more soundstage. Gain is 26, best compromize, gain of 33 give "bad" mid and hight but good bass..
Cf and Rf2 (schematic post one) give better detail for example.
I try Caddock, and I found that is not a good choice for price.
For R1 (330 Ohm), I use carbon Takmann.
Grouding and heatsink are very sensitive with LM, in my case heatsink stay very cold about 28°C (inroom 20°C).
I love this amp, better than my others, but powered with Li Ion battery, it's just amazing, but not comfortable due by recharching !
What is lenght of your cables, are they not too capacitive ?
Can you post few picts of your amp ?
Phil.
Just saw, your dac is M1, it can give around 2,2 v at output.
How do you mamage level, at player in digital, with pot ?
Cause LM don't like hight voltage at input, maybe 1 to 1,5 v is good value.
I use passive pot with 12 db pre-attenuator (dac is MF Vdac 2).
Whith 2 v output, mid and hight are not so good, with lower, it's more "natural" and well balanced.
Phil.
How do you mamage level, at player in digital, with pot ?
Cause LM don't like hight voltage at input, maybe 1 to 1,5 v is good value.
I use passive pot with 12 db pre-attenuator (dac is MF Vdac 2).
Whith 2 v output, mid and hight are not so good, with lower, it's more "natural" and well balanced.
Phil.
Doug Self came up with an easy method that only requires a typical DVM capable of reading AC voltage over the audio band.
Connect a 600 ohm resistor in series with the speaker and drive this from a source (amplifier) set to give 6 volts RMS. You then read the voltage in millivolts across the speaker, the voltage corresponding to the equivalent impedance at that frequency. 40 millivolts would 4 ohms, 100mv 10 ohms and so on.
What you want is a really slow sine wave sweep covering 10Hz to 20Khz over a couple of minutes or more, something which you can make in minutes using Audacity. Even better, two or three tracks covering say 10 to 100Hz, 100Hz to 1Khz etc.
http://www.diyaudio.com/forums/soft...ing-using-audacity-get-you-started-guide.html
the National chipamps are all current crippled.
They cannot drive large curents into typical 4ohms speakers.
The Audiosector circuit was developed by Peter Daniel and he makes the case very clearly that his implementation is very much aimed towards mid and treble performance.
He used ~95dB/W 4ohms speakers during the development.
The low current demands of his very efficient speakers may hide the inadequate current capability of the National chipamps.
I am not surprised that discrete amplifiers from some decades ago can outperform an Audiosector implementation.
They cannot drive large curents into typical 4ohms speakers.
The Audiosector circuit was developed by Peter Daniel and he makes the case very clearly that his implementation is very much aimed towards mid and treble performance.
He used ~95dB/W 4ohms speakers during the development.
The low current demands of his very efficient speakers may hide the inadequate current capability of the National chipamps.
I am not surprised that discrete amplifiers from some decades ago can outperform an Audiosector implementation.
If your supply rails are sourced from two 20,000uF caps then you'll have way too high inductance on your supplies and hence a lot of HF noise. Use an array of much smaller caps and slap the chip directly on top of the array with the shortest wires you can - it'll reduce the '2D' soundstage effect you got.
thanks guys..
I guess all in all i am a little disappointed that all that effort, energy and time was spent in building (what i thought would be THE GREATEST diy amp I ever built, and it is), was undone with a bargain basement amp costing £23.. 🙁
Atleast I learnt something about building amps in the process 😀
I guess all in all i am a little disappointed that all that effort, energy and time was spent in building (what i thought would be THE GREATEST diy amp I ever built, and it is), was undone with a bargain basement amp costing £23.. 🙁
Atleast I learnt something about building amps in the process 😀
IIRC, my lm3875 runs off 4,400uf per rail @ 22v - basic National implementation sheet stuff and has reasonable imaging and bass, but then it's used with a pair of speakers that present it with an easy load.
I've also run it from batteries.... there was a gain in performance But not enough to warrant the effort.
lm3875 is a bargain IF the builder does not go mental with 'audiophile' components and more importantly Manages His Expectations it can match or even get the better of 'Budget' amps which is why a good friend suggested it to me all the way back in 1993 AFTER I went through 3 budget amps.
I've also run it from batteries.... there was a gain in performance But not enough to warrant the effort.
lm3875 is a bargain IF the builder does not go mental with 'audiophile' components and more importantly Manages His Expectations it can match or even get the better of 'Budget' amps which is why a good friend suggested it to me all the way back in 1993 AFTER I went through 3 budget amps.
Weems' uses a similar method
Use a 1k resistor in series with the speaker. Better for 20-40w amp range.
Using a 5 or 10 Ohm resistor in place of the speaker, measure the AC mV across this load resistor. Adjust volume to get a convenient value e.g. 10 mV with a 10 Ohm calibration.
Then each mV is one Ohm.
I have an Audiosector LM4780 pcb or two sat here, which I will build sometime. I think discrete is capable of better then these chipamps. But, I think their limitation is two fold, both current and voltage capability.
I'm planning to use them in parallel, to better the current capability. I could bridge those paralleled boards together, that would help the other problem (voltage).
I also have some LME49810/11/13 (I think I recall the numbers...)
I think these predriver ICs and discrete OPs could be special (maybe with a nested driver opamp pike LME49720?)
Use a 1k resistor in series with the speaker. Better for 20-40w amp range.
Using a 5 or 10 Ohm resistor in place of the speaker, measure the AC mV across this load resistor. Adjust volume to get a convenient value e.g. 10 mV with a 10 Ohm calibration.
Then each mV is one Ohm.
I have an Audiosector LM4780 pcb or two sat here, which I will build sometime. I think discrete is capable of better then these chipamps. But, I think their limitation is two fold, both current and voltage capability.
I'm planning to use them in parallel, to better the current capability. I could bridge those paralleled boards together, that would help the other problem (voltage).
I also have some LME49810/11/13 (I think I recall the numbers...)
I think these predriver ICs and discrete OPs could be special (maybe with a nested driver opamp pike LME49720?)
Last edited:
I built an Audiosector LM3875 a few years back. It still runs strong. I'm only using the stock capacitance (almost nothing compared to "upgrades" that some people do)
I have fairly large toroids that rectify to approx 37v rails IIRC.
I've always liked these pair of mono blocks I built from day 1. They've been driving some old TLs that my Dad built in the 70's.
Anywho, by comparison, it replaced an H/K AVR 247 receiver. It's the low end model. I always thought the lm3875 let my Oppo CD player sound a little more open and wider. The biggest audible difference was in the bass though. Initially I thought this was because I built dual-mono each with it's own powersupply. Later I discovered a handicap in the H/K, it has a 40Hz high pass built in that can not be turned off! I've tried every menu.
That's what I like about the lm3875, signal in -> speakers out, and nothing more.
Next I'm going to build Tom's Modulous 86. I've slowly been working on speakers with digital crossover, which will be tri-amped by 6x LM3886 in Mod86 flavor. BTW the speakers I'm building are 95dB 8ohm. I think it will be a great amp for them.
I have fairly large toroids that rectify to approx 37v rails IIRC.
I've always liked these pair of mono blocks I built from day 1. They've been driving some old TLs that my Dad built in the 70's.
Anywho, by comparison, it replaced an H/K AVR 247 receiver. It's the low end model. I always thought the lm3875 let my Oppo CD player sound a little more open and wider. The biggest audible difference was in the bass though. Initially I thought this was because I built dual-mono each with it's own powersupply. Later I discovered a handicap in the H/K, it has a 40Hz high pass built in that can not be turned off! I've tried every menu.
That's what I like about the lm3875, signal in -> speakers out, and nothing more.
Next I'm going to build Tom's Modulous 86. I've slowly been working on speakers with digital crossover, which will be tri-amped by 6x LM3886 in Mod86 flavor. BTW the speakers I'm building are 95dB 8ohm. I think it will be a great amp for them.
He, he. And i thought the chip-amp mass delusion had ended 10 years ago. 🙂
These things can drive Lowther type high efficiency drivers in a crossoverless setup, but presented with real world loads they just crumble. No bass, no dynamics. Nada. Massively paralleling helps but at the expense of their only worthwhile quality: midrange immediacy.
Maybe it's time to move on and build some real amps.
These things can drive Lowther type high efficiency drivers in a crossoverless setup, but presented with real world loads they just crumble. No bass, no dynamics. Nada. Massively paralleling helps but at the expense of their only worthwhile quality: midrange immediacy.
Maybe it's time to move on and build some real amps.
Philfr:
Sorry, I don't mean to thread hijack, but a very quick question:
Would you please tell me the make/model of the terminals you have on TOP of your Chipamp boards? I wasn't sure anything would fit there. Does anyone know if the same terminals will fit on a Chipamp.com LM3886 board?
Thanks
Sorry, I don't mean to thread hijack, but a very quick question:
Would you please tell me the make/model of the terminals you have on TOP of your Chipamp boards? I wasn't sure anything would fit there. Does anyone know if the same terminals will fit on a Chipamp.com LM3886 board?
Thanks
Just to chime in here .. I have a Peter Daniel and a Chipamp.com, both with standard parts. I find they sound just the opposite to the OP's findings. Both sound very open and have plenty of bass. I can compare them to a Pioneer A717Mk2 and a modified Quad 306. They are driving a pair of KEF LS50 speakers connected with low capacitance cables...
Just my 2 cts.
Greetings ;-)
Just my 2 cts.
Greetings ;-)
- Status
- Not open for further replies.