Help choosing a measurement amplifier

[imix500]

Hello all,
I have been tasked with designing an impedance/FFT sweep rig using an AP System 2 at work. We have literally hundreds of very good amplifiers but they are all rental items and subject to, well, rental. So the idea is to build an amplifier for in-house measurement use. I have built several amps for listening and MI, but what I'm looking for an amp that will measure and perform really well.
The amp can be any class, but must fulfill the following:

FR from near DC to at least 20kHz, probably needs to be DC coupled.
Power output of around 10-20W into 8ohms.
Extremely low THD at specified output power.
Must fit into a 2 space rack chassis.

Candidates include:
Mini-A
LM1875
LM3886
DX Amps
41Hz Amp3,6 (Not sure if they can be DC Coupled)

I'm open to suggestions.
Thanks!

[jackinnj]

The F5 with Cviller's boards will do a very nice job -- the '1875 and '3886 are great chips but I don't think they'll meet your THD criteria. LM4702 (LME49830 etc.) do a really good job.

[Shaun]

Originally Posted by jackinnj

LM4702 (LME49830 etc.) do a really good job.

I can never remember the name of the project, but someone here has made a circuit using these chips, enclosed in the (control) loop of a high performance op amp, which reportedly takes it a step higher, performance-wise. Should fit your requirements.

[Mooly]

Might I suggest Doug Selfs Class B "Blameless" design which is quite happy run on lower supply rails.

Power output: 50 W rms into 8 ohm
Distortion: Below 0.0006% at 1 kHz and 50W/8 ohm
Below 0.006% at 10 kHz
Slew-rate: Approx. 35 V/μsec
Noise: 91 dBu at the output
EIN: 117 dBu (referred to input)
Freq Response: +0, –0.5 dB over 20 Hz–20 kHz

By using Dougs two pole compensation the 10khz THD can be pushed to 0.0015%

[Shaun]

I have built some Blameless amps. Subjectively, noise is very low at the lower frequencies; was the best I had used in my loudspeaker measurement setup. IMO, mediocre amplifiers fall short mostly in terms of noise levels in the lower frequency range.

[jcx]

the real question is what are you willing to give up in size, power, distortion performance, output protection, reliability, useability, cost, design/build time, certification...

if you can compete in most of the categories your company should be building, selling, renting out the amps you make
because the companies that are building the amps you rent have experience and in a competitive market can't charge monopoly profits - so their prices are a fair reflection of value



if you can make enough trade-offs: "diy" time is "free", buy the most recent, highest level of development integrated Si for the application, large, "ugly", low efficiency are a few categories you can make some gains re established products


if still wanting to build rather than buy then forget about "optimum Class B", or "low/no feedback" audiophoolish nonsense

wrap today's best op amps around a Class A or really heavy bias AB output stage with 60 dB or more 20KHz loop gain

at lower current demands "grounded bridge" with floating supplies boost Vout so you can use CFA chip amps - paralleled LT1210 1 A output in TO-220 CFA op amp would be one output candidate

next on my short list would be discrete CFP output stages with 2-4x V gain and composite op amp input/driver


project difficulty is critically dependent on what you mean by "8 Ohm load" do you mean:

a 8 Ohm resistive load

a range of "8 Ohm" individual dynamic drivers with complex impedance from mass-spring resonance impedance peak and 4-6 Ohm series R

any "8 Ohm" multidriver loudspeaker where passive crossovers can give large complex impedance swings and drop to <<1/2 of "8 Ohms" at some frequencies

if a complex load do you have to drive the "worst case" signal that peaks the current draw by 2-5x for a given nominal load impedance


while +/-20 V, 2.25 A will give you ~ 20 Wrms into 8 Ohms resistive load you could need 5-10x current capability depending on assumptions about the load and signal

if you can just turn down the V to keep the current in range then maybe ~60-70 W continuous dissipation Class A could work (to simplify design reserve V,I "headroom" to keep output stage devices away from saturation, starvation caused nonlinear operating regions) but in 2U that may need a fan, certainly if sizing for higher current peaks forced air cooling, switching power supply will be required in such a small space

[imix500]

Thanks for the suggestions so far. I have started making comparisons between some of the amps mentioned.

jcx,
many really good points made. I'll try to address as many as I can.
First, as this will only be for in-house use, no certification is necessary for this beyond normal common sense building practices.

I'm looking for:

Very good performance. I should clarify here. We need about 20W at 8ohms and the ability to drive 4ohms. Dips in impedance below 4ohms will certainly be a consideration. (In the case of the F5, I had already planned on populating all output devices)

Ability to handle the possibility of a faulty driver at full power

Decent reliability- 10 year life span would be nice

Size needs to be 2RU if possible, although I might be able to do 3RU

Cooling can absolutely be forced air. It's not going to live in a living room.

Cost isn't the larger issue I don't think. We will already have many of the build supplies- PSU, Chassis, Heatsinks, Etc (We have stacks of dead console PSU's)

Usability will be for measurement purposes and probably subjective listening in a repair lab environment.

Build time- I probably have about 50Hours to build it. I think being able to buy PCB's will be a factor.

As far as renting out amps we make- Even though we rent many pieces we build, I don't think amplifiers will be one of them. There's no way we could build an amp such as a LabGruppen for less than just buying a pallet or two at a time. Even the smallest amps we use, D-45's- there's no way we could build one for what we pay for them.

Thanks!

[Mooly]

I still think Dougs designs fit your requirements. PCB's available.


The Signal Transfer Company: Load-Invariant Power Amplifier

http://www.signaltransfer.freeuk.com/index.htm

[Samuel Groner]

FR from near DC to at least 20 kHz, probably needs to be DC coupled.
Power output of around 10-20 W into 8 Ohms.
Extremely low THD at specified output power.
Must fit into a 2 space rack chassis.

First get the specifications right before stepping further. "Extremely low THD" is not a meaningful specification; the solution to the problem will change a lot if "extremely low" means -80 dB or -140 dB. Also if you apply DC coupling you might need to think about DC precision. How much offset can be tolerated?

Samuel

[pinkmouse]

With those load requirements, have you thought about a KrellKlone? With forced cooling it would easily fit in a 2 or 3u case. Something like the one of the old TAC PSUs with the big heat tunnels down the centre would be perfect.